基于图像分析的纤维直径和曲率的测量方法  

Measuring Diameter and Curvature of Fibers Based on Image Analysis

作　　者：任帅杰[1,2] 张文生[1] 何源[1] 王颖[1] 

机构地区：[1]中国科学院自动化研究所,北京100080 [2]中国科学院研究生院计算与通信工程学院,北京100080

出　　处：《中国图象图形学报》2008年第6期1153-1158,共6页Journal of Image and Graphics

基　　金：国家基础研究发展计划973项目(2004CB318103);国家自然科学基金项目(60033020);中国科学院海外杰出人才研究计划项目(06S3011S01)

摘　　要：纤维直径(细度)和曲率的测量在羊毛产业和纺织工业中具有举足轻重的作用。提出了一种基于图像分析的全自动测量方法,该方法首先对采集的图像进行图像增强,利用Canny边缘检测提取纤维边缘,根据纤维图像的特点用聚类的方法对增强后图像进行二值化,并对二值化图像进行一系列后处理,然后提取出纤维的骨架,进而通过距离变换和特定的搜索算法计算骨架上每个点对应处的纤维直径和曲率。实验结果表明,该方法与基于直线拟合的方法相比,不仅计算速度快,而且绕开了很多可能产生误差的环节,测量准确率较高。Measurement of the diameter and curvature of fibers plays a very important role in wool and textile industry. In this paper, an automatic measurement method based on image analysis is presented, In this method, the image processing and analysis is fully used, Firstly, this method enhances the captured images. Secondly, it uses canny operator to detect the edges of the fibers, And then, it acguires the binary images by using clustering. It does post-processing operation to the binary images and extracts the skeletons of the fibers. Finally, it computes the value of diameter by Euclid-distance-transformation, and calculates the value of curvature. As experiments demonstrate, this method is not only more quickly than the method based on line fitting, but also generates more accurate results than the method based on line fitting because we have avoided many steps which may introduce errors.

关 键 词：直径 曲率 骨架 距离变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]