Researches on Change Detection of Multitemporal Remote Sensing Image
|Keywords||Remote Sensing Image Multi-band KL Transform Band Combination Gray Difference Texture Difference Fusion|
The change detection technique of multi-temporal remote sensing image is getting the change information of the research objects by comparing and analyzing remote sensing images at different temporal. With the development of remote sensing technology, the change detection of multi-temporal remote sensing image has been widely used in the national economy and national defense fields, mainly for data update and use of urban resources, environmental monitoring, disaster prevention and mitigation, analysis of battlefield damage, assessment results and so on. After analyzing the change detection in detail, this paper proposes two different change detection algorithms for two different sources of remote sensing separately.Firstly, for the multi-band TM images, this paper presents a new improved change detection algorithm based on multi-band KL transform. The KL transform can compress the image data. After this transform, some minority components of image contain the useful information of the original multi-band image and the new components are unrelated with each other. The method of the optimal band combination can select a few bands but ensure the accuracy of the premise to detect the changed area. It reduces the time of computation and avoids many shortcomings of the traditional KL transform. The results show that the front few components contain the main message of the two images and the last few components reflect the difference between the two images. So the changed area can be extracted through choosing some components. Experimental results show that the method joined the information of two images, made the changed information obvious, improved the detection accuracy and was less affected by the noise.Secondly, for the high-resolution SPOT images, this paper presents a new change detection algorithm based on integration the gray difference image and texture difference image. Image difference is the most direct change detection method for change area extraction, but the gray difference only based on spectral feature is difficult in describing the structure change of an object. This paper considers the shortage of the gray difference image and combines with the structural features of the texture difference image, fuses these two images. Then, using OSTU method to select the threshold and detecting the change area. Experimental results show that the detection accuracy of the algorithm is better than the simple gray difference and the texture difference.