| Fringe Projection Profilometry(FPP)is a commonly used three-dimensional profile measurement technology,with fast,non-contact characteristics,and widely used in many fields.In this technology,the physical quantity to be measured mainly exists in the phase of the fringe patterns.Therefore,extracting an accurate phase is the key to the research and application of FPP technology.Fourier fringe projection profilometry is one of the commonly used methods to extract phase,In Fourier fringe projection profilometry,the extraction of fringe parts is a key step.Based on the method of machine learning,this thesis proposed a new single-frame fringe projection phase extraction method.Combined with the variational image decomposition(VID),a high-precision single-frame fringe projection phase extraction method is proposed.Aiming at the problem of dynamic measurement of the shape of three-dimensional objects with patterns on the surface,a method based on modified FCM,VID and multi-scale Retinex(MSR)is proposed,and the proposed method is also applied to the measurement of dynamic hand posture with patterns on the surface.The specific research work and innovation are as follows:(1)A single-frame fringe projection phase extraction method based on modified FCM algorithm is proposed.This thesis introduced machine learning algorithm into FPP technology,and proposed a modified FCM algorithm for phase extraction of single-frame fringe projection.The experiments proved its feasibility by applying the proposed method to two simulated patterns and one experimental pattern,and comparing with Fourier transform method,two-dimensional empirical mode decomposition method based on morphology and variational decomposition TV-Hilbert-L2 model.(2)A single-frame fringe projection phase extraction method based on modified FCM algorithm and variational decomposition TV-Hilbert-L2 model is proposed.In order to further improve the accuracy of phase extraction of single-frame fringe projection,this thesis combined the modified FCM algorithm with TV-Hilbert-L2,and used it in the phase extraction of single-frame fringe projection.Experimental results proved that the proposed method achieved similar accuracy to the phase shift method,and could also obtain better results on fringe projection patterns with uneven fringe contrast and multiplicative speckle noise.(3)A three-dimensional shape dynamic measurement method with patterns on the surface is proposed.As far as this thesis knows,there is no literature report on the use of Fourier’s method to measure dynamic three-dimensional objects with patterns on the surface.The modified FCM,variational decomposition TV-Hilbert-L2 model and multi-scale Retinex(MSR)algorithm are combined to measure three-dimensional objects with patterns on the surface.In order to verify the effectiveness of the proposed method in dynamic measurement,this thesis also did an experiment of dynamic shape measurement with patterns on the hand.Experiments showed that the proposed method in this thesis could measure the shape of dynamic three-dimensional objects with patterns on the surface. |
