融合Canny算子和形态学的齿轮图像边缘检测研究

Edge detection of gear image based on Canny operator and morphology

针对含噪齿轮图像边缘检测中存在的难以有效抑制噪声和准确检测出更多真实边缘等问题,将改进的Canny算子和数学形态学算法应用到含噪齿轮图像边缘检测中,提出了一种融合Canny算子和数学形态学的含噪图像边缘检测算法。首先利用了改进的Canny算子边缘检测,接着运用了多尺度多结构数学形态学边缘检测;然后对两幅边缘图像进行了小波分解,得到各层子图像;最后分别对子图像采用了自适应加权融合,并使用小波逆变换重构图像得到了最终的边缘检测图像。实验及研究结果表明:融合算法比单独使用改进的Canny算子、数学形态学去噪效果好、定位精度高、边缘连续清晰,并且当噪声浓度升高时依然具有良好的去噪效果,是一种可行的无监督融合算法。

Aiming at the problem that it is difficult to effectively suppress noise and accurately detect more real edges in the edge detection of noisy gear images,the improved Canny operator and mathematical morphology algorithm were applied to the edge detection of noisy gear images,and a fusion was proposed based on Canny operator and mathematical morphology of the noisy image edge detection algorithm.Firstly,the improved Canny operator edge detection was used,then the multi-scale multi-structure mathematical morphology edge detection was applied.Then,the two edge images were wavelet-decomposed to obtain the sub-layer images.Finally,the sub-images were adaptively weighted and used.The image was reconstructed by inverse wavelet transform to obtain the final edge detection image.The experimental results show that the fusion algorithm is better than the improved Canny operator,the mathematical morphology denoising effect,the positioning accuracy is high,the edge is continuous and clear,and it has good denoising effect when the noise concentration increases.It is a feasible unsupervised fusion algorithm.