基于神经网络的绝缘瓷瓶裂缝类型识别研究  

Recognition of Porcelain Bottles Crack Types Based on Neural Network

作　　者：陆斌斌[1] 刘国海[1] 

机构地区：[1]江苏大学,江苏镇江212013

出　　处：《电测与仪表》2006年第12期11-15,共5页Electrical Measurement & Instrumentation

基　　金：国家863基金(2005AA420064)

摘　　要：采用CCD摄相头等硬件模拟机器人的视觉。首先完成对绝缘瓷瓶裂缝图像的采集,然后运用软件完成对图像的滤除噪声、图像分割等预处理操作,再将的图像分割成的子块图像,使其成为的二值图像,提取五个特征值,并通过对十幅图像统计分析,得出这五个特征值能够反映图像的信息。最后分别设计基于BP以及RBF神经网络的类别分类器,实现对绝缘瓷瓶裂缝五种状态：横向、纵向、块状、网状,无裂缝的分类识别,给出十幅图像的识别结果,并分析比较,得出RBF神经网络识别效果更为突出的结论。The hardwares such as CCD,pan controls and PC are used to simulate the vision system of the robot.Firstly,the images of porcelain bottles crack are collected by the sys- tem.Secondly,the software is used for image preprocessing as smoothing and segmenting the object from the background by the threshold.Third,the image of is divided into pixels sub- images and turned into twain value image of pixels.Five features are extracted,which reflect the information of the whole image through analyzing the results of ...

关 键 词：图像预处理 特征提取 分类识别 神经网络 

分 类 号：TM85[电气工程—高电压与绝缘技术]