自动目标识别系统的区间评估分值模型及算法  

Interval score evaluation model and its algorithms for ATR system

作　　者：何峻[1] 赵宏钟[1] 刘峥[1] 付强[1] 

机构地区：[1]国防科技大学电子科学与工程学院ATR实验室,湖南长沙410073

出　　处：《信号处理》2009年第5期802-806,共5页Journal of Signal Processing

基　　金：武器装备预研重点基金资助项目(6140522)

摘　　要：针对自动目标识别(ATR)系统评估的迫切需求和特点,提出了一种区间评估分值模型,借助多属性决策分析理论构建了适用于ATR系统的评估模式,并给出了两个具体的区间分值评估算法。基数区间分值评估算法基于价值理论,通过确定评估对象属性的价值函数并加权形成基数区间评分值,定量评价ATR系统满足决策者主观价值需求的程度;序数区间分值评估算法基于TOPSIS法思想,通过计算评估对象与理想点和负理想点的距离差距,综合生成序数区间评分值,定量指示各个ATR系统距离理想系统的相对差距程度。运用所提出的两种评估算法对6个空地ATR系统进行了实例操作,评估结果表明所提模型及算法合理、直观。With respect to the requirements and characteristics of automatic target recognition （ATR） system evaluation, a new interval score evaluation model is proposed. This paper founds an appropriate evaluation mode for ATR systems by using multi-attribute decision making （ MADM ） theory, and also introduces two interval evaluation algorithms. Based on the value theory, the cardinal interval score evaluation algorithm attempts to quantify the satisfaction degree of an ATR system by cardinal interval score via establishing the value function of each attribute and weighting them;while the ordinal interval score evaluation algorithm calculates the distances of each ATR system to the ideal point and minus-ideal point that is originally proposed by TOPSIS method,and then aggregate two distances to get the ordinal interval scores to quantify their relative difference. We use these two algorithms to evaluate six air-to-ground （A-G） ATR systems, and the application results shows that the interval scored evaluation model and the two algorithms are clear and reasonable.

关 键 词：自动目标识别 评估 多属性决策分析 区间数 

分 类 号：TN959.1[电子电信—信号与信息处理]