基于图像处理的原棉疵点识别方法  

Raw cotton defect recognition method based on image processing

作　　者：黄仰东 邓中民 HUANG Yangdong;DENG Zhongmin(College of Textile Science and Engineering,Wuhan Textile University,Wuhan 433074,Hubei,China)

机构地区：[1]武汉纺织大学纺织科学与工程学院,湖北武汉433074

出　　处：《上海纺织科技》2020年第7期47-50,共4页Shanghai Textile Science & Technology

基　　金：湖北省技术创新专项(2019AAA005)。

摘　　要：设计了一种基于图像处理的特征识别算法。图像采集部分设计了多角度、多光源的图像采集系统,减少阴影和光照不匀对后续处理的不良影响;图像分割部分选择基于Mean-shift的自适应阈值算法,可以较好地将疵点分割出来;轮廓提取部分采用形态学扫描法进行提取,并对轮廓断裂处进行修复处理;特征值提取部分选择了4个特征值进行提取并分析其相关性,认为离心率和面积周长比的综合计算结果可以有效地对原棉疵点进行识别和分类。将设计算法得到的检测结果与实际结果进行比对,结果显示,识别准确率达到了92.6%,证明所采用的算法有效。A recognition method based on image processing is designed.In the part of image acquisition,a multi-angle and multi-light source image acquisition system is designed to reduce the adverse effects of shadows and illumination unevenness on subsequent processing.In the part of image segmentation,an adaptive threshold segmentation algorithm based on Mean-shift could effectively segment cotton defects from the image;In the part of contour extraction,the morphological scanning method is used to repair the contour fracture.In the part of eigenvalue extraction,the four selected eigenvalues are extracted and their correlation is analyzed.Compared with the final test results and the actual ones,the accuracy rate reaches 92.6%,which proves the feasibility of processing method.

关 键 词：测试 原棉 疵点 图像处理 自适应阈值 特征值 识别 

分 类 号：TS101.9[轻工技术与工程—纺织工程]