红外目标跟踪技术研究

众多的目标跟踪算法中，Mean—Shift跟踪算法有良好的实时性，对遮挡、目标变形具有一定的适应性，是公认的效果比较好的跟踪方法。但它也存在不足，传统的Mean—Shift算法当背景的直方图分布和目标的直方图分布类似时，或者目标受到光照、阴影等影响，或有干扰物体靠近目标时，在跟踪时很容易发生目标丢失。鉴于此，提出最先使用Kalman滤波器对距离相对比较远的红外弱小目标的大致运动位置做出目标估计，接着使用Mean—Shift跟踪算法在先前目标估计出的区域内做目标的跟踪匹配，并保证精度。实验结果指出，文中提出的算法对于跟踪系统的观察噪声扰动具有较强的鲁棒性。

Research of Infrared Target Tracking Technology

In many of the target tracking algorithms, Mean-Shift tracking algorithm has good real-time quality, and has certain adaptability with shelter, target deformation. It is recognized as better effect tracking method. Traditional Mean-Shift tracking algorithm very easy occurrence tracking error when the shift algorithm of histogram distribution similar with target background of histogram distribution or exposed to light and shade or have distractions object close to target. First Kalman filter for long distance by infrared object motion position roughly estimate, then by the Mean-Shift target do accurately match in the estimation algorithm since - area. The experimental data show that the algorithrn for tracking system observation noise disturbance has strong robustness.