运动目标捕获跟踪方法研究

Research of moving object tracking method

对于光学测量丰富背景下的运动目标跟踪过程中,通常的算法存在跟踪稳定性和识别准确率无法同时满足的问题,本文对于空中飞行目标跟踪提出了一种识别跟踪方法,基于图像差分算法结合改进的目标特征不变矩匹配,将整个过程分为捕获与跟踪状态,差分目标算法进行捕获提取;跟踪过程中,构造特征不变矩,特征模版更新采用动态极值匹配法;最后通过回归投影滤波方法达到抗干扰与遮挡效果,确保准确稳定的跟踪目标。试验证明,该算法利用较少的系统资源可以稳定的跟踪目标,对发生变化的目标运动姿态具有很强的适应性。

In order to identify the moving target tracking in optical measurement, in the complex background, the traditional algorithm can not meet the stability and accuracy of problems at the same time, and puts forward the method of moment matching invariant an improved image difference algorithm based on combining the characteristics of the target, for the air flying target tracking is proposed to capture and track two state, the differential target extraction algorithm in capturing the state, in the tracking state constructed a feature invariant moments; then through dynamic extremum matching method, dynamic update feature template; finally through regression projection filtering method of anti-interference and occlusion, to ensure that the goal of accurate, stable tracking. Tests show that, the algorithm changes the target motion has very strong adaptability, ensure the stability of target tracking using the system resources can be less.