为提高X射线图像缺陷自动识别的能力与图像分割的效果, 提出了一种基于主成分分析法的X射线焊缝缺陷图像增强与分割算法。该算法首先通过计算图像的协方差矩阵特征值与其对应的特征向量, 并根据特征向量分布, 选择感兴趣区域即图像中的焊缝部分, 从而减少图像处理的计算量; 其次通过分析特征值累计百分比和试验结果, 筛选出最佳的特征向量, 对图像进行基于主成分的重构; 最后采用Otsu阈值分割法, 对重构后的图像进行分割。试验结果表明, 该算法在对比度低、噪声严重的X射线缺陷图像分割中有很好的应用效果。

In order to improve the automated recognition and segmentation in X-ray image of weld defects, an algorithm of X-ray image enhancement and segmentation based principal component analysis (PCA) was proposed. Firstly, the eigenvalue and its corresponding eigenvector of the image covariance matrix were calculated, according to the distribution of eigenvalue, the region of interest (ROI), just as weld, was located, the calculation capacity was reduced; Secondly, through analyzing the eigenvalue cumulative percentage and experimental results, the optimum eigenvector was selected to reconstruct the image based on PCA; Finally, the Otsu thresholding segmentation approach was employed to segment the reconstructed image. The results showed that this algorithm was effective in segmenting the X-ray image which was low contrast and noise severely.