Research On Single Image Calibration Method In Confocal Binocular Vision System

Existing surface detection methods for large-scale targets are difficult to meet the requirements of high-resolution technical indicators.In order to solve the bottleneck problem,the research group proposed a binocular vision measurement technology based on confocal scanning to improve the system's resolution.The confocal binocular vision system needs to calibrate the internal parameters of its two optical systems and the structural parameters between the two before m easurement application.The accuracy of the calibration affects the accuracy of the final measurement.The first difficulty of the calibration is that the collected image data cannot be directly extracted by the existing feature point detection algorithm.The second difficulty is that the introduction of the confocal scanning system makes the distortion of the entire system much more complicated than that of a single camera,and it cannot be calibrated simply by introducing mathematical models such as radial distortion.The third difficulty is that the depth of field of the confocal binocular vision system is small,and the collected images are distributed in the focal plane and its vicinity,so the existing calibration methods cannot be applied.Aiming at the above problems,this paper proposes a calibration algorithm based on a single image.First,The paper performs denoising processing based on region processing and enhancement processing based on the priori theory of dark primary colors on the collected confocal images.It can get an image that is easier to extract feature points without changing the details of the image.This paper also proposes a method based on morphological processing to detect the feature points of the grid calibration target,which has a high initial inspection efficiency and is ready for the subsequent distortion correction and parameter calibration.Finally,a small precision target is used with a large-range displacement structure to correct the distortion of the system,and a single image is used to achieve the calibration of the binocular system without considering the distortion model.The specific research content completed is as follows:(1)The paper uses the related theories in vision measurement and the imaging model of the vision system to build a mapping model from three-dimensional points in space to two-dimensional pixel coordinates to determine the parameters that need to be calibrated.It also conducts detailed analysis and derivation of common imaging distortion causes and mathematical models,and completes the derivation analysis and summary of the calculation process,advantages and disadvantages,and applicable premises of traditional calibration algorithms.(2)An optical grid small target combined with a large-range displacement structure is designed as a distortion correction target and a calibration target,which solves the problems of insufficient precision and high cost of large-scale calibration targets.A denoising method based on area processing is proposed,which can effectively denoise the image.and enhance the image based on the dark primary color prior theory.In this paper,the image is enhanced based on the dark primary color prior theory which can protect the tiny details of the original image,adjus t the image from the brightness to make the image details more prominent,and fade the colors of irrelevant components.After image preprocessing,morphology-based methods are used to extract the image line structure for the grid calibration target,and the convolution operation is used to extract the feature points.This method has strong anti-noise ability,and also has the advantages of fewer redundant points and false detection points,and fast calculation speed.(3)The paper proposes a distortion correction technology based on a precision moving target,which can theoretically perform pixel-level correction for the full field of view.After the distortion correction is completed,the operation process is simplified,and a calibration algorithm based on a single image is proposed,which is simple to operate and has good applicability to the instrument in this paper.(4)A binocular vision experimental system based on confocal scanning is built,and the instrument is used to scan multiple moving precision small targets to collect data.Moreover,we perform calibration experiments after preprocessing the images and analyze the experimental results.The paper proposes a calibration method based on single image according to the characteristics of confocal binocular vision system.After the distortion correction of the whole field of view is realized,the system can be calibrated with high precision using only one image,and it provides effective guarantee for the accurate measurement of confocal binocular vision system.