研究目的
To solve the problem of unifying the coordinate system of distinct cameras in a multi-camera system for complete object measurement by proposing a global calibration method using a three-dimensional cube calibration object and a correction method to minimize errors.
研究成果
The proposed multi-camera calibration method using a 3D calibration object effectively unifies camera coordinate systems and improves measurement accuracy through correction. Experiments demonstrate feasibility and effectiveness, with reduced errors and enhanced point cloud quality. Future work will focus on 360° surface reconstruction with multi-perspective systems.
研究不足
The method requires prior knowledge of camera information for registration, and when extended to systems with more than two perspectives (e.g., four-perspective), issues like error transfer between multiple cameras may arise, necessitating additional optimization methods.
1：Experimental Design and Method Selection:
A multi-camera measurement system is built using fringe projection profilometry (FPP) for non-contacting measurement. The system consists of multiple measurement units (each with a camera and projector), calibrated using a pinhole model and phase-shifting algorithms. A 3D cube calibration object with chessboard patterns on four sides is designed to unify world coordinates. A correction method based on deviation between world and reconstructed coordinates is applied to improve point cloud accuracy.
2：Sample Selection and Data Sources:
Feature points on the chessboard patterns of the cube calibration object are used, with known spatial coordinates. Images are captured by cameras during calibration and measurement phases.
3：List of Experimental Equipment and Materials:
Basler acA1300–30um cameras, BenQ GP-30 projectors, a Dell-PC computer, a precisely designed cube calibration object, and a Venus plaster model for verification.
4：Experimental Procedures and Operational Workflow:
Adjust camera positions to capture one side of the calibration object; move the object repeatedly; project fringe patterns; capture images; extract feature points; compute phase maps; calibrate cameras and projectors; reconstruct 3D coordinates; apply correction parameters; merge point clouds; verify accuracy with additional experiments using a chessboard planar pattern.
5：Data Analysis Methods:
Use least squares method for plane fitting, Rodrigues' rotation formula for rotation matrix calculation, and root mean square (RMS) error analysis to evaluate accuracy before and after correction.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
