中文摘要：

针对人工陶瓷水阀片缺陷检测效率低、劳动强度大、容易漏检的问题，提出一种基于支持向量机的陶瓷水阀片表面缺陷图像检测方法。先对陶瓷水阀片进行预处理，然后针对缺边和划痕两种缺陷分别提取陶瓷水阀片样本的Hu不变矩和Gobar纹理特征，最后输入训练好的支持向量机模型进行分类识别。搭建了陶瓷水阀片的缺陷检测实验装置并进行了可行性和对比性实验。结果表明本文构建的系统对缺边和划痕缺陷识别率高，同时对第二类型陶瓷水阀片样本的检测具有较好的适用性。

英文摘要：

Manual defect detection of ceramic valve plate has such problems as low detection efficiency and large labor strength. Furthermore, some minor defects can be skipped over. To solve the problems, an image detection method for surface defects of ceramic water valve plate based on support vector machine （SVM） is proposed. Firstly, the preprocessing was carried out for the ceramic valve plate. Secondly, Hu invariant moment and Gobar texture features were extracted from the ceramic valve plates according to different defects, respectively. Finally, the extracted features were inputted into the trained SVM to classify. The detection experimental facility was constructed. The feasible and comparative experiments were carried out. The results show that the proposed method has high recognition rate of the missing edge and scratches and it has good applicability for the second type of ceramic water valve plates.