Research On Particle Field Measurement Technique Based On Digital Holographic Microscopy

Digital holographic microscopy is a real-time,quantitative,non-contact and high-precision three-dimensional(3D)measurement technique.It has been widely applied in the fields,such as microscopic observation of living cells,3D morphology detection of micro-electronic components,position tracking of microscopic particles and microchannel detection.At present,the detection of microscopic particles is closely related to the development of science and technology.The measurement of particle parameter has the important scientific value for industrial smelting,agricultural production,air quality and other aspects.Therefore,the theoretical and experimental research on particle field measurement is carried out based on digital holographic microscopy technique in the thesis,respectively.A convolutional neural network for the target recognition of adherent particles and single particle is proposed in particle field.A set of optical system based on off-axis digital holography is built for the particle field measurement,in which the spatial position measurement and morphology parameter extraction of particles can be realized.The main research contents of this thesis are as follows:(1)To solve the recognition problem of adherent particles and single particle in particle field,an improved Faster R-CNN network model is proposed in the thesis.Based on Faster RCNN network framework,the balanced feature pyramid,deformable convolutional network and efficient pyramid split attention mechanism are adopted to enhance the network.The network may be suitable for the automatic particle recognition under the conditions of dense distribution,deformation squeezed,adhesion and overlap.Taking public data set as the research object,the target recognition capability of the network model is verified.The improved network model is compared with other network models.Their results show that the improved Faster R-CNN network proposed is better than the original network model and the single-stage mainstream network models in terms of the evaluation index of mean average precision,which lays a foundation for the subsequent particle field measurement.(2)The 3D space coordinate detection of particle field includes two-dimensional(2D)plane coordinate measurement and 3D longitudinal coordinate measurement of particle field.In the 2D plane coordinate measurement of particle field,the particle image preprocessing,binarization,edge detection,Zernike moments edge detection with sub-pixel presicion and Hough transform circle detection are carried out to obtain the 2D plane coordinate of particle.On this basis,the optimal reconstruction distance of the particle is determined by the digital autofocusing method,thus obtaining the Z-axis coordinate of the particle.The 2D particle field and 3D particle field are simulated respectively to verify the feasibility of the 3D coordinate measurement method for particle field,which provides a guidance for the spatial coordinate measurement of the actual particle field.(3)The particle field measurement optical system based on off-axis digital holography is designed and built.The USAF1951 standard resolution plate is used to calibrate the experimental system.In the experiment,10μm polystyrene microspheres are selected as samples,and their holograms in the suspension liquid are captured.The automatic recognition of adherent particles and single particle in the particle field is performed by the improved Faster R-CNN network model.Subsequently,the 3D spatial coordinates of the sparse and dense particle fields are measured,respectively,thus obtaining the 3D spatial coordinates of the particles in the actual particle field.At the same time,the 3D morphology of particles is reconstructed to realize the measurement of particle diameter.The particle field measurement method based on digital holographic microscopy adopted in this thesis has the advantages of high precision and automatic detection,which can be applied in biomedicine,industrial detection,processing and manufacturing and so on.