Abstract: Traditional methods of change detection for remote sensing images are highly relied on high-precision relative-registration, and mostly cost many broken pieces of polygons which are redundantly or wrongly extracted. In this paper, a rapid low-precision registration dependent method for remote sensing images changing detection is extracted. The method is based on the Perceptual Hash algorithm, which uses both the spectral and texture features as the detection feature sets, makes grid partition of the images as the detection object, calculates the similarity-index of each pair of the grid using the Perceptual Hash algorithm to extract the changing area. Result shows that, the method makes effective promotion of reducing the dependent of the relative-registration precision. The accuracy of correct-extraction can stay 95%, and the incorrect-extraction rate is lower than 5%, with a registration-shift within 15 pixels, which is robust for the registration precision.

Key words: change detection, remote sensing images, spectral features, texture features, Perceptual Hash