Research On Calibration And 3D Reconstruction Of Long-focus Free Binocular Stereo Vision System

The two cameras in the free binocular stereo vision system are independent of each other and can rotate freely.It is suitable for the 3D reconstruction of large scene targets and is a heated research topic in the field of computer vision.The calibration of the initial parameters of the system and the real-time estimation of the relative pose of the camera are important steps to realize the 3D reconstruction of the free binocular stereo vision system.To improve the resolution of the 3D reconstruction,a long focal length lens is required.However,the long focal length system has a small depth of field,and the imaging is susceptible to vibration and atmospheric turbulence disturbances.During the initial parameter calibration of the system,the collected target images have different degrees of blur,which affects the calibration accuracy.The field of view of the long focal length system is small.When the relative pose of the camera is estimated in real-time,it is difficult to shoot the same calibration object from multiple positions to establish the relationship between the relative pose of two cameras and the mechanical movement of the system.This paper studies the calibration method of the binocular stereo vision system with high adaptability to fuzzy noise and the real-time estimation of the camera's relative pose of the long focal length system.The main work and innovations are as follows:1.Proposing a calibration method based on color-coded phase shift stripes.Using a liquid crystal display panel as the calibration target,horizontal and vertical color-coded phase shift stripes are displayed in sequence.Through color channel separation,orthogonal phase shift stripes are obtained.Based on the phase shift theory,the intersection point of the orthogonal phase truncation line is regarded as the feature point.The target position and posture are changed multiple times to extract the feature point for realizing the calibration of the binocular stereo vision system with the application of the calibration theory based on the planar two-dimensional target.Experiment's result shows that this method has high calibration accuracy and good robustness when the image is blurred.2.The real-time estimation method of camera relative pose based on 2D-3D features is studied.The initial parameters of the system are known,and the world coordinates of the reference points in the area to be measured can be obtained by the least squares three-dimensional reconstruction theory.After the system rotates,respectively solving the poses of the two cameras relative to the world coordinate system by rotating the world coordinates of reference points in front and rear public view,as well as coordinates of their corresponding images.Experiment's result shows that this method has high precision of relative pose estimation.3.The long-focus binocular stereo vision system was built outdoors.Through the calibration and reconstruction experiments at distances of 50 m,75m,and 100 m,the experimental results show that the proposed method is suitable for the calibration of long-focus systems at long distances.The scanning reconstruction method based on the parallax method is studied,and a long focal length free binocular stereo vision system is built to scan and reconstruct the scene to be tested,and the reconstruction effect is good.