Study On Phase Unwrapping Technology In Optical 3-D Profilometry Based On Minimum-norm

The technology of optical three-dimensional (3-D) profilometry, with an ability of high-accuracy non-contact measurement, has a wide application in industrial production and CAD/CAM. Phase unwrapping is one of the key technologies that enable optical 3-D profilometry for surface profile measurement. The reliability and accuracy of the measurement depend to a great extent on the solution of the phase unwrapping problem. For many years people made a lot of the research on phase unwrapping. However, the phase unwrapping problem is far from being resolved. On the basis of referring to existing phase unwrapping methods, the in-depth researches on the minimum-norm phase unwrapping methods are carried out in this thesis. The main work of the thesis is as follows.1.According to phase unwrapping based on whether the branch settings, minimum-norm phase unwrapping methods include path-dependence algotithms and path-independence algorithms, and the latter include least squares algorithm based on the matrix equation. This thesis completed the derivation of discrete poisson equation and weighted discrete poisson equation.2.For phase discontinuity arising from the hole and edge steep side, an improved quality map integrating edge detection information is proposed based on the study of traditional quality maps to guide the phase unwrapping algorithms(CG algorithm, PCG algorithm and Picard algorithm). Computer simulation and experimental results have verified that CG algorithm has the slowest convergence speed and the lowest iterative accuracy, the convergence speed of Picard algorithm is faster than CG algorithm, and that PCG algorithm has the fastest convergence speed and the highest iterative accuracy. And the results also have verified that the improved quality map is better than traditonal quality maps, and that PCG algorithm effectively avoids phase discontinuity regions and has higher accuracy.3.System calibration converts the two-dimensional unwrapped phase data into three-dimensional data, and is the key of 3-D measurement profilometry. For the problem of large amount of computing and poor sensitivity in explicit system calibration methods, in this thesis a new implicit system calibration method which has relatively high scaling precision and efficiency is brought up based on combination of BP and RBF neural networks, with using traditional calibration methods for reference and using the pixel method to extract pixel coordinates. Experimental results show that this new method has good promoting accuracy and learning accuracy.Experimental results have verified that phase unwrapping algorithms referring to improved quality map improve phase unwrapping accuracy, and that system calibration method based on combination of BP and RBF neural networks improves scaling accuracy.