Research On Key Techniques Of 3D Measurement Based On Digital Fringe Projection

With the merits of non-contact,high resolution,high precision,high efficiency,high universality,low cost and full field measurement,three-dimensional(3D)measurement technology with digital fringe projection is widely used in automobile industry,aerospace,new energy industry,heavy industry manufacturing,transportation and other fields.It is the hot spot of scientific research and product development in the field of optical 3D measurement.At present,the existing measurement products are expensive,and most of them use the dual frequency heterodyne phase resolution technology,which takes a long time in image acquisition,limiting its application in the field of high-speed dynamic measurement.This paper focuses on the key technologies of digital fringe projection measurement,such as phase demodulation,phase nonlinear error correction,system calibration,mapping model establishment and so on,to explore an efficient,high-precision and low-cost 3D measurement method of digital fringe projection.In digital fringe projection measurement,the accurate calculation of the phase is an important prerequisite to ensure the measurement accuracy of the system.The standard N-step phase-shifting algorithm has strong anti-interference characteristics to the measurement environment,and it is the main method to calculate the phase at present.However,the value range of the phase obtained by the phase shifting method is[-?,?],which changes periodically along the direction of the fringe projection and is called the wrapped phase.It is necessary to determine the fringe orders of the wrapped phase to obtain the absolute phase.In order to meet the requirements of high efficiency and high accuracy of measurement,two time phase demodulation methods are put forward in this paper:phase demodulation based on dual variable-frequency coded fringes and phase demodulation based on phase coded fringes.The two methods mentioned above only need to project two additional coded patterns to complete the phase orders decoding.Among them,the method based on the dual variable-frequency coded fringes is to encode frequency of the coded fringe in equal period,and the order of sinusoidal fringe in different periods corresponds to the frequency of the coded fringe one by one.Fast Fourier Transform(FFT)algorithm is used to calculate the frequency of the coded fringe with complete period,and then the phase order decoding of the whole measurement area can be realized by combining the consistency of the phase orders in the fringe direction and the continuity in fringe projection direction.The method based on phase coded fringes is to encode phase of the coded fringe in equal period.To improve the maximum decodable orders and the stability of decoding,the coded phase is arranged in terms of a certain sequence.Before the phase demodulation,the connected regions of the coded pattern is identified and classified,which can simplify the matching process of the coded phase and the phase order and improve the decoding efficiency.The digital fringe projection measurement system is constructed to decode the phase order of objects with different geometric structures.The experimental results show that the two proposed phase demodulation methods can achieve accurate phase demodulation.Since the wrapped phase is consistent with the decoding interval of the coded fringes,therefore,the absolute phase calculated by the phase demodulation method based on the dual variable-frequency coded fringes has no phase error of 2? at the boundary of the fringe in different periods.And the phase demodulation method based on phase coded fringes is more stable when the coded numbers from four consecutive periods are used for phase order matching.In order to meet the human visual perception,the digital projectors on the market will undergo gamma correction during production,resulting in the sinusoidal fringe projected by the projectors showing non-sinusoidal characteristics.Therefore,the calculated phase based on acquired sinusoidal patterns has non-linear error compared with the actual phase.At present,many methods have been developed to modulate the projected fringe directly or to compensate the calculated phase to reduce the nonlinear phase error caused by the gamma effect of the projector.In view of the fact that there is still high frequency errors in the phase difference after phase compensation,this paper propose a phase difference extraction method based on wavelet transform(WT)and orthogonal matching pursuit(OMP)algorithm.Wavelet transform is used to decompose the phase difference in multiple levels,and then orthogonal matching tracking algorithm is used to determine the sparse solution that makes the difference between the reconstructed phase difference and the original phase difference meet the set threshold value.Finally,the reconstructed phase difference is obtained based on the calculated decomposition coefficient and sparse solution.The experimental results show that the reconstructed phase difference based on WT-OMP method retains the effective low-frequency component of the original phase difference and reduces the amplitude of the high frequency error component.Using the phase difference extracted by this method to reconstruct the object in three dimensions can effectively reduce the measurement error.When establishing the fringe projection measurement model,on the one hand,the measurement accuracy should be considered,on the other hand,the flexibility of model calibration and sample data acquisition should be considered.In this paper,two kinds of 3D coordinate mapping models based on the geometric relationship of measurement system,camera imaging and projector projection principle are proposed.One is through the analysis of measurement principle to establish the mapping model of pixel coordinate(u,v)and phase difference ?? to Z coordinate.And then,based on Z coordinate,phase difference ?? and pixel coordinate(u,v),the mapping model of X coordinate and Y coordinate are built respectively.The parameters to be calibrated in the model are all constants related to the system structure.The structure of the measurement model is simple and the mapping relationship is clear,so nonlinear least square algorithm is used to calibrate the model parameters.The other is based on the mapping model of ideal measurement system,which is established by generalizing the world coordinates and imaging plane.Due to the distortion effect of the camera,the relationship of the mapping model is complex,but the input(u,v,?(p)and output(X,Y,Z)of the model are clear.Therefore,ELM network with good generalization performance is selected to calibrate the model parameters.The experimental results show that the two measurement models can achieve 3D measurement of the object with high precision.The measurement model based on nonlinear least square algorithm has a simple structure,and only a small amount of sample data is need to complete the system calibration.The measurement model based on ELM network needs a certain amount of sample data for system calibration,but it has a higher universality.In this paper,two phase demodulation methods based on few coded patterns are proposed,which ensure the accuracy of demodulation and reduce the time of image acquisition by the camera,and provide ideas for improving the measurement speed of digital fringe projection.Using machine learning method to calibrate the measurement system can simplify the calibration process and improve the measurement accuracy,which provides theoretical and experimental basis for the research of 3D measurement model.