节点文献

图像识别在羊绒毛检测中的应用

The Application of Image Recognition in Wool Fiber Detection

【作者】 班志杰

【导师】 高光来;

【作者基本信息】 内蒙古大学 , 计算机应用技术, 2004, 硕士

【摘要】 羊绒和羊毛外表相似,但性能差别很大。由于羊绒保暖性能比羊毛好,价格也就比羊毛昂贵,因而常常出现羊毛和羊绒相混的现象。为了保证产品质量和商家利益,必须采用一种检测方法将其区分。以往检测人员是用手工方法来测量羊绒毛的直径和鳞片长度,不仅耗时耗力,且主观影响大,还得培训专业人员,所以迫切需要一个识别羊绒毛的系统。为了实现这样的系统,检测羊绒毛直径是最重要的工作之一。 本文主要研究光镜羊绒毛的细度提取。羊绒毛具有很好的几何图形特征,从图像的局部来看,羊绒毛图像可以近似为平行直线。这样,可以通过检测图像中平行直线来测量羊绒毛直径。为了能够准确的检测出羊绒毛图像中存在的平行直线,采用了以下几个步骤: 1、用Sobel算子进行边缘检测,以便增强羊绒毛的边缘。 2、采用全局阈值中的最小误差方法对图像进行二值化,以区别开背景和目标。 3、去掉麻点噪声,主要是去掉目标中的背景噪声和背景中的目标噪声。 4、对图像进行细化,从而找出羊绒毛的边缘骨架特征。 5、去掉小的连通区域。 6、用最小二乘法检测图像中的直线,并根据点到直线的距离计算羊绒毛的平均直径。

【Abstract】 Although the figure of wool and cashmere are similar, their performance is great different. The price of cashmere is higher than that of wool because the warm capability of cashmere is better than wool’ s, so that cashmere and wool are often interblended. In order to ensure product quality and manufacturer profits, it is urgent to find a method to distinguish them. Technician usually measures wool fiber’ s diameter and squama’ s length by hand. The technique is waste of time and energy, and the result is greatly affected by subject. What’ s more, professional workers must be trained. As a result, it is stringent to need a measure system of detecting wool fiber. The measure of wool fiber diameter plays an important role in the system.The diameter measure of wool fiber magnified by optical microscope is mostly content of this paper. Wool fiber has very nice geometrical character. The image of wool fiber can be approximately regarded as parallel beelines in partial image. So we can measure wool fiber’ s diameter by detecting parallel beelines. For exactly detecting parallel beelines in the wool fiber image, the following processes are adopted:1 Using Sobel edge detector to intensify edge.2 Using minimum error method .which belongs the global threshold , to attain binary image, so that background and object are distinguished.3 Deleting dot noise to reduce background noise in object and object noise in background.4 Thinning to discover wool fiber’ s edge framework.5 Deleting small connected area.6 Using least-square technique to detect beeline, then calculating average wool fiber’ s diameter based on the distance from point to beeline.

  • 【网络出版投稿人】 内蒙古大学
  • 【网络出版年期】2004年 04期
  • 【分类号】TP274.4
  • 【被引频次】3
  • 【下载频次】214
节点文献中: