李忠健 董龙 倪海云 王静安 王鸿博 邹专勇

摘要： 毛羽是衡量纱线质量好坏的重要指标，然而现有毛羽检测中难以快速获得其三维长度（毛羽空间轨迹长度）。文章提出了在获得单幅图像的基础上准确表征纱线毛羽空间长度的方法。采用远心镜头获取纱线毛羽图像，经过同态滤波、直方图均衡化、边缘增强、水平集分割等处理，获得包含空间信息的毛羽分割图像。结合散焦成像原理，建立毛羽宽度、灰度与其空间深度之间的模型关系，实现毛羽空间三维长度的快速测量。基于单幅散焦信息的毛羽空间长度测量，为实时三维检测奠定基础。实验结果表明，文章所提方法可有效实现清晰与模糊区域毛羽的分割，所测毛羽三维长度与人工测量结果相近。

关键词： 毛羽三维长度;单幅图像;散焦信息;图像分割;深度恢复;毛羽检测

中图分类号： TS103.7

文献标志码： A

Abstract： Hairiness is an important indicator to measure the quality of yarn. However， it is difficult to quickly obtain its three-dimensional length（space trajectory length of hairiness） in the existing hairiness test. In this paper， we propose a method to accurately characterize the three-dimensional length of yarn hairiness based on single images. A telecentric lens is used to obtain the hairiness image of yarns， and after homomorphic filtering， histogram equalization， edge enhancement， level set segmentation， etc.， a segmentation image of hairiness which contains spatial information is obtained. Combined with the principle of defocus imaging， the model relationship between hairiness width， gray scale and its spatial depth is established to achieve the quick measurement of the three-dimensional length of hairiness in space. The spatial length measurement of hairiness based on the defocus information of single images lays a foundation for real-time 3D detection. The experimental results show that the proposed method can effectively achieve the segmentation of hairiness in clear and fuzzy areas， and the measured three-dimensional length of hairiness is close to the results of manual measurement.

Key words： 3D length of hairiness; single image; defocus information; image segmentation; depth recovery; hairiness detection

毛羽作為评定纱线质量最重要的指标之一，其数量及长短不仅直接决定了纱线的外观、质量和等级，而且当毛羽分布不匀或者长毛羽较多时，会使织机开口困难，染色不匀，降低织造和染色工序生产率，造成织物质量下降，使企业难以生产出优质的纺织品[1-3]。目前毛羽的检测主要采用光电式分级统计法的毛羽根数分布指标和光电式全毛羽法的毛羽H值指标进行表征，是当前占主导地位的测量方法。然而由于光电传感器本身的限制，其检测精度较低，测量结果为统计指标，无法表征单根毛羽的信息，更不能直观反映毛羽的空间形态[4-8]。……