随着计算机技术的发展，机器视觉已成为无损检测技术中颇具生命力的一个分支，并开拓出无损检测技术崭新的应用领域。陶瓷材料有许多优良性能，然而它断裂韧性很低，生产工艺比较特殊，成批生产时质量不易控制，因此，研究陶瓷管缺陷识别技术非常有意义。 本文按照模块化设计思想，借助于数字图像处理技术、机电一体化技术、无损检测及评价技术、计算机技术等，深入研究了基于机器视觉图像的实时处理算法，并在此基础上完成了陶瓷管缺陷识别的关键技术研究与系统实现。对实现产品的自动分选、提高生产效率、降低生产成本、减少劳动强度等具有很重要的意义。 通过对陶瓷管特征和应用特点进行分析，阐述了陶瓷管缺陷识别原理，提出了陶瓷管缺陷识别系统的总体方案，重点研究了陶瓷管缺陷识别方法。采用中值滤波对图像平滑，改善了图像质量；通过对多种边缘检测算子进行实验比较，分析了各个边缘检测算子的优缺点，选择了适合于陶瓷管图像处理的边缘检测算子；最后用迭代式阈值分割法求出的最佳阈值对图像进行二值化，通过计算二值缺陷图像的面积和周长后，采用圆形度指标判断陶瓷管是否为合格品，实现正次品的分类。在利用数字图像处理技术研究和分析缺陷图像的过程中，借助于Visual C++6.0来开发检测系统软件，采用模块化编程方法，代码具有很好的可读性和可修改性，软件界面友好，运行稳定而且效率高。

With the development of computer technology, machine vision has become a branch of much more life-force and exploiting a new application area of nondestructive testing. The ceramic material has many excellent characteristics, but its fracture toughness is very low, and the process of production is rather special, so the quality of ceramic tube is difficult to control in batch production, therefore it’s very significant to study detecting recognition of ceramic tube. According to the idea of modularizing design, in virtue of the technologies such as image processing, mechatronics, nondestructive testing and evaluating, PC and so on. A real time processing-algorithm based on machine vision was further studied, on this basis, the paper completed the research on the key technology and realize the system for detecting recognition of ceramic tube. It is very significant to realize auto classifier, enhance productivity, reduce production costs and reduce the labor. Through analyzing the features and application characteristics of ceramic tube, the principle of ceramic tube defects identified has been described, the general scheme of ceramic tube defect identification system has been came out. Using median filter to eliminate the image noise, and the image quality is greatly improved. After analyzes the advantages and disadvantages of each edge detection operator, this paper choose a suitable operator. Finally using the threshold segmentation method for binary image processing, calculate the area and perimeter of the binary defect image, and use the circular degree to judge and realize classifier. During the course of using the digital image processing to research and analyze defect image, the system software is developed by Visual C++6.0 and the modularization programming method, and the code has good readability and modifiability. The software is user-friendly, steadily and efficiently.