Design and implementation of road target detection and tracking system based on YOLOv8
Zuowen Chen, Yuyao Yang, Weisu Li, Yahong Ma, Zhe Liu, Jing Li, Yuhong Xing · 2024
In the field of intelligent driving, cars have higher and higher requirements for the precision and speed of road condition detection. To solve this problem, this paper adopts the YOLOv8 algorithm, which has higher speed improvement and precision improvement in target detection algorithm than previous generations of algorithms. At the same time, this paper integrates the data enhancement algorithm into the original YOLOv8 network model to further improve the generalization and recognition ability of the model, with an average precision improvement of 94% at [email protected] and an average time of less than 0.1s for single-frame image recognition. However, the training and prediction strategies of YOLOv8 require the modification of complex configuration files, which brings challenges to YOLO novice developers; large-scale data in the training set and real-time road data need to be manually labeled one by one for classification, which increases the labor cost and burden. In order to solve these two problems, this paper develops a visual road target detection and tracking system based on PyQt5 and OpenCV, which helps developers to efficiently complete the training of YOLOv8, automatically complete the detection and recognition of actual road conditions, and can store data, improving the development efficiency of the model and the practicality of the YOLOv8 algorithm in the field of intelligent driving.