A CNN-BASED CHANGE DETECTION METHOD FOR SQUATTER STRUCTURE RECOGNITION FROM AERIAL IMAGES AND DSM

Abstract. Squatter structures have been a serious threat to human safety and health for long time. And monitoring their changes is important facilitate government management of squatters. However, existing methods are still not automatic, accurate fast enough meet the actual needs practical applications. In this paper, we propose novel deep learning-based method detect squatter structure from bi-temporal remotely sensed (RS) images digital surface models (DSMs). The proposed convolutional neural network (CNN) takes advantages spectral information high resolution image height DSM, so as more accurately in type structures. Moreover, create data set learning model training, covering variety Hong Kong. Compared with three representative methods, Our performs best, Kappa 0.6786 0.6458 detection results two test regions, respectively, which indicates that it has application potential.