The use of Siamese multi-task neural network for building change detection from VHR remote sensing Images

Abstract

Building change detection is essential to many applications such as monitoring of urban areas, land use management, and illegal building detection. This paper proposes a Siamese multi-task neural network, labeled as SMTNet, to detect building changes from high-resolution remote sensing images. We combine the advantages of the multi-task learning method and Siamese neural networks to improve the geometric accuracies of detected boundaries. We applied the proposed method to a very high-resolution (VHR) remote sensing dataset that is a GF2 image-pair in Fuzhou City. We also compared the proposed method with two other existing methods, i.e., Obj-SiamNet, STANet. Our results show that the proposed method using SMTNet performed better than the existing method from the perspective of geometric and attribute accuracies. We conclude that the proposed SMTNet method has a high potential for extracting changed boundaries of buildings from high-resolution remote sensing images.