基于改进YOLOv5与Deep SORT的行人多目标跟踪算法研究  

Research on pedestrian multi-target tracking algorithm based on improved YOLOv5 and DeepSORT

作　　者：赵建光[1] 范晶晶[1] 韩泽山 ZHAO Jianguang;FAN Jingjing;HAN Zeshan(Hebei University of Architecture,Zhangjiakou 075000,China)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《长江信息通信》2023年第6期29-31,36,共4页Changjiang Information & Communications

基　　金：河北建筑工程学院硕士研究生创新基金项目《基于YOLOv5s算法的吸烟行为检测研究》(XY2023024)。

摘　　要：目标跟踪算法在计算机视觉领域一直是研究的难点和热点,但是受行进中人员自身和周边因素的影响跟踪效果一直不佳。文章采用基于检测器的跟踪框架对目标进行跟踪。首先将YOLOv5s算法进行改进,为降低模型的计算量引入GhostNet轻量化模型,在添加P-CBAM注意力机制以增强检测器的有效特征提取;然后跟踪信息使用DeepSORT算法实现行人的跟踪。Target tracking algorithm has always been a difficult and hot topic in the field of computer vision,but the tracking effect is not good because of the influence of people's own and surrounding factors.This paper uses the tracking framework based on detector to track the target.Firstly,the YOLOv5s algorithm was improved,and GhostNet lightweight model was introduced to reduce the calculation amount of the model.P-CBAM attention mechanism was added to enhance the effective feature extraction of the detector.Then the tracking information uses DeepSORT algorithm to track pedestrians.

关 键 词：多目标跟踪 YOLOv5s GhostNet DeepSORT 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]