Research On Calibration Techniques For Trinocular Vision Measurement Systems

The rapid development of computer vision and artificial intelligence has promoted the continuous upgrading of visual inspection technology,making visual inspection systems capable of handling more complex and diverse image data,providing strong technical support for aerospace telemetry,military reconnaissance,medical imaging and many other fields.When conducting range tests,the parametric test system built by two high-speed cameras is usually required to record the flight of projectiles,fragments,etc.to obtain their speed,angle,attitude and other information to make an assessment of the effectiveness of destruction.The traditional binocular measurement system is limited by the baseline length and camera resolution,which makes it unable to meet the measurement needs of larger scenes,and cannot get accurate measurement results for overlapping or blocking situations between fragments.Therefore,the study of trinocular measurement system with larger field of view and higher measurement accuracy is particularly important.This paper focuses on the calibration technology of trinocular vision measurement system,aiming to improve the accuracy and stability of the measurement system by optimizing the calibration method.The specific research contents include:(1)In-depth analysis of the trinocular vision measurement system’s trinocular calibration principle and method.After the calibration of a single camera is completed,the system calibration parameters of the trinocular system are obtained by combining the binocular stereo imaging model and using the multi-view geometry theory to calculate the system transformation matrix between cameras.At the same time,an aberration model suitable for the trinocular vision measurement system is introduced,and the initial aberration coefficient is estimated approximately by the aberration parameter estimation method based on the single-strain matrix,followed by a more accurate aberration coefficient using nonlinear iterative optimization.(2)The detection method of the feature points of the calibrator is improved.Aiming at the special diagonal characteristics of the checkerboard grid,this paper proposes an algorithm for detecting the corner points in the checkerboard grid based on the image gray-scale features.Firstly,the corner points are pre-screened to eliminate the non-corner point regions,and the refined set of inner corner points is obtained based on the grayscale features of the corner points of the checkerboard grid and combined with the BW operator.The good performance of the algorithm is proved by the experiments under different light intensity and noise environment,and it can meet the demand of high accuracy of camera calibration.(3)A method for solving the 3D coordinates of the trinocular system is studied.The method makes full use of the conversion relationship between the internal and external parameters of the three cameras,unites the structural parameters and other information between the three camera pairs to solve the initial values,and unifies the coordinates of the three groups under the reference camera coordinate system through the conversion,and obtains the coordinates of the measured point by the method of center of gravity for optimal estimation.(4)Test experiments were conducted on the Matlab experimental platform for the trinocular vision measurement calibration scheme proposed in this paper.Firstly,the measurement accuracy experiments of the trinocular measurement system were conducted.In the test depth of 3.0m,the average reprojection error of the system is 0.12 pixel;when the physical measurement is carried out,the relative error between the reconstructed length and the real length is less than 2.3%,which can achieve high accuracy;to verify the influence of different measurement depths on the accuracy and stability of the system,the test experiments of the object being measured at 1.5m,3m and 7m from the reference camera are conducted;through the comparison experiments with the The comparison experiment with binocular measurement system proves that the trinocular vision measurement system of this paper has higher accuracy and stability compared with the traditional binocular system,and can accurately extract the internal and external parameters of the camera and correct the distortion,and also verifies the reliability and validity of the calibration method of this paper,which provides the basic guarantee for the high accuracy measurement in practical applications.