Improved phase-to-height mapping method combine with device attitude

Area of Science:

• * Optics and Metrology
• * Computer Vision and Machine Learning

Background:

• * Traditional 3D phase measurement profilometry faces accuracy limitations due to device attitude variations.
• * Existing methods require extensive mapping equations and memory, impacting efficiency.

Purpose of the Study:

• * To develop an improved phase-to-height mapping method that accounts for device attitude.
• * To enhance measurement accuracy and reduce computational complexity in 3D profilometry.

Main Methods:

• * Obtaining unwrapped phase data of the target.
• * Employing a generalized regression neural network (GRNN) to mitigate phase offsets caused by random device attitudes.
• * Implementing an improved phase-to-height mapping by integrating the GRNN-processed phase data.

Main Results:

• * The proposed method achieves high-quality phase-to-height mapping with fewer mapping equations and reduced memory usage.
• * Demonstrated significant accuracy improvements: 44.30% over probabilistic neural network (PNN) and 39.58% over radial basis function neural network (RBFNN) methods.
• * Successfully compensated for phase offsets due to device attitude variations.

Conclusions:

• * The novel method effectively enhances the accuracy and efficiency of 3D phase measurement profilometry.
• * GRNN-based attitude compensation offers a superior alternative to existing nonlinear mapping techniques.
• * This approach is valuable for applications requiring precise 3D surface reconstruction.