Research On Distortion Correction And Target Reconstruction Technology Of Binocular Fisheye System

With the continuous development of machine vision,conventional pinhole cameras can no longer meet the demands of many scenarios due to their limited field of view.The emergence and development of fisheye cameras is the general trend.Recently,fisheye cameras have shown promising prospects in fields such as defense technology,healthcare,autonomous driving,and navigation.Compared to conventional binocular stereo systems,using a binocular fisheye system for 3D object reconstruction is more efficient but also poses significant technical challenges.Not only does it need to overcome the problem of excessive imaging distortion brought by fisheye cameras,but it also has to deal with the difficulty of image matching caused by the large field of view.This paper focuses on the calibration of fisheye cameras,the correction of fisheye image distortion,and the target reconstruction technology based on binocular fisheye systems.The main research contents and achievements of this paper are as follows:1)The imaging principle of the fisheye camera and four classic fisheye projection models are studied.Investigated two universal fisheye imaging models: odd polynomial model and omnidirectional camera model.The calibration of fisheye cameras based on these two models was achieved using a circular array calibration plate.2)The binocular fisheye system model was established.The system calibration was carried out based on the odd polynomial model.To address the issue of significant system calibration error,the parameters of the system were optimized using Bundle Adjustment,thereby improving the calibration accuracy of the system.3)The distortion correction method of fisheye images was studied.Four calibration methods were employed to correct the distortion of fisheye images,including distortion correction methods based on the odd polynomial model and the omnidirectional camera model,as well as distortion correction methods based on the double longitude mapping model and spherical perspective projection model.The distortion correction method based on the odd polynomial model performed best,with the corrected image distortion rate of 0.7351%.4)The method of 3D reconstruction using binocular fisheye system is studied.Firstly,stereo rectification of the target object image pair was accomplished using system calibration parameters.Then,the SGBM algorithm was used to perform stereo matching on the fisheye image to obtain the disparity map.Finally,the 3D coordinates of the target object points were calculated and point clouds were generated based on the parameters of the rectified system and the disparity map.The reconstructed object was visualized using open3 D,with an error of8.0160% between the dimensions of the reconstructed target and the actual size.