高斯混合模型在印花织物疵点检测中的应用  

Application of Gaussian mixture model on defect detection of print fabric

作　　者：李敏[1] 崔树芹[1] 谢治平[1] 

机构地区：[1]武汉纺织大学数学与计算机学院,湖北武汉430072

出　　处：《纺织学报》2015年第8期94-98,共5页Journal of Textile Research

基　　金：国家自然科学基金资助项目(61103085);湖北省教育厅中青年基金项目(Q20111613);湖北省教育厅重点项目(D20131604)

摘　　要：为实现印花织物疵点的自动检测,提出了一种基于改进的高斯混合模型的疵点检测方法。该方法针对传统高斯混合背景模型应用于疵点检测中所出现的精度不高的问题,充分利用印花织物图像像素间强相关性的特点,引入自适应分块建模的思想来实现印花织物疵点的检测。实验结果表明,使用该方法进行疵点检测,正确率可达到94%。不仅如此,该方法还能有效处理检测过程中出现的光照不均和噪声等问题,是一种非常适合于对印花织物进行疵点检测的方法。A new approach based on improved Gaussian mixture model was developed to detect defects on print fabric. Considering the detection accuracy is not high by using the conventional Gaussian mixture model based method,fully utilizes the characteristic of strong correlation among print fabric image pixels,this paper proposed a defect detection algorithm with self-adaptive partition block modeling. Experimental results indicate that by using the proposed method, the detection success rate can reach 94%.Furthermore,the proposed method can effectively handle problems in the detection process on nonuniform illumination and noise,and it is suitable for the defect detection of print fabrics.

关 键 词：印花织物 疵点 高斯混合模型 分块建模 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]