一种运动目标多特征点的鲁棒跟踪方法研究

提出了一种基于特征光流分割和卡尔曼滤波估计的鲁棒性的运动目标跟踪方法。该方法具有很多特点：首先在特征光流的计算中采用由粗到细的层级匹配算法，因而能够计算大的运动速度和具有更好的匹配精度；其次采用了有效的遮挡判决算法，该算法综合利用了先验的信息，对噪声的干扰不敏感；最后建立了线性卡尔曼滤波模型，当特征点被遮挡或丢失时，能够预测它们的位置，这使得跟踪更具有主动性。实验表明，该方法具有高精度、快速跟踪和很好的鲁棒性。

Robust Tracking Method for Multiple Feature Points of Moving Target

A robust tracking method for a moving target based on the segmentation of optical flow velocities of feature points and Kalman filtering estimation is presented. The advantages of this method are as follows: firstly, the flow velocities of feature points are computed by a hierarchical coarse to fine matching algorithm. The algorithm can attain the computation of high motion velocities and has a higher matching accuracy. Secondly, an effective discrimination algorithm of occlusion is applied, which is insensitive to the noise because the prior information is roundly applied. Finally, a linear Kalman filtering model is set up and positions of feature points can be predicted when they are occluded or lost, which makes the tracking more active. Experimental results show that the tracking method improves the tracking accuracy and has a high tracking velocity and a better robustness.