Change Detection in SAR Images Based on Superpixel Segmentation and Image Regression
Shangluo University,Peng Guo-Hua,Yan Wei-dong,Pan Lu-Lu,Wang Li-Ya
DOI: https://doi.org/10.1007/s12145-020-00532-y
2020-01-01
Earth Science Informatics
Abstract:Change detection (CD) is one of the most important application in remote sensing domain. The difference image (DI) generated by traditional change detection methods are sensitive to several factors, such as atmospheric condition changes, illumination variations, sensor calibration, and speckle noise, greatly affecting the detection performance. To avoid the aforementioned problem, in this paper, a novel approach based on superpixel segmentation and image regression is proposed to detect changes between bitemporal synthetic aperture radar (SAR) images. Specifically, the bitemporal images are firstly divided into a number of superpixel pairs under the guidance of segmentation result of a pre-DI. Next, each pixel in pre-event image is reconstructed utilizing its nearest neighbor to reduce the influence of noise. Then, a set of preselected unchanged sample are selected to learn the local regression model and to estimate the post-event image. After that, the final DI can be obtained by measuring the difference between estimated post-event image and the actual one. Finally, the fuzzy c-means (FCM) clustering algorithm is adopted to generate the binary change map. Adequate experiments on four SAR datasets have been tested, and the experimental results compared with the state-of-the-art methods have proved the superiority of the proposed method.