一种新的快速鲁棒多摄像头目标跟踪方法

A Fast and Robust Multi-Camera Object Tracking Algorithm

多摄像头(视角)目标跟踪系统中,由于视角间存在大量信息冗余可有效提高跟踪鲁棒性.但在传统基于模板匹配方法中,由于视角不同导致匹配不准,会带来较大跟踪误差.针对这个问题提出了一种基于卡尔曼滤波的在线目标信息时空交互算法,利用多摄像头几何限制,实现多个摄像头的信息交互,减少了模板匹配的搜索范围,进而降低了多摄像头目标跟踪算法的计算复杂度.同时通过在线估计卡尔曼滤波模型中噪声功率,并且自适应调整卡尔曼增益将信息交互过程中误差传递降至最小.仿真结果表明,该方法可以实现更鲁棒的目标跟踪.

The tracking robustness can be greatly improved by using multi-camera system because of the redundancy of object and background information.However,traditional template matching based multi-camera object tracking suffers from tracking failure brought by different views.To solve this problem,a new algorithm is proposed,which uses the geometrical constraints to communicate object location between different camera views.This approach greatly reduces computation cost.Experiments show that the proposed algorithm is more robust than other methods.