| Phase shifting fringe projection profilometry(PSFPP)is one of the most widely used optical 3D imaging methods for its strong anti-interference ability and high measurement accuracy.By combining with monocular structured light imaging or stereo structured light imaging,it can be applied to many emerging fields such as virtual reality,intelligent manufacturing,nondestructive testing,microscopic imaging,etc.For static or quasi-static scenes in these fields,PSFPP shows superior imaging performance due to time-domain sequence projection,dense fringe coding,and inter-frame consistent features.However,when applied to dynamic scenes,the PSFPP-based structured light imaging technologies encounter the following difficulties and challenges.First,PSFPP obtains the wrapped phase information and fringe order distribution through long-time projection and shooting.The long data acquisition period results in a low frame rate of 3D acquisition in dynamic scenes.Secondly,the dense periodic stripes make the global phase distribution obtained by PSFPP show ambiguity.The expansion of the imaging depth range in the dynamic scenes makes the phase ambiguity increasingly aggravated,thus limiting the depth range of the 3D perception.In addition,unknown motion introduces motion errors into image frames,which greatly weakens the similarity of texture and phase shift between frames,and reduces the accuracy of 3D imaging in dynamic scenes.Finally,the differentiated imaging requirements of different dynamic scenes make the structural light imaging system based on PSFPP unable to be applied,and it is difficult to show excellent dynamic three-dimensional imaging performance.Focusing on difficulties and challenges such as low-frame-rate acquisition,narrow perception range and weak frame similarity in dynamic 3D imaging,the specific research contents and innovations of this dissertation are as follows:1.A high-frame-rate dynamic 3D imaging method with spatial geometric prior constraints is proposed.Fully excavate spatial geometric prior and compress the length of the time-domain acquisition sequence.Specifically,the mathematical model of the spatial distribution of the photoelectric sensing end,modulation projection end,and dynamic 3D scene was established in active structured light imaging technology,and the shadow contour guidance and virtual perspective perception strategies were proposed.The fringe order distribution was obtained by using spatial geometric prior constraints,and two dynamic 3D imaging methods with high frame rate based on N-step PSFPP were established,i.e.,shadow-assisted dynamic 3D imaging and multi-threaded view matching dynamic 3D imaging.The two dynamic 3D imaging methods fully integrate the potential spatial geometric prior.While ensuring ultra-fast and high-fidelity modulation projection of digital fringe patterns,the length of fringe pattern sequence and fringe image sequence is significantly compressed,and the data acquisition cycle of Nstep PSFPP is shortened,so that high frame rate imaging performance can be achieved in dynamic 3D imaging tasks.2.A wide-depth-range dynamic 3D imaging method based on quasi-periodic coding is proposed.Quasi-periodic coding mechanism avoids phase ambiguity and broadens the range of dynamic absolute 3D imaging depth.Specifically,periodic phase coding is a general coding mechanism in active structured light imaging technology based on PSFPP.For dynamic 3D imaging,because of the periodic phase distribution(i.e.,phase ambiguity),short-sequence periodic coding can only reconstruct the absolute 3D surface profile in a narrow depth range,that is a narrow motion range and a narrow topography range.A quasi-periodic encoding and decoding mechanism with global phase specificity was designed,and a progressive phase recovery framework based on quasiperiodic encoding and decoding was constructed to solve the phase ambiguity problem in wide-depth-range imaging with short sequence fringe projection.The dynamic depth perception range of structured light 3D imaging technology was expanded,and widedepth-range imaging was realized in dynamic 3D imaging tasks.3.A high-precision dynamic 3D imaging method with motion artifact compensation is proposed.The motion artifacts diffused into fringe sequences are compensated and the feature consistency between frames of fringe images is restored.Specifically,motion destroys the inter-frame feature consistency of N-step PSFPP in an unknown form and degree,which is manifested as the texture misalignment and phase offset between the fringe images.An error model of motion artifacts combining the intensity domain and phase domain was established to reveal the error diffusion of the intensity domain and phase domain in the N-step PSFPP.Aiming at the difference in texture richness,two kinds of high-precision dynamic 3D imaging techniques under motion artifact compensation were proposed:dynamic 3D imaging based on composite fence image tracking and dynamic 3D imaging based on surface texture tracking.In the complex structured light illumination environment,the two dynamic imaging technologies can extract the subpixel displacements of 3D surfaces,accurately invert the intensity domain and phase domain error distribution,and effectively compensate for motion artifacts,so as to restore the feature consistency between frames,and demonstrate the high precision imaging capability in the dynamic 3D imaging task.4.Carry out optimization research of structured light 3D imaging systems for different dynamic scenarios.Jointing front-end design and back-end optimization improves the 3D imaging performance of structured light under different dynamic scenes.Specifically,aiming at the ultra-fast wide-depth reconstruction requirements of transient scenes,a unified 3D imaging frame rate constraint model involving modulation projection speed,time-domain coding sequence length and photoelectric sensing speed was established,and the inadequate part of the frame rate constraint model was completed.Meanwhile,a high-frame-rate wide-depth 3D imaging system for transient scenes was constructed by using the striped image sequence reuse strategy.The complete 3D reconstruction of time-critical events was reconstructed with a time resolution of 0.001 s.In addition,considering the need of optical imaging system for sufficient exposure,incompressible exposure time causes the motion blur of fringe image.To meet the highprecision measurement requirements of challenging moving surfaces,it is proposed to adjust the exposure time of projection and imaging to actively obtain the motion information,and establish a motion error compensation model integrating motion artifacts and motion blur,so as to improve the dynamic 3D measurement accuracy to 0.043 mm.To sum up,guided by the 3D imaging requirements of actual dynamic scenes and N-step PSFPP as the research object,this dissertation proposes dynamic 3D imaging methods with faster data acquisition speed,wider depth perception range and higher imaging accuracy,solving a series of difficulties and challenges in actual 3D imaging.Finally,the overall optimization and the performance verification of the structured light imaging system for transient scene reconstruction and challenging moving surface measurement are completed,which proves the effectiveness,robustness,and applicability of the proposed 3D imaging methods and system optimization methods. |
