基于改进Kalman滤波与Camshift结合的嵌入式跟踪系统设计  

Design of embedded tracking system based on improved Kalman filter combined with Camshift

作　　者：邱晓欢[1,3] 郑尚坡 刘俊峰 徐诗康[2] 廖丁丁 Qiu Xiaohuan;Zheng Shangpo;Liu Junfeng;Xu Shikang;Liao Dingding(Guangzhou Railway Polytechnic,Guangzhou,Guangdong 510430,China;School of Automation Science and Engineering,South China University of Technology;Joint Laboratory of Energy Saving and Intelligent Maintenance for Modern Transportations)

机构地区：[1]广州铁路职业技术学院,广东广州510430 [2]华南理工大学自动化科学与工程学院 [3]现代交通节能控制和智能运维技术联合实验室

出　　处：《计算机时代》2023年第11期41-45,51,共6页Computer Era

基　　金：广东省普通高校重点领域专项(新一代信息技术)(2021ZDZX1136)。

摘　　要：针对传统Camshift算法难以在运动目标受遮挡的情况下有效跟踪的问题,提出一种基于改进Kalman滤波与Camshift相结合的目标跟踪算法,通过对Kalman滤波器的状态变量Xk增加高维特征宽高比a与高度h参数,并将P0、Q、R与窗口高度h相关联,同时在受遮挡时自适应改变Xk参数,使Kalman滤波器能够代替Camshift算法输出足够大的跟踪框标注受遮挡目标位置。实验表明,改进算法在有效帧率方面提高了42.6%,且平均BH距离降低了0.27,显著提高了跟踪的准确性和鲁棒性。Aiming at the problem that traditional Camshift algorithm is difficult to track moving targets effectively when they are obscured,a target tracking algorithm based on the combination of improved Kalman filter and Camshift is proposed.The state variable Xkof Kalman filter is added with high dimensional feature aspect ratio a and height h parameters,the window height h is associated with P0,Q,and R,and the parameter Xkis adaptively changed in the case of occlusion,which make Kalman filter able to replace Camshift algorithm in outputting a sufficiently large tracking box to mark the position of the occluded target.Experimental results show that the effective frame rate of the improved algorithm is increased by 42.6%,and the average BH distance is reduced by 0.27,which significantly improves the tracking accuracy and robustness.

关 键 词：目标跟踪 CAMSHIFT算法 KALMAN滤波 目标遮挡 状态向量 嵌入式系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]