LiDAR From the Sky: UAV Integration and Fusion Techniques for Advanced Traffic Monitoring

Abstract

Light detection and ranging (LiDAR) technology's expansion within the autonomous vehicles industry has rapidly motivated its application in numerous growing areas, such as smart cities, agriculture, and renewable energy. In this article, we propose an innovative approach for enhancing aerial traffic monitoring solutions through the application of LiDAR technology. The objective is to achieve precise and real-time object detection and tracking from aerial perspectives by integrating unmanned aerial vehicles with LiDAR sensors, thereby creating a potent Aerial LiDAR (A-LiD) solution for traffic monitoring. First, we develop a novel deep learning algorithm based on pointvoxel-region-based convolutional neural network (RCNN) to conduct road user detection. Then, we implement advanced LiDAR fusion techniques, including raw data fusion and decision data fusion, in an endeavor to improve detection performance through the combined analysis of multiple A-LiD systems. Finally, we employ the unscented Kalman Filter for object tracking and position estimation. We present selected simulation outcomes to demonstrate the effectiveness of our proposed solution. A comparison between the two fusion methods shows that raw point cloud fusion provides better detection performance than decision fusion.