Abstract: Aiming at the defects such as long time consumption and easy effect by noise of traditional ant colony algorithm,an image sub-pixel edge measure method by combining improved ant colony and Zernike moment was proposed.Two-dimensional gray histogram was used to solve the clustering-center,and Laplace operator clustering,divided image edge points,target points and noise points with the image edge could be extracted with global adaptive pheromone updating.Further,sub-pixel image edge based on Zernike moments was subdivided to improve edge segmentation accuracy.Type 32308 J2/Q bearing was taken as research subject to detect inner and outer ring edge of bearing image with proposed algorithm,and bearing inside and outside diameters size was measured by least square fitting and coordinates calibration.Experimental results showed that the proposed method had superior measured accuracy by comparison with improved Hough transformation.

Key words: ant colony algorithm, Zernike moment, sub-pixel, image edge, bearing

摘要： 针对传统蚁群算法计算耗时、易受噪声影响等缺点,提出一种改进蚁群算法与Zernike矩细分的图像亚像素边缘测量方法。该算法采用二维灰度直方图求解聚类中心、拉普拉斯算子聚类、划分图像边缘点、目标点和噪声点等,利用全局自适应信息素更新方式提取图像边缘,进而通过Zernike矩快速算法细分图像亚像素级别边缘,提高了边缘分割精度。以SKF 32308 J2/Q轴承为研究对象,采用该方法检测了轴承图像的内、外圈边缘,并通过最小二乘拟合,采用标准件进行轴承的坐标标定,测量了轴承内、外径等几何参数,将该测量方法与改进Hough变换的测量结果相比较,证明了该算法具有较高的测量精度。

关键词: 蚁群算法, Zernike矩, 亚像素, 图像边缘, 轴承