摘要
实现AGV长期准确跟踪对AGV系统实现无人化管理,提升工作效率有着重要意义。但目前工厂AGV跟踪过程中存在的光照、背景的变化以及部分遮挡等问题会导致跟踪失败。针对这些问题,本文提出一种基于YOLO v3改进的LK光流AGV跟踪算法。该算法利用基于YOLO v3训练好的AGV检测模型对LK光流跟踪器进行初始化;利用LK光流法正向跟踪这些特征点,并预测下一帧的特征点;反向追踪,并分别计算两组特征点的相似度和特征点集生成的跟踪框的距离。当相似度或距离超出阈值范围时,调用AGV检测模型对特征点集进行修正,最终实现对AGV的有效跟踪。
It is of great significance to realize the long-term accurate tracking of AGV for the realization of unmanned management and improvement of work efficiency of AGV system.But at present,the problems of illumination,background change and partial occlusion in the process of AGV tracking will lead to the failure of tracking.To solve these problems,this paper proposes an improved LK optical flowtarget tracking algorithm based on Yolo v3.Firstly,the algorithm uses the method based on Yolo V3 trained AGV detection model initializes LK optical flowtracker,where 100 feature points in the detection frame are sampled.Then LK optical flowmethod is used to track these feature points forward and predict the feature points in the next frame.Later reverse tracking is carried out and the similarity of the two groups of feature points is calculated and the distance of the tracking frame is generated by the feature point set.When the similarity or distance exceeds the threshold range,the AGV detection model is called to modify the set of feature points,and finally the AGV is tracked effectively.
作者
陈俊廷
刘翔
CHEN Junting;LIU Xiang(School of Electric and Electronic Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《智能计算机与应用》
2020年第7期180-182,共3页
Intelligent Computer and Applications
