基于Mahalanobis距离与BP网络的图像小波标定识别技术  

Wavelet image recognizing and calibrating technology based on Mahalanobis distance and BP network

作　　者：刘晓初[1] 梁忠伟[1] 

机构地区：[1]广州大学先进制造技术研究院,广东广州510006

出　　处：《广州大学学报（自然科学版）》2008年第6期11-17,共7页Journal of Guangzhou University:Natural Science Edition

摘　　要：针对计算机芯片检测,提出了基于Mahalanobis距离与BP人工神经网络的小波图像识别技术.在对IC芯片图像进行二维小波分析后,通过Mahalanobis距离在图像的低频小波系数矩阵上进行目标图元的区域搜索与位置标定,而后根据待定目标图元像素分布特征建立特征参数向量,建立BP网络对目标图元进行模式识别.实验证明此方法能够有效地对IC图像进行目标图元的位置标定与模式识别,获得准确的IC标定识别结果.For purpose of IC chip detection, a new image recognizing and calibrating technology based on Mahalanobis distance and BP network is proposed. After IC chip＇ s micro-image has been given wavelet analysis, Mahalanobis distance is calculated and used to implement IC object＇ s region searching and position calibrating, then its feature parameters are calculated, BP network is established for object＇ s pattern recognition. Experiment proves that this method can achieve IC image＇ s position calibration and pattern recognition, and get accurate image detecting result.

关 键 词：MAHALANOBIS距离 BP网络 小波分析 芯片检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]