机载光电系统的行人跟踪方法研究

The Research of Pedestrian Track Based on Airborne photo-electric System

行人目标,作为计算机视觉领域内最有价值、最受关注的目标,一直以来都是目标跟踪领域内的研究热点和难点。机载光电系统的行人跟踪技术一般用于无人机搜索、边境巡逻等非战争军事领域。为了克服形变、遮挡因素对行人跟踪的影响,文章提出一种基于上下文辅助的行人跟踪方法,它利用局部区域的所有像素点与目标组成上下文关系,将跟踪问题转化为计算目标可能出现位置的置信图,置信图响应最大的地方即为当前帧估计出的目标中心。通过仿真试验对比,验证本文方法在运算速度和稳定性方面都有较大提高。

Pedestrian target, as the most valuable and most concerned in the field of computer vision, has always been a hot re-search topic in the field of target tracking. The pedestrian tracking technology of airborne photo-electro system is generally used in the field of non-war military fields such as UAV search, border patrol and so on. This paper proposes a context-based aid target track method, which uses all the pixels of the target composition of the context of the local area relationship, the track problem into possible locations confidence map for the calculation of the target, where the greatest confidence map in response to the cur-rent frame is estimated center of the target. Compared through simulation experiment, the method on the speed and stability are improved greatly.