基于区域生长型PCNN模型的织物疵点分割  

FABRIC DEFECT SEGMENTATION BASED ON REGION GROWING PCNN MODEL

作　　者：张清[1] 杨关辉[1] 高婷婷[1] 郑宏[1] 罗骏尧[1] 

机构地区：[1]武汉大学电子信息学院,湖北武汉430079

出　　处：《计算机应用与软件》2011年第11期171-175,210,共6页Computer Applications and Software

基　　金：中央高校基本科研业务费专项资金(6081001)

摘　　要：提出一种新的织物疵点自动分割的方法。该方法将待检测织物图像的像素点映射为脉冲耦合神经网络中的神经元,根据织物疵点图像的特点将改进的脉冲耦合神经网络模型同区域生长的理论结合起来,同时综合无疵点样本织物图像的统计信息完成了图像疵点区域的自动分割。最后,通过对TILDA数据库中疵点图像的检测实验,说明了该方法在织物疵点检测中应用的有效性和可行性。This paper presents a new automated segmentation method for fabric defects. In this method, the pixels of image of the fabric to be detected are mapped onto the neurons in pulse coupled neural network （PCNN）. The improved PCNN model and the regional growing theory are combined together in light of the features of fabric defects image. And at the same time the statistic information of defect-free images of sampling fabric are integrated to have implemented the automated segmentation of defect area of the image. At the end of the paper, the validity and feasibility of the proposed method applying in fabric defect detection are illustrated through tests experiment of the defect images of TILDA database.

关 键 词：脉冲耦合神经网络 区域生长 织物疵点检测 纹理分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]