Research On Transient Three Dimensional Tracer Particle Field Reconstruction Techonology Based On Synthetic Aperture Imaging

The visualization of flow field has important theoretical significance and application value in aerospace,energy engineering,aerodynamics,hydrodynamics,medicine science and many other fields.Benefitting by the advantages of non-invasive,transient and non-interference optical measurement technology,synthetic aperture particle image velocimetry(SAPIV)has ability to observe three dimensional(3D)high resolution and large scale velocity in complex fluid.SAPIV can further present the flow structure,reveal flow pattern,and provide experimental basis for theoretical fluid mechanics research.This Ph.D thesis mainly focus on the problems and requirements of the particles extraction from refocused particle field and particle images refocusing method in 3D transient tracer particle field reconstruction technology.The main works are summarized as follows.Particle images intensity values correction in the camera array.The location difference of different cameras relative to the laser results in the difference of the particle image on camera imaging sensors.In order to alleviate the influence of this problem on particle field reconstruction quality,the sliding minimum subtraction method(SMS)is used to process the scattering particle images.The desirability of the particle scattering image preprocessing and the feasibility of the particle image correction using SMS method are verified by numerical simulation and practical experiment.Research on the adaptive threshold based particle field reconstruction method.The intensity distribution of focused particle centers is studied herein to solve the problem of large deviation in gray-level threshold estimation based on Gaussian distribution model.The iterative process of the combination of particle image and refocused particle field reprojection is implemented.The correlation coefficient is further calculated between the projected particle images and the camera captured particle images.The optimal threshold is obtained adaptively using this method,and the 3D particle field can be reconstructed eventually.Research on the convolutional neural network(CNN)based particle field reconstruction method.The refocused particle is hourglass shaped in the refocused image stacks.A combination of deep learning and SAPIV technique is proposed to reconstruct the focused particles.This deep learning network is used to calculate the probability of the focused particle according to the intensity distribution of the refocused particle field.The refocused particle field is reprojected onto different camera sensors to calculate the structural similarity(SSIM).Thereafter the optimal probability threshold for extracting focused particles can be obtained,and the transient 3D particle field can be reconstructed.This CNN based method resolves the problem of focused particles extraction from the perspective of structural features.This method develops a new train of thought for the study of SAPIV.Research on hybrid refocusing particle field reconstruction.With regard to the isolated noise interference in particle images refocusing process,minimum line-of-sight(Min LOS)refocusing and additive refocusing are performed respectively.The gray-level information from the Min LOS refocused image stacks is extracted as the candidate particle field.The local maximum searching is subsequently used to locate the focused particles.During the 3D particle field reconstruction process,the corresponding intensity values of the focused particles are derived from the additive refocused image stacks.SAPIV experimental research on the flow around the circular cylinder.In this paper,the wake flow observation experiment and the Karman vortex detection experiment are carried out under different illumination conditions.Five different methods are used to reconstruct the three dimensional tracer particle field,and the reconstruction qualities of these five reconstruction methods are compared.The results verify the feasibility of the research work in this paper.The application scope of different methods is also illustrated in this paper.