Aimed at efficient processing of the massive image data obtained by multi-sensors in planetary exploration missions,the authors proposed a solution to automatic accurate geo-rectification of the remote sensing images by matching them to the available reference images,which are already mapped into certain coordinate frame. Considering the large resolution gap between some of the images, this paper recommended an indirect matching strategy. The authors also tackled the problem concerning the choice of the appropriate reference image according to the characteristics of the typical lunar and Mars images. Using CCD images of CE-1 and CE-2 images from Chang'E missions and HiRISE images from MRO mission,with reference images from LROC WAC and THEMIS VIS respectively,the authors conducted the experiments. The accuracy of the corrected image were evaluated with manually selected checkpoints. The results show that, by matching them with referenced images, the control points can be chosen automatically and the subsequent rectification can achieve a significant accuracy improvement compared with that without any control points. The proposed solution is effective and practical for automatic geometric processing of planetary remote sensing images.
With the development and matureness of the satellite remote sensing technology, the use of satellite images for mapping production has become more and more important. As satellite images and traditional aerial photographs are different, during the utilization of existing mapping techniques to enlarge the image to the pixel level to seek for control points, the exact pixel points and the point coordinates can't be found in a series of mixed pixels because of the division generated randomly by the sensor element and the influence of the surrounding surface features. As a result, the image can't be accurately measured at a point, which will cause aerial mapping error. Therefore, a new symbol used in digital photogrammetry is presented in this paper, which is different from the ones in relative standard, and an algorithm is put forward based on the grayness of pixels of this kind of symbol in the digital image, which can find points and simultaneously improve the single point positioning accuracy. Finally, with a GeoEye-1 stereo pair through the actual encryption work, the practicability and the accuracy of the method are verified.
The adoption of information from two images can improve resolution of images; nevertheless, the method based on interpolation is apt to make fussy boundary or cause partial loss of detailed information. To solve this problem, the authors employed an approach to super-resolution reconstruction based on staggered pixels and non-uniform B-spline surface. Two frames of images with low resolution,which were matched and geometrically registered at sub-pixel level,were adopted,and then these two images were staggered and re-sampled to double times of the origin grid;the positions with non-value were interpolated and filled with tri-B-spline surface interpolation,and non-uniform node parametric method was adopted. And the curved surface was composed of 36 known neighborhood pixels. For solving the value of interpolated points,the authors introduced parallel and golden section methods to iterate searching for the optimal value,which made interpolation more accurate; the last but not unimportant, the interpolated images were restored to reconstruct better visualization high resolution images. The assessment of the results demonstrates that this approach can considerably improve definition, quantity of information, signal-to-noise ratio and resolution.
Aimed at tackling the fast solver problem for the large-scale and nearly pathological close-range block sparse bundle adjustment normal equation,the authors propose a solution method based on the preconditioned conjugate gradient(PCG)sparse algorithm. Firstly,the normal equation coefficient matrix corresponding to diagonal matrix square root is selected as the preconditioning matrix by changing the coordinate base of the parameter vector to be estimated, which can improve the behavior of the normal equation coefficient matrix so as to achieve the purpose of improving the convergence rate of the conjugate gradient method. Then, through the application of the sparse matrix, the efficiency of storage can be improved and the adjustment of normal equation coefficient matrix can be achieved. Experiment results show that the method proposed in this paper has the advantage that any scalar change in variables has no effect on the range of convergence of the iterative technique,and hence it has not only high accuracy of calculation but also faster speed.
With its fast development,the unmanned aerial vehicle (UAV) technology has become an important method for obtaiing the remote sensing image data. Nevertheless, this flexibility,rapid acquisition method for remote sensing image has poor stability in the platform in comparison with the traditional way of large aircraft aerial potography. The acquisition process of UAV image is affected by its counterweight,real-time flight environment and other external factors,and all of these factors lead to a host of difficulties in image registration. In this paper,firstly,the authors used the POS data to estimate the overlapped area of the UAV image,utilized the Forstner operator to extract feature points,and segmented the images based on the entropy information. After that, the matching feature points were found with the rotation model based on normalized cross-correlation(NCC). Finally, the registration of the UAV images was realized. The experimental results show that the method proposed in this paper is effective and maintains a better robustness.
Edge lines extraction from remote sensing images is a classic problem, and different edge extraction algorithms are applicable to different types of images. The road shape is not very regular, the contrast is low and the impact of noise is serious in actual remote sensing image because the road might be blocked by buildings and trees, and the road edge lines are likely to be broken; therefore road edge lines extraction from high-resolution remote sensing image is always a hot research topic. In this paper,the authors propose a new method for extraction of the road lines from remote sensing image so as to solve the problem that it is difficult for the methods available to extract clear and continuous road edge lines. Firstly, the direction templates are introduced to detect the edge points and search for the sub-segments in block image; then the sub-segments are extended and the line segment voting is taken to connect straight line segments in the curved edge lines, and the edge lines whose length is greater than a given threshold are output; finally, the spur and bifurcation are removed and the union of edge lines in eight directions is taken as the final road network. Experiment results show that the method proposed in this paper can be used to extract the road edge lines which have a certain curvature and low contrast and are affected by noise seriously from high-resolution remote sensing images.
Airborne LiDAR data can be used to monitor ground collapse in the vegetation-covered area effectively. A progressive triangulation filtering DEM-construction method based on region segmentation is proposed in this paper. In this method, the raw point clouds are re-organized so as to improve the efficiency of points calculation; combined with the regional statistical value of elevation difference, the authors conducted segmentation of ground points and non-ground points according to survey area's terrain, and then used ground points to build the initial sparse TIN model. Following the calculation of the distance between other points and TIN, the authors obtained progressive encryption triangulation and extracted ground points. Finally the authors eliminated isolated points, thus generating a DEM. This method was applied to airborne LiDAR data obtained in Hunan Province. The experiment results show that the proposed method is promising. The DEM constructed by this method conveys more refined topographical information. Especially in the vegetation-covered area, the extraction of high-precision DEM can be achieved. Meanwhile, the location and range of ground collapse can be shown.
Soil moisture is a very important part of earth ecosystem and plays an important role in global water cycle. Passive microwave has advantages of all-weather and high temporal resolution, and its data processing is simple; therefore soil moisture index extracted from passive microwave data greatly promote the repeated observations of soil moisture in large areas. 8 kinds of microwave remote sensing soil moisture indices were extracted from AMSR-E data, half of which were put forward in the past and half of which were newly raised. And then their variation trends were compared with each other at Miyun and Hanzhong, the two meteorological stations, and the data obtained showed that PIV,6.9 and DIV,10.7 were respectively related to the precipitation. Afterwards, the precipitation monitoring of PIV,6.9, DIV,10.7 and MPDI10.7 at two 10 pixels×12 pixels rectangle areas, including Miyun and Hanzhong respectively, were comparatively studied. Finally, precipitation on August 21th was interpolated in the whole country, and distributions of precipitation and three soil moisture indices were comparatively analyzed, which were PIV,6.9, DIV,10.7 and MPDI10.7. The result shows that PIV,6.9 seems to be the best index for soil moisture monitoring, and also the best choice in soil moisture monitoring in China at present.
The determination of parameters for texture analysis is crucial to remote sensing image classification. In this paper, the Hongze Lake wetlands were taken as the study area and the texture was calculated based on gray level co-occurrence matrix. The effect of the window size, moving step and direction in computing texture upon the separability of freshwater lake wetlands was discussed. The classification of wetlands was carried out based on decision tree classification by using texture and spectral features. The classification accuracy was assessed based on error matrix. It is shown that the parameters of 3 pixel ×3 pixel in the direction of 90° are the optimal ones. Mean, entropy, correlation are used for the classification of wetlands in the study area. The classification accuracy is 83.24% with Kappa of 0.788. The results show that the effect of texture parameters upon the classification of freshwater lake wetlands is significant.
Land surface temperature (LST) and outgoing long wave radiation (OLR), which are commonly used in seismic monitoring, were compared and analyzed from their own characteristics and seismic applications. The analytical results of the global data show that LST and OLR at high latitudes and mid-latitudes have the consistency in spatial distribution, but show a significant difference in equatorial and low-latitude regions, and this difference is closely related to the global total cloud amount. The results of feature points selected according to the cloudiness distribution in China's mainland show that LST and OLR have poor synchronization in the region whose cloud amount is greater than 65% and show better synchronization in the region whose cloud amount is less than 65%. On such a basis, the authors selected Qinghai region where the synchronization is relatively good and mid-south China where the synchronization is poor as the test areas. The results achieved show that the spatial, temporal and intensity characteristics of two types of data can be either identical or different, as shown by the comparison between the two computing results using the vorticity method. LST mainly reflects the warming temperature phenomenon whereas OLR is focused on a comprehensive reflection of the whole earth-atmosphere system.
In order to demonstrate the correlativity of the vegetation indexes between Landsat TM and HJ CCD imageries, the authors first conducted the conversion from gray values of multi-spectrum bands to apparent reflectance for several pairs of Landsat TM and HJ CCD imageries. The quantitative relationship of normalized differential vegetation indexes between the two imageries was established through regression analysis in the light of different land cover types. At last, the conversion equation was calculated and the difference between the two kinds of data was analyzed. The result shows that vegetation indexes between Landsat TM and HJ CCD have significant linear positive correlation, and the conversion precision of the transformation equation is reasonably high.
The quadratic difference method, as an improved water vapor signal extraction algorithm, is employed in "clear sky region" from FY-2E infrared channel. By means of both split window and temporal difference calculation from infrared cloud mask images, the method can weaken the surface temperature interference and help trace the weak signal of water vapor in "clear sky region", regardless of the order of the two calculations. Application examples show that this method can trace the weak signal of water vapor in "clear sky region" more effectively and make up for the lacking wind field data in clear sky with high water vapor content values as compared with the obvious limitation of deriving cloud motion wind by the traditional method. A comparison between the wind fields using this technique and that obtained from the NCEP reanalysis data shows a good relative accuracy.