基于形状特征的棉花异性纤维图像分割方法  

Segmentation of Foreign Fibres Based on the Geometric Features

作　　者：夏心怡[1] 苏真伟[1] 李国辉[1] 

机构地区：[1]四川大学制造科学与工程学院,成都610065

出　　处：《农机化研究》2010年第7期13-16,共4页Journal of Agricultural Mechanization Research

基　　金：国家自然科学基金项目(60875022/F030410)

摘　　要：棉花中的白色异性纤维与背景相近,现有的算法很难识别出来。为此,在使用透射光增加异性纤维和棉花背景的灰度差的基础上,提出了一种利用形状特征来分割异性纤维的算法;同时,在自适应局部阈值处理的基础上,利用骨架的走向对间断的纤维进行修补,最后根据几何特征过滤图像中的噪声。实验结果表明,该方法能够有效地分割出异性纤维。It is difficult to detect white color foreign fibres in cotton by the existing methods.To solve this problem,an optical transmission imaging method which enhances the grayscale differences between cotton and foreign fibres,and a segmenting algorithm based on foreign fibres geometric features are proposed in this paper.On the basis of local adaptive threshold method,the discontinuous fibre segments are linked according to the orientation of their skeletons.Then the image noises are removed based on their geometric features.Experimental results indicated that this method can effectively detect the white color foreign fibres in cotton.

关 键 词：棉花 异性纤维 机器视觉 图像分割 形状特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]