基于YOLOv5算法的足球训练识别系统

Soccer training recognition system based on YOLOv5 algorithm

随着全民健身时代的发展,体育运动分析越来越重要。文章设计了一个基于YOLOv5目标检测算法的足球识别系统,该系统通过预先准备的足球训练集对YOLOv5训练得到初步目标检测算法模型之后,再通过K-means算法进行二次算法模型分析,对图像聚类分析改进算法对图像特征。在算法训练结束后,文章对此目标检测算法模型进行测试,测试结果表明:聚类分析后的目标算法模型对足球漏检的情况显著变少,而且对足球的实时检测效果表现良好,提升了算法识别速度和识别准确度;其mAP平均精度达到了98.6%,与原改进前提升了1%。

With the development of the era of national fitness,sports analysis is becoming more and more important.In this paper,a soccer recognition system based on YOLOv5 object detection algorithm is designed,which obtains a preliminary object detection algorithm model by training YOLOv5 through a pre-prepared soccer training set,and then conducts quadratic algorithm model analysis through K-means algorithm,and improves the image features of the image clustering analysis.After the algorithm training,the object detection algorithm model is tested,and the target algorithm model after cluster analysis has significantly fewer missed detection of football,and the real-time detection effect of football is good,which improves the algorithm recognition speed and recognition accuracy,the average accuracy of mAP reached 98.6%,which is an improvement of 1%compared with the original improvement.