Abstract: Remote sensing image change detection is widely used in land use,urban expansion,deforestation,as well as disaster assessment etc.This paper presents a novel object-level change detection method for remote sensing images based on the combined utilization of change vector analysis and spectral gradient difference.Firstly,we utilize the entropy rate segmentation method to segment the image and consider the heterogeneity of super-pixels.By changing the number of super-pixels to obtain the multi-layer super-pixel regions with different sizes and the segmentation results are corresponding to the bi-temporal images.Afterwards,in the spectral space and gradient space,we extract the change intensity image and gradient difference image,and make a fusion with these two change information under different weights.At last,the OTSU threshold segmentation algorithm is applied to get the changed/non-changed results.Experimental results on the sets of SPOT5 multi-spectral images show that the fusion strategy can effectively integrate the advantages of the two methods utilized in our process.Compared to using only one single change information,the stability and applicability of the change detection process are improved.

Key words: change detection, super-pixel segmentation, change vector analysis, spectral gradient difference