Evaluation of Flash LiDAR in Adverse Weather Conditions Toward Active Road Vehicle Safety

Flash light detection and ranging (LiDAR) is a cost-effective sensing technology and is seen as a potential complement to RADAR and point-cloud LiDAR sensors. This study explores the performance of a low-density 2-D flash LiDAR in various weather conditions for active road vehicle safety. The influence of different weather conditions, such as fog, rain, and clear sky, on the maximum detection range for different categories of vehicles, such as motorcycles, cars, and heavy vehicles, are studied, and the results are presented. Furthermore, some interesting observations noticed in wet conditions, namely, water splashing effects and false detections, are discussed along with the potential solutions to handle these challenges. Sensor fusion of flash LiDAR and camera along with the object detection technique is proposed to improve the detection distance in adverse weather conditions. Furthermore, the maximum detectable distance of various vehicle categories in different weather conditions is compared for you only look once (YOLO), region-based convolutional neural networks (faster R-CNNs), and single shot detector (SSD) object detection techniques. The results show that implementing sensor fusion using YOLO and R-CNN increase the detection distance of cars and heavy vehicles, and on the other hand, the object detection technique needs to be further improved for motorcycles.