Studies On Change Detection Methods Based On Probability Statistics For Remote Sensing Images

The detection of changes in remote sensing image is defined as the procedure of quantitatively analyzing and identifying changes occurred on the earth surface from the multi-temporal remote sensing images at different times. Such a problem has played a key role in the wide range of applications in which change detection methods can be used, such as investigations of forest resources, studies on land use/land cover dynamic detection, assessments of environment disaster, arrangements of urban growth, and monitoring of national defense, etc. Change detection technique is of urgent demands and has great potential in scientific applications.In this paper, we focus on the issues related to how to automatically extract change information rapidly and effectively from muti-tempral remote sensing images. And we have finished three main tasks as following:(1) A change detection method based on difference edge and spatial-consistency of joint probability for multi-temporal remote sensing images is proposed. Firstly, sptial-consistency of joint probability is used to locate the edge of noise pixel sets from misregistration. Then, the the edge of noise pixel sets is used to smooth the corresponding pixels in the original imags. Finaly, the smoothed images are used to create difference image with sptial-consistency image of joint probability and adaptive spatial neighborhood, and the change detection result will be obtained according to Otsu threshold. Experiments carried out on the simulated and the real multi-temporal remote sensing images show that the proposed approach is efficacious.(2) A new change detection method based on image optimalizing segmentation and spatial conditional probability fusion in multi-temporal SAR image is presented. Firstly, the original SAR images are segmented into two classes with the proposed segmentation method based on canny edge match. Then, the two segment result images are fused with conditional probability rule and a fusion image considering spatial context is obtained. Finaly, the fusion image is segmented with threshold and change detection result can be obtained.Experimental results obtained on simulated and real SAR images confirm the effectiveness of the proposed approach.(3) Another novel approach based on adaptive spatial neighborhood analysis and difference histogram for change detection from multi-temporal remote sensing image is proposed. Firstly, a new difference image based on adaptive spatial neighborhood anylysis is created. Then, the difference image is segmented according to the new threshold selected by comparing original histogram, the histogram of membership degree and the histogram of uncertainty image.Experimental results obtained on simulated and real SAR images confirm the effectiveness of the proposed approach .