Methods For Improving Measuring Accuracy And Speed In Fourier Transform Profilometry

Three dimensional surface measurement has been widely used for machine vision, industry monitoring, biomedicine, real object visualization etc. Fourier transform profilometry based on fringe projection is one of the popular non-contact full-field 3-D measurement methods. This method requires only one frame of deformed fringe pattern reflected by a diffuse object surface, so it is very suitable for real time data acquisition and 3-D measurement of dynamic process. In this paper, we research the filtering problem in frequency domain and an improvement on FTP by using bicolor fringe projection. The main work include :We analyze how to select filtering window to improve the measuring accuracy in FTP when noise exists and does not exist. We present the error distribution caused by different filtering window, and conclude that using flattened Gaussian window gets the best measuring accuracy at low levels of added noise and without noise, while using Hanning window gets the best measuring accuracy at high levels of added noise.We research an improved fast FTP based on bicolor fringe pattern projection. We can eliminate the zero frequency of the deformed fringe and extract a binary mask representing the object profile by acquiring one frame of bicolor sinusoidal fringe. Using the binary mask we expand the fringe data for Discrete Fourier Transform to improve the frequency resolution, and avoid the phase unwrapping error due to theuncertain pixel outside of the object profile.