基于三维面阵激光成像系统的目标跟踪算法研究

Research on target tracking algorithm based on three-dimensional array laser imaging system

摘要:

针对现有的RGB目标跟踪算法难以有效应对光线变化、尺度变化、遮挡等问题，我们提出了一种基于三维面阵激光成像系统的感知遮挡相关粒子滤波目标跟踪算法。首先提出基于Depth深度数据及其HOG特征的目标遮挡判断机制，然后融合相关最大似然估计粒子滤波算法形成目标预测-跟踪-校准-再检测的跟踪机制。与现有的目标跟踪算法进行对比实验的结果表明，基于三维面阵激光成像系统的感知遮挡相关粒子滤波目标跟踪算法可以很好的检测目标遮挡并进行目标跟踪，降低了光线变化带来的影响，在目标预测-跟踪-校准-再检测方面，计算量较小，实时性较好，精确度达到85.7%。

Abstract:

As the existing RGB target tracking algorithms are difficult to cope with the problem of illumination changing and target occlusion effectively, we presents a sensing occlusion correlation particle filter target tracking based on three dimensional array laser imaging system. Firstly, the target occlusion judging mechanism based on depth information and HOG feature is proposed. Then the correlation maximum likelihood estimation particle filter algorithm is used to form the target prediction-tracking- calibration-redetection tracking mechanism. The experimental results show that the sensing occlusion correlation particle filter target tracking based on three dimensional array laser imaging system can well detect target occlusion and target tracking, reduce the impact of illumination changing, and has less calculation in target prediction-tracking-calibration-redetection, has better real-time performance and accuracy up to 85.7%.