判别相关滤波融合深度信息的目标跟踪算法

Object Tracking Algorithm Based on Discriminative Correlation Filter and Depth Information

基于相关滤波的目标跟踪算法已经取得了较好的性能并引起关注.在RGB图像序列的跟踪中,遮挡、背景与前景相似纹理的情况下会出现跟踪失败.本文提出了一种在判别相关滤波框架中融合深度信息的跟踪算法.由深度图分割获得空间可靠性图,根据可靠性图计算约束滤波器,避免传统判别相关滤波的边界效应.在跟踪阶段,通过对通道响应进行可靠性加权求和获得目标位置.通过目标的深度信息估计尺度,根据目标区域的深度分布和相关滤波器的响应来检测遮挡.在遮挡期间不更新模型,减少漂移问题.最后,在Princeton RGBD跟踪数据集中进行实验,结果表明,加入深度图分割与基准算法相比效果有提升.文中方法在遮挡以及尺度变化情况下能够有效地跟踪目标.

Object tracking algorithms based on correlation filters have achieved good performance and draw n attentions.In RGB tracking,false results occur in case of occlusion and the background and foreground have similar texture properties.In this paper,we proposes a tracking algorithm that combines depth information in discriminative correlation filter framework.To avoid the boundary effects of conventional discriminative correlation filter,the spatial reliability map is constructed with the help of depth image segmentation,constrained correlation filter is computed according to spatial reliability map.In tracking stage,the object is localized by summing perchannel responses weighted by the channel reliability scores.Scale is estimated by depth information of target.Occlusion is detected according to depth distribution of target region and responses of correlation filters.The model is not updated during occlusion,reducing drift problems.Finally,the experiments in Princeton RGBD Tracking Benchmark show that adding depth map segmentation method improves the performance compared with the baseline.The proposed approach can effectively track target in scale variation,occlusion scene.