Research On Camera Calibration Method Based On Circular Mode Plane Target

Machine vision is a low-cost,highly effective and contact-free technology that simulates human eyes with machines for measurement and judgement.It can serve the needs of many fields,including video surveillance,manufacturing and medical imaging.Camera calibration,as an important part of machine vision,its calibration accuracy has direct impact on the margin of error of visual inspection.In this sense,it is of great significance to research on the improvement of camera calibration method.The planer circular pattern is widely employed since it is easy to produce and can recognize low-quality images.However,drawbacks still exist in the traditional camera calibration method based on the planer circular pattern.For starters,the traditional ellipse feature coordinates need to be calculated through a set of pre-calculated contour points,thus their accuracy is largely affected by external factors such as illumination.Furthermore,considering the ellipse eccentricity error engendered by lens distortion and projection deviation,the center of the imaged ellipse is not the corresponding feature point of the center of the planar pattern.Finally,it is easy to fall into local optimum when using the Levenberg-Marquardt optimization algorithm for camera parameter optimization.These problems may all damage the accuracy of camera calibration.This research aims to improve the camera calibration method based on the geometric features of eclipse and the characteristics of camera model.The main steps are as follows:(1)Leveraging the properties of dual conics,the parameters of the dual conic is fit according to the gradient vector field at the edge of the ellipse.It replaces the calculation of contour points of the eclipse with dual conics,which drives up the accuracy of the coordinates of the feature points.This method is tested and evaluated in simulation experiment,and compared with other ellipse fitting methods.It turns out that the Euclidean distance error of the eclipse center coordinates is reduced by 94.9%.This anti-noise and accurate method delivers a highly precise detection of the ellipse center.(2)The inclined ellipse is calibrated back to an approximate circle after calibrating the perspective projection and lens distortion of the pattern.The approximate circle is refitted,while iteration delivers more accurate imaging point coordinates of the circle center.Iterative calibration compensates the eccentricity error and lower its influence to a minimum.In other words,the calibration results turn out to be more accurate without the help of eccentricity error model.Compared with traditional calibration methods,this method is of greater accuracy and efficiency since its results are 81.8% more accurate than before,and convergence can be realized after simply four iterations.(3)For the optimization of camera parameters,the improved gray wolf optimization algorithm is used,chaotic sequence is employed to improve the initialized results of the gray wolf population,transform the linear convergence factor into a nonlinear one,and add fitness weight to upgrade the position update equation to avoid local optimum.It turns out that the improved algorithm can be reasonably applied to the optimization of camera parameters.The calibration accuracy is 52.6% more accurate than that of LM optimization,while the repetitive experiments also report less fluctuation.in summary,the improved optimization performs better in accuracy and stability.