摘要
为了应对目标跟踪过程中出现的目标小、遮挡以及形变等问题,本文实验算法采用了基于核相关滤波的二次跟踪策略。第一次粗略跟踪将颜色特征和卷积特征融合在一起,构建了一个全局核相关滤波模型。第二次精确跟踪将粒子滤波的重采样和FHOG特征结合在一起,构建了多个局部核相关滤波模型。最后,经过实验对比表明本文的二次目标跟踪算法在性能上优于比较的其他四种算法,实现了对运动目标的精准跟踪。
In order to deal with the problems of small target,occlusion and deformation in the process of target tracking,the experimental algorithm in this paper uses a secondary tracking strategy based on kernel correlation filtering.For the first rough tracking,color features and convolution features are fused together to construct a global kernel correlation filtering model.The second accurate tracking combines the resampling of particle filter with FHOG feature,several local kernel correlation filtering models are constructed.Last,the experimental results show that the performance of this algorithm is better than the other four algorithms,the accurate tracking of moving target is realized.
作者
王伟东
罗莹
王坤
吕俊峰
WANG Wei-dong;LUO Ying;WANG Kun;LV Jun-feng(Xinjiang 2nd Medical College,Karamay Xinjiang 834000)
出处
《数字技术与应用》
2021年第7期87-89,共3页
Digital Technology & Application
关键词
核相关滤波
特征融合
粒子滤波
Kernel correlation filtering
Feature fusion
Particle filter