Research on Vehicle Detection Algorithm Based on Improved YOLOv4

Abstract

Abstract The check of vehicle fact is a significant technology in the intelligent control of transport brightness. A vehicle diagnosis and identification algorithm based on Yolov4 is proposed to encounter the dare of flow vehicle information monitoring technology in exactness, rate and firmness. The K-means++ gathering manner is used to get the mooring frame organize points appropriate for the data set. The CIOU loss function is used to optimize the training process, and then it’s trained through the CSPDarkNet53 network framework to improve the YOLOv4 network structure. The DenseNet module is applied instead of the feature pyramid. The 5-time convolution module in this paper simplifies the feature network. It is found that the experimental results have achieved good results, which can be able to achieve the goal in the practice.