摘要
利用深度学习算法检测非机动车交通违法行为有助于加快我国交通智能化发展,保障交通安全。为此,设计一种基于改进YOLOv5算法的电动车骑手头盔佩戴自动检测方法。该方法在YOLOv5算法的基础上利用Inception卷积减少特征提取网络参数量,引入注意力机制优化目标检测结果。在自建电动车头盔数据集QCKJ-MH上的实验结果表明,该方法的识别精度均值达96.4%,检测速度达82FPS,模型大小为12.9 MB,能够精准、快速地对电动车骑手头盔佩戴情况进行识别。
The use of deep learning algorithms to detect non motorized vehicle traffic violations can help accelerate the development of intelli-gent transportation in China and ensure traffic safety.To this end,design an automatic detection method for helmet wearing of electric bike rid-ers based on the improved YOLOv5 algorithm.This method is based on the YOLOv5 algorithm and utilizes Inception convolution to reduce the parameters of the feature extraction network,introducing an attention mechanism to optimize the object detection results.The experimental re-sults on the self built electric vehicle helmet dataset QCKJ-MH show that the average recognition accuracy of this method reaches 96.4%,the detection speed reaches 82 FPS,and the model size is 12.9 MB.It can accurately and quickly identify the wearing situation of electric vehicle riders'helmets.
作者
刘超
高健
LIU Chao;GAO Jian(School of Electrical and Information Engineering,Jiangsu University,Zhenjiang 212013,China)
出处
《软件导刊》
2024年第6期143-149,共7页
Software Guide
基金
江苏省六大人才高峰项目(XXRJ-012)。
