Unsupervised Change Detection in Optical Satellite Images using Binary Descriptor

Abstract

In this paper, a novel unsupervised technique is proposed to get the change analysis of multitemporal satellite images. The proposed technique is based on the local binary similarity pattern (LBSP) concept. In this binary descriptor, inter-LBSP is used to detect the changes. In this approach, the main challenge is to calculate the threshold which is used to generate the binary feature vectors. Here, an effective solution has been found, where the neighbourhood information is used for calculation of threshold. Both images are partitioned into overlapping image blocks which are used to calculate the threshold. This calculated threshold is used to obtain binary feature vectors for each pixel of both images. To get the binary feature vectors, difference between neighbouring pixels and center pixel of each block is compared with the calculated threshold. Hamming distance is used as a similarity metric to compare the binary vectors of each image for each pixel position which gives the binary change map of changed and unchanged region. To obtain this binary change map, calculated hamming distance is compared with empirically chosen minimum similarity value. Optical satellite images acquired by Landsat satellite are used to perform the experiments. Experimental results show that the proposed method provides better results compared to earlier reported techniques like expectation maximization and kernel k-means methods