【作者】 张琴；

【导师】 卢新明；

【作者基本信息】 山东科技大学 ， CAD与图形图形处理， 2011， 硕士

【摘要】 数字图像是信息时代人们获取信息的最主要和重要的途径之一,因此对图像处理技术的研究和应用就颇具意义和价值,也一直是国内外专家和学者的重要研究方向。NSCT具有完全的平移不变性、多尺度和多方向等特征,是较小波和Contourlet变换更优越的图像表示方法。NSCT在2005年才被提出,已迅速被许多学者应用在图像去噪、图像增强和边缘检测等方面,并取得了较好的效果。鉴于目前白酒瓶检测技术中的图像处理研究颇少,本文就基于机器视觉的白酒瓶瓶口图像处理技术进行了研究和实验,着重研究瓶口图像处理的方法,以达到自动检测瓶口的目的,而图像处理所涉及的关键技术则为研究的重点。本文首先对图像处理的内容和方法进行了简单介绍,重点研究了特征提取相关的理论、算法和技术,主要进行了基于小波变换和NSCT的图像处理的研究与实验。在对Gabor变换,小波包变换、NSCT等进行研究和实验之后,利用NSCT方法实验了对纹理图像进行特征提取并达到检索目的的新方法。为实现瓶口自动检测,本文提出并且实验了对白酒瓶瓶口图像进行适当分割、特征提取并达到检测优劣目的的新方法,并利用VC编程实现了瓶口图像自动检测功能。具体方案是首先对原图像使用Hough变换检测圆心位置,然后根据瓶口的内外圆半径将瓶口图像从整个图像中分割出来,再分割成面积相等方向一致的若干个子图像。将每个子图像利用NSCT提取特征值,再求出特征值的均值和方差作为子图像特征向量,然后求出无损子图像的特征向量的均值作为标准值,根据每个子图像特征向量与标准特征向量的相似度,判断瓶口是否完好。实验结果表明该方法速度快,检测正确率高,有很高的实用价值。NSCT理论正趋于完善,其应用研究正处于起步阶段,发展前景极为广阔。更多还原

【Abstract】 In the information age digital image is the one of the main and important ways to us to obtain the information. Researches on image processing technology and application are meaningful and valuable. And it’s an important direction to the most of the domestic and international experts and research scholars.The NSCT is based on a nonsubsampled pyramid structure and nonsubsampled directional filter banks. It is a fully shift-invariant, multiscale, and multidirection expansion that has a fast implementation. The NSCT was proposed in 2005 and have been studied and used by many scholars, which has been quickly applied in image denoising, image enhancement and edge detection and so on, and good effect has been obtained. The research and experiment of the finish image processing technology have been made, the methods of the finish image processing is the key point, for the purpose of bottle automatic detection and the image processing key technologies involved are well studied.In the paper, first, the content and methods of image processing is simply introduced, with the focus on the feature extraction related theory, algorithm and technology, the research and experiment of the image processing based on wavelet transform and NSCT are mainly studies. After the research and experiment of the Gabor transform, the wavelet packet transform and NSCT, we obtain the numbers of the feature extraction and reach the purpose of the retrieval using NSCT. The new method is been obtained using the method of experimental NSCT grain image feature extraction for the purpose of the retrieval.To realize the automatic detection of the finish, the paper puts forward and tests the new methods to the finish image processing through the image segmentation,feature extraction,which can detect merit from demetirt, the finish image automatic testing function is realized using VC programming too. The scheme,first,detects the central position of the circle to the original image using Hough transform,then segments the bottle mouth image from the whole image according the inside and outside circle radius of the bottle mouth,and segments to several subimage with equal areas and equal direction. The characteristic value of each subimage is extracted using NSCT method,then the average value and variance of characteristic value is obtained to be the subimage feature vector,and nondestructive subimage feature vector is obtained to be the standard value,so the good or bad of the bottle mouth can be judged according the similarity of the subimage feature vector and standard feature vector. The experimental results show that this method is fast, has high detection accuracy and has good practical value.NSCT theory is more perfect, but its application of research is in the initial stage. So its development is bright and expected.更多还原