基于多光源图像融合的织物纱线分割研究

项目名称： 基于多光源图像融合的织物纱线分割研究

项目编号： No.61502425

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 计算机科学学科

项目作者： 郑德均

作者单位： 浙江工业大学

项目金额： 20万元

中文摘要： 织物图像分割是织物结构分析的基本问题。近年来，织物图像分割受到了纺织界和计算机界国内外研究者的广泛关注。由于织物微观结构和纹理的复杂性，其结构参数的分析还依赖于人工进行，存在耗时费力等缺点。织物图像的分割模型与样品的成像条件、材料表面特征以及样品面积大小有关，因此，理论模型的构建应从织物图像获取方式出发。本项目提出一种新的织物纱线微观结构图像获取方法和理论分割模型。首先，利用一种基于多光源成像的图像融合方法获取新的织物纱线微观结构图像，以降低纱线表面的光影对分割模型的消极影响。然后，根据织物图像特点提出一种基于混合颜色空间模型的颜色表征方法和纱线区域算子，并将图像结构-纹理分解模型耦合到分割理论模型，提出基于模糊区域竞争的彩色纱线分割理论模型。最后，设计模型的快速求解算法，通过数值实验验证建立精确的纱线分割系统。项目的研究将有助于提升纺织生产信息化技术水平，促进纺织行业技术进步。

中文关键词： 识别系统；识别算法；图像分割；织物纱线；图像分解

英文摘要： Fabric image segmentation is a basic issue in fabric structure analysis. In recent years, fabric texture segmentation has received wide attention from researchers in both computer and textile communities. Due to the complexity of the micro-scale structure and texture of fabrics, the fabric structure analysis is still conducted by manual inspection, which is very time-consuming and tedious work. Fabric image segmentation models are dependent on the imaging conditions, the material surface characteristics as well as the size of the sample areas. For this reason, the modeling process of the theoretical segmentation models should be based on the fabric image acquisition methods. This project proposes a new image acquisition method and theoretical segmentation models for fabric yarn segmentation at the micro-scale level. First, the micro-structure image of fabric yarns is obtained by using an image fusion technique through multiple light sources, which is used to reduce the interferences of the light shadows on the yarn surface to the segmentation models. Second, a color mixture model and a yarn region descriptor are proposed according to the characteristics of the fabric images to describe the color and region of yarns. New segmentation models for color yarn separation are then developed by coupling between the image structure-texture decomposition model and the segmentation models based on fuzzy region competition. Third, the fast numerical algorithms are designed for the new models and the experimental validation is conducted to develop an accurate yarn segmentation system. This study will help to improve the level of information technology in fabric production and promote technological progress of textile industry.

英文关键词： Recognition system;Recognition algorithm;Image segmentation;Fabric yarn;Image decomposition