Research On Circle Fitting Algorithm Based On Robust Regression

Circle detection is one of the common target detection problems in image processing and pattern recognition area.It is of great practical value in the field of machine vision industry detection.Such problems can be summarized as the design algorithm to find the target circle center and radius from the edge points.Circle fitting is a kind of circle detection algorithm,which is also a numerical optimization problem.The traditional circular fitting algorithm includes geometric circular fitting,algebraic circle fitting,and fitting algorithm based on statistical features or circle properties.They calculate the objective function of the loss error using the least squares model,and get the target by minimizing the objective function.The drawback of this method is that when the partial edge points deviate from most of the contours to be fitted,the resulting fitting results will deviate significantly from the ideal target circle.Upon the above shortcomings,this paper presented a robust regression algorithm based on robust regression,the main research results and innovation are as follows:(1)Affected by environmental factors such as lighting and internal factors surface scratches,the contour of the components surface is rather complex.On the basis of existing edge detection algorithm,this paper proposed an edge extraction algorithm based on gradient direction.(2)In contrast to the traditional circular fitting algorithm,the robust regression theory was applied to the geometric distance error of circular fitting,and a circle fitting algorithm based on robust regression was proposed to solve the problem that the outliners made the fitting result bad.(3)A numerical solution based on the gradient descent method was proposed,solving the problem by minimizing the objective function and the stability and convergence of the algorithm were verified to satisfiy the robustness requirements of algorithms in industrial applications.(4)Design experiments to determine the initial the parameters of the function,took the strategy proposed by Barzailai and Borwein to calculate the step length.The test shown that the algorithm decreased the normal average error from 3.5 down to 0.44,and the number of iteration times from 21 to 12,circle fitting accuracy and efficiency was improved significantly.Based on the above research results,the algorithm proposed in this paper solves the problem that the edge of the outliers has a large influence on the target circle fit in the industrial application,which effectively solves the problem of irregular bifurcation or outliers.The algorithm proposed meet the requirements of precision testing,with academic and practical significance.