基于Mean Shift算法的运动平台下红外目标跟踪

Airborne infrared target tracking based on Mean Shift

运动平台下,图像的运动包括目标、背景和平台的运动。复杂的运动关系,加上运动平台下成像质量差,增加了目标跟踪的难度。提出了一种有效的运动平台下前视红外(FLIR)成像目标跟踪算法。对于每一个被检测出的目标,计算灰度和局部标准差的分布,通过计算Mean Shift向量,最小化当前帧目标与模板的核密度分布,实现对目标的跟踪。采用自动更新模板的策略克服目标特征分布发生改变的问题,该策略同样取决于得到的模板与目标分布相似性度量。实验仿真证明,该算法能有效地、准确地跟踪红外成像序列中的运动目标,计算量小,可以满足实时性要求高的场合。

On airborne platform, image movements involve target movement, background movement and platform movement. Movement complexity and bad quality of imaging on airborne platform increase the difficulty of target tracking. An efficient approach is proposed for tracking targets in FLIR （Forward Looking Infrared） imagery taken from an airborne platform. For each detected target, distributions of intensity and local standard deviation are computed, and tracking is performed by computing the Mean Shift vector that minimizes the distance between the kernel density distribution of the target in the current frame and the template. To overcome the problems related to the changes in the target feature distributions, the strategy is used to automatically update the target template. Selection of the strategy updating new target template is based on the distance measure computed from the likelihood of target and candidate distributions. Experimental results show that the proposed algorithm can track the moving target in airborne infrared image sequence efficiently and precisely, and also can meet high real-time situation for its small calculation.