改进的Canny算法在直焊缝图像缺陷边缘检测中的应用

对含有夹渣、气孔和裂纹缺陷的直焊缝X射线图像进行缺陷的边缘检测时,传统的Canny算法虽然可以检测出缺陷的边缘,但具有自适应差和容易出现伪边缘的缺点。提出了一种改进的Canny算法,对经过非极大值抑制后的梯度直方图,利用像素最值梯度和像素最值梯度方差自适应获取连接缺陷边缘所需的高、低阈值;同时也改进了计算梯度图像所需的模板。实验结果证明,改进的Canny算法不仅进一步抑制了噪声,也避免了伪边缘的出现,在保留缺陷的边缘细节方面具有良好的效果。

Application of Improved Canny Algorithm in Defect Edge Detection of Straight Weld Image

When defect edge detection is conducted to the X-ray image of a straight weld with clamp slag,pores and cracks,the traditional Canny algorithm,though being able to detect edge defects, is poorly adaptable and prone to false edges. An improved Canny algorithm is presented in this paper, suppressing gradient histogram of pixels by non-maximal value,and the required high and low threshold are adaptively acquired to connect defect edges by using the maximal pixel gradient and the maximal gradient variance;it also improves the template for image gradient calculation. The experimental results show that the improved Canny algorithm not only restrains the noise,but also avoids false edges,and has favorable effect on retaining details of defect edge details.