Evaluation of the quality of Lidar measurements for foggy scenes

Abstract

Light Detection and Ranging (LiDAR) is a remote sensing technology. The ability to perform 3D mapping of its surroundings with good resolution using Time of Flight (TOF) information made LiDAR a promising candidate for use in autonomous vehicles. Inefficient imaging in adverse weather conditions stands as a major obstacle for its sustenance in this field. The TOF information can be visualized as TOF images and the processing of TOF images to remove fog, is a way to overcome this problem. The performance of such processing techniques can be evaluated by analyzing the quality of the TOF images generated after processing. The project uses three methods to compare LiDAR TOF images for different fog visibility values. The first method is a novel and simple approach and is named 'Foreground Pixel Co-occurrence'(FPC) method, the second method is an adaptation of a well-known method in the literature, mean structural similarity index (MSSIM). The product of these two methods (PRODUCT), is also used to analyze the quality of LiDAR measurements taken in fog. These methods have been implemented for four different scenes and are found to give consistent results.