Building extraction based on multiscale segmentation

Building extraction algorithm based on multiscale segmentation using high resolution remote sensing images and LiDAR data is presented. Image objects are obtained by multiscale image segmentation. Following that, extraction is fulfilled using DSM and spectrum features of building objects. Optimized scale is selected with unsupervised method, and segmentation is optimized with multiscale method. The algorithm has a simple structure and low computation complexity. Results of experiment prove it is efficient.