尺度自适应的多特征融合相关滤波目标跟踪算法

Scale adaptation and multi-feature fusion correlation filtering object tracking algorithm

目标跟踪一直以来都是计算机视觉领域中的关键问题,核相关滤波算法(KCF)可避免在时域中进行目标跟踪,通过傅里叶变换将时域的计算转换到频域中进行,可大量简化计算,不但提高了跟踪速度,而且在跟踪精度上也有很大的提升.针对复杂条件下的目标跟踪问题,在确保算法实时性的前提下,在KCF的基础上对其特征、尺度以及模型更新机制进行3处改进:提出一种多特征融合算法,针对每种特征在不同环境下的优势,将其进行融合;提出一种分类树形尺度自适应的算法,通过树形搜索方式对目标尺度的大小进行判断,找到最佳响应位置;提出一种自适应模型更新策略的算法.实验结果表明,在公开数据集OTB-2013中算法整体的跟踪精确度达到87.4%,成功率也达到67.1%,可很好地实现复杂条件的目标跟踪,综合性能在已公开发表的跟踪算法中排名第2.尤其是在尺度变化、目标遮挡和图像模糊情况下,所提出算法的跟踪精确度和成功率排名第1.

Object tracking has always been a key issue in the field of computer vision.The kernel correlation filtering(KCF)transforms the calculation from the time domain into the frequency domain by Fourier transform,which greatly simplifies the calculation,not only improves the tracking speed,but also greatly improves the tracking accuracy.This article addresses the problem of object tracking under complex conditions.On the premise of ensuring the real-time performance of the algorithm,the features,scales and model updating mechanism are improved on the basis of KCF:This paper proposes an multi-feature fusion algorithm in order to combine all the advantages from different features;This paper proposes an adaptive algorithm of classification tree scale,which can judge the size of target scale and find the position of optimal response;This paper proposes an algorithm of adaptive model updating strategy.The experimental results show that,the public dataset OTB-2013,the average tracking accuracy and the average success rate of the proposed method are 87.4%and 67.1%respectively,which means that it can track the target under complex conditions very well.The comprehensive performance of the proposed algorithm ranks second among the published tracking algorithms.Especially in the case of scale change,target occlusion and image blurring,the tracking accuracy and success rate of the proposed algorithm rank first.