污染云团报警中一种抑制虚警率的有效技术  

An Effective Technique of False Alarm Rate Rejection for IR Spectra Auto-Discrimination Alarm of Pollution Clouds

作　　者：张孝霖[1] 严峻 刘宁 

机构地区：[1]华北光电技术研究所,北京100015

出　　处：《激光与红外》1997年第5期280-281,共2页Laser & Infrared

摘　　要：污染云团光谱自动鉴别报警中的虚警是十分重要而必需研究解决的问题。理论上，满足分类方程的解是无限的。对未参与训练的干扰谱，特别是背景与系统噪声，由于其波形的随机性，有很大的概率满足污染云团光谱的判决条件，从而造成完全随机的虚警。根据对光谱频谱的分析，本文提出了一种判决算法。经系统动态训练选择出判决阈值，解决了虚警问题，从而在外场鉴别试验中获得圆满成功。Theoritically, the resolution is limitless for adapting the classifying equation in the statistic pattern recognition. For those unknown spectra,they are not concerned in the training process. Especially, for the background and system clutter or noise, because of their random characteristic, there are very large probability to meet the pollution cloud's decision equation, so as to cause some entire random false alarm. According to the spectra analysis, a decision algorithm is proposed in this paper.The decision threshold is obtained from the dynamic training. Then, the false alarm problem is solved.This technique provides a successful discrimination for the interested pollution cloud with negligible false alarm rate in the passive IR remote sensing.

关 键 词：污染云团 红外遥感 鉴别告警 环境遥感 

分 类 号：X87[环境科学与工程—环境工程] TP722.5[自动化与计算机技术—检测技术与自动化装置]