Key Technology Research Of Curved Surface Bionic Compound Eye 3D Detection

Unlike the single-aperture camera that can only acquire the image information of the target,the bionic curved compound eye system has attracted a lot of attention from scientists at home and abroad because of its ability to acquire the depth information of the target and its characteristics of large field of view and sensitivity to fast moving targets.In this paper,based on the bionic curved compound eye system developed by the group,the application study of bionic curved compound eye system in spatial 3D detection is carried out.In order to achieve high-precision localization of spatial targets,a mathematical model for target localization based on the imaging principle of the compound eye camera is established in this paper;a spot localization experiment is conducted to verify the localization accuracy of the compound eye system;in addition,a 3D reconstruction algorithm is established based on the localization mathematical model and the feature point matching algorithm;experiments on 3D reconstruction of spatial targets are conducted using the compound eye camera.Specific research includes:(1)The imaging principle of the bionic surface compound eye camera was analyzed in detail,the camera calibration parameters were determined according to the imaging process,and the internal camera parameters(including focal length,principal point coordinates,radial aberration,tangential aberration)of each subeye were calibrated using CALibration Tag self-identification calibration board combined with MATLAB stereo calibration toolbox,and the reprojection was calculated according to the calibration results The errors were between0.24-0.28 pixels,indicating a high calibration accuracy.For the calibration of the external parameters of the subeyes,because there is a field of view overlap between the neighboring subeyes,the positional relationship between the neighboring subeyes is first calibrated,and then the external parameters of the non-adjacent subeyes are calibrated by establishing the mapping relationship between the central subeye and the other subeyes.(2)Based on the imaging principle of the camera,a mathematical model of target localization based on the laboratory homemade bionic curved compound eye camera is established.The field-of-view overlap between adjacent subeyes of the bionic curved compound eye system was analyzed in detail.The self-made bionic curved compound eye camera can have at most four non-coherent adjacent subeyes capable of field-of-view overlap when the working distance exceeds 43.69 cm,and the target can be localized using the quadruple vision measurement system.MATLAB software is used to write the image segmentation,spot mass extraction and target localization programs for target localization experiments.The spot localization experiments were conducted using the binocular,trinocular and quadranocular measurement systems,respectively.The experimental results illustrate that the more the number of subeyes involved in localization,the smaller the target localization error.The localization error of the bionic curved compound eye camera is less than 2% within the working distance of 4meters,which proves that the bionic curved compound eye camera can achieve high precision localization of spatial targets.(3)For the bionic surface compound-eye camera,the 3D reconstruction algorithm is proposed by combining the mathematical model of target localization and the feature point matching algorithm.First,the feature point detection is performed on the sub-images of different sub-eyes using the SIFT(Scale Invariant Feature Transform)feature detection algorithm,and after that,the coarse matching of feature points is performed,and the RANSAC(RANdom SAmple Consensus)optimization algorithm is used to reject the incorrectly matched feature points.According to the target localization algorithm,the spatial coordinates of the feature point in the world coordinate system are calculated from the 2D coordinates of the feature point in the subocular pixel coordinate system according to the target localization model and the internal and external parameters calibrated by the camera,and so on,to find the 3D coordinates of all points in space to obtain the complete target 3D point cloud.A 3D point cloud reconstruction program was written using MATLAB software,and reconstruction experiments were conducted.The experimental results illustrate that the bionic surface compound eye camera can perform better3 D point cloud reconstruction of spatial targets.