诚成设计,基于计算机视觉的染色品色差检测.doc

论文题目: 基于计算机视觉的染色品色差检测 学科专业： 计算机技术 作者姓名： 指导教师： 递交日期： 2012年 月 日 基于计算机视觉的染色品色差检测Color-difference evaluation for dyeing products based on computer vision摘要 人类社会约百分之七十的信息是通过人类视觉进行的，通过眼睛来获取图像具有其它方法所不可比拟的优势，包括：直观性、易理解性、快速性等计算机视觉即是用计算机实现人类的视觉功能随着数字图像技术、计算机技术、模式识别技术的快速发展，计算机视觉已经广泛地应用于医药卫生，国防建设，航空航天、兵器、制造检测、质量控制等社会应用领域而人民生活水平的日益提高促使着纺织品印染行业的快速发展，色差检测技术作为其中的一项基本要求渐渐成为相关公司及科研单位迫切需要解决的问题。

在基于计算机视觉的染色品色差检测系统中，需要解决的问题主要包括：待测目标的采集、颜色空间的选择、颜色信息的提取、色差评价模型的建立、颜色图像的光照较正及色差信息的分类等关键技术问题采用高性能的工业彩色相机对染色品的表面颜色进行检测，使用相关处理软件实现实时监控，这样便于生产管理的透明化、智能化、快速化此系统不仅可以很好地提高染色品染色的满意度并且可以降低工厂的人力成本，提高生产效率让产品质量管理透明化，获得良好的经济效益，从而增强企业的市场竞争力为了实现精准的染色品色差检测，本文对样品图像的光照校正问题进行了深入研究本文提出了一种基于小波理论的染色品光照校正算法，同时，本文采用一种染色品色差检测的评价指标——一致性指标和均匀性指标为了对采集的样本数据进行很好的分类，本文对几种经典的分类算法进行研究，包括：基于支持向量机的分类算法、基于级联弱分类器的AdaBoost分类算法，最终本文根据染色品本身的纹理特性与颜色特性，提出了一种基于LBP-GMM的支持向量机分类算法，并将分类算法对染色品的色差质量进行判别分类试验结果表明，该模型对染色品颜色数据具有很好的分类效果，可以实现染色品色差的定量化和自动化评价，并且对色差检测时光照不一致的情况具有较好的鲁棒性，为实现工业现场染色品色差的检测系统奠定了良好的基础。

关键字：计算机视觉；色差检测； 基于LBP-GMM模型的支持向量机分类算法；光照校正Color-difference evaluation for dyeing products based on computer visionAbstractThrough the eyes, human gets almost 70％ information. This method has advantages which other methods cannot unmatched, such as intuitive and easy to understand. Computer vision is to realize the human visual function by computer. With the rapid development of digital image technology and computer technology, computer vision has been widely used in medicine and health, national defense construction, aerospace, and other social areas. In addition，increasing living standards of people promotes the rapid development of textile printing and dyeing industry, and as one of the basic requirements, color detection technology gradually become the urgent problem which need to solve between the related companies and research institutes.In the color difference detection system based on computer vision, we need to solve the color image acquisition, building of color evaluation model, illumination correction of color images and other key technical issues. Using high-performance industrial camera to monitor the surface color of the dyed goods and using processing software to achieve real-time online monitoring, it is easy to manage production transparent, intelligent and fast. This system not only improve dyeing quality of the dyeing products，but，reduce the factory labor costs and raise production efficiency. Transparency of product quality management can obtains good economic efficiency, thereby enhance the market competitiveness of enterprises.In this thesis, the light correction algorithms are studied during the color-difference detection. Key research contents and findings are as follows: firstly, the evaluation indexes of color detection for dyeing products, which are the consistency index and the uniformity index, are put forward. In addition, a method of light correction based on wavelet theory is proposed.To make a perfect classification, several important classification methods are studied, including Adaboost algorithm and SVM algorithm. SVM algorithm color evaluation model based on LBP-GMM improves the evaluation accuracy of the color, and laid the foundation for the design and dyeing products color detection system based on computer vision. In addition, in the process of dyeing color detection, there are a variety of factors causing the color evaluation errors. This article will also analyze related factors which may appear in the color detection scene. The reasonable control of these factors is to reduce the color evaluation error. The reasonable control of these factors can achieve its goal which reduce evaluation errors.Keywords: Computer vision; color detection; LBP-GMM SVM algorithm; illumination correction 目录摘要………………………………………………………………………………………………IAbstract…………………………………………………………………………………………..III目录………………………………………………………………………………………………VI第一章 绪论……………………………………………………………………………………..11-1计算机视觉概述……………………………………………………………………………..11-2基于计算机视觉的染色品色差检测………………………………………………………..11-3 本文研究背景、目的及意义……………………………………………………………….31-4本文的研究内容及思路………………………………………….………………………….41-5 本文章节目录及概要……………………………………………………………………….4第二章 染色品色差检测相关理论及系统设计………………………………………………..62-1 图像预处理基本算法……………………………………………………………………….62-2 图像颜色空间……………………………………………………………………………….102-3 色差计算原理……………………………………………………………………………….132-4 色差检测系统结构………………………………………………………………………….142-5 照明系统和光源……….. …………………………………………………………………..152-6 建立染色效果评价模型…………………………………………………………………….162-7 本章小结…………………………………………………………………………………….29第三章 染色品色差检测中颜色分类算法的研究……………………………………………..303-1 支持向量机算法…………………………………………………………………………….303-2 AdaBoost分类算法…………………………………………………………………………..313-3 一种基于LBP-GMM模型的支持向量机色差检测分类算法…………………………….333-4 本章小结…………………………………………………………………………………….36第四章 染色品色差检测中的光。