【摘要】 为实现精准的图像多目标捕获与跟踪,提出基于双视线特征感知编组的多目标跟踪方法。利用自由形状线特征涵盖的丰富信息,使用双视线特征感知编组方法将线特征匹配结果当作计算条件,推算已匹配的每对线特征左右视差均值,将方位相近、左右视差均值接近的线特征视作一组,明确各组线特征作用窗口,代入时差收敛均值,实现多目标特征匹配;代入联合概率数据关联算法,构建改进粒子滤波模型,让粒子权值呈现出测量值与目标轨迹的关联概率,确保各目标跟踪轨迹持续性,并完善轨迹初始点与终点,增强跟踪时效性。仿真结果表明,所提方法能很好地处理车辆交叉引发的错误匹配问题,在交叉或遮挡等复杂情况下,依旧可以快速准确地同时对多个目标车辆进行跟踪。更多还原

【Abstract】 In order to achieve accurate image multi-target capture and tracking, a multi-target tracking method based on perceptual grouping of the dual line of sight features is proposed. Using the rich information of free-form line features, using the perceptual grouping method of the double line of sight features to use the matching result of line features as the calculation condition, we calculated the average value of left and right parallax of each pair of matched line features. The line features with similar azimuth and similar average value of left and right parallax were regarded as a group, and the action window of each group of line features was defined, and the mean value of time difference convergence was substituted to realize multi-objective feature matching. The improved particle filter model was constructed by substituting the joint probability data association algorithm to make the particle weight present the correlation probability between the measured value and the target trajectory, so as to ensure the continuity of each target tracking trajectory, improve the initial point and end point of the trajectory, and enhance the tracking timeliness. Simulation results show that the proposed method can deal with the problem of wrong matching caused by vehicle crossing, and can track multiple target vehicles quickly and accurately in complex situations such as crossing or occlusion.更多还原