Google Earth Engine is a cloud-based, global-scale geospatial analysis platform that makes full use of Google Earth’s rich data resources and cloud computing power to store and process petabyte-level data, being an effective and convenient tool for remote sensing research. Based on the introduction of Google Earth Engine system architecture, the authors firstly sorted out the research fields of Google Earth Engine. 291 related articles on CNKI and Web of Science published from 2011 to 2019 were analyzed, and some results were concluded such as publication time, research field, research area, the first author’s institution and journal of the article. Then the authors analyzed Google Earth Engine’s application and research trends of land use and land cover. The authors found that Google Earth Engine is widely used in the field of land cover remote sensing information extraction and has advantages in global or large-scale study. Based on the advantages of Google Earth Engine in remote sensing information extraction, the authors divided the study fields into agricultural remote sensing mapping, vegetation extent mapping and dynamic monitoring, building extraction, hydrological information extraction and land cover classification mapping. The research and application progress of Google Earth Engine was elaborated from two aspects: large-area mapping and multi-temporal dynamic monitoring. Finally, the authors discussed the Google Earth Engine’s problems and the development potential in land use and land cover. This paper is intended to serve as a basis for further understanding the advantages, application status, trends and potential of Google Earth Engine as well as for further understanding and using Google Earth Engine in the future.

With the fast development and wide application of remote sensing technology, remote sensing images with higher quality are needed. However, it is difficult to directly acquire high-resolution, multispectral remote sensing images. To obtain high-quality images by integrating the information from different imaging sensors, pansharpening technology emerged. Pansharpening is an effective method used to obtain multispectral images with high spatial resolution. Many scholars have studied this method and obtained fruitful achievements. In recent years, deep learning theory has developed rapidly and has been widely applied in pansharpening. This study aims to systematically introduce the progress in pansharpening and promote its development. To this end, this study first introduced the traditional, classical pansharpening methods, followed by commonly used remote sensing satellites. Then, this study elaborated on the pansharpening methods based on deep learning from the perspective of supervised learning, unsupervised learning, and semi-supervised learning. After that, it described and analyzed loss functions. To demonstrate the superiority of the pansharpening methods based on deep learning and analyze the effects of loss functions, this study conducted remote sensing image fusion experiments. Finally, this study presented the future prospects of the pansharpening methods based on deep learning.

With a special geographical location and abundant marine resources, Zhoushan is the first prefecture-level city composed of islands in China. Therefore, the acquisition of dynamic information on the coastline is of great significance to this area. However, the large amount of suspended sediments, the tortuous coastline, the numerous tidal flats and some other factors have brought a lot of challenges to coastline extraction and the analysis of the spatial-temporal dynamics in Zhoushan Islands. In order to solve this problem, the authors have developed a method for extracting coastline remote sensing information based on the tasseled cap transformation and used long time series satellite remote sensing data to carry out the analysis of the temporal and spatial evolution of the coastline. The experimental results show that the proposed method can effectively remove the influence of suspended sediments, winding coastline and shoals on the extraction of coastline information, and make its position accurate. From 2000 to 2018, the total length of the coastline of Zhoushan Islands increased by about 327.36 km, the average growth length was 18.19 km, the average growth rate was 0.72%, the total area of Zhoushan Islands increased by about 112.26 km², the average growth area was 6.24 km², and the average growth rate was 0.49%. The constructions of reclamation and marine projects seem to have been the main reasons for Zhoushan’s coastline changes. This study is of great significance for improving the accuracy of coastline remote sensing information extraction as well as coastal development and protection in complex marine environments.

The quantitative characterization of water body color can provide important reference data for the comprehensive water quality assessment of inland lakes and reservoirs. The Guanting Reservoir is a large inland lake in North China. Based on FUI inversion using the seasonal-scale Sentinel-2 and Landsat 8 OLI reflectance data during 2016—2020, this study quantitatively analyzed the heterogeneous characteristics of Forel-Ule Index (FUI) of the Guanting Reservoir on the spatial, intra-annual, and inter-annual scales. To explore the coupling relationship between the FUI and the nutrient status of the water body, models were built using both hue angle α and FUI and the trophic status index (TSI). Moreover, this study demonstrated the comparability of FUI among different sensors and its application potential. The results are as follows. ① On the spatial scale, the FUI value was low at the center but high on the edge of the reservoir. ② On the seasonal scale within a year, the FUI value showed a trend of reaching the highest in winter, slightly decreasing in spring, reaching the lowest in summer, and rising again in autumn. ③ On the interannual scale, the FUI value in the latest three years was lower than that in the first two years during 2016—2020 and the water color changed accordingly from yellowish brown to yellowish green. These may be attributable to the effective governance of the Guanting Reservoir by the Beijing Municipal Government. ④ The Pearson correlation coefficient between TSI and α and that between TSI and FUI were -0.85 and 0.80, respectively, indicating a strong correlation between FUI and TSI. ⑤ The FUI values obtained through the inversion based on the Sentinel-2 and Landsat 8 OLI images of the same day were very approximate and were 13.04 and 13.16, respectively. This indicates that FUI is comparable between the images from different sensors. Therefore, the inversion of FUI can be achieved using the long time-series remote sensing data from multiple sensors. Meanwhile, FUI possesses notable application potential and advantages in the assessment of water quality and trophic status.

The Yellow River basin is an important ecological safety barrier, an agglomeration area of resource and energy, and an area with highly intensive production activities in China. Therefore, its ecological change directly affects the sustainable development of the ecological environment and economy in the basin. This paper aims to quantitatively assess the ecological vulnerability and analyze the spatial heterogeneity in the Yellow River basin. To this end, an evaluation system was established using the partition-integration assessment method by selecting indicators such as water resources, climate, soil, vegetation, and human activities. Meanwhile, a multiplication model was introduced. The assessment results are as follows. The overall ecological environment in the basin is moderately vulnerable, with moderately vulnerable areas accounting for 42.37% of the total area of the basin. Meanwhile, the areas with a highly vulnerable ecological environment in the basin are mainly distributed in the urban economic belt along the upper mainstream of the Yellow River. From 2000 to 2018, the ecological vulnerability of the basin first decreased and then increased. During this period, ecological problems were the most notable in 2000 and ecological vulnerability was the lowest in 2015, with the Comprehensive Vulnerability Index (CVI) of 2.28 and 2.00, respectively in 2000 and 2015. The ecological vulnerability and its evolution trend in the basin significantly varied in space. In detail, the ecological vulnerability notably increased in the plateau areas in the upper reaches, slightly changed in the urban belt along the river, and significantly decreased in the middle and lower reaches.

Remote-sensing-based time series data of vegetation have been increasingly available with the accumulation of remote sensing data. These data are vital for ascertaining the changes in an ecosystem and analyzing relevant driving factors. However, some factors (e.g., cloud cover and instrument errors) restrict the observation quality of the vegetation products of remote sensing, creating data gaps in continuous and high-quality observation data. The data gaps can be filled based on the spatio-temporal dependence of the earth’s surface characteristics, which is called the spatio-temporal reconstruction of time series data. High-quality spatio-temporal reconstruction of time series data is an important prerequisite for the accurate extraction of changes in time series data. Taking the remote-sensing-based time series data of vegetation indices as examples, this study briefly reviewed the widely used reconstruction methods of time series data firstly. These methods generally include two steps: interpolation and smoothing. The interpolation can be divided into three major types, namely time-based, space-based, and spatio-temporal interpolation. Then, taking the simulated normalized vegetation index (NDVI) time series and actual GIMMS NDVI time series as examples, different proportions of data gaps in the two time series were created. Then, this study compared the effects of four types of data reconstruction methods (i.e., linear interpolation, singular spectrum analysis (SSA), Whittaker, and time series harmonic analysis (HANTS)) on the reconstruction results of the two time series. The results show that the four methods have their own advantages and disadvantages, and the Whittaker method showed relatively good performance overall. However, the performance of interpolation methods might vary within different regions, and thereby the data reconstruction methods need to be further verified.

With the substantial improvement in spatial resolution, spectral resolution, temporal resolution, and data coverage, domestic satellites have been widely used in natural resources supervision and geological surveys. Taking ZY-1 02C, GF-1, GF-2, and ZY-3 satellites as examples, this paper explores and studies their interpretation applications in basic geography, basic geology, land resources, mineral resources development, hydrogeology, engineering geology, and geological disasters. Furthermore, this paper compares, assesses, and summarizes the ability of the domestic satellites in the interpretation of geological survey elements. All these will provide guiding suggestions and scientific references for more extensive and in-depth applications of domestic satellites.

Hongyashan Reservoir is located in northwestern China where water resources are lacking. Reservoirs are an important support for the ecosystem in this region. Analyzing the changes in the area of the reservoir can effectively help Minqin County Government to make overall plans for water ecological protection and restoration as well as rational use of water resources and can also provide support for its decision-making. Through the extraction and analysis of the water area and vegetation coverage of the Landsat series data and GF-2 data from 2000 to 2019 and in combination with the surrounding meteorological data and the collection of local data, the authors comprehensively analyzed the influencing factors of the water area change and explored the spatial and temporal changes of the water area as well as the driving force. The results show that, on the whole, the water area of Hongyashan Reservoir has continued to increase in the past 20 years, the total area has increased by 8.98 km², and the area change rate is as high as 42.6%, and that, in terms of monthly changes, the change in water area has an inverted “normal distribution” curve. The trend is that the wet season is mainly concentrated in March and September-October in the spring and autumn seasons, and the dry season is mainly concentrated in June in the summer. In terms of interannual variability, the water area is greatly affected by the seasons, so it is divided into spring, summer, autumn and winter. Interannual analysis shows that the water area in spring and winter continues to rise, with average annual growth rates of 5.03% and 5.22%, the lowest average annual growth rate in autumn is only 2.42%, and the average annual growth rate of summer water area is 22.19%, which is the season with the largest variation amplitude, exhibiting “V” fluctuation and rising. According to the meteorological data such as temperature, precipitation and evaporation, the correlation analysis of vegetation coverage and water area, and the analysis of related hydrological data, the following conclusions can be drawn: the direct driving forces are the change in precipitation, the increasing project expansion, and the change of runoff into the reservoir, whereas the indirect driving forces include changes in temperature, changes in vegetation coverage, the industrial, agricultural and domestic water use, and the restoration of the ecological environment.

With the inland waters in Chongming Island, Shanghai as the study area, this study researched the color changes of waters and the identification of suspected polluted waters using unmanned aerial vehicle (UAV) hyperspectral remote sensing images. First, reflectance calibration was carried out for the radiance signals detected by the hyperspectral sensor carried by UAVs. Compared with on-site observations, this calibration method was more accurate, the average unbiased absolute percentage differences of various bands were 13.34% on average and the average determination coefficient R² was 0.83. Afterward, the inversion of hue angle and apparent visible wavelength (AVW) was conducted using the hyperspectral reflectance of the inland waters according to the CIE-XYZ color space and weighted harmonic mean. Then an inversion model of water quality parameters was constructed based on measured data, and the water colors in the study area were classified by setting the threshold of hue angle. As indicated by the results, there exist many abnormal yellowish-brown inland waters in the Chongming Island in dry seasons and it is necessary to strengthen the supervision and governance of the aquatic environment of major shipping rivers. Finally, the suspected polluted inland waters were identified and analyzed by integrating the inversion results of the parameters of water color and water quality. This study shows that UAV hyperspectral remote sensing can be used to obtain the inversion results with high temporal-spatial continuity of the parameters of water color and water quality, which will provide credible technical support for the aquatic environment investigations of inland waters while saving costs.

Deep learning technology provides technical means for hyperspectral image classification due to its unique advantages in deep mining of features. However, in the pixel-level feature classification of hyperspectral images, the number of deep learning layers is limited due to the influence of the sample input size, and the depth features in the hyperspectral images cannot be fully mined. The classification of hyperspectral image based on feature fusion of residual network is proposed in this paper. First, the principal component analysis (PCA) method is used to extract the first principal component in the original hyperspectral image, and the residual network is used to effectively extract the spatial spectrum features of the ground objects; then the feature map is expanded by the deconvolution algorithm, and after deconvolution, features of different dimensions are fused with multi-scale features to fully mine the depth feature information in the hyperspectral image, thus further improving the classification accuracy of the hyperspectral image. The ground feature classification experiment was conducted on the two areas of Taihu Lake in Jiangsu and Chaohu Lake in Anhui captured by the “Zhuhai-1” satellite. The results show that, compared with other methods, this method can effectively solve the problem of insufficient depth feature extraction in hyperspectral image classification, thus showing better classification performance.

The ZY1-02D satellite is the first hyperspectral operational satellite in China. To test the application ability of ZY1-02D hyperspectral loading data in geological and mineral survey, the authors identified lithologic and mineral information on the basis of data pre-processing, and the results were compared with GF-5 data. The application ability of the data was analyzed effectively in combination with the results of field survey. The results are as follows: the coincidence degree of ZY1-02D hyperspectral data reflectivity spectrum curve and geological body spectrum curve is high in shape, which can meet the requirements of rock and mineral information identification; through the identification of rock and mineral information in combination with the geological and mineral data of the study area, it is shown that the lithological information of marble, monzogranite, calcite and dolomite and alteration mineral information of chlorite and limonite are consistent with the measured results. The results show that the data has good recognition effect on the information of rocks and minerals, and can provide data guarantee for the application of hyperspectral technology in the field of geology and mineral exploration.

The rapid and accurate estimation of soil water content at different spatial and temporal scales is key research content in the fields of hydrology, environment, geology, agriculture, and climate change. However, it is still a challenge to obtain accurate soil water content presently. In the past, the traditional point-based soil sampling and analysis methods were time-consuming and laborious. By contrast, retrieving soil water content using remote sensing images has the advantages of a wide range, high timeliness, low cost, and strong dynamic contrast. In hyperspectral remote sensing, soil water content is related to the wavelength range of soil reflectance. So far, many methods have been used to describe the relationships between soil water content and hyperspectral remote sensing. This paper summarized existing methods for estimating soil water content based on hyperspectral reflectance and divided them into four categories: spectral reflectance methods, function methods, model methods, and machine learning methods. Moreover, this paper compared and analyzed the potential and limitations of different methods in terms of accuracy, complexity, auxiliary data requirements, operability under different modes, and the dependence on soil types. Finally, this study put forward corresponding suggestions for future research on the relationships between soil water content and hyperspectral reflectance.

The extraction of impervious surface (IS) is very important for the development of cities, and linear spectral mixture analysis is commonly adopted to calculate the fraction of IS in the mixed pixel to improve the extraction of the urban IS at the subpixel scale. Owing to errors in the spectra of pure pixels selected from remote sensing images, incorrect fractions of different land cover types often emerge after unmixing. In this paper, the modified endmember selection was proposed to improve the accuracy of the spectral information of endmembers. Sentinel-2A images were applied to selected endmembers to get the spectral, which was used to modify the spectral information of the endmembers from Landsat8. In addition, the optimization scheme of LSMA results in which the normalized differential vegetation index (NDVI) and dry bare-soil index (DBSI) thresholds are used to optimize the mixing results was applied to improve the accuracy of LSMA results. With the WorldView-2 remote sensing image for sample verification, the results showed that the accuracy of IS fraction extracted by the method in this paper was 20% higher than that of the traditional method, providing reliable theoretical support for endmember selection and IS extraction.

The researches on the effects of band parameters on the biophysical parameters of vegetation estimation using the normalized difference vegetation index (NDVI) have great significance for the improvement in the application accuracy of NDVI in vegetation dynamic monitoring. Based on the hyperspectral images of artificial grassland obtained from a Resonon, Inc. Pika XC2 Hyperspectral Imaging Camera loaded by an unmanned aerial vehicle (UAV), this study analyzes the effects of the positions and width of red and near-infrared bands on NDVI and assesses the sensitivity of NDVI to fractional vegetation cover and the estimation accuracy. The results are as follows. When band positions were fixed, the width expansion of red and near-infrared bands had little effects on NDVI and its sensitivity, and the accuracy of fractional vegetation cover estimated using narrowband NDVI is higher than the accuracy based on broadband NDVI. When the red and near-infrared bands moved towards long waves, the NDVI and its sensitivity were affected to different extents. With an increase in the sensitivity, the anti-disturbance performance of NDVI decreased, and the estimation accuracy of fractional vegetation cover decreased. The sensitivity coefficient of narrowband NDVI and the R² determined by the linear fitting of the sensitivity coefficient and the fractional vegetation cover greatly fluctuated, and the estimated fractional vegetation cover at various locations was unstable. High estimation accuracy of fractional vegetation was obtained at different locations using the 10 nm NDVI, with the maximum R² value of 0.83. The broadband NDVI calculated using four popular satellite images can be well applied in the inversion of the fractional vegetation cover in areas with high vegetation cover. However, its inversion accuracy of fractional vegetation cover still suffered some attenuation compared with narrowband NDVI (10 nm). These results will serve as scientific references and bases for accurate inversion of vegetation parameters using NDVI.

With the rapid development of remote sensing technology, the research on the classification methods of hyperspectral remote sensing images has received widespread attention. However, existing studies on the classification of hyperspectral remote sensing images conduct image segmentation using a single-scale superpixel method. As a result, the optimal superpixel number cannot be determined, image details are liable to be omitted, and a single kernel matrix cannot characterize multiple feature information, thus leading to a decrease in the classification precision. Therefore, this study proposes to perform multiscale superpixel segmentation of the first principal component of hyperspectral images. Then it conducts hyperspectral image classification using the composite kernel obtained by coupling the multiscale spatial-spectral kernel with the original spatial-spectral kernel according to weights. Finally, it tests and analyzes the proposed method using the hyperspectral images of the National Mall in Washington, D.C. as experimental data. The test results show that the effective classification precision of this method is 6.93% higher than that of the compared methods. As proved by the results, this method can be used to effectively solve the problems such as the lack of self-adaption of image spectra and incomplete spectrum information acquired, thus significantly improving the classification accuracy of hyperspectral images.

Soil salinization is one of the important factors that affect the soil health in the arid area, so it is very important to obtain the information of soil salinity and monitor the change of soil salinity for the rational use of land resources and soil restoration in the arid area. Based on 52 soil samples collected in the field and Landsat 8 OLI remote sensing images obtained at the same time, the correlation and curve regression analysis were used to quantitatively analyze the correlation and fitting degree between the soil salinization evaluation index based on multispectral remote sensing data and the measured soil Electrical Conductivity (EC). The results are as follows: ① The soil salinity in the study area is relatively light, and the total proportion of non-salinized and slightly salinized soil samples is 82.68%; ② The correlation between salinity index and soil EC is higher than that of vegetation index. The correlation between salinity index S₃ (S₃), salinity index S₅ (S₅), salinity index S₆ (salinity index, S₆) and salinity index Si (salinity index, SI) is above 0.50; ③ Salinity indexes S₂ (S₂), S3, S5 and Si have the highest fitting degree with soil EC in the whole sample, among which S5 has the best performance (R² = 0.41). The fitting degree of index and soil EC increases significantly with the increase of soil salinity under different salinity levels. The highest fitting degree of salinity index and soil EC is S₁ (R² = 0.73) and S₂ (R² = 0.72); ④ In the fitting model, the evaluation index and soil EC calculated based on cubic model, quadratic model and S model has a high fitting degree. This study has analyzed the applicability of various soil salinization evaluation indexes in soil salinity monitoring of Yinbei irrigation area, and the preliminary conclusions can provide reference for remote sensing monitoring of soil salinity in Yinbei irrigation area of Ningxia.

Automatic extraction of buildings from satellite remote sensing images has a wide range of applications in the development of economy and society. Due to the influence of mutual occlusion, illumination, background environment and other factors in satellite remote sensing images, it is difficult for traditional methods to achieve high-precision building extraction. This paper proposes an attention enhanced feature pyramid network (FPN-SENet) and constructs a large-scale pixel-wise building dataset (SCRS dataset) by using multi-source high-resolution satellite images and vector data to realize the automatic extraction of buildings from multi-source satellite images, and compares it with the other full convolution neural networks. The results show that the accuracy of building extracted from SCRS dataset is close to the world’s leading open source satellite image dataset, and the accuracy of Pseudo color data is higher than that of true color data The accuracy of FPN-SENet is better than that of other full convolution neural networks. The extraction of building can also be improved by using the sum of cross entropy and Dice coefficient as the loss function. The overall accuracy of the best classification model is 95.2%, Kappa coefficient is 79.0%, and F1-score and IoU are 81.7% and 69.1% respectively. This study can provide a reference for building automatic extraction from high-resolution satellite images.

The ancient city of Pingtao, Zhengzhou City, Henan Province, was an important city in the Eastern Zhou Dynasty and has important historical value. Due to the problems of time-consuming, heavy investment and heavy workload in traditional archaeological investigations, the settlement layout and relic distribution of the old city of Pingtao are still unclear. In this study, the authors selected Corona, Google Earth historical images and aerial thermal infrared images, comparatively analyzed the tonal and texture features on images of different loads, phases and scales, and extracted the archaeological anomalous areas of the Pingtao City site and Dianjuntai site. Suspected ruins such as city walls, gates, corner platforms and rectangular building foundations were discovered, and the spatial structure of the ruins was initially reconstructed based on the identification results. The results of the study show that Corona imagery helps to identify the early appearance of the site, Google Earth historical imagery provides assistance for the detection and extraction of tiny suspected relic features, and aerial thermal infrared imagery can reveal such archeological features as indistinct burial on the ground or optical image. The research proves that the comprehensive utilization of multi-source high-score data can investigate, predict and reconstruct the distribution and spatial structure of the relics, thus providing a reference for further archaeological research and site protection.

The task of extracting buildings with high-resolution remote sensing image plays an important role in urban planning and urbanization. In view of the problems of existing deep learning extraction methods, for example, the shallow features can’t been used effectively and small target information is easily lost, this paper proposes a multi-level perceptual network. This network uses dense connection mechanism to fully extract feature information, and constructs parallel structure to retain spatial information of different feature resolution and enhance feature information of different depth and scale in order to reduce the loss of detail feature. At the same time, the ASPP module is used to obtain the information of different receptive fields and extract the deep architectural features at different scales. The experimental results show that the overall accuracy of the proposed method is 97.19%, intersection over union is 74.33% and theF1 score is 85.43% in the buildings extraction of GF-2 remote sensing image, all of which are higher than those of the traditional method and other deep learning methods. In addition, buildings with multi-source remote sensing images still have good extraction effect, which reflects the practicability of the method presented in this paper.

Remote sensing interpretation of the Siliguri Corridor, West Bengal, India was carried out based on 33 scenes of multispectral remote sensing images from GF-1 and GF-2 satellites, which cover an area of 154 814 km². As a result, the mileage, density, and distribution of highways at all levels in the Siliguri Corridor were obtained, and the overall characteristics of the transportation in the area were ascertained. Then this paper assessed the trafficability in the selected key areas using the weighted scoring method from the aspects such as landform, lithology, geologic disasters, and road conditions. Furthermore, the factors such as the variation and relative decrease rate of whole network’s efficiency (ΔE and e) of 19 pivotal nodes were calculated using the complex network theory. They can be used to characterize the importance of pivotal nodes relative to the overall trafficability of the road network. For the four most important pivotal nodes, the geological environment characteristics (i.e., important targets, slope, and engineering rock and soil masses in the peripheries of the nodes) were analyzed and potential disasters and risks were proposed.

The distribution of wind farms is an important basis for the monitoring and early warning of wind power investment, the analyses of land occupation, and the assessment of clean energy consumption capacity. Remote sensing technology serves as an effective method for extracting wind farm distribution on a large scale. As the remote sensing interpretation marks of wind farms, wind turbine towers are a kind of multi-scale targets in high-resolution images. However, their characteristics greatly differ due to the effects of image acquisition time, illumination conditions, and surface coverage. Therefore, it's difficult to automatically detect wind turbine towers in remote sensing images. Aiming at the above problems, this paper proposed an automatic detection method of wind turbine towers based on the YOLOv3 model, and the steps are as follows. Firstly, determine the sample construction conditions and the target scale of wind turbine towers according to the analyses of the remote sensing characteristics of a wind farm. Secondly, optimize the depth of the feature extraction network of the YOLOv3 model to improve the characterization capacity of multi-scale targets. Finally, suppress false detections using the DBSCAN density clustering algorithm according to the density difference between noise and wind turbine towers. The experimental results show that the proposed method exhibits superiority over the benchmark models such as Faster R-CNN and FPN. With a detection accuracy rate of 96%, a recall rate of 94%, and F1 of 95%, the proposed method has good effects for the detection of small targets in the remote sensing images with complex background.

Tidal flat topography is closely related to the structure and function of the ecosystem in intertidal wetlands. Therefore, it is significant for the analyses of tidal flat dynamics and the monitoring of the diffusion process of saltmarsh vegetation to obtain high-precision topography data. However, owing to limited ebb time, muddy tidal flats, and saltmarsh vegetation, traditional geographic observation techniques suffer the shortcomings such as low accuracy and efficiency, high cost, and limited coverage. In this study, unmanned aerial vehicle (UAV) low-altitude remote sensing was employed to obtain aerial images and their band information. Then the 3D and spectral information with precise coordinates were extracted based on the structure obtained using motion technology. They were used to construct a high-precision digital surface model (DSM) of the study area. The DSM of bare flats can be directly used as the digital elevation model (DEM) of the tidal flat. In the areas with saltmarsh vegetation, the information of red, green, and blue bands was used to calculate the visible-band vegetation index (VDVI), which was combined with field surveys to build an inversion model for vegetation height. Finally, vegetation was filtered out from the DSM using the height inversion model to obtain accurate DEM. In this way, the elevation of the vegetation zone in the tidal flat can be reflected. As indicated by the results of this study, the method that combines UAV low-altitude remote sensing with field surveys can realize precise inversion of tidal flat topography. The root mean square error (RMSE) of the topography in bare flat obtained using the method was 0.07 m and the accuracy was close to the terrestrial laser scanner (TLS). For areas with saltmarsh vegetation, the RMSE was reduced to 0.14 m and the geographical accuracy can be improved by 60% after the vegetation was filtered out. Therefore, the method is superior to traditional point cloud filtering. Overall, this study provided an inversion method of tidal flat topography based on UAV remote sensing and field surveys, which can effectively monitor large-scale natural tidal flat systems. The method can be applied to other similar natural tidal flat systems or coastal areas, providing important technological support for the protection and management of coastal tidal flat wetlands.

At present, the common methods for extracting tidal flats using remote sensing images tend to estimate tidal flat boundaries. Therefore, it is difficult to ensure high extraction accuracy. This study combined remote sensing cloud computing platform Google Earth Engine with the geographic information system (GIS) technology and selected 77 Landsat images during 1990—2021. Meanwhile, the mean high-tide line was set to the artificial coastline obtained through visual interpretation, and the mean low-tide line was determined through the fitting of the shoreline. Based on these, this study extracted the tidal flat in the Cixi section on the south bank of the Hangzhou Bay and estimated its area. Furthermore, this study analyzed the spatio-temporal changes in the area of the tidal flat. The results are as follows. During 1990—2021, the area of the tidal flat in the Cixi section on the south bank of the Hangzhou Bay was roughly maintained in the range of 20 000~24 000 hm², and the tidal flat migrated from south to north at a speed of 286.9 m·a^-1. The main driving force behind the spatial and area changes of the tidal flat was local policies.

This study aims to compare and analyze the effects of social control and industrial shutdown induced by the COVID-19 epidemic on the particulate matter and aerosol types in Wuhan City, Hubei Province. To this end, the aerosol optical depth (AOD) and fine mode fraction (FMF) data of Wuhan City from December 1, 2019 to April 30, 2020 were obtained based on the data of atmospheric particulate matter (PM₁₀ and PM_2.5) and the data from MODIS aerosol products. Then the models of four types of aerosols (urban/industrial, sand-dust, clean marine, and mixed types) were established, obtaining the following results. During the period of social control and industrial shutdown, the concentration of atmospheric particulate matter showed a downward trend owing to the reduction in anthropogenic emissions. Meanwhile, the proportion of urban/industrial aerosols also showed a downward trend, while the proportion of dry and clean marine aerosols increased to 13.4% in the period except for the Spring Festival holiday. In contrast, the atmospheric particulate matter and the aerosols of the above types showed opposite trends after the ordered resumption of work and production. Compared with the same period during 2017—2019, the concentration of atmospheric particulate matter and aerosol parameters were also lower during the continuous control and shutdown after the Spring Festival. It can be inferred that MODIS aerosol products can be used to effectively obtain the characteristics of regional aerosols and thus provide data for the monitoring and governance of the regional atmospheric environment.

Shadow is a common interference factor in remote sensing image interpretation in mountainous and hilly areas. The study of shadow detection in hyperspectral remote sensing images is helpful to removing shadow and giving full play to its advantage of hyperspectral resolution. Taking the multi-angle hyperspectral image PROBA/CHRIS as the data source, this paper tries to increase the spectral differences among three typical ground objects, namely, bright area vegetation, shadow area vegetation and water area, selects the characteristic bands by using the sequential projection algorithm (SPA), and analyzes the spectral characteristics of typical ground objects in the original band of CHRIS image and normalized difference vegetation index. Therefore, the normalized shaded vegetation index (NSVI) of the image is constructed. The reasonable threshold is set based on the step-size method, and the images are classified. The ability of NSVI to detect CHRIS shadow is evaluated from two aspects of classification accuracy and spectral difference enhancement effect. The results show that B₉ and B₁₅ can be used as the characteristic bands for constructing NSVI of CHRIS images by using SPA to select the band subset with the smallest root-mean-square error (RMSE) and discard the edge bands. CHRIS multi-angle images are classified based on NSVI threshold method. The classification accuracy of three kinds of land in each angle image is above 94%, and the total Kappa is higher than 0.89. The classification effect of 0° image is the best. The sub-images of the three classified land objects are obtained through the mask, and the spectral mean values of the sub-images are different. However, considering the standard deviation, it is found that the spectral overlap phenomenon is obvious, which indicates that NSVI can enhance the spectral differences among typical land objects and improve the separability between spectral confusion pixels. By further comparing the shadow detection effects of NSVI with NDUI and SI, it also proves the shadow detection ability of NSVI, which shows that the constructed NSVI can be applied to shadow detection of PROBA/CHRIS hyperspectral image and can provide important support for shadow removal and shadow information restoration of this image.

Gross Domestic Product (GDP) is commonly regarded as the best measure of a country's economic health. In 2020, China suffered from a relatively serious COVID-19 epidemic, which had a huge impact on economic development. This paper aims to accurately analyze the spatial and temporal variation pattern of the GDP contributed by the second and tertiary industries in Guangdong Province, China in the first quarter under the background of the epidemic. To this end, the remote sensing data of night-time light was taken as an indicator of GDP contributed by the secondary and tertiary industries (GDP 23). By combining the real-time monitoring data of the epidemic and point of interest (POI) data of Guangdong Province, the authors firstly determined that the epidemic was the factor that caused the decrease in urban total night light intensity (TNLI). Then they analyzed the fitting of various night light indices and different regression models to the GDP 23 of Guangdong Province. Based on this, the optimal index and model were selected for the spatial grid partition of GDP 23 and the comparison of GDP 23 with that in 2019. Afterward, the authors analyzed the impacts of COVID-19 on GDP 23 of Guangdong Province in the first quarter and the reasons from spatial-temporal perspectives according to the spatial simulation results of GDP 23. For the cities most affected by the epidemic, the most affected industries were obtained through the statistical analysis of POI data, aiming to scientifically guide the precise resumption of work and production in Guangdong Province. The results are as follows. The spatial distribution of GDP 23 in 2019 was highly consistent with that in 2020, and the heart of Guangdong's economic development consisted of Guangzhou, Shenzhen, Dongguan, and Foshan cities in the two years. In terms of temporal distribution, 21 cities in Guangdong Province were affected by COVID-19 at different degrees in 2020 compared to 2019. Among them, the cities with relatively developed economies were affected the most, including Shenzhen, Guangzhou, Dongguan, and Foshan. According to POI data and the spatial distribution difference of GDP 23 between 2019 and 2020, the cities having suffered the biggest economic impacts were Guangzhou and Zhongshan, where the leading industries included shopping, real estate, and enterprises and companies, while the cities with the highest increased amplitude of GDP 23 included Shaoguan and Shenzhen, where the leading industries consisted of finance, real estate, and shopping. Therefore, the provincial and municipal governments should formulate corresponding policies on the financial industry, life service industry, and shopping consumption in Guangzhou and Zhongshan cities, in order to accurately assist enterprises and companies to early resume work and production.

In order to better carry out the ecological protection and restoration of the wetland in the lower reaches of the Yangtze River, the authors selected the Landsat images of Shengjin Lake in different seasons from 2000 to 2019 as the research data with the support of Google Earth Engine (GEE) cloud platform. The land surface temperature (LST) was retrieved by a batch program using radiative transfer equation method. The spatio-temporal variation of LST and its responses to land cover in Shengjin Lake during the past 20 years were comprehensively analyzed. The results are as follows: ① From the perspective of space, the spatial distribution of different temperature grades has shown obvious differences with the seasonal changes. The high-temperature region is dispersed in spring, generally located in the northwest in summer and autumn, and mostly in the south in winter. The area of the lake varies with the seasons, but its temperature belongs to very low or low temperature grade at all seasons. ② From the perspective of time, in the past 20 years, affected by forest and water, the Shengjin Lake wetland has always been dominated by the medium and low temperature grades that account for a large proportion of 70%-85% or so. The area proportion of temperature grades varies with the time trend such as seasons and years. ③ There exist seasonal differences in the responses of LST to land cover. It is basically presented in the form of a descending order of artificial surface> cultivated land> forest and mudflat> water. ④ Non-urbanization factors have a certain impact on the surface temperature of natural wetland. The research results are certainly significant for the reasonable development of Shengjin Lake.

Aiming at the omission in the ship target detection from remote sensing images with complex background caused by the arbitrary and dense arrangement of ships, this study, based on the rotation region generation network, proposes a ship target detection algorithm using the multi-scale feature enhancement of remote sensing images. The detailed steps are as follows. Firstly, improve the feature pyramid network using the receptive field module with dense connection at the feature extraction stage. Then obtain the characteristics of multi-scale receptive fields using the convolution of different dilate rates. In this way, the expression of high-level semantic information can be enhanced. Then design a feature fusion structure based on attention mechanisms to restrain noise and highlight the target characteristics. Afterward, fuse all layers according to the spatial weight value of each layer to obtain a feature layer that takes both semantic and position information into account. Then conduct attention enhancement to the features of this layer, and integrate the enhanced features into the original feature layer in the pyramid network. Consequently, pay more attention to target locations by increasing attention loss and optimizing the attention network according to the classification and regression loss. As indicated by the experiment results of DOTA remote sensing dataset, the average precision of this algorithm is as high as 71.61%, which is higher than the latest ship target detection algorithm based on remote sensing images. In this manner, the omission in ship target detection can be effectively solved.

The spectral analysis method can be used to qualitatively and quantitatively research water quality parameters using the characteristics that the molecules or ions of substances in the solution can absorb the full spectrum of ultraviolet-visible light. It enjoys the advantages such as high detection speed, low cost, in-situ measurement, no secondary pollution, and the simultaneous online monitoring of multiple water quality parameters. Based on the statement of the theoretical basis of water quality spectrum analysis, this paper systematically analyzes the principles and characteristics of various measurement methods. By comparing domestic and foreign full-spectrum water quality online monitoring devices, this paper points out the key technological difficulties in the establishment of high-precision online inversion models of water quality parameters and further proposes the development trends of multi-parameter online monitoring systems of water quality using the spectral analysis method. Therefore, this paper will provide a reference for the research on water environment monitoring technologies and the development of instruments for water quality parameter detection based on the theories of spectral analysis.

The clarification of the dynamic change trend of cultivated land and its driving factors is an important basis for ensuring national food security, rationally developing and utilizing soil and water resources and adjusting land use structure. Taking Alar reclamation area in southern Xinjiang as an example and based on Landsat satellite remote sensing images, population, GDP and other data of seven important periods from 1990 to 2019, the authors selected the best algorithm to interpret remote sensing images by comparing the accuracy of five classification algorithms comprising SAM-CRF, ANN-CRF, MDC-CRF, MLC-CRF and SVM-CRF. Next, the characteristics of cultivated land area change, type transformation and spatial dynamic change were analyzed by using the interpretation results, and then the main driving factors, action path and intensity of cultivated land area change were discussed. The results show that the SVM-CRF algorithm has the highest classification accuracy among the five classification algorithms, with the overall accuracy of 0.95 and the Kappa coefficient of 0.94. The overall accuracy of the other four algorithms is between 0.65 and 0.89, and the Kappa coefficient is between 0.58 and 0.86. The area of cultivated land in the study area has continued to increase in the past three decades, and the net increase in cultivated land area is 729.97 km ² (312.21%). Cultivated land transfer-in and transfer-out has shown a trend of outward expansion and inward contraction, respectively. Total population, GDP, Total Investment in Fixed Assets, gross agricultural product and cotton price are the main driving factors for the change of cultivated land area,among which GDP has the greatest direct impact on the change of cultivated land area, while cotton price has the least impact. Except that GDP has a negative effect on cultivated land area, the other four factors have a positive effect on cultivated land area, and the overall performance of the five factors is a positive effect.

For the classification of remote sensing images, traditional feature extraction methods frequently ignore their intrinsic properties and the multi-scale local characteristics of the images. As a result, only a small amount of image information can be acquired. Given this, this study proposed a model of multi-scale gray level and texture feature fusion (Ms_GTSFF ) for the feature extraction of remote sensing images, and the extraction steps are as follows. Firstly, extract the gray-level features of the images at different scales. Then obtain the local texture features of the images using the local binary pattern (LBP) algorithm and meanwhile, obtain the image features of a larger receptive field using a multi-scale method. Afterward, obtain the gray-level attributes corresponding to various codes using the obtained multi-scale LBP histograms. Finally, code and fuse multi-scale feature information obtained from the above steps to constitute the Ms_GTSFF feature extraction model, to which multiple machine learning classifiers are connected for classification and recognition. Taking the aerial hyperspectral remote sensing images of Xiongan New Area (Matiwan Village) as the test dataset, the feature extraction and classification tests were performed following the data preprocessing by blocks. The classification accuracy was up to 99.44%, indicating a great improvement in the recognition capability compared with traditional methods. This verified the effectiveness of the proposed model in enhancing the feature extraction capability and improving the classification and reorganization performance of remote sensing images.

This study aims to explore the differences and characteristics of the impacts of coal mining activities at different stages on various land use types in mining areas. Taking Yijin Huoluo Banner-a major coal-producing area in China-as the study area and multi-stage Landsat remote sensing images of nearly 30 years during 1990-2019 as the main data source, this study extracted land use distribution information using the random forest classification method on the Google Earth Engine platform. Based on this as well as coal mining statistical data, this paper analyzed the characteristics of land use changes at three stages of coal mining using the intensity analysis theory. The results are as follows. ① The intensity change theory can be used to comprehensively analyze the land use change from the aspects of intervals, categories, and transformation and to more systematically exhibit the characteristics of land use changes and the impacts of human activities in the study area. These are greatly significant for the in-depth understanding of the land use change process. ② Coal mining produces different impacts on different types of land, and it primarily affects the vegetation, water areas, and bare land. ③ Coal mining imposes different impacts on various types of land at different stages. It produces slight impacts on various types of land at the initial stage. It produces increasing impacts on various types of land at the high-speed development stage, during which it mainly affects vegetation, bare land, and water areas in and around the mining area. Then the impacts decrease at the steady development stage of coal mining. The results of this study can serve the implementation of precise protection plans for different types of land at different coal mining stages and provide a scientific basis for the protection of the ecological environment in the mining area.

As an important indicator reflecting the surface ecological environment, vegetation is widely used in the study of regional resources and environmental carrying capacity. Taking Erhai Lake basin as an example and based on the Google Earth Engine remote sensing big data cloud computing platform, the authors obtained the annual maximum normalized difference vegetation index (NDVI) value of Erhai Lake basin in 1988—2018 by using nearly 455 Landsat series images with 30 m resolution. The pixel binary model was used for quantitative estimation of fractional vegetation cover (FVC), and the spatial-temporal change characteristics of FVC in Erhai Lake basin were comprehensively analyzed through the linear regression model. Additionally, the internal relationship between FVC and geological lithology was investigated. The results are as follows: (1) From 1988 to 2018, the vegetation coverage of Erhai Lake basin showed a trend of continuous fluctuation growth, with a growth rate of 0.38%/a. (2) The basin was dominated by high vegetation coverage, of which 82.54% of the regional vegetation coverage continued to be improved, whereas the area of continuous degradation accounted for only 3.27%, which was mainly distributed in the area with significant urbanization. (3) the FVC varied in different types of lithological areas, among which the highest was metamorphic rock, and the lowest was dolomite and volcanic rock.

Soil moisture (SM) plays an irreplaceable role in agricultural production, and agricultural water use, yield estimation, and drought monitoring are all closely related to SM. Therefore, it is of great significance to monitor the changes in SM. At present, the remote sensing technique is an effective tool for the monitoring of the changes in SM in large areas. Optical remote sensing is sensitive to the composition of surface vegetation, while microwaves can penetrate vegetation to obtain the information of SM under vegetation. Meanwhile, the sensitivity of synthetic aperture Radar (SAR) backscattering to the changes in SM is greatly affected by the vegetation canopy. In areas covered by vegetation, microwave remote sensing will be affected by both surface roughness and vegetation. Therefore, the joint application of optical and SAR remote sensing can well remove the impacts of vegetation and surface roughness, thus improving the inversion accuracy of SM. This paper summarizes the remote sensing models and retrieval methods commonly used in the research on the cooperative inversion of SM using optical and SAR remote sensing. Meanwhile, it proposes the difficulties in the research and the future development of the cooperative inversion.

The identification and extraction of mineralized alteration information play an important role in the ore prospecting using remote sensing technology. Taking the Beiya gold polymetallic deposit as an example, this study designed an alteration information extraction scheme using the principal component analysis technique according to Landsat8 OLI data and the spectral characteristics related to mineral alteration. Specifically, the extraction scheme consists of the removal of interference information (vegetation, water, and shadows), extraction of abnormal information, anomaly gradation, median filtering, and anomaly screening successively. According to the anomaly information extracted, as well as geological interpretation of remote sensing data (lithology and structures) and field surveys, three prospecting areas were delineated in the study area. This will provide basic data and decision-making bases for the ore prospecting in the Beiya area.

The monitoring of surface subsidence in mining areas can provide key information for local production safety protection and mining planning and management. Based on the Sentinel-1A images from September 2018 to October 2019, this study characterized the surface subsidence in the mining area of the Shadunzi Coal Mine in Hami City, Xinjiang, China using the combined small baseline subset (SBAS) and interferometric synthetic aperture radar (InSAR) analysis. The InSAR measurement results revealed a subsidence funnel with a maximum subsidence rate of about -150 mm/a to the northwest of the main shaft of the coal mine. As indicated by the displacement time series, the subsidence funnel showed a significant linear subsidence pattern from September 2018 to June 2019 and gradually stabilized thereafter. Then the surface deformation inversion was conducted using the Okada rectangular dislocation model to obtain the parameters of the working face of the coal mine. The modeling results showed that the working face had a depth of about 349.89 m, a length of about 1 001.27 m, and a width of about 211.80 m. Based on the inversion results as well as the apparent density of the coal seams, the annual mining capacity of the coal mine was estimated to be about 3.18 Mt during 2018—2019, which is consistent with the reported annual production capacity of the coal mine. This paper provides a feasible way to conduct the parameter inversion of coal mine working face under the constraints of InSAR measurements and to infer the relationship between the working face parameters and the mining capacity according to the apparent density of coal seams.

With the improvement of the spatial resolution of remote sensing images, the imaging features of ground objects have become increasingly complex. As a result, the change detection methods of remote sensing images based on texture expression and local semantics are difficult to meet the demand. To improve the change detection accuracy of high-resolution remote sensing images, this study constructed a large-scale remote sensing-based human activity change detection dataset (HRHCD-1.0) with a high resolution of 0.8~2 m. Moreover, this study designed an attention-based Siamese change detection network with a strong capability to extract contextual semantic features by introducing spatial attention and channel attention mechanisms. In the model comparative experiment, the attention-based Siamese change detection network proposed in this study increased the mean intersection over union on the validation set by 24% and showed more complete detection results compared to the models using non-attention mechanisms, effectively alleviating the problems of poor boundary, local omission, and holes of models using non-attention mechanisms. The post-processing method allows for small polygon removal, hole filling, and graphic smoothing of the detection results, improving the processing graphic effects of polygons. Furthermore, the increase in the sample size in the training of change detection significantly improves the application accuracy and generalization ability of the attention-based Siamese change detection network proposed in this study.

The information value model (IVM) is a statistical prediction method derived from information theory, which is widely used in natural hazard risk assessment. The problem as to how to formulate a suitable factor classification method to maximize the advantages of pre- single factor statistical analysis remains a key issue. In order to solve this problem, the authors processed a method of factor classification by combining symmetrical intervals. Statistical knowledge related to normal distribution was referred, the factors was pre-segmented by 1/2 standard deviation, and the intervals were merged symmetrically from outside to inside. After that, factors approximately fitting normal distribution, such as slope angel and topographic wetness index (TWI), were classified based on this method, and IVM was built, which was later used in landslide hazard susceptibility analysis in Wenchuan area. Meanwhile, 5 standard classification methods were selected and tested as comparative experiments for rationality verification, namely equal quantile (EQ) classification method, natural break (NB) classification method, geometric break (GB) classification method and standard deviation (SD) classification method. The results show that the IVM using symmetrical method as factor classification method stands out among the rests. The actual landslide area ratio in the high and extremely high-risk areas in the susceptibility map reached 80.87%, higher than that obtained by other standard classification methods. This proves that the symmetrical classification method performs well.

When a satellite is in transit, the presence of clouds or fog will cause shadows on some remote sensing images, and this accordingly directly affects the quality of image and the extraction, interpretation and recognition of the feature information. The authors firstly counted the data of 2017 MODIS11A1 in Gansu Province, and found that the data pixels values of 2017 MODIS11A1 are void to a large extent. Mainly because it is difficult for the remote sensing image to penetrate the cloud to obtain the feature information, the image pixel value is 0. Then the authors explored and compensated the missing value based on the phenological solar term as the time period, proposing the method of historical average value. After using the historical average method to compensate the data, the authors found that the effective utilization ratio of pixels could be greatly improved. The image information basically reflects the real feature information, and the compensation result can meet the demand of remote sensing images.

A farmland drought remote sensing dynamic monitoring system has been established on the Android mobile platform in order to meet the actual needs of users for observation of agricultural conditions such as farmland drought. For the problem of inefficiency in traditional manual field recording by users, the system combines the advantages of portable mobile devices and global positioning system (GPS) to realize the digital management of farmland data, and completes a set of processing flow from field data entry, processing to export. With the purpose of real-time drought dynamic monitoring, the system uses the massive remote sensing data management and powerful calculating ability advantages provided by the Google Earth Engine remote sensing cloud computing platform, utilizes multi-source remote sensing data such as Landsat, MODIS and Sentinel, applies the Flask framework to implement the Google Earth Engine platform Python service interface access scheme, and completes the function of dynamic drought monitoring for farmland, which provides users with a technical application platform for selecting the remote sensing data source, calculating the drought monitoring model and finally generating the grade thematic map of drought.

As a central city in the Huaihai Economic Zone, Xuzhou has a long way to go in terms of environmental protection. A map of land use change during 2005—2015 in Xuzhou was prepared according to the remote sensing images of this period. Based on this as well as relevant statistic yearbooks, the land use dynamic degree and land use transfer matrix of Xuzhou during 2005—2015 were calculated using the method of GIS spatial statistics. Then relevant correction coefficients were determined according to the specific conditions of the study area using the equivalent factor method, and the spatial-temporal changes in the ecosystem service values in Xuzhou were quantitatively analyzed. Meanwhile, the relationship between the land use change and ecosystem service value change was investigated. The results are as follows. ① The land use types in Xuzhou are dominated by cultivated land. During 2005—2015, the area of the cultivated land, forest land, water areas, and grassland decreased, the unused land slightly increased, and the construction land considerably increased. Meanwhile, different land use types were drastically converted. In detail, a large area of cultivated land was converted into forest land, grassland, and construction land, and the increased area of the construction land was mainly converted from the cultivated land. ② Among the second-order ecosystem services, the hydrological regulation and waste treatment services possessed the highest values, while the raw material production service showed a low value. During the study period, the value of each individual ecosystem service showed a downward trend, which led to a continuous decrease in the overall ecosystem service value of Xuzhou City. Specifically, the values of the second-order ecosystem services decreased by 2.9×10⁹ yuan in total in the ten years. For the first-order ecosystem service types, the values of cultivated land and forest land ecosystem services decreased by 1×10⁹ yuan and 1.43×10⁹ yuan, respectively, the sum of which accounted for more than 80% of the total reduced value of first-order ecosystem services. ③ The sensitivity index values of the ecosystem services to various land use types were all less than 1 in different stages, indicating an inelastic relationship between the ecosystem value coefficients and the ecosystem service values of various land use types during the study period. Therefore, the ecological value coefficients and calculation method used in this paper are reasonable and reliable, and thus the calculation results are credible.

This study proposed a K-means clustering-guided threshold-based approach to classifying the high-resolution remote sensing images obtained using unmanned aerial vehicles (UAVs). The steps of the approach are as follows. First, calculate the average silhouette of the UAV remote sensing image dataset as the optimal number of clusters in the K-means clustering. Then perform K-means clustering on the original images, and manually remove non-target areas in the initial segmentation results. Afterward, perform threshold-based segmentation and image optimization on the new objects obtained to extract objects. Finally, combine all the feature tags obtained to realize the recognition and classification of remote sensing images. The abovementioned processing steps were integrated using the MATLAB/GUI platform. Based on this, a classification processing system of UAV remote sensing images was developed. It can quickly process UAV remote sensing images and achieve semi-automatic interpretation. The accuracy of the classification results was verified, obtaining an overall accuracy of 91.09% and a Kappa coefficient of 0.88. This indicates that the approach proposed in this paper can obtain high-quality segmentation results of UAV remote sensing images.

Secchi disk depth (Z_sd) is an important parameter for describing the optical properties of water bodies. With high spatial and temporal resolution, satellite remote sensing technology has become an important method of Z_sd observation. Using the in-situ measured data and GOCI images of Jiaozhou Bay (JZB) on May 16, 2017, the authors used semi-analytical algorithms Doron11 and Lee15 to retrieve the Z_sd. It is shown that the Lee15 performed better than Doron11, with the decision coefficient of 0.976 and the root mean square error of 0.02 m between the estimated values and in-situ measured values. Selecting eight GOCI images from 8: 16 to 15: 16, the authors used Lee15 algorithm to get the spatial and temporal distribution characteristics of the diurnal variation ofZ_sd on the JZB. On the spatial distribution, the overall Z_sd level of the JZB is low (0~4 m), and gradually increases from the inside to the outside of the Bay. On the time variations, the Z_sd at the Bay mouth is obviously affected by the tides. The changes between the Bay mouth and the Bay outside are dominated by the solar zenith angle (SOLZ). The change of averageZ_sd of the JZB is mainly caused by the joint effect of the SOLZ and the tide. According to the respectively statistical analysis between the in-situ Z_sd at each sampling station and simultaneously measured other environmental factors, the change in the Z_sd of the JZB is the result of the joint action of multiple environmental factors, and has a strong positive correlation with the water depth, with correlation coefficient reaching 0.84, but it is negatively correlated with other environmental factors.

The studies on the hyperspectral inversion of salt lakes are still scarce due to the limitations of geographical conditions at present. This study explores the inversion ideas and methods of the water quality parameters of salt lakes by taking the dissolved oxygen inversion of a salt lake as an example. Based on the analyses of the hyperspectral data of the Chaerhan Salt Lake in Qinghai Province and the hyperspectral inversion technology of water quality parameters, this study determined the hyperspectral inversion model of the dissolved oxygen in the salt lake by means of waveband combination using the unique spectral information of the water body of the lake. The results show that the correlation coefficient between various wavebands of the original spectrum curve and the dissolved oxygen content was less than 0.3, while that between the band combination data in the unique spectral information of the water body and the dissolved oxygen content was greater than 0.75. According to the precision verification of the finally established band ratio model using the measured value, the inversion result of the dissolved oxygen content was roughly consistent with the measured value. It is impossible for the water quality parameters to significantly change with time owing to the relatively stable nature of the water body of the salt lake. Therefore, the verification using the measured data of November 2019 can also indicate that the waveband ratio model established based on the spectral characteristics of the salt lake enjoys high precision for a long term. Therefore, the hyperspectral inversion model can meet the precision requirements for the large-area monitoring of the dissolved oxygen in the lake area. Meanwhile, this study also proposed a new idea for the establishment of the inversion model of plateau salt lakes, which lays a foundation for the establishment of the monitoring system of plateau lakes in the future.

Urban land use (ULU) reflects urban functions and structures, and the study of ULU classification can provide guidance for the sustainable development of cities. This study conducted the ULU classification of the main urban area of Harbin City using the object-oriented and random forest methods by integrating multi-source geospatial data including Sentinel-2A remote sensing images, OpenStreetMap (OSM) data, point of interest (POI) data, and nighttime light data from the Luojia-1 satellite. The results are as follows. The overall accuracy of the first-level land use type was 86.0%, with a Kappa coefficient of 0.75. The overall accuracy of the second-level land use types was 73.9%, with a Kappa coefficient of 0.69. The introduction of POI data can significantly improve the classification accuracy of residential land, industrial and mining storage land, and educational land. Meanwhile, night light data can effectively improve the classification accuracy of commercial office land and business land. This study shows that the combination of remote sensing images with multi-source geographic data is effective for ULU classification.

In order to facilitate the low-cost, ultra-light weight and easier operation, the authors constructed a smartphone based unmanned aerial vehicle (UAV) low altitude oblique photogrammetric system by integrating DJI Phantom 4 UAV flight platform with good flight traits and Nokia 808 PureView mobile phones with good image-taking functions. In this system, relative functions of multi-camera imaging system with mobile phones were optimized, and the module design method was adopted for the system which includes the measurement of improving image quality, and the design of flight control module used for automatically image-taking control developed by open source flight control system, the design of the POS module and some other means. The integrating mode by the multi-camera system adopted as payload and flight platform was discussed, and then the working flow of the integrated system was concluded. The system was used for different applicable fields, i.e., real estate surveying, open-pit mine monitoring, and 3D reconstruction of urban buildings. The application results assessed by check points measured with field work and manual vision inspect indicate that the real-world 3D model has better texture quality, and the digital survey and mapping products, real-world 3D model and digital linear graph as well as some other means have higher geometric accuracy with centimeter level. The proposed system will be very important for boosting the development of UAV low altitude oblique photogrammetry in terms of practical demands.

In order to further study the method of obtaining high-resolution soil moisture by downscaling FY-3B soil moisture and make it more suitable for agricultural and hydrological simulation, the authors constructed a comprehensive ATI and TVI by using MODIS data in Naqu area. Combined with low resolution FY-3B soil moisture products, the coefficients of soil moisture inversion model under high resolution were obtained by using soil moisture downscaling method, and the high-resolution soil moisture was obtained. Compared with the ground observation data, the R² of the downscaling soil moisture and the measured data is above 0.4, and the RMSE is between 0.055 and 0.103 cm³/cm³, indicating that the downscaling soil moisture can better reflect the spatial distribution and change of soil moisture.

The crop planting structure consists of information such as crop species, quantity structure, and spatial distribution characteristics, and it serves as the basis for agricultural scientific management. Taking the Shijin irrigation area, Hebei Province as the study area and on the premise of not considering the optimal window period of crop time series, this study calculates and analyzes the ability of texture features in crop classification and identification based on GF-1WFV images. Meanwhile, the vegetation index is introduced into the time phase in which the classification effects based on texture features are poor, in order to make up for the defects of texture in the expression of crops. According to the comparison of the classification results of various groups, the classification accuracy of individual texture features reached greater than 80% in April and August when the crop structure is obvious but was still less than 80% in May, June, July, and September when crops are the most complex. After combining the texture features with the vegetation index, the classification results of the crops in these four months were greatly improved. In detail, the overall classification accuracy was greater than 80%, which basically meets the need for agricultural dynamic monitoring. Meanwhile, the accuracy was improved by 2.27%~9.75 % and the Kappa coefficient was increased by 0.02~0.16 compared to the individual texture features. As verified using summer maize samples, the recognition accuracy reached up to 98%, the recognition effects were relatively complete, the fragmentation degree was the minimum, and the optimal discrimination from other crop categories was achieved. Meanwhile, it also proved that the texture features based on GF-1WFV images can be applied to the extraction of the crop planting structure, especially in the months when the crop structure is relatively obvious, and they can provide some effective information for the information extraction of crops base on images.

The occurrence and development themselves of loess have recorded abundant historical information, and the macro element content of loess can accurately reflect the environmental evolution. Hyperspectral remote sensing technology enjoys the advantages of being multi-band, continuous, and high-resolution. Therefore, it can be used to detect subtle differences in soil attributes and thus provide technical support for the fast and effective acquisition of basic loess information. In this paper, the loess profile of Zaoshugou Village, Zhengzhou City is studied. Combining the hyperspectral technology, the correlation between the spectral data and the macro elements of the loess was analyzed according to smoothed original spectra, first-order differential (FD), second-order differential (SD), de-envelope (CR), and reciprocal logarithm (Log(1/R). A partial least square regression (PLSR) model was established using the wave band with a larger correlation coefficient R as the characteristic band. The main conclusions are as follows. The variations in Ga, Fe, and Mg elements in the loess profile indicate that the study area has experienced a cold dry - warm wet - cold dry climate cycle since the Middle Holocene about 5400 aBP. The reflectance spectra of the loess in different stratigraphic units show the characteristics with similar trends. However, their spectral reflectance is in the order of L_0-2>L_0-1>L_t>S_0-1>T_S. According to the method of partial least squares, the optimal inversion models of Fe₂O₃, CaO, and CaO/MgO are the PLSR model with FD spectral transformation as the independent variable, while the best inversion model of MgO is the PLSR model with CR spectral transformation as the independent variable. The optimal inversion model of Fe₂O₃, CaO, and CaO/MgO can effectively distinguish different climate zones and indicate palaeoclimate cycle changes in the region where the study area falls. The optimal inversion model of MgO can better indicate the palaeoclimate evolution law of the region where the study area falls and thus has a certain reference value.