Vehicle Detection in Traffic Monitoring Scenes Based on Improved YOLOV5s
Xiaomeng Liu, Jun Feng, Peng Chen
Abstract
Accurate vehicle detection plays an important role in road traffic monitoring. Aiming at the problem of false detection and missed detection caused by complex scenes and large differences in target scales, an improved vehicle detection algorithm based on YOLOv5s is proposed. Firstly, a detection layer is added to better learn the multi-level features of the vehicle, and then the Spatial Pyramid Pooling(SPP) module of the original YOLOv5s algorithm is replaced with the Atrous Spatial Pyramid Pooling(ASPP) module to increase the receptive field of images of different sizes and extract multi-scale context information. The experimental results on UA-DETRAC dataset show that the precision, recall and average accuracy of the proposed algorithm are improved compared with the original YOLOv5s algorithm, which achieves the purpose of improving the vehicle detection accuracy and reduces the phenomenon of missing and false detection of vehicles to a certain extent.