厚钢管X射线图像中焊缝区域的检测

发布时间：2018-03-02 16:47

本文选题：厚钢管　切入点：X-ray图像　出处：《湖南大学学报(自然科学版)》2017年04期 　论文类型：期刊论文

【摘要】：由于传统焊缝区域检测算法难以准确提取模糊和对比度低的厚钢管焊缝区域,提出一种新的基于鲁棒PCA模型的焊缝区域检测算法,该算法能克服传统方法的不足,并能准确提取焊缝区域.首先,收集一序列X射线图像,并对其进行空域对齐及亮度归一化预处理.其次,计算得到系列图像的背景图像,并将背景图像与待测试X射线图像张成一个观测矩阵.最后,使用鲁棒PCA算法对观测矩阵进行低秩与稀疏分解,测试图像中的不均匀强度及噪声被消除,焊缝区域被凸显出来,通过全局阈值可将焊缝区域较好地分割出来.实验结果表明,该算法能较大地消除厚钢管X射线图像中噪声及不均匀强度分布带来的干扰、同时增强模糊的焊缝边缘及对比度低的区域,相比传统焊缝区域检测算法,具有更高的检测灵敏度(0.952)和精度(0.989),能更好地将模糊和对比度低的焊缝区域完整检测出来.
[Abstract]:Because it is difficult to extract the weld zone of thick steel pipe with fuzzy and low contrast accurately, a new weld area detection algorithm based on robust PCA model is proposed, which can overcome the shortcomings of the traditional method. Firstly, a sequence of X-ray images is collected and processed with spatial alignment and brightness normalization. Secondly, the background images of the series images are calculated. Finally, using robust PCA algorithm to decompose the observation matrix to low rank and sparse, the non-uniform intensity and noise in the test image are eliminated, and the weld area is highlighted. The weld zone can be well segmented by global threshold. The experimental results show that the proposed algorithm can eliminate the interference of noise and uneven intensity distribution in the X-ray image of thick steel tube. At the same time, the fuzzy weld edge and the low contrast area are enhanced. Compared with the traditional weld area detection algorithm, it has higher detection sensitivity (0.952) and higher precision (0.989), which can better detect the blur and low contrast weld area.
【作者单位】：上海交通大学电子信息与电气工程学院;宝山钢铁股份有限公司研究院;宝山钢铁股份有限公司钢管条钢事业部;
【基金】：国家自然科学基金资助项目(61375020) 国家“973”计划资助项目(2013CB329401)~~
【分类号】：TP391.41

【相似文献】