Abstract: In order to effectively identify and classify weld defects of thin-walled metal canisters in industrial production, a weld defect detection and classification algorithm based on machine vision is proposed in this paper. By using the Gaussian mixture model, a modified background subtraction method is proposed to extract the feature areas of the weld defects. On this basis, it proposes an algorithm for weld detection and classification according to the extracted features, such as the defect areas, the defect brightness and the gray-value curves. Experimental results show that the proposed algorithms can identify and classify the thin-walled weld defects with more than 96% of accuracy rate and can meet the requirement of the real-time and continuous weld defect detection.

Key words: machine vision, weld defect detection, weld defect classification, Gaussian mixture model, background subtraction, curve detection method

摘要： 对于在工业生产中如何有效地识别薄壁金属罐焊缝的缺陷及其类型判别的问题，提出了一种基于机器视觉技术的自动化焊缝缺陷检测及分类算法。利用混合高斯模型，提出了一种改进的背景差分法，主要用来提取焊缝缺陷的特征区域。在此基础上，以不同缺陷类型的缺陷面积、亮度及波形特征等差别作为依据，对焊缝缺陷进行了分类。实验检测结果表明，算法可以对主流的薄壁金属制罐焊缝缺陷类型进行准确的识别和归类，达到了96%以上的精确度。同时，算法的运算时间也能够满足在实际生产中的高实时性需求。

关键词: 机器视觉, 焊缝缺陷检测, 焊缝缺陷类型识别, 混合高斯模型, 背景差分法, 波形检测法