【作者】 刘会敏；

【导师】 张瑞林；

【作者基本信息】 浙江理工大学 ， 计算机应用技术， 2012， 硕士

【摘要】 随着计算机技术的飞速发展和图像处理技术的日益成熟，图像检测技术已经成功应用于工业检测，人工智能，生物医学，空间探测以及遥感等许多领域，不仅取得了很多研究成果，并促使这些学科产生了新的发展。蚕丝被工艺质量中填充物均匀度的检测是蚕丝被质量检测中最重要的检验项目之一，目前基本上依靠人工检测，传统的检测方法劳动强度大，耗费时间长，且容易受检验人员主观因素影响，重复一致性偏差大，而且是破坏性测试。因此，寻找一种能有效地代替人工检测的方法具有重要的理论与现实意义。本文正是在这种背景下，提出了基于图像处理的蚕丝被工艺质量中填充物均匀度的检测研究这一课题。本文对传统的蚕丝被填充物均匀度检测方法和图像检测相关技术进行了详细的分析和研究，在分析了蚕丝被国家标准(GB/T24252-2009)中关于蚕丝被填充物均匀度检测的相关检测方法，以及蚕丝被厚薄差异率和角质量差异率的计算方法的基础上，从图像处理的角度建立了厚薄差异率和角质量差异率的计算公式，具体研究工作：1.搭建合适的光照环境利用数码相机近距离大幅度的采集蚕丝被图像，然后进行图像的预处理，消除噪声和畸变。主要研究了图像几何畸变的校正算法。本课题利用双线性内插算法校正图像采集过程中造成的几何畸变，首先通过坐标变换对图像像素位置进行校正，然后利用双线性内插法进行像素灰度值的估算。2.研究了图像光照不均匀的校正算法。针对图像采集过程中背景光照的不均匀对图像灰度值提取所造成的影响；提出了基于小波的光照不均匀校正算法。首先通过基于小波的非线性回归方法来估计光照分布，然后从原图中减去光照分布图，实现消除不均匀背景光照的目的。3.研究了蚕丝被透光率(亮度)偏差与厚度均匀性，厚薄差异率的关系，结合光学知识建立了图像灰度，光密度及重量或体积的线性关系，然后利用蚕丝被图像积分光密度代替蚕丝被的重量求得厚薄差异率和角质量差异率。具体推导过程见附录。根据蚕丝被国家标准进行等级分类。实现蚕丝被均匀度检测的非破坏性测试。本文从图像处理的角度建立蚕丝被图像灰度值与蚕丝被厚度或者重量的关系，结合光学知识提取蚕丝被填充物含量的测量参数，然后由填充物含量测量参数代替蚕丝被的重量计算厚薄差异率H和角质量差异率J。实现蚕丝被工艺质量中填充物均匀度的非破坏性检测。本文的研究促进图像处理技术的在工业检测方面的应用更加广泛，同时也为蚕丝被工艺质量中厚薄均匀度的检测开辟新思路。更多还原

【Abstract】 With the development of modern computer technology especially the development ofimage manipulation technology; computer technology has become the research hotspot in manyfields. Is applied widely on many fields such as automatic target recognition, intelligenttechnology, biomedicine, medical image analysis, remote sensing and space exploration. And alot of researching results produced, these studies have pushed the new development in everysubject.The wadding cover evenness detection is one of the most important areas in the detection ofthe quality of silk quilts. At present, the detection of the quality of silk quits mainly rely onhuman experience. And this traditional approach has become the bottleneck of textiles’ rapiddesign and production. Therefore, in order to replace artificially detection, it is boththeoretically and practically significant to develop an effective system which can detect silkquilts automatically. Based on this background, this paper presented a study regarding to thenondestructive examination of the homogeneous degree of the fill for silk quilts based on imageprocessing technique.This paper analyses and studies the existing evenness detection methods of silk quits andthe existing image processing methods. According to the computational methods of the weightof silk(H) and the homogeneous degree of the fill for silk quilt(J) in the state principal of thesilk quilts(GB/T24252-2009), evenness detection of wadding cover from the view of imageprocessing was discussed. The calculating formula of Thick and thin was given and algorithmmodel was established. The research points and achievements are listed as follows:(1) In the system. It utilizes CCD for gathering digital image of silk quilts, and thenprocesses those images by computer image processing technology; this method can reducenoises and the correction of the geometrical distortion and imaging non-uniformity of theprojected image; then the gray scale of pixels are acquired in terms of gray distribution.(2) Research on Digitalize Imaging non-uniformity Correction Methods, The limitation of imaging condition in actual environment causes sometimes the non-uniform background lighton the images when capturing. To resolve this problem, a new method for the correction ofunevenness elimination technique based on wavelet transference is proposed. Using nonlinearwavelet transform for obtaining distribution characteristic of background light. Experimentresults demonstrated that the method achieved a satisfactory effect of removing the unevenillumination.(3) In order to improve the efficiency and veracity of silk quilt evenness detection based onimage processing. A new method that is based on optical theory and image processing has beenproposed for analyzing the relationship between pixel gray values, the optical density in animage and weight of silk(H). The homogeneous degree of the fill for silk quilt (J) can bedetermined by using pixel gray values replace weight to calculate deviation. After that, thesedata are classified based on the Chinese National Standards (GB) by H and J. To compare withtraditional method, it can build the Non Destructive Testing (NDT), which has been proved tobe a practical methodThis paper presented a study regarding to the nondestructive examination of thehomogeneous degree of the fill for silk quilts based on image processing technique. A methodthat is based on optical theory has been proposed for analyzing the weight of the fillingparameter measurement. The homogeneous degree of the fill for silk quilt can be determined byusing pixel gray values replace weight to calculate deviation. This study has an importantpractical significance for image processing technology and the development of thenondestructive examination of the homogeneous degree of the fill for silk quilts.更多还原