摘要
针对目前道路目标检测算法因存在网络结构复杂、计算量大而不利于在嵌入式平台部署的问题,提出了一种改进的轻量级YOLO v5s道路目标检测算法,将YOLO v5s骨干网替换为MobileNetV3进行特征提取,降低了网络的参数量和计算量。实验结果表明:在自制的常见道路目标数据集上,改进后的算法在保证较高检测精度的情况下,使模型体积减小49.3%,参数量减少50.3%,从而降低了硬件部署成本,可满足在嵌入式端部署后对道路目标检测的准确率和实时性要求。
Aiming at the problem that the current road target detection algorithm has a complex network structure and a large amount of calculation that is not conducive to deployment on an embedded platform,an improved YOLOv5s lightweight road target detection algorithm is proposed,and the YOLOv5s backbone network is replaced by MobileNetV3 for feature Extraction reduces the amount of parameters and computation of the network.The experimental results show that:on the self-made common road target data set,the improved algorithm can reduce the model volume by 49.3%and reduce the number of parameters by 50.3%while ensuring high detection accuracy.The improved algorithm reduces the cost of hardware deployment,which can meet the accuracy and real-time requirements of road target detection after deployment on the embedded side.
作者
赵浙栋
张成涛
ZHAO Zhedong;ZHANG Chengtao(School of Mechanical and Transportation Engineering,Guangxi University of Science and Technology,Liuzhou Guangxi 545006,China)
出处
《汽车零部件》
2023年第8期67-71,共5页
Automobile Parts
基金
广西创新驱动发展专项资金项目(桂科AA19182006-2)
作者简介
赵浙栋(1995—),男,硕士,研究方向为自动驾驶感知技术;通信作者:张成涛(1978—),男,博士,副教授,研究方向为新能源自动驾驶技术
