摘要
对同类目标畸变不变的正确识别率与不同类目标分类误识别率是衡量一个自动目标识别 (ATR)系统的两个最重要性能指标。但在实际应用中 ,ATR系统所获取的外场的目标与背景总是处于随时间不断变化的条件下 ,与系统所存储的参考目标通常都不会一致 ,从而导致相关识别SNR劣化。特别对于多目标识别与不同类目标的区分 ,常规的相关门限判决方法会造成很大的误识别 ,大大影响了ATR系统的识别可靠性。本文采用人工神经网络 (ANN)与模糊逻辑技术 ,对相关信号与噪声进行实时数字后处理 ,通过对信号与噪声强度分布等高线而不仅仅是强度的识别 ,大大提高了ATR系统的识别可靠性 ,改善了识别效率。
The recognition possibility for intra class target distortion version and misrecognition possibility for inter class target are two main performances of an automatic target recognition(ATR) system.In application,the acquired target and background by a ATR system are always changed temporarily,causing the disaccord of the reference and acquired target and the degradation of correlation SNR.As a result,the general threshold judgement of correlation peak will fail,misrecognition occurs,and the recognition possibility was decreased unavoidably,especially for the multi target recognition and inter class target partition tasks. In this paper,a artificial neural network(ANN) and fussy logic technique was proposed to make real time digital post processing of correlation signal and noise.The recognition was performed by means of the intensity contour mapping of correlation signal and noise.Experimental results show a dramatic improvement in recognition possibility and recognition reliability of the ATR system.
出处
《激光杂志》
CAS
CSCD
北大核心
2000年第2期43-44,共2页
Laser Journal
关键词
模糊逻辑
自动目标识别
神经网络
可靠性
ATR(Automatic Target Recognition),ANN(Artificial Neural Network),Intra class Recognition,Inter class Recognition
Distortion Invariant
