Convolutional Neural Networks for Interpreting Unclustered Radar Data in Automotive Applications
Yung-Chi Kung, Xingyu Zhou, Heran Shen, Jin Ahn, Junmin Wang · 2023
Many of the sensors used for increasing accuracy, precision, and safety on ground vehicles like cameras and LiDAR have reduced efficacy in severe weather conditions such as heavy rain and snow. Radars are generally far less affected by these conditions and most contemporary vehicles are equipped with enough sensors to provide 360-degree coverage around the vehicle. In challenging cases where the other supporting sensors are reporting confidence values that are too low to be acceptable, it may be preferable to shift computational resources on extracting as much useful information as possible from available radar detection. Driving under these extreme conditions is dangerous so information the radar provides that the driver does not normally have access to should be presented in a non-intrusive and easily understood way. This paper introduces a framework to allow Convolutional Neural Networks to process unclustered radar detections quickly enough to operate in real-time and to output the results in a human-understandable way.