An evaluation method for MEMS-based 3D reconstruction systems

Abstract

Three-dimensional reconstruction technology based on fringe projection profilometry is widely used in industrial measurement, defect detection, and other fields. The lateral and longitudinal resolution of 3D reconstruction is mainly determined by the camera's resolution, while the axial resolution along the z-axis is primarily determined by the accuracy of phase retrieval. In industrial component inspection, the 3D measurement system's ability to distinguish the small height differences is crucial. In this paper, we propose a method to evaluate the system's axial resolution based on plane fitting. Firstly, the proposed method employs a multi-frequency 12-step phase-shifting method to generate the point cloud of a step-like standard part. Then, we generate an image mask by setting the threshold of the modulation intensity to filter out abnormal point clouds. To address the multi-plane extraction problem, we propose a multi-plane fitting method based on the RANSAC framework. This method constructs a model score using orthogonal distances and sequentially extracts planes from the point cloud using the least squares method. Finally, our method determines the system's axial resolution by calculating the distances between planes. Given the importance of axial resolution in industrial inspection, our proposed method has significant practical application value for any given structured light system.