Influence of Fog Weather on Automotive Vision Target Detection

The target detection based on vision will be seriously affected by bad weather (rain, snow, fog, etc.), which causes safety of the intended functionality (SOTIF) problems. The implementation of autonomous driving systems still faces huge challenges. It is of great significance to quantify and evaluate the influence of bad weather on automotive vision target detection. This study focuses on the analysis of the influence of fog weather on automotive vision target detection. The fog weather simulation platform is built in 51 Sim-one simulation software, with which the image data of different targets, different distances, and different scenarios are collected. The numbers of the scale-invariant feature transform (SIFT) feature points and the oriented FAST and rotated BRIEF (ORB) feature points in the rectangular region where the target is located are taken as the indicators to quantify the influence of fog weather on target detection. The results show that the SIFT and ORB feature points decrease by more than 50% under all fog visibilities when the distance is from 5m to 15m. The extraction of SIFT feature points has better adaptability to fog weather. While the overall number of ORB feature points is much higher than the number of SIFT feature points, which is more conducive to target feature matching. The proposed evaluation scheme can reliably quantify the influence of fog weather on automotive vision target detection, which is of great significance to solving the SOTIF problem of the autonomous driving system and improving the safety of autonomous driving in fog weather.