MeanShift结合粒子滤波在游泳运动目标检测中的自动跟踪研究  

Research on Automatic Tracking of MeanShift and Particle Filter in Swimming Target Detection

作　　者：程浩 CHENG Hao(Tianjin university of sport,Tianjin 301617,China)

机构地区：[1]天津体育学院,天津301617

出　　处：《自动化与仪器仪表》2023年第10期31-35,共5页Automation & Instrumentation

基　　金：天津体育学院青年科研基金(ZR-0911)。

摘　　要：随着各领域学者对跟踪技术的不断研究和创新性改进,目标跟踪被普遍应用于智能交通、体育运动等多种领域。研究基于MeanShift和粒子滤波优缺点互补的特性展开了游泳运动目标检测中的自动跟踪算法研究,提出了基于多特征融合与MeanShift的粒子滤波跟踪算法。实验结果表明,分别使用颜色、纹理和结构特征进行跟踪的算法在融合MeanShift后的重采样次数分别降低了52.81%、44.71%和46.24%,而使用多特征并融合MeanShift后重采样次数减少了44.71%,X坐标偏差减少了16.67%,Y坐标偏差减少了30.23%。实验结果证明了所提出算法提高了运动目标跟踪鲁棒性和时效性,能够使用更少数量的粒子进行稳定跟踪。With the continuous research and innovative improvement of tracking technology by scholars in various fields,target tracking is commonly used in various fields such as intelligent transportation and sports.Based on the advantages and disadvantages of MeanShift and particle filter,the research on automatic tracking algorithm in swimming target detection is carried out,and a particle filter tracking algorithm based on multi-feature fusion and MeanShift is proposed.The experimental results show that the resampling times of the algorithm using single feature for tracking are reduced by 52.81%(color),44.71%(texture)and 46.24%(structure)after merging MeanShift,while the resampling times are reduced by 44.71%,the X coordinate deviation is reduced by 16.67%,and the Y coordinate deviation is reduced by 30.23%after using multiple features and merging MeanShift.Experimental results show that the proposed algorithm improves the robustness and timeliness of moving target tracking,and can use fewer particles for stable tracking.

关 键 词：目标自动跟踪 Mean Shift 粒子滤波算法 多特征融合 游泳目标检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP277[自动化与计算机技术—计算机科学与技术]