织物颜色配准到标准色卡的计算机识别与仿真  

Computer recognition and simulation of fabric color from matching to standard color chip

作　　者：陈孝之[1] 谢莉青[1] 

机构地区：[1]青岛大学纺织服装学院,山东青岛266071

出　　处：《纺织学报》2016年第5期150-154,166,共6页Journal of Textile Research

摘　　要：为克服人工目测织物的颜色易受到光线和经验的影响产生误差的问题,提出一种基于计算机视觉和图像分析技术的织物颜色与纺织专用标准色卡自动配准的颜色识别与仿真的方法,实现了针对1 925种潘通色卡的织物颜色自动配准系统。该方法利用扫描仪获得潘通色卡棉布版的扫描图像,然后从色卡图像中提取颜色有效特征信息,构造色卡图像的色度特征数据库;设计了颜色分层模型和基于"一对一"支持向量机(SVM)与色卡图像数据的配准模型,经过对SVM模型参数优化和识别训练,系统与色卡匹配的正确率达96.89%。另外使用296种未知色号的织物样本,将它们与系统进行配准实验,匹配正确率为98.85%,为客观、快速地测色和数字仿真颜色提供了参考工具。To reduce the errors of fabric color caused by light and experience when measured by human eyes,a color recognition and simulation method for automatically matching fabric color with special standard color card was put forward. This method is developed based on computer vision and image analysis technique,and fabric color automatic registration system for 1 925 Pantone TCX color swatches was realized. The method used the scanner to acquire Pantone TCX color scan images,and then extracted effective color characteristic information from color image for constructing database of color image feature.Color layered model and registration model of chromatic value and Pantone TCX color card based on ″one to one″ support vector machine（ SVM） were designed. After optimization and implement training of color identifying for SVM model parameters,the accuracy is 96. 89%. Using 296 previously unknown samples for verification,accuracy is 98. 85%. This would provide a reference tool for objective,fast color measurement and color of digital simulation.

关 键 词：织物颜色 图像处理 支持向量机 配准 潘通色卡 

分 类 号：TS107.4[轻工技术与工程—纺织工程]