Abstract: To achieve visual detection of small signal in X-ray defect-image of precision weldment,the segmentation method of weal image information with strong noise and large gray gradient background was studied.An adaptive gradient-threshold method based on the scan line was proposed to quickly identify suspicious defect-region,which significantly reduced the amount of data and the segmentation difficulty.The algorithm to extract small signal in X-ray image of weld defects was developed based on the anti-geometric diffusion model.Through a series of adaptive threshold surfaces-formed by anti-geometric diffusion process,the weal image information of defects was marked and classified,and the noise was identified according to the improved classification rules.The experimental results indicated that the method could accurately extract weal signal in X-ray defect-image with uneven brightness,low contrast and strong noise.The distortion was low and the efficiency was high.

Key words: X-ray image, welding seam, defect detection, anti-geometric diffusion, weal signal

摘要： 为实现精密焊件内部缺陷的X射线微弱信号视觉检测,研究在强噪声、大灰度梯度背景下微弱图像信息的分割方法。首先提出一种基于扫描线的自适应梯度阈值方法来快速确定可疑缺陷区域,以减少数据处理量、降低分割难度;然后提出基于反几何扩散模型的焊缝缺陷X射线微弱信号提取算法,利用在反几何扩散过程中形成的一系列自适应阈值面,根据改进的分类规则对缺陷微弱图像信息进行标记分割,同时识别噪声。实验表明,该方法能在亮度不均匀、边缘模糊、强噪声的X射线图像中准确提取出焊缝缺陷微弱信号,失真小且效率高。

关键词: X射线图像, 焊缝, 缺陷检测, 反几何扩散, 微弱信号