Multitemporal Remote Sensing Images Change Detection Based On Wavelet Transform And ICA

Through change detection between multitemporal remote sensing images of thesame region, one can determine the change characters and process quantificationally forthe remote sensing images obtained from different time . This plays a very importantrole in civil economy and national defense widely, mostly being used to update the dataof resource and environment, analyze the situation of the battle field, assess the battledamage and so on. With the development of remote sensing, change detection hasbecome an important research direction of remote sensing images's application.In this paper, through the research and analysis of the traditional method of changedetection problems, we studied the approach using wavelet transform and independentcomponent analysis (ICA) on remote sensing images change detection, focusing on boththe multi-resolution and great internal information of remote sensing images. In thetime-frequency domain, wavelet transform has a unique multi-scale analysis technique,it applys the multiscale analysis on functions or signals by computing functions ofdilation and translation, etc, in which wavelet coefficients represent most of the dignalinformation. In this paper, we take the approximate coefficients as estimates of thesource signal through using the two-dimensional discrete stationary wavelet transform(2D-SWT). On this basis, we extract the change information by using the independentcomponent analysis (ICA) method, and use the multi-scale fusion algorithm to get thefinal change detection results. Experimental results confirm the effectiveness of thestudied technique.