Abstract: Traditional spectral-based remote sensing image change detection methods usually use spectral mean, brightness, standard deviation and other features for difference calculation, but these features have a certain degree of correlation, resulting in a lot of false detection of change detection results, and the selection of threshold has a greater impact on the change detection results. The traditional detection methods based on spectral features are mostly based on pixels, and the detection results will have salt and pepper phenomenon. In order to solve these problems, an object-oriented spectral component substitution change detection method is proposed in this paper, which mainly uses the single-band spectrum of two-phase images for combination substitution, so as to highlight the change information. The change information is extracted by classification method. The experiment shows that the change detection accuracy of this method can reach 90%. About%. This method is simple and effective, and can effectively assist land use change monitoring, geographical and national conditions cover change monitoring and other production projects.

Key words:  remote sensing； component substitution, multiscale segmentation; , classification, change detection