Yolo Target Detection Algorithm in Road Scene Based on Computer Vision

Abstract

Using artificial intelligence, computer vision and other technologies, cars can monitor driving conditions to ensure safety or assist driving. As a driver’s primary task, finding a parking space is equivalent to “eyes”, which need to identify cars and pedestrians in front of the road. The purpose of this paper is to study the Yolo object detection algorithm for road scenes based on computer vision. This paper expounds the research basis and importance of road object detection, as well as the status quo of traditional road object detection, and deeply studies the progress of road object detection optimization. For the analysis of related technologies, firstly, the grayscale of the image is introduced, then the binarization of the image is introduced, and finally the YOLOv4 is analyzed in detail. The software and hardware environment used in the experiment is introduced, the image collection and label labeling process in the data set used in the experiment is described, and the evaluation indicators commonly used to evaluate target detection algorithms are introduced. Through the comparison of FasterR-CNN, SSD and YOLOv4, The average accuracy rate (mAP) of YOLOv4 reaches 87.2%, which meets the requirements of autonomous vehicles.