基于尺度自适应和遮挡重定位机制的KCF改进算法  

Improved KCF algorithm based on scale adaptation and relocation mechanism

作　　者：孙静 刘凌 亢春燕 师超 孟雪 陈伟[3] Sun Jing;Liu Ling;Kang Chunyan;Shi Chao;Meng Xue;Chen Wei(School of Mechanical and Material Engineering,Xi'an University,Shaanxi Xi'an,710065,China;China National Petroleum Corporation Logging Co.,Ltd.,Shaanxi Xi'an,710077,China;Xi'an Institute of Applied Optics,Shaanxi Xi'an,710065,China)

机构地区：[1]西安文理学院机械与材料工程学院,陕西西安710065 [2]中石油测井技术有限公司,陕西西安710077 [3]西安应用光学研究所,陕西西安710065

出　　处：《机械设计与制造工程》2023年第5期107-111,共5页Machine Design and Manufacturing Engineering

基　　金：西安市科技计划项目高校院所科技人员服务企业项目(22GXFW0110);西安文理学院三年行动计划项目(2021XDJH16);陕西省教育厅青年创新团队建设科研计划项目(21JP106)。

摘　　要：针对传统KCF跟踪算法在实际场景中存在特征提取不准确、尺度不能自适应以及遮挡适应能力差等问题,对该算法进行了三方面的改进:一是特征融合改进,在FHOG特征和CN特征的基础上,加入边缘特征,提升了目标定位的准确性;二是尺度自适应改进,通过尺度的调整动态改变目标定位搜索框的大小,提高了算法对目标尺度变化的适应性;三是抗遮挡改进,在出现遮挡时,通过遮挡判断和重定位机制,使得算法具有一定的抗遮挡能力,实现了对目标的长时间跟踪。将改进KCF跟踪算法与传统KCF跟踪算法在OTB-50数据集下进行对比测试,结果表明,改进算法的跟踪精确度提升了55.21%,跟踪成功率提升了74.33%。The traditional KCF tracking algorithm has some application problems in the actual scene,such as inaccurate feature extraction,un-adaptive scale and low occlusion adaptability.The traditional KCF tracking algorithm is improved in three aspects:firstly it improves feature fusion.Based on the feature of FHOG and CN,edge features are added to improve the accuracy of target location.Then it improves scale adaptive.By changing the size of the target location search box dynamically through the adjustment of the scale,the target scale is improved by the adaptability of the algorithm.Thirdly,it improves anti-occlusion.Through the occlusion judgment and relocation mechanism,the algorithm has a certain anti-occlusion ability when occlusion occurred,and tracks the target for a long time.Comparing the improved KCF tracking algorithm,this traditional KCF tracking algorithm under the OTB-50 dataset shows that the tracking accuracy is improved by 55.21%and the tracking success rate is improved by 74.33%.

关 键 词：核相关滤波 尺度自适应 重定位机制 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]