Wang Congyi, Gao Xiangdong, Ma Nvjie, Liu Qianwen, Liu Guiqian, Zhang Yanxi
Opt Express. 2022 Jan 17;30(2):1280-1292. doi: 10.1364/OE.446015.
This paper develops an integrative scheme combining new image acquisition, filtering and enhancement methods specified for orthogonal weld defect detection based on magneto-optical imaging (MOI) technique. For image acquisition, a controllable magnetic system enabling rotation of magnetic angles is invented to accurately collect MO images. Multiple images are acquired, yet few are utilized for further processing in the conventional method based on human subjective preferences, bearing chances that images containing defects are discarded. Therefore, we turn to an automated-filtering system to scrutinize MO images and filter effective images through Bhattacharyya coefficient screening method. This not only elevates efficiency and objectivity but also eliminates missed inspection. For image enhancement, normalization method is used to balance the image intensity, followed by image fusion and edge extraction by a two-dimensional gradient method. Our pre- and post-processing approaches significantly improve accuracy in defect recognition and precision in MO images.
本文基于磁光成像（MOI）技术，开发了一种集成方案，该方案结合了用于正交焊缝缺陷检测的新图像采集、滤波和增强方法。对于图像采集，发明了一种能够旋转磁角的可控磁系统，以精确采集MO图像。采集了多个图像，但在基于人类主观偏好的传统方法中，只有少数图像被用于进一步处理，这就存在丢弃包含缺陷图像的可能性。因此，我们转向自动滤波系统来仔细检查MO图像，并通过巴氏系数筛选方法过滤有效图像。这不仅提高了效率和客观性，还消除了漏检。对于图像增强，使用归一化方法来平衡图像强度，然后通过二维梯度方法进行图像融合和边缘提取。我们的预处理和后处理方法显著提高了缺陷识别的准确性和MO图像的精度。
