Please wait a minute...
 
Remote Sensing for Land & Resources    2018, Vol. 30 Issue (1) : 22-29     DOI: 10.6046/gtzyyg.2018.01.04
Orginal Article |
The typical object extraction method based on object-oriented and deep learning
Yongtao JIN1,3(), Xiufeng YANG1,3, Tao GAO2, Huimin GUO2, Shimeng LIU1
1.North China Institute of Aerospace Engineering, Langfang 065000, China
2.Space Star Technology Co., LTD,Beijing 100086, China
3.Collaborative Innovation Center of Aerospace Remote Sensing Information Processing and Application of Hebei Province,Langfang 065000,China
Download: PDF(1149 KB)   HTML
Export: BibTeX | EndNote | Reference Manager | ProCite | RefWorks    
Abstract  

The object-oriented method solves the problem of segmentation of objects, divides different features into different objects and to a great extent separates the cultivated land, forest land, water, roads, buildings and other typical objects which are inseparable; nevertheless, the object oriented method for features such as shape, texture description is not comprehensive, the amount of information is not enough to support the whole classification and recognition. In this paper, a new method of combining object-oriented and deep learning is proposed, in which the Caffe framework of convolution neural network is used to study the training sample data in depth and, by grasping the texture of different objects and forming deep learning model, guides the classification of objects. The experiment shows that the new method can effectively solve the problem of the low classification accuracy.
Keywords object-oriented      deep learning      convolution neural network      object recognition     
:  TP79  
Issue Date: 08 February 2018
Service
E-mail this article
E-mail Alert
RSS
Articles by authors
Yongtao JIN
Xiufeng YANG
Tao GAO
Huimin GUO
Shimeng LIU
Cite this article:   
Yongtao JIN,Xiufeng YANG,Tao GAO, et al. The typical object extraction method based on object-oriented and deep learning[J]. Remote Sensing for Land & Resources, 2018, 30(1): 22-29.
URL:  
https://www.gtzyyg.com/EN/10.6046/gtzyyg.2018.01.04     OR     https://www.gtzyyg.com/EN/Y2018/V30/I1/22
Fig.1  GF-2 image of the selected sample area
分类方法 优点 缺点 两者结合点
面向对象方法 将不同的地物分割到不同的对象之中,对分类起了至关重要的作用,可以建立各种对象特征规则集来提取地物 对于形状、纹理特征描述得不够全面、准确,信息量还不足以支撑地物分类和识别 通过面向对象构建特征规则集可以进行前期的地物提取,并由地物对象构建深度学习所需要的数据训练集; 由于深度学习往往都是通过RGB图像进行训练,没充分考虑遥感图像多波段的特性,结果会有错误和偏差,因此,可以在面向对象通过地物特征集中进行修正
深度学习方法 可以掌握不同地物的形状、纹理、背景等特性,比较准确地区分不同地物,并按照训练结果来分割地物 一方面对于图像分类模型需要大量的数据来训练,特别对于图像分割模型来说,需要大量的标签来标识不同地物,人工标识的话工作量太大; 另一方面,深度学习的结果往往是栅格图像,最终结果很难修正
Tab.1  Advantages and disadvantages of object-oriented and in-depth learning methods
Fig.2  The flow chart of typical objects recognition technology
Fig.3  Schematic diagram of multiscale segmentation
序号 层次 类别 判断依据 规则 备注
1 大尺度分割层(分割尺度100,颜色0.7,平滑度0.5) 建筑物 亮度值 亮度>398 建筑物光谱较为高亮,方差较大
2 道路 宽度和长宽比 宽度<17.8,长宽比>12.5 呈条带状,长宽比较大
3 水体 NDWI NDWI>0.5 NDWI用来提取影像中的水体信息,效果较好,NDWI= (B2-B4)/(B4+B2)①
4 植被 NDVI NDVI>0 NDVI是植被生长状态及植被覆盖度的最佳指示因子。NDVI=(B4-B3)/(B4+B3)
5 小尺度分割层(分割尺度65,颜色0.7,平滑度0.5) 农作物 提取于植被,形状规则 边界指数>1.5,同质性<12 考虑内部父对象与子对象关系、对象的边界指数、紧致度、同质性等
6 林地 提取于植被,颜色更绿 NDVI>0.45,同质性>32
Tab.2  The rule set of objects
序号 样本名称 样本 样本数
1 建筑物 612
2 水体 202
3 道路 645
4 农作物 700
5 林地 687
Tab.3  The training sample of typical objects
Fig.4  Convolution neutral network structure
Fig.5  Imagine convolution process diagram
Fig.6  Pooling operation
Fig.7  Typical class selection result of Longhuzhuang township by Object-Oriented method
Fig.8  Typical class selection result of Longhuzhuang township by deep learning method
Fig.9  Random point distribution of Longhuzhuang township
地物类别 建筑物 农作物 林地 水体 道路 其他 分类总数
建筑物 32 1 3 1 4 1 42
农作物 2 65 22 0 0 0 89
林地 5 20 105 1 1 0 132
水体 1 0 1 12 1 0 15
道路 5 1 1 2 9 0 18
其他 0 0 0 0 1 3 4
实际总数 45 87 132 16 16 4 300
Tab.4  Confusion matrix of typical classification by object-oriented method(个)
地物类别 错分误差 漏分误差 制图精度 用户精度
建筑物 23.81 28.89 71.11 76.19
农作物 26.97 25.29 74.71 73.03
林地 20.45 20.45 79.55 79.55
水体 20.00 25.00 75.00 80.00
道路 50.00 43.75 56.25 50.00
其他 25.00 25.00 75.00 75.00
Tab.5  Commission and omission error,production and user precision by object-oriented method(%)
地物类别 建筑物 农作物 林地 水体 道路 其他 分类总数
建筑物 38 0 2 0 2 0 42
农作物 0 69 5 1 1 0 76
林地 5 18 125 0 1 1 150
水体 1 0 0 13 0 0 14
道路 1 0 0 1 12 0 14
其他 0 0 0 1 0 3 4
实际总数 45 87 132 16 16 4 300
Tab.6  Confusion matrix of typical classification by deep learning method(个)
地物类别 错分误差 漏分误差 制图精度 用户精度
建筑物 9.52 15.56 84.44 90.48
农作物 9.21 20.69 79.31 90.79
林地 16.67 5.30 94.70 83.33
水体 7.14 18.75 81.25 92.86
道路 14.29 25.00 75.00 85.71
其他 25.00 25.00 75.00 75.00
Tab.7  Commission and omission error,production and user precision by deep learning method(%)
[1] 郑毅,武法东,刘艳芳.一种面向对象分类的特征分析方法[J].地理与地理信息科学,2010,26(2):19-22.
[1] Zheng Y,Wu F D,Liu Y F.A feature analysis approach for object-oriented classification[J].Geography and Geo-Information Science,2010,26(2):19-22.
[2] 赵英时. 遥感应用分析原理与方法[M].北京:科学出版社,2003.
[2] Zhao Y S.Theory and Methods of Remote Sensing Application Analysis[M].Beijing:Science Press,2003.
[3] 黄慧萍. 面向对象影像分析中的尺度问题研究[D].北京:中国科学院研究生院遥感应用研究所,2003.
[3] Huang H P.Scale Issues in Object-oriented Image Analysis[D].Beijing:Institute of Remote Sensing Applications Chinese Academy of Sciences,2003.
[4] Bengio Y.Learning deep architectures for AI[J].Foundations and TrendsR○ in Machine Learning,2009,2(1):1-127.
[5] Deng L.An overview of deep-structured learning for information processing[C]//Proceedings of Asian-Pacific Signal & Information Processing Annual Summit & Conference (APSIPA-ASC).Xi’an:Chinese Information Processing Society of China,2011:301-313.
[6] Arel I,Rose D C,Karnowski T P.Deep machine learning-a new frontier in artificial intelligence research[research frontier][J].IEEE Computational Intelligence Magazine,2010,5(4):13-18.
[7] Do V H,Xiao X,Chng E S.Comparison and combination of multilayer perceptrons and deep belief networks in hybrid automatic speech recognition systems[C]//Proceedings of Asian-Pacific Signal & Information Processing Annual Summit & Conference(APSIPA-ASC).Xi’an:Chinese Information Processing Society of China,2011.
[8] Farabet C,Couprie C,Najman L,et al.Learning hierarchical features for scene labeling[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1915-1929.
[9] Kavukcuoglu K,Sermanet P,Boureau Y L,et al.Learning convolutional feature hierarchies for visual recognition[C]//Proceedings of the 23rd International Conference on Neural Information Processing Systems.Vancouver,British Columbia,Canada:ACM,2010:1090-1098.
[10] Charalampous K,Kostavelis I,Amanatiadis A,et al.Sparse deep-learning algorithm for recognition and categorisation[J].Electronics Letters,2012,48(20):1265-1266.
[11] Olshausen B A,Field D J.Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J].Nature,1996,381(6583):607-609.
[12] Olshausen B A,Field D J.Sparse coding with an overcomplete basis set:A strategy employed by V1?[J].Vision Research,1997,37(23):3311-3325.
[13] Goodfellow I,Courville A,Bengio Y.Large-scale feature learning with spike-and-slab sparse coding[J].arXiv preprint arXiv:1206.6407,2012.
[14] 王卫红,何敏.面向对象土地利用信息提取的多尺度分割[J].测绘科学,2011,36(4):160-161.
[14] Wang W H,He M.Multi-scale segmentation in land-use information extraction based on object-oriented method[J].Science of Surveying and Mapping,2011,36(4):160-161.
[15] Woodcock C E,Strahler A H.The factor of scale in remote sensing[J].Remote Sensing of Environment,1987,21(3):311-332.
[16] Alain G,Bengio Y,Rifai S.Regularized auto-encoders estimate local statistics[J].arXiv:1211.4246,2012:1-17.
[17] Rifai S,Bengio Y,Dauphin Y,et al.A generative process for sampling contractive auto-encoders[J]. arXiv:1206.6434,2012:2.
[18] Zhan Q M,Molenaar M,Tempfli K,et al.Quality assessment for geo-spatial objects derived from remotely sensed data[J].International Journal of Remote Sensing,2005,26(14):2953-2974.
[19] 宫鹏,黎夏,徐冰.高分辨率影像解译理论与应用方法中的一些研究问题[J].遥感学报,2006,10(1):1-5.
[19] Gong P,Li X,Xu B.Interpretation theory and application method development for information extraction from high resolution remotely sensed data[J].Journal of Remote Sensing,2006,10(1):1-5.
[20] 曹雪,柯长青.基于对象级的高分辨率遥感影像分类研究[J]. 遥感信息, 2006(5):27-30,51.
[20] Cao X,Ke C Q.Classification of high-resolution remote sensing images using object-oriented method[J]. Remote Sensing Information, 2006(5):27-30,51.
[1] XUE Bai, WANG Yizhe, LIU Shuhan, YUE Mingyu, WANG Yiying, ZHAO Shihu. Change detection of high-resolution remote sensing images based on Siamese network[J]. Remote Sensing for Natural Resources, 2022, 34(1): 61-66.
[2] GUO Xiaozheng, YAO Yunjun, JIA Kun, ZHANG Xiaotong, ZHAO Xiang. Information extraction of Mars dunes based on U-Net[J]. Remote Sensing for Natural Resources, 2021, 33(4): 130-135.
[3] FAN Yinglin, LOU Debo, ZHANG Changqing, WEI Yingjuan, JIA Fudong. Information extraction technologies of iron mine tailings based on object-oriented classification: A case study of Beijing-2 remote sensing images of the Qianxi Area, Hebei Province[J]. Remote Sensing for Natural Resources, 2021, 33(4): 153-161.
[4] FENG Dongdong, ZHANG Zhihua, SHI Haoyue. Fine extraction of urban villages in provincial capitals based on multivariate data[J]. Remote Sensing for Natural Resources, 2021, 33(3): 272-278.
[5] LIU Wanjun, GAO Jiankang, QU Haicheng, JIANG Wentao. Ship detection based on multi-scale feature enhancement of remote sensing images[J]. Remote Sensing for Natural Resources, 2021, 33(3): 97-106.
[6] WU Yu, ZHANG Jun, LI Yixu, HUANG Kangyu. Research on building cluster identification based on improved U-Net[J]. Remote Sensing for Land & Resources, 2021, 33(2): 48-54.
[7] LU Qi, QIN Jun, YAO Xuedong, WU Yanlan, ZHU Haochen. Buildings extraction of GF-2 remote sensing image based on multi-layer perception network[J]. Remote Sensing for Land & Resources, 2021, 33(2): 75-84.
[8] AN Jianjian, MENG Qingyan, HU Die, HU Xinli, YANG Jian, YANG Tianliang. The detection and determination of the working state of cooling tower in the thermal power plant based on Faster R-CNN[J]. Remote Sensing for Land & Resources, 2021, 33(2): 93-99.
[9] CAI Xiang, LI Qi, LUO Yan, QI Jiandong. Surface features extraction of mining area image based on object-oriented and deep-learning method[J]. Remote Sensing for Land & Resources, 2021, 33(1): 63-71.
[10] PEI Chan, LIAO Tiejun. Multi-scale architecture search method for remote sensing object detection[J]. Remote Sensing for Land & Resources, 2020, 32(4): 53-60.
[11] LIU Zhao, LIAO Feifan, ZHAO Tong. Remote sensing image urban built-up area extraction and optimization method based on PSPNet[J]. Remote Sensing for Land & Resources, 2020, 32(4): 84-89.
[12] ZHENG Zhiteng, FAN Haisheng, WANG Jie, WU Yanlan, WANG Biao, HUANG Tengjie. An improved double-branch network method for intelligently extracting marine cage culture area[J]. Remote Sensing for Land & Resources, 2020, 32(4): 120-129.
[13] DU Fangzhou, SHI Yuli, SHENG Xia. Research on downscaling of TRMM precipitation products based on deep learning: Exemplified by northeast China[J]. Remote Sensing for Land & Resources, 2020, 32(4): 145-153.
[14] CAI Zhiling, WENG Qian, YE Shaozhen, JIAN Cairen. Remote sensing image scene classification based on Inception-V3[J]. Remote Sensing for Land & Resources, 2020, 32(3): 80-89.
[15] Wenya LIU, Anzhi YUE, Jue JI, Weihua SHI, Ruru DENG, Yeheng LIANG, Longhai XIONG. Urban green space extraction from GF-2 remote sensing image based on DeepLabv3+ semantic segmentation model[J]. Remote Sensing for Land & Resources, 2020, 32(2): 120-129.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
京ICP备05055290号-2
Copyright © 2017 Remote Sensing for Natural Resources
Support by Beijing Magtech