Research On Real-time 3d Measurement Method Based On Super-resolution Image Reconstruction

The 3D measurement technology based on fringe projection and phase-shifting fringe analysis is widely used in biological recognition and other fields due to its good accuracy and robustness.3d reconstruction technology of static objects has become increasingly mature.Nowadays,more and more requirements are for real-time reconstruction of objects in a certain scene.The image super-resolution reconstruction technology estimates a high-resolution image from multiple low resolution images with relative displacement in the same scene,which can adapt to complex application environments such as uncertain noise interference,and has a good application prospect in high-precision online 3D measurement.This paper proposes a real-time three-dimensional measurement method based on super-resolution image reconstruction to improve the measurement accuracy.The main work of this paper includes:First,take multiple low resolution fringe images in the same scene,and reconstruct high resolution images based on super-resolution image reconstruction technology.Based on the maximum a posteriori(MAP)method,the Markov Gibbs random field model is used to build a posteriori probability model of high-resolution deformed fringe patterns,and the optimal estimation of high-resolution fringe patterns is obtained by minimizing the objective function.After obtaining the high-resolution fringe pattern,the truncated phase of each point in the image is calculated according to the single frame raster projection model.The diamond phase unwrapping algorithm is used to solve the unwrapping phase,and the 3D data of the measured object under a single camera system is calculated according to the results of camera depth calibration.Then,the left and right camera 3D data are spliced according to the 3D surface splicing technology based on face features to obtain the3 D data under the entire system.Second,analyze and compensate the errors in the real-time 3D process.The look-up table method is used to compensate the error caused by the nonlinear response of the system,the crosstalk coefficient compensation method is used to compensate the error caused by the crosstalk of the color channel,and the background correction method based on the least square method is used to correct the background.The absolute phase obtained by the eight step phase shifting method is taken as the standard phase,and the compensation effect is verified by the standard deviation and other indicators.Finally,the qualitative and quantitative analysis of the three-dimensional data reconstructed by the system is carried out.Two groups of comparative experiments are carried out on the traditional single frame grating projection model and the single frame grating projection model used in this paper,as well as the nearest neighbor interpolation method and the MAP based super-resolution image reconstruction algorithm in this paper to verify the effectiveness and accuracy of the methods proposed in this paper.