当前位置>主页 > 期刊在线 > 计算机技术 >

计算机技术22年1期

基于 Yolov5 和 DeepSort 的视频行人识别与跟踪探究
张梦华
(安徽理工大学计算机科学与工程学院,安徽 淮南 232001)

摘  要:视频监控在信息化时代尤其是交通系统中占据重要地位,文章提出一种基于 Yolov5 和 DeepSort 在可见光环境下将行人识别和行人跟踪两大模块相结合的多目标跨镜头跟踪算法。首先使用 Yolov5 算法通过保存视频号、行人序号和位置信息给视频中行人赋予标签,得到视频中所有行人的信息;然后根据信息用 DeepSort 实现行人跟踪。经过测试和训练可以快速准确地完成任务,有一定的理论探索意义和实用价值。


关键词:Yolov5;DeepSort;行人识别;行人跟踪



DOI:10.19850/j.cnki.2096-4706.2022.01.024


中图分类号:TP391.4                                      文献标识码:A                                      文章编号:2096-4706(2022)01-0089-04


Exploration of Video Pedestrian Recognition and Tracking Based on Yolov5 and DeepSort

ZHANG Menghua

(College of Computer Science and Engineering, Anhui University of Science & Technology, Huainan 232001, China)

Abstract: Video surveillance plays an important role in the informatization age, especially in traffic system. This paper proposes a multi-target cross-shot tracking algorithm, which combines two modules of pedestrian recognition and pedestrian tracking in the visible light environment based on Yolov5 and DeepSort. Firstly, Yolov5 algorithm is used to label the pedestrian in the video by saving the video number, pedestrian serial number and location information, and obtain the information of all pedestrians in the video. Then, according to the information, DeepSort is used to achieve pedestrian tracking. After testing and training, it can complete the task quickly and accurately, which has a certain theoretical exploration significance and practical value.

Keywords: Yolov5; DeepSort; pedestrian recognition; pedestrian tracking


参考文献:

[1] 宋艳艳,谭励,马子豪,等 . 改进 YOLOV3 算法的视频目标检测 [J]. 计算机科学与探索,2021,15(1):163-172.

[2] 张咏,李太君,李枚芳 . 利用改进的背景差法进行运动目标检测 [J]. 现代电子技术,2012,35(8):74-77.

[3] 杨阳,唐慧明 . 基于视频的行人车辆检测与分类 [J]. 计算机工程,2014,40(11):135-138.

[4] SUN S J,HAYNOR D,KIM Y M. Motion estimation based on optical flow with adaptive gradients [C]//Proceedings 2000 International Conference on Image Processing (Cat. No.00CH37101). Vancouver:IEEE,2002:852-855.

[5] 王亮,胡卫明,谭铁牛 . 人运动的视觉分析综述 [J]. 计算机学报,2002(3):225-237.

[6] 侯跃恩,李伟光 . 时间连续贝叶斯分类目标跟踪算法 [J].计算机工程与设计,2016,37(8):2125-2131.

[7] DAVID G L. Distinctive Image Features from Scale-Invariant Keypoints [J].International Journal of Computer Vision,2004,60(2):91-110.

[8] 杨陈晨,顾国华,钱惟贤,等 . 基于 Harris 角点的 KLT跟踪红外图像配准的硬件实现 [J]. 红外技术,2013,35(10):632-637.

[9] HARRIS C,STEPHENS M. A Combined Corner and Edge Detector [C]//Proceedings of the 4th Alvey Vision Conference. Manchester:Alvety Vision Club,1988:147-151.

[10] KASHIF M,DESERNO T M,HAAK D. Feature description with SIFT,SURF,BRIEF,BRISK,or FREAK? A general question answered for bone age assessment [J].Computers in Biology and Medicine,2016,68(C):67-75.

[11] 梁锡宁,杨刚,余学才,等 . 一种动态模板匹配的卡尔曼滤波跟踪方法 [J]. 光电工程,2010,37(10):29-33.


作者简介:张梦华(1996—),女,汉族,山西临汾人,硕士在读,研究方向:计算机视觉。