高斯过程模型圆心定位算法

Center Location Algorithm Based on Gaussian Processes Model

在对圆形目标进行图像处理以实现精密测量的各种任务中,圆心定位的精度直接决定了测量结果的精度。目前,成熟的圆心定位算法包括重心法、Hough变换法、高斯拟合法、圆拟合法以及椭圆拟合法等。这些常用算法各具优缺点,因此有不同的适用场景。本文提出了一种新的研究思路,利用高斯过程模型解决圆心定位的问题,并且提出了分别基于标准圆和椭圆方程的高斯过程圆心定位算法。实验结果表明,高斯过程标准圆中心定位算法与圆拟合算法的精度相当,而且在残缺圆的圆心定位中抗噪能力更好,精度更高;高斯过程椭圆中心定位算法比椭圆拟合法的抗噪能力更强,精度更高。

In various tasks of image processing of circular objects to achieve accurate measurement,the accuracy of the center positioning directly determines the precision of measurement results.At present,mature algorithms of center positioning include center of barycentre model,Hough transformation model,gaussian fitting model,circle fitting model and ellipse fitting method.These commonly used algorithms have their own advantages and disadvantages to apply to different application scenarios.In this paper,a new research idea about center positioning based on gaussian processes model is proposed and gaussian process center positioning algorithms based on standard circle and the ellipse equation are developed respectively.The experimental results show that the gaussian process standard circle center positioning algorithm achieves the same sub-pixel precision as the circle fitting method and is more accurate and robust in dealing with defect circles.The gaussian process ellipse center positioning algorithm achieves higher accuracy and is more robust than the ellipse fitting method.