SAR图象自动目标识别研究  

A Survey of Researches on SAR ATR

作　　者：匡纲要[1] 计科峰[1] 粟毅[1] 郁文贤[1] 

机构地区：[1]国防科技大学电子科学与工程学院,长沙410073

出　　处：《中国图象图形学报（A辑）》2003年第10期1115-1120,共6页Journal of Image and Graphics

摘　　要：目前,SAR已经成为一种不可或缺的对地观测和军事侦察手段.面对不断增长的SAR图象数据收集能力,如何对这些图象进行自动或半自动快速、准确地解译已经越来越引起人们的关注和重视.自动目标识别(ATR)是自动或半自动SAR图象解译研究的一个重要方面.SAR ATR过程可概述为:从观测得到的SAR图象中,找到感兴趣的区域(ROI),并计算出每个ROI的种类.为此,介绍了SAR ATR的含义及其一般流程,对SAR ATR系统按照它所采用的分类方法进行了归纳分类,分析了SAR ATR的难点,介绍了国内外SAR ATR的研究现状和发展趋势.Recently, SAR (Synthetic Aperture Radar) is evolving to become an indispensable tool for earth observation and military reconnaissance. The collection capacity for SAR images is growing rapidly , and along with that growth is the expanding need for automated or semi-automated exploitation of SAR images accurately and efficiently. Automatic target recognition (ATR) is an important aspect of automatic or semi-automatic SAR imagery interpretation. An ATR algorithm finds target like regions (regions-of-interest (ROIs)) within a SAR image and computes a class for each ROI. The meaning of SAR ATR is explained, and the general architecture of SAR ATR system is introduced in this paper. SAR ATR systems are analyzed and classified according to their classification methods. The main difficulty of SAR ATR is analyzed, the present state and the future direction of SAR ATR research are also introduced.

关 键 词：SAR图象 自动目标识别 ATR 合成孔径雷达 主动传感器 

分 类 号：TN957.52[电子电信—信号与信息处理]