Research Remote Sensing Image Registration Based Building Change Detection Algorithm

In remote sensing image processing,building change detection has always been a hot topic in the field of remote sensing,and plays a vital role in urban planning,urban management,identification of illegal buildings,etc.It takes many manpower and material resources to detect the changes of buildings in field investigation,and data acquisition cycle is long and makes mistakes easily.However,using remote sensing technology for building change detection not only can solve the above problems,but also improve the detection efficiency and accuracy of the change detection.In the process of acquiring remote sensing images,the remote sensing images may suffer from low overlapping rate and non-rigid distortion due to the changes of imaging angle,imaging time,etc.Moreover,remote sensing image pairs(e.g.multi-temporal or multi-viewpoint)collected in the same scene are not in the same coordinate system,and change detection cannot be directly performed.Therefore,the acquired images need to be corrected,and the remote sensing image registration can correct the two images to the same coordinate system by the feature point correspondence between the two images.Therefore,image registration is a key step of change detection,and the image registration performance directly affects the result of the change detection.The paper mainly studies the building change detection algorithm based on remote sensing image registration.The study is comprised two phases: one is the research of remote sensing image registration algorithm,and the other is the research of building change detection algorithm.The main contributions in the paper are as follows:(1)we present an unmanned aerial vehicles(UAVs)image registration method to address the inherent severe outlier caused by low image overlap ratio and non-rigid distortions.The key idea of the proposed method is to maintain an accurate alignment on overlapping areas while taking advantage of outliers to guide the formation of gridded images in non-overlapping areas;(2)A robust building change detection framework is proposed to detect multi-temporal remote sensing image pairs acquired by UAVs or satellites.The framework uses a combination of CNN features and superpixel segmentation for change detection.In this paper,60 pairs of small UAV building remote sensing images and 60 pairs of satellite building remote sensing images are used for image registration and change detection experiments.The experimental results show that the proposed registration method has the best performance compared with other several state-of-the-art methods in most cases.Due to the good registration performance,the change detection method used in this paper has also achieved good experimental results.