改进K-means聚类算法的自适应Canny算子工件边缘检测

Edge Detection of Workpiece Based on Improved K-Means Clustering Algorithm and Adaptive CannyAlgorithm

针对传统Canny算子在复杂工件边缘检测中存在识别相似度低,工件边缘连续性与鲁棒适应性差的问题,提出了一种改进K-means算法的自适应Canny算子工件边缘检测技术。首先,进行相关复杂工件的形态学预处理操作,通过改进K-means算法进行分割与融合图像;其次,改进Canny算子的梯度方向与自适应度,Otsu阈值分割处理,使用最小二乘法拟合工件孔洞锯齿边界;最后,将得到的边缘结果与传统Canny算子图像进行对比,结果表明改进后图像的峰值信噪比(PSNR)与相似度(SSIM)有所提升。实现了复杂相似工件的自适应边缘检测,为算法结合图像边缘处理技术提供了一定的参考价值。

In order to solve the problems of low recognition similarity,edge continuity of complex workpiece and bad adaptability of robustness in edge detection with traditional Canny algorithm,an improved K-means algorithm based on adaptive Canny operator is proposed.First,the image is segmented and fused by the improved K-means algorithm,which is used to pre-process the related complex mathematical morphology.Secondly,the gradient direction and self-adaptive degree of Canny algorithm are improved,the image of workpiece is optimized by Otsu threshold,and the hole boundary of workpiece is fitted by least square method.Finally,comparing the results with the traditional Canny algorithm image,the results show that the improved image has higher Peak signal-to-noise ratio and similarity,the adaptive edge detection and processing of complex similar workpiece is realized,which provides certain reference value for the algorithm combined with image processing technology.