ATR算法识别率的区间估计与样本量分析  

Interval estimation and sample size calculation for ATR algorithm classification accuracy

作　　者：何峻[1] 赵宏钟[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：《系统工程与电子技术》2007年第7期1021-1026,共6页Systems Engineering and Electronics

基　　金：武器装备预研重点基金资助课题(6140522)

摘　　要：针对雷达ATR性能评估中样本量的特点,从贝叶斯分析角度出发,选取最小长度准则对ATR算法的识别率进行区间估计,主要分析不同估计精度要求下的最小样本量。分析了识别率区间估计的影响因素,分别给出了无先验信息和两类典型有先验信息情况下的识别率区间估计方法和样本量确定准则,得到了识别率估计精度和所需最小样本量的关系。Classification accuracy estimation is a primary issue in automatic target recognition （ATR） algorithm performance evaluation. An interval estimator of the classification accuracy is more exact than a traditional point estimator. Several factors are analyzed in classification accuracy interval estimation to the question of the sample size in radar ATR performance evaluation. Then, the estimation methods and the corresponding criteria for three different conditions （lacking a priori information and having two types of representative apriority information） are proposed respectively. The relationship between the estimation precision and the minimum sample size is shown, and eventually directions for the ATR algorithm performance test design are given.

关 键 词：自动目标识别 性能评估 样本量 区间估计 贝叶斯分析 

分 类 号：TN955[电子电信—信号与信息处理] O212.6[电子电信—信息与通信工程]