Based On The Study On The Method Of Canonical Correlation Analysis Of Remote Sensing Change Detection

Change detection in remotely sensed images is defined as the procedure of quantitatively analyzing and identifying changes occurred on the earth’s surface from remotely sensed images acquired at different times. As a key element for many applications of earth observation such as resource inventory, environment monitoring, update of fundamental geographical database, etc., change detection technique is of urgent demands and has great potential in scientific applications. To get the change information correctly from multi-temporal and multispectral remote sensing images, we focus on two key problems including production of temporal difference images and concentration of changed areas in this dissertation. We conduct the research using multivariate statistical analysis in following three aspects:1) To eliminate impact of inter-channel correlations, canonical correlation analysis (CCA) and the so-called multivariate alteration detection (MAD) method based on CCA are introduced into multi-channel change detection. According to MAD method, two multi-channel images covering the same geographic location and acquired at different times are taken as two sets of random variables, then MAD transformation is performed on these random variable sets. Correlations between channels can theoretically be removed as much as possible, so that the actual changes in all channels can be detected in the resultant difference image.2) To improve the effectiveness of the MAD method, we use autocorrelation or signal-to-noise ratio (SNR) instead of variance as a measurement for change information distribution. Multivariate statistical transformation based on maximum autocorrelation factors (MAF) or minimum noise fraction (MNF) is introduced as a post-processing step for MAD transformation. Change information can be separated from the noise to the greatest extent, which can lead to concentration of change information and production of difference images.3) Considering the disadvantage of traditional liner transformation method used in change detection applications, we apply the kernel theory to the change detection, and introduce the KMAD algorithm based on kernel canonical correlation analysis into our research. Experimental results indicate that the above methods are able to extract change information effectively from multi-temporal multi-channel remotely sensed images and improve the accuracy and the reliability of change detection.