Toward a Deep Learning Approach for Automatic Semantic Segmentation of 3D Lidar Point Clouds in Urban Areas

peer reviewedSemantic segmentation of Lidar data using Deep Learning (DL) is a fundamental step for a deep and rigorous understanding of large-scale urban areas. Indeed, the increasing development of Lidar technology in terms of accuracy and spatial resolution offers a best opportunity for delivering a reliable semantic segmentation in large-scale urban environments. Significant progress has been reported in this direction. However, the literature lacks a deep comparison of the existing methods and algorithms in terms of strengths and weakness. The aim of the present paper is therefore to propose an objective review about these methods by highlighting their strengths and limitations. We then propose a new approach based on the combination of Lidar data and other sources in conjunction with a Deep Learning technique whose objective is to automatically extract semantic information from airborne Lidar point clouds by enhancing both accuracy and semantic precision compared to the existing methods. We finally present the first results of our approach