研究目的
To develop a high-speed 3D shape measurement method that uses Fourier transform profilometry and stereo vision to reconstruct 3D surfaces without the need for phase unwrapping, thereby improving speed and accuracy.
研究成果
The proposed method successfully enables high-speed 3D shape measurement using a single fringe pattern without phase unwrapping, achieving higher accuracy and faster processing times compared to existing methods like LSSM and MFH. It is effective for complex and discontinuous objects, with potential applications in industrial settings.
研究不足
The method may be sensitive to the quality of fringe patterns and could have challenges with very complex or highly reflective surfaces. It relies on stereo vision calibration accuracy and may require optimization for different environmental conditions.
1:Experimental Design and Method Selection:
The method combines Fourier transform profilometry (FTP) with stereo vision to avoid phase unwrapping. It uses dithering fringe patterns to mitigate gamma effects and includes original image matching, phase matching, and sub-pixel parallax optimization for accurate point correspondence.
2:Sample Selection and Data Sources:
A white mask and a white house with complex surfaces are used as test objects. Data is captured using two CCD cameras and a projector.
3:List of Experimental Equipment and Materials:
Projector (Samsung SP-P310MEMX), digital CCD cameras (Daheng MER-500-14U3M/C-L), lenses (Computar M1614-MP with 16 mm focal length).
4:Experimental Procedures and Operational Workflow:
Steps include generating dithering patterns, stereo vision calibration, capturing images from left and right cameras, applying FTP to get wrapped phase, using original image matching for rough parallax, phase matching to find candidate points, average phase calculation for edge points, and sub-pixel parallax optimization for precise matching. Height is calculated based on calibrated stereo vision.
5:Data Analysis Methods:
MATLAB 2015a is used for processing. Metrics include average height, RMS error, average error, and maximum error, compared with LSSM and MFH methods.
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