一种基于阈值的射线焊接缺陷图像特征提取算法

A Threshold Based Feature Extraction Algorithm for X-Ray Welding Defects

摘要: 本文对来源不同的两部分射线图片数据进行灰度处理，同时去除图片中的标记信息；通过滤波处理，去除绝大多数噪音信号；接着对图像进行Otsu阈值分割算法进行初次特征提取，结果发现靠近边缘的缺陷特征出现了部分丢失，针对这一问题，本文采用闭操作，膨胀处理对图像进行修复。针对焊接缺陷图像背景区域占比远大于特征区域且图像尺寸较大的特征，本文提出了一种基于特征阈值的焊接缺陷特征提取的方法。该方法进一步对焊接缺陷进行特征提取，并构建数据集。通过实验发现，该方法能够很好地提取焊缝缺陷特征，达到预期效果，能够满足后期模型训练的要求。

Abstract: In this paper, two parts of ray image data from different sources were processed in gray scale, and the marking information in the image was removed. Most noise signals are removed by filtering. Then the Otsu threshold segmentation algorithm was used for the initial feature extraction of the image, and it was found that the defect features close to the edge were partially lost. In view of this problem, the paper used the closed operation and expansion processing to repair the image. In view of the fact that the proportion of background area in welding defect image is much larger than that of feature area and the image size is large, a method of welding defect feature extraction based on feature threshold is proposed in this paper. The method is used to extract the features of welding defects and construct a data set. Experiments show that this method can extract the weld defect characteristics well, achieve the expected results, and meet the requirements of later model training.