基于改进Canny算子的焊缝图像边缘提取研究

Research on edge extraction of weld image based on improved Canny operator

针对焊缝图像边缘提取过程中,由于工件材质表面特性和复杂光照环境影响产生噪声、伪边缘干扰的问题,提出一种基于改进Canny算子的焊缝边缘提取方法。在边缘检测阶段,首先采用应用改进的双边滤波器进行滤波降噪,降低噪声的同时保留焊缝边缘梯度信息;其次应用改进的Sobel算子,增加了45°、135°方向进行梯度幅值计算,以提高边缘的定位精度。在边缘连接阶段,通过融合Otus与梯度直方图分析法自适应获取双阈值对图像焊缝进行边缘连接。实验表明,该方法在检测精度等方面优于传统Canny算法,提取的中心曲线与相对于实际焊缝中心曲线标准误差降低了46%,平均绝对偏差降低了35%,能够准确地提取焊缝边缘,同时通过自适应获取双阈值,增强了算法自适应性。

Aiming at the problems of noise and pseudo-edge interference caused by the surface characteristics of workpiece material and complex lighting environment during the process of edge extraction of weld image,a weld edge extraction method based on improved Canny operator is proposed.In the edge detection stage,firstly,the improved bilateral filter is used to preprocess the weld image noise,then the different weights are applied from the spatial domain and the pixel range domain to reduce the noise while retaining the weld edge gradient information;Secondly,the improved Sobel operator is applied to calculate the gradient amplitude in 45°and 135°directions to improve the edge positioning accuracy.In the stage of edge connection,the edge connection of the image weld is performed by fusing Otus and gradient histogram analysis method to obtain a double threshold adaptively.Experiments show that the proposed method is superior to the traditional Canny algorithm in terms of detection accuracy.The standard error between the extracted center curve and the center curve relative to the actual weld seam is reduced by 46%,and the average absolute deviation is reduced by 35%.It can accurately extract the edge of the weld seam and obtain dual thresholds through self-adaptation,which enhances the adaptability of the algorithm.