A review of research on vehicle detection in adverse weather environments
Sheng Feng, Xueying Cai, Limin Li, Weixing Wang, Ying Shi
Abstract
Real-time vehicle detection in adverse weather environments is highly important for improving the level of vehicle intelligence and reducing the occurrence of traffic accidents. Multi-sensor fusion plays an irreplaceable role in the detection process, helping autonomous vehicles to obtain more comprehensive and reliable environmental information in adverse weather. In this paper, vehicle detection and denoising methods under adverse weather conditions using different sensors are comprehensively reviewed. First, we analyze the influence of different atmospheric particles on image quality and summarize the methods for selecting detection models for various weather conditions. Second, we discuss the advantages of different sensors and commonly used sensor fusion methods. Third, we review vehicle detection algorithms based on cameras and LiDAR, as well as the image denoising and point cloud denoising methods. Fourth, we introduce relevant datasets for vehicle detection under adverse weather. Fifth, we introduced roadside cooperative perception technology, which helps autonomous vehicles obtain richer detection information under adverse weather. Finally, we discuss the challenges and future trends of vehicle detection, including sensor fusion, real-time performance constraints, and evaluation criteria. This review covers a wide range of topics and is designed to help readers understand vehicle inspection methods, especially in inclement weather. • Analyzes atmospheric particles' impact on vehicle detection, with adaptive sensors. • Introduces multi-sensor fusion and image preprocessing for adverse weather. • Explores V2X to overcome sensor limitations through vehicle-infrastructure interaction.