基于MobileVit轻量化网络的车辆检测方法  

Vehicle detection method based on MobileVit lightweight network

作　　者：熊李艳[1] 涂所成 黄晓辉[1] 余俊英 谢云驰 黄卫春[2] Xiong Liyan;Tu Suocheng;Huang Xiaohui;Yu Junying;Xie Yunchi;Huang Weichun(School of Information Engineering,East China Jiaotong University,Nanchang 330013,China;Network Information Center,East China Jiaotong University,Nanchang 330013,China;Dept.of Transport,Traffic Monitoring&Command Center of Jiangxi Provincial,Nanchang 330036,China)

机构地区：[1]华东交通大学信息工程学院,南昌330013 [2]华东交通大学网络信息中心,南昌330013 [3]江西省交通厅交通监控指挥中心,南昌330036

出　　处：《计算机应用研究》2022年第8期2545-2549,共5页Application Research of Computers

基　　金：江西省交通厅科技资助项目(2021X0011,2022X0040);国家自然科学基金资助项目(62067002,61967006,62062033);江西省自然科学基金资助项目(20212BAB202008);江西省教育厅资助项目(GJJ190317)。

摘　　要：针对车辆检测模型参数量大,以及对小目标和遮挡目标漏检问题,提出了一种基于MobileVit轻量化网络的车辆检测算法。首先,在数据预处理阶段使用GridMask图像增强方法,提升模型对遮挡车辆目标的检测性能;其次,使用基于MobileVit网络作为模型的主干特征提取网络,充分提取特征信息且使得模型轻量化;最后,在预测层网络中,使用基于PANet实现多尺度的车辆检测,提升模型对小目标车辆的检测能力。实验结果表明,该模型的平均检测精度达98.24%,检测速度达每张图片0.058 s,模型大小为136 MB,与对比算法相比综合性能更好。In view of the large amount of vehicle detection model parameters and missing detection of small targets and occluded targets,this paper presented a vehicle detection algorithm based on lightweight MobileVit.Firstly,in the data preprocessing stage,the method used GridMask image enhancement to improve the performance of occluded target detection.Secondly,the method used MobileVit network as the backbone feature extraction network of the model to fully extract the feature information and make the model lightweight.Finally,in the prediction layer network,the method used multiscale vehicle detection and recognition based on PANet network to improve the detection performance of the model for small targets.The experimental results show that the average detection accuracy of this algorithm is 98.24%,the detection speed is 0.058 s per picture,and the model size is 136 MB,compared with the comparison algorithm,the comprehensive performance is better.

关 键 词：车辆检测 MobileVit 轻量化 图像增强 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]