基于BP网络与Mahalanobis距离的图像识别技术(英文)  

Image Recognizing Technology Based on BP Network and Mahalanobis Distance

作　　者：梁忠伟[1] 叶邦彦[2] 刘晓初[1] 

机构地区：[1]广州大学机械与电气工程学院,广州510006 [2]华南理工大学机械与汽车工程学院,广州510640

出　　处：《科学技术与工程》2009年第11期2966-2973,共8页Science Technology and Engineering

基　　金：广东省科技计划项目(2008B12080633)资助

摘　　要：提出基于BP人工神经网络与Mahalanobis距离的图像识别技术。在对IC芯片图像进行二维小波分析后,通过Ma-halanobis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 artificial neural network is proposed. After IC chip＇s micro-image has been wavelet analyzed, 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 built up 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 图像识别 芯片检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]