Research On Real-time Three-dimensional Measurement And Reconstruction Technology Of Structured Light Based On Stereo Phase Unwrapping

In recent years,structured light three-dimensional shape measurement technique based on Phase Measurement Profilometry(PMP)has been widely applied in quality-test,virtual reality and machine vision,with the advantages of high precision,high-efficiency and low-cost.The acquisition of three-dimensional data is becoming more and more important among dynamic scenes,while numerous methods can be only compatible with static ones.The PMP needs no less than three frames to extract the phase information ranging from -? to ?.It is no doubt that the measurement will be more sensitive to the motion,as extra images are essential to remove the discontinuities of the wrapped phase.This thesis presents a novel real-time three-dimensional measurement method based on stereo phase unwrapping technique,and develops a set of measurement system.1.Remove the nonlinear output response of the system using a LUT,and pick up the phase information by means of three-step phase-shifting algorithm.Before the application of stereo phase unwrapping technique,calibration parameter should be prepared in advance.2.The final purpose of this unwrapping technique is to ascertain the period order of wrapped phase in each pixel.In this scheme,an additional camera is introduced,and then we traverse every possible period of fringe patterns and re-project the pixel in the left camera to the other.Still,geometry constraint and axial restraint are adopted to overcome the little-probability candidates,and the final period will be judged with the matching degree between the pixel pair of binocular system.To improve the recognition of each pixel,we also re-design the phase-shifting patterns through embedding the random speckle.Distinguished from the traditional methods,only the area where gray scale is not full use of will be embedded,which will avoid reducing signal to noise ratio(SNR)as well as contrast noise ratio(CNR)of fringe patterns.3.Considering that the compute in each pixel is immune to others in each steps of the algorithm above,we introduce the graphics processing unit based on parallel computing architecture to optimize the procedure.Finally,our system achieves real-time 3D measurement at 26 frames per second with the resolution of 644*484 and the accuracy(RMSE)below 0.1991mm.