Enhancing sparse LiDAR data captured on an airfield using 3D aircraft models

Abstract

During aircraft ground operations, pilots have to look out for obstacles and maintain a safe distance of separation. Failure to do so can result in collisions between aircraft and obstacles. Noticing obstacles and correctly judging distances can be quite challenging in certain situations such as in fog and in darkness. This problem can be addressed through the use of sensors that can detect obstacles even in low visibility and low illumination conditions. One such sensor is a LiDAR that can reconstruct 3-dimensional point clouds of the surroundings. There are various types of LiDARs and this paper focuses on the use of low-resolution LiDARs such as the Velodyne VLP-16. This type of LiDAR produces sparse point clouds which contain gaps between the LiDAR points corresponding to each laser. This sparseness is a problem when detecting other aircraft on the airfield because their wingtips can be missed and, hence, if the wingtips are pointing towards the ownship, the separation distance can be overestimated. The proposed solution is to prepare a database with polygon mesh models of different types of aircraft, and then fit these models onto the LiDAR data. The proposed algorithm for fitting the polygon mesh models is evaluated on scenarios with a target aircraft. These scenarios are repeated for two different LiDARs: the HDL-64E and VLP-16 LiDARs. The results show that the proposed algorithm fits the polygon mesh onto the obstacle point cloud correctly.