Research on image enhancement and vehicle detection system for adverse weather condition

Abstract

This paper aims to solve a difficult problem: how to detect vehicles in bad weather (such as fog). They came up with a system that combines image enhancement with target detection. The system uses a network called FFA(Feature Fusion Attention) to make the image clearer, and also uses an improved version of YOLOv5 algorithm to find the target, so that vehicles can be found accurately even in complex weather. The experimental results show that this system can greatly improve the clarity of the image and maintain a strong detection ability. In addition, this system can work in real time and can be extended, so it is very suitable for intelligent transportation and automatic driving.