Calibration Technique For Fringe Projection Three-Dimensional Measurement System

Fringe projection profilometry is one of the widely used 3D shape measuring techniques, which plays an important role in many areas, such as industry, security, health, entertainment and so on. It includes phase measurement, camera calibration and system calibration, all of which have significant influence on the measurement accuracy. In this dissertation, we focus on the system calibration section, which is vital to map the 2D image coordinates and the 3D world coordinates, and then we can measure the 3D shape of objects. The main work in this dissertation is outlined as follows:1. A new model of fringe projection 3D shape measurement system is proposed. Compared to the strict constraints of spatial location of the conventional model, this new model relaxes the position requirement of the CCD camera and the projector, so it can be easy to adjust the measurement system, and this can be used for field measurement, which improves the practicability of fringe projection 3D shape measurement system. Based on the new model, a new phase-height mapping relation formula can be acquired. The undetermined coefficients of this formula has no relationship with the pixel coordinate system, so the amount of sampling points for coefficients calibration is reduced, and it can save time of system calibration. The experiments verify that this new method has high accuracy, and can measure the objects with complex shapes.2. To solve the complex nonlinear phase-height mapping issue, a neural network is employed to solve this problem. There are 42 undetermined coefficients in the phase-height formula when considering the distortion of the camera lens, which cannot be calculated effectively, and this can cause the instability of measurement system. So the powerful ability of function approximation of a neural network can be used to get the mapping relationship. The structure of the neural network and the training principle is introduced. The experiments verify that the neural network method can achieve higher measurement precision and instability.