Abstract: With the development of flying UAV platforms, sensor technologies, and corresponding data processing methods, low-altitude oblique photography becomes a powerful method to obtain true three-dimensional urban spatial information. In existing region-growing methods however, corners within urban dense point clouds are not fully grown. In this paper, we propose an improvement over the previous mesh-growing algorithms for fast surface reconstruction, a PTE (point to edge) data structure and traversing algorithm. Experimental results show that the improved algorithm is able to handle dense true three-dimensional urban point clouds, and can solve most of the incomplete growing problems in corners, with better performance than the exiting methods.