基于改进CamShift算法的纱管跟踪方法研究  

Study on bobbin tracking method based on improved CamShift algorithm

作　　者：徐健[1] 陈倩倩 黄磊 孙泽维 XU Jian;CHEN Qianqian;HUANG Lei;SUN Zewei(College of Electronic Information,Xi'an Polytechnic University,Xi'an 710048,Shaanxi,China)

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《河南理工大学学报（自然科学版）》2021年第2期152-158,共7页Journal of Henan Polytechnic University（Natural Science）

基　　金：陕西省重点研发项目(2019GY-173);陕西省高校人才服务企业项目(2019KJ-036);陕西省科技攻关研发项目(2019KJ-038)。

摘　　要：为了提高对纱管的实时精准跟踪,将Camshift算法应用于纱管分拣喂纱跟踪领域。传统MeanShift算法需要手动选择初始跟踪窗口,纱管分拣喂纱过程中单一颜色特征的跟踪算法很难满足实时跟踪要求。针对此问题,在CamShift跟踪算法基础上加入颜色识别,有效降低了误跟踪现象;加入卡尔曼预测机制,解决自动框选初始跟踪窗口的问题。通过3组对比实验可知:改进后的算法使目标的质心坐标偏差大约维持在4个像素点,验证了改进方法的实时性和精准性。In order to improve the real-time and accurate tracking of bobbins,the CamShift algorithm was applied to the field of bobbin sorting and feeding tracking.The traditional MeanShift algorithm needs to manually select the initial tracking window,which makes it difficult for the tracking algorithm with a single color feature in the bobbin sorting and feeding process to meet the real-time tracking requirements.In response to the above problems,color recognition was added to the CamShift tracking algorithm to effectively reduce the mistracking phenomenon.The Kalman prediction mechanism was added to solve the problem of automatically selecting the initial tracking window.Through three sets of comparative experiments,it could be concluded that the improved algorithm kept the deviation of the target’s center of mass coordinates at about 4 pixels,which verified the real-time and accuracy of the proposed method.

关 键 词：MEANSHIFT算法 CAMSHIFT算法 卡尔曼预测机制 目标跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]