基于背景感知相关滤波的深度提取跟踪算法研究

背景感知相关滤波算法将目标背景和前景一起建模1],利用包含背景信息的负样本进行训练,得到高鲁棒性核相关滤波器,实现跟踪目标与背景的分离。这种单目实现的相关滤波跟踪算法能够做到实时跟踪,却以丢失深度为代价。本文提出了一种将双目立体视觉深度提取与基于单目的背景感知相关滤波相结合的算法,该算法在做到实时跟踪的同时,能够对视频序列中的若干帧利用改进的双目立体匹配SGM算法得到视差图,反馈出跟踪目标的深度信息。实验结果表明该算法具备实时性,且三维坐标定位准确。

Research on Depth Extraction Tracking Algorithm Based on Background-Aware Correlation Filtering

Background-aware correlation filter algorithm models the background and foreground together, uses negative samples containing background information for training a high robustness kernel correlation filter to achieve the separation of tracking target and background. Although this kind correlation filter algorithm based on monocular vision can achieve real-time tracking, the depth information will be lost. This paper designs an algorithm which combined binocular stereo vision and background-aware correlation filter algorithm together. In addition to real-time tracking, the improved binocular stereo matching SGM algorithm uses several frames in video sequence to obtain parallax map then feed back the depth information of the tracking target. The experimental results show that this algorithm has the characteristics of real-time and accurate location.