Font Size: a A A

High-precision Target Scene Reconstruction Of Mobile Robot Based On PSO-GSA And GPR

Posted on:2023-07-07Degree:MasterType:Thesis
Country:ChinaCandidate:D W YangFull Text:PDF
GTID:2558307040482424Subject:Engineering
Abstract/Summary:PDF Full Text Request
3D reconstruction technology is the main component of indoor target scene reconstruction,industrial visual inspection,and unmanned driving technology,and it is also an important research direction of computer vision.The 3D reconstruction implementation method mainly uses stereo vision to obtain the three-dimensional information of the object.At present,the most common 3D reconstruction methods are active 3D reconstruction technology and passive 3D reconstruction technology.The main research content of this thesis is the 3D reconstruction of indoor target scenes based on RGB-D cameras based on the principle of active measurement.The RGB-D camera is influenced during the reconstruction process by the camera calibration algorithm,external environment interference,and system error,resulting in low reconstruction accuracy.To solve the above problems,this thesis proposes a 3D visual reconstruction technology based on the PSO-GSA algorithm and Gaussian process regression compensation method.The specific contents are as follows:(1)Aiming at the situation that the reprojection error obtained by the Zhang Zhengyou camera calibration method is too large,a camera calibration method based on the Particle Swarm Optimization-Gravitational Search Algorithm(PSO-GSA)is proposed.The method combines the development ability of the Particle Swarm Algorithm(PSO)with the search ability of the Gravity Search Algorithm(GSA),and performs secondary optimization on the internal parameters,external parameters and distortion coefficients of the solved camera,which improves the accuracy of camera calibration.By comparing and analyzing the calibration method based on the PSO-GSA algorithm,the calibration method based on the PSO algorithm,the calibration method based on the GSA algorithm and the Zhang Zhengyou calibration method,it is finally verified that the calibration method based on the PSO-GSA algorithm has the best reprojection effect and the highest accuracy,and the reprojection error is reduced by about 20% on average.(2)RGB-D cameras can get color images and depth images.The depth image describes the position relationship between the camera and the measured object,and its measurement accuracy directly affects the subsequent 3D reconstruction.In order to improve the measurement accuracy of the RGB-D camera,a systematic error correction model based on the principle of pinhole imaging is proposed to correct images on the depth map at different positions.Aiming at the measurement error,a depth error compensation model based on Gaussian process regression method is proposed.The compensation model and the error look-up table method are used to compensate the nonlinear measurement distance to further improve the measurement accuracy of the camera.(3)First,the camera calibration method based on the PSO-GSA algorithm and the compensation model based on the GPR algorithm are applied to the mobile robot(Turtlebot2)carrying the RGB-D camera.Then control the mobile robot to move to different positions and use the RGB-D camera to obtain the depth map and color map under each frame.The ICP algorithm is used to calculate the pose relationship between the current frame and the previous frame,and the point cloud of the current position is fused into the TSDF network model.Finally,the 3D reconstruction of the target object is realized.Through the reconstruction experimental analysis of the test object,the compensated point cloud reconstruction error can be effectively reduced by about 67%,and the reconstruction accuracy is effectively improved.
Keywords/Search Tags:3D Reconstruction, RGB-D Camera, PSO-GSA, Camera Calibration, Error Compensation, Gaussian Process Regression
PDF Full Text Request
Related items