针对红外序列复杂度对目标跟踪性能的影响问题, 提出基于多属性决策评估红外序列复杂度.采用修正逼近理想解多属性决策和熵权方法, 综合7种图像度量尺度, 评估红外序列各帧图像复杂度；基于加权和多属性决策及熵权方法, 综合3种度量尺度, 评估红外序列整体复杂度.采用归一化相关模板匹配算法、基本均值偏移算法和方差比算法进行跟踪实验.采用复杂度不同的红外序列, 验证提出的红外序列复杂度评估方案的有效性.结果表明: 提出的红外序列复杂度评估方案能够真实显示各种红外序列目标跟踪任务困难度的差异, 与跟踪性能指标之间的相关性强, 并能准确反映目标跟踪任务的主要影响因素.

To analyse the influences of infrared sequence complexity on the target tracking performance, the infrared sequence complexity evaluation had been modeled as a multi-attribute decision making problem. The each frame complexity of the infrared sequence had been evaluated with seven image metrics based on the modified technique for order preference by similarity to ideal solution method and entropy weights. The whole infrared image sequence complexity had been evaluated with three metrics based on weighted summation method and entropy weights. The normalized correlation template matching algorithm, basic mean shift algorithm, and the variance ratio algorithm had been used to implement tracking experiments. Infrared sequences with different complexity had beed used to validate the effectiveness of the presented infrared sequence evaluation method. The experiments showed that: the proposed infrared sequence complexity evaluation solution could truly indicate the differences of the tracking task difficulties for diverse infrared sequences, there was strong correlation with the tracking performance, and could accurately reflect the major influencing factors for target tracking task.