The Research Of Phase Extraction From Fringe Patterns Based On Frequency Domain Transform

Three dimensional shape measurement technique based on structured light projection has promising prospects in the fields of industrial inspection,reverse engineering,machine vision,and biometric identification etc.Based on frequency domain transform,phase extraction technique can calculate phase from single fringe pattern,thus is widely used in rapid object shape measurement.Fourier transform is prone to cause frequency aliasing and sensitive to noise.To solve this problem,this thesis introduces the windowed Fourier transform and wavelet transform method,and presents the simulation experiment and actual experiment on the application of frequency domain transform method for phase extraction.In order to study the phase extraction algorithm and provide a quantitative analysis method,this thesis presents a virtual fringe projection system,which can simulate the modulated fringe pattern distributed on object surface and the captured image by a camera.A generally applicable explicit calibration method has been put forward to establish the mapping relationship between the phase of fringe pattern and the height of the object.This virtual fringe projection system provides an ideal platform for fringe analysis,calculation of wrapped and unwrapping phase,error analysis,etc.Beginning with the defects of the Fourier transform algorithm,this thesis explains the theory of windowed Fourier transform algorithm and wavelet transform algorithm for phase extraction.In the proposed virtual fringe projection system,effectiveness of windowed Fourier transform algorithm and wavelet transform algorithm are verified by analyzing error of reconstruction of object height and defects of Fourier transform algorithm.Discussion and research on the window size of windowed Fourier transform and selection of wavelet function for wavelet transform under variety of environments are also presented.The experimental results show that the windowed Fourier transform and wavelet transform can reach the accuracy of 0.2mm within measurement range of 100 mm for complex objects surface,which are more accurate than Fourier transform.Through the actual experiment,rapid and accurate three dimensional shape measurement is preliminarily achieved by using windowed Fourier transform and wavelet transform.