Abstract: Aiming at the problem that image correlation analysis using single spectral features is poor in remote sensing change detection applications, a correlation coefficient calculation model combining texture features and spectral features is proposed. Firstly, the multi-temporal image is segmented by the same scale, and the correlation coefficients in each corresponding segmentation window are calculated. Then the coordinates of the center point of the window and the correlation coefficient value are taken as a feature point, and the spatial distribution of the correlation coefficient of the whole region is obtained by interpolation in the three-dimensional space. Finally, the change information is extracted by density slicing. The paper carries out the change information extraction experiment with two GF-1 image data. The results show that the change detection result of the combined feature correlation coefficient is obviously better than the single spectral correlation coefficient change detection result. The application research of combined correlation coefficient provides a new idea for remote sensing change information extraction using correlation analysis method.

Key words: image correlation, change detection, checkerboard segmentation, combined correlation coefficient, density slicing