基于KCF和SIFT特征的抗遮挡目标跟踪算法

Anti Occlusion Target Tracking Algorithm Based on KCF and SIFT Feature

摘要:

为了解决KCF目标跟踪中由于目标遮挡和目标尺度的变化造成跟踪目标丢失的问题,对核相关滤波器(KCF)目标跟踪的框架进行了研究,提出了一种基于KCF和SIFT特征的抗遮挡目标跟踪算法,引入了一种目标跟踪丢失后重新搜索定位目标的策略。利用尺度金字塔估计出目标的尺度,实现跟踪框自适应目标尺度大小,通过核相关滤波器(KCF)跟踪算法对目标进行跟踪。跟踪过程中对目标遮挡情况进行判断,当目标遮挡时,对当前帧跟踪框内的目标提取SIFT特征,生成模板特征。提取下一帧视频图像的SIFT特征并与模板特征进行匹配,框出与模板特征相匹配的目标,对目标继续进行跟踪。通过TB数据库标准视频序列和实际环境拍摄的视频序列进行测试。实验结果表明,跟踪框能适应目标的大小,在目标发生遮挡的情况下,能够重新找到目标并进行准确跟踪。

Abstract:

In order to solve the KCF target tracking target tracking problem caused by the loss due to the change of target occlusion and the object scale of nuclear related filter (KCF) target tracking framework is studied, this paper presents an improved anti occlusion target tracking algorithm based on KCF and SIFT feature, introduced a target tracking lost re search target by strategy. Pyramid scale estimate the scale, realize the adaptive target tracking frame size by kernel correlation filter (KCF) tracking algorithm for target tracking. The tracking process of target occlusion judgment, when occlusion happens, the current frame tracking box target extraction SIFT feature, SIFT feature extraction feature template. The next frame of video image and matching the feature template, frame matching and template feature target for target tracking. To test the video sequence through the TB database standard video sequence and the actual shooting environment. The experimental results show that the tracking frame can adapt to the size of the target, the target is occluded, be able to find new targets and accurate tracking.