基于稀疏约束与双线索选择的目标跟踪算法  

Target Tracking Algorithm Based on Sparse Constraint and Dual-cue Selection

作　　者：吴捷[1,2] 马小虎[2] WU Jie;MA Xiaohu(College of Information Technology,Taizhou Polytechnic College,Taizhou 225300,China;School of Computer Science and Technology,Suzhou University,Suzhou 215006,China)

机构地区：[1]泰州职业技术学院信息技术学院,江苏泰州225300 [2]苏州大学计算机科学与技术学院,江苏苏州215006

出　　处：《火力与指挥控制》2023年第2期19-25,共7页Fire Control & Command Control

基　　金：国家自然科学基金(61402310);江苏省自然科学基金(BK20141195);泰州职业技术学院重点科研项目(1821819039)。

摘　　要：为了解决现有DCF类跟踪器存在边界效应及时间滤波器退化的问题,进一步增强其在复杂场景下跟踪的准确性,基于LADCF算法,提出了一种有效的双线索目标跟踪框架,用于鲁棒视觉跟踪。将结构化稀疏约束应用到多通道滤波器,并通过提取9维HOG特征与11维CN特征构建新的跟踪线索,与原有线索协同跟踪目标。建立可靠性评估策略,在每一帧中选择合适的线索进行跟踪。在OTB-50、OTB-100基准数据集上进行了定性和定量评价,实验结果表明,所提出的方法跟踪准确度相比LADCF算法提升了2.4%,相比ECO_HC提升了3.7%,优于现有的主流跟踪算法,且跟踪速度达到21.1帧/s,可以实现实时跟踪。In order to solve the issues of boundary effect and temporal filter degradation of existing type DCF trackers,and to further enhance their tracking accuracy in complex scenes,an effective dualcue target tracking framework for robust visual tracking is proposed based on LADCF algorithm.The structured sparse constraint is applied to the multi-channel filter,and a new tracking clue is constructed by extracting 9-dimensional hog feature and 11-dimensional CN feature,which cooperates with the original clue to track the target.The reliability evaluation strategy is established,and the appropriate clue is selected for tracking in each frame.The qualitative and quantitative evaluation is carried out on OTB-50 and OTB-100 benchmark datasets.The experimental results show that the tracking accuracy of the proposed method is 2.4%,higher than LADCF algorithm and 3.7%,higher than ECO_HC algorithm.It is superior to the existing mainstream tracking algorithms,and the tracking speed reaches 21.1 frames/s,the real-time tracking can be realized.

关 键 词：双线索选择 目标跟踪 稀疏约束 可靠性评估 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]