基于互信息测度红外目标图像相关匹配跟踪算法

Infrared Image Object Tracking Based on Multi-pattern Algorithm

在红外地面目标跟踪过程中,目标在图像系统中呈现多样化,任何一种单一的跟踪方式都无法满足系统的要求。对此本文采用一种多模跟踪方式,针对具体的目标,采用不同模式的跟踪方法。对于小目标,采用互信息测度的相关跟踪方式;对于大目标,采用形心跟踪方式。通过自适应波门来决定跟踪模式,实验证明该算法具有稳定而且精确的跟踪性能,算法执行时间短,完全可以满足实时性要求。

For the infrared ground moving object in image sequence emerges diversification in the course of tracking the object, any single tracking pattern can not satisfy the requirement of an object tracking system. The multi-pattern tracking algorithms were adopted, in which the correlation matching algorithm based on mutual information similarity measure is used for small abject and the centroid tracking one for large one. The experimental results of two different infrared image sequences show that the algorithm has stable and precise tracking performance, shorter runtime and can fully satisfy real-time requirement.