A Study Of Change Detection Algorithms Based On Feature Analysis Of Remote Sensing Images

The technology of change detection based on remotely sensed images means that given the remotely sensed multi-temporal images, panchromatic or multi-spectral, one's aim is to detect the changes of ground feature or target by means of image processing and analyzing. This technology plays an important role in the field of civil and military affairs, and a large volume of algorithms have been put forward. Most of these algorithms were studied based on pixels, while only a few are synthetically studied from other patterns of image such as texture and shape. The aim of this paper is to realize the change detection of ground feature or target in remotely sensed images by image feature analysis. On the one hand, using image texture feature analysis by wavelet transformation is to detect change of images; on the other hand, using feature invariant of target image is to analyze the shape of targets, then extract targets and detect those changes. Main works of this paper are as follows:1) It analyzes the background and current status of change detection, points out the main problems of the existing algorithms, brings forward the range and basic ideas of this paper.2) Aiming to common process of change detection based on remotely sensed multi-temporal images, it introduces the data source and the work of image preprocessing in favor of change detection, such as geometric correction and radiometric correction. And according to the applied objects, it summarizes the existing algorithms, detailed expounded some classical methods.3) It studies an algorithm of change detection based on texture feature by wavelet. This method applies wavelet theory to extract different features from wavelet approximate constituent and detail constituents. Compared to conventional methods, this method has better effect, and has faster speed than directly extracting texture feature. So it improves change detection effect and efficiency.4) The segmentation algorithm of iterative threshold is developed, which is rather suitable to segment targets with small proportion of the image. And compared to other existing methods, this method is simple and has better effect of segmentation.5) This paper also puts forward a method of target extraction and its change detection based on multi-invariant features. It mainly aims at the isolated target, realizing the quantitative and positional analysis with fine accuracy for mutative targets.