EXPERIMENT ON PRODUCING DISPARITY MAPS FROM AERIAL STEREO IMAGES USING UNSUPERVISED AND SUPERVISED METHODS

Abstract. Recent advancement in hardware and software provides the possibility of realizing full automation stereo-image tasks. This paper investigated disparity map generation from aerial images with different methods: unsupervised method supervised methods. The datasets were stereo dense matching benchmark dataset for deep learning ISPRS 2021: Vaihingen WHU MVS/Stereo Dataset released CVPR 2020. Two neural networks: GC-net PSMnet have been trained Dataset. With methods, block matching(StereoBM) Stereo Semi-Global Matching (StereoSGM) methods OpenCV studied. We selected seven image pairs six testing evaluation. Difficulty scenes such as textureless areas, reflective surfaces, repetitive patterns also included our study. performance was compared by both visualization quantitative means. advantages disadvantages are presented.