Structured Light Unwrapping Technology And Multi-view Registration For Measurement Data

The structured-light3D shape measurement technology based on phase shifting hasbeen used in many kinds of fields such as reverse engineering, product quality control,heritage repair and so on, for the advantages of non-contact, fast, and high precision.Among the technologies for3D shape measurement based on phase shiftingstructured light, gray code combined with phase-shifting method or multi-frequencyheterodyne principle was generally adopted to complete phase unwrapping. Becausethe multi-frequency heterodyne can get higher precision, it has been widely studied.However, because of the nonlinear influence of the system, the phase containsfluctuation error. Error compensation method has attracted much attention of theresearchers. Because of its stability， high precision and easy operation, theregistration algorithm based on fiducial marker has been widely used in multi-viewdata registration. But the matching efficiency and accuracy based on fiducial markersin the one view point and in different view point is still an important and openproblem. To solve these problems mentioned above, two different phase unwrappingalgorithms, compensation for fluctuation error and the coarse registration algorithmbased on fiducial marker have been studied in this paper.In terms of unwrapping phase we adopt normalization method to solve the unstableproblem of area partition by Gray code when the ambient light is too strong. And weexpend the normalization to modify the sinusoidal of capture fringe image, reduce theimpact of ambient light, noise, reflecting light and the color of the object. The wholemeasurement accuracy is improved by the processing. We propose a method forunwrapping process based on multiple frequency technology and obtain a good result.In terms of error compensation, we analyze the impact of the number of phaseshifting, the phase step and the defocus through simulation experiment. Based on theanalyze result we adjust the parameter of fringe image and the parameter of imageprocessing. The result shown the surface quality of data is improved. Furthermore wecalibrate the system with the method proposed by zhang xu and get the parameter ofGamma for the system. Then the grayscale response curve is modified by the parameter. In terms of data registration, base on the colors and sizes of fiducialmarker, we propose two simple criterions for marker detection, and the markers arewell matched by projecting assisted encoded images on target object. The wrongstereo matching was almost eliminated. Base on Geometric invariance principle wepropose an algorithm to find corresponding common fiducial marker in differentviewpoints. The efficient registration of multi-view data is accomplished.At last the accuracy of the measured data in one viewpoint and multi-view dataregistration results were evaluated. The average error of the measured data in oneviewpoint is0.0452mm. We measured a rotator which called seven viewpoints to getthe complete data. The average error between adjacent viewpoints is0.2mm. Theaverage error of the complete rotator is0.4mm...