TL;DR
This paper introduces an improved Circular Fringe Fourier Transform Profilometry (CFFTP) method that enhances accuracy and robustness in 3D surface measurement. By reformulating the displacement calculation from a quadratic to a linear equation and incorporating Gerchberg iteration for phase error correction, the method achieves more reliable results, especially in non-telecentric systems.
What Problem Does This Paper Address?
Traditional Fringe Projection Profilometry (FPP) is a powerful technique for 3D surface measurement, but it often faces challenges with phase ambiguity and system errors. Circular Fringe Fourier Transform Profilometry (CFFTP) offers a way to obtain absolute phase with fewer fringes by utilizing the circular nature of the projected pattern. However, existing CFFTP methods rely on solving a quadratic equation to determine pixel displacement, which can be prone to errors due to phase noise and non-uniform fringe periods, particularly in non-telecentric setups. This can lead to inaccuracies, especially in regions close to the circular center. The paper aims to address these limitations by proposing a more robust and accurate CFFTP method.
Methods
The core of the proposed method lies in two key improvements:
- Linear Equation for Displacement Calculation: Instead of solving a quadratic equation, the authors introduce an additional circular fringe with a lateral shift. This allows the calculation of pixel displacement (related to object height) by solving a linear equation. This simplification avoids the potential root-seeking issues associated with quadratic equations, especially in the presence of phase errors.
- Gerchberg Iteration for Phase Error Correction: To mitigate phase errors that occur in regions close to the circular center (due to spectral leakage), the Gerchberg algorithm is employed. This iterative technique helps to refine the phase information in these critical areas.
- Plane Calibration for System Error Elimination: A crucial step involves establishing a displacement-to-height look-up table through plane calibration. This process accounts for systematic errors inherent in the measurement system, particularly in non-telecentric configurations where fringe periods are not constant. A cubic polynomial fitting method is used to create this table.
- Coordinate Transformation: As with traditional CFFTP, a transformation from Cartesian to Polar coordinates is necessary to process the circular fringes. The paper details this transformation and the interpolation techniques used to handle the gridded sampling in polar space.
Datasets
Not discussed in the paper. The paper focuses on simulations and experimental validation using self-generated fringe patterns and measured objects.
Key Insights and Ideas
- Simplifying the Math, Improving Robustness: Reformulating the displacement calculation from a quadratic to a linear equation is a significant simplification that directly addresses a known source of error and instability in existing CFFTP methods.
- Targeted Error Correction: The application of Gerchberg iteration specifically for the problematic central regions of circular fringes is a clever way to improve accuracy where it's most needed.
- Non-Telecentric Systems are Key: The paper explicitly addresses the challenges of non-telecentric systems, where fringe periods vary. The proposed method, combined with plane calibration, demonstrates its ability to handle these more complex scenarios.
- Look-up Tables for System Calibration: The reliance on a displacement-to-height look-up table derived from plane calibration highlights the importance of system-specific calibration for achieving high accuracy in 3D metrology.
Key Results
- Improved Accuracy: Simulations and experiments show that the improved CFFTP method achieves significantly higher accuracy compared to existing CFFTP methods, as evidenced by lower Root Mean Square Errors (RMSE). For example, in simulations with heights of 20, 30, and 40 mm, the improved method's RMSE was considerably lower than the existing method (e.g., 0.0623 mm vs. 0.1875 mm for 20 mm height).
- Robustness in Non-Telecentric Systems: The method demonstrates good performance even in non-telecentric systems, where the period of the captured reference fringe is not constant. The RMSE distributions at different optical axis angles show that the method remains effective within a reasonable range (10-20 degrees).
- Effective Phase Error Reduction: The Gerchberg iteration successfully eliminates phase errors in the middle region of the reconstructed surfaces.
- Accurate Reconstruction of Objects: Experimental results on flat planes, a gourd-shaped object, and a mask demonstrate the method's ability to accurately reconstruct the 3D shape of out-of-plane objects.
Limitations and Future Work
- Limitations:
- The paper acknowledges that the measurement error of CFFTP might still be larger than linear fringe projection FTP with marker points due to the additional coordinate transformation operations.
- The accuracy is affected by the included angle between the camera optical axis and the projector optical axis; a larger angle leads to a bigger error. The authors suggest setting this angle to an appropriate value (10–20 degrees) for experiments.
- The paper mentions that the period of the encoded fringe should be set to avoid frequency overlapping, similar to traditional FTP.
- Future Work:
- The authors plan to further explore error compensation methods to improve accuracy.
- Specific areas for future research include eliminating errors caused by the non-uniform reference fringe period in non-telecentric systems.
- Investigating insufficient sampling rates in certain areas during the coordinate transformation operation is also a future direction.
Summary updated