Research On 3D Particle Location Reconstruction Method Based On Light Field Camera

This paper proposes an improved inversion algorithm which is easy to implement,aiming at solving problems of time-consuming calculation of weighting matrix and complicated definition steps for 3D spatial position reconstruction of particles.This method meets the accuracy requirement,while significantly reducing calculation time,saving cost of equipment.It provides a basis for locating much smaller particles with a light field camera combined with a microscopic device.It has good practical application value.The thesis verifies feasibility of reconstructing position of particles by improved MART algorithm through program simulation and experiments.Research work and content of the thesis mainly include the following parts:(1)The biplane parameter method is used to establish a ray forward tracing model to simulate the imaging effect of particles at different positions.There are differences in imaging of the same particle at different depth positions.The image of particles which are in focal plane converges clearly,while the image of particles which are in defocus plane diverges.When particles are farther from the focal plane,the divergence degree of image will be greater,and the number of non-zero pixels of macro pixel will decrease.(2)A tomographic inversion model for reverse ray tracing is established.It is proposed to use dimensionality reduction method to calculate weighting matrix of MART algorithm to reduce the operation time.But it will bring problems of large depth position calculation error and low spatial positioning accuracy.Therefore,the algorithm is improved: The particle depth position is calculated by similar triangles formed in light path.The position in other two coordinate axes is still implemented by MART algorithm.Thus,an improved MART algorithm is formed,which can improve accuracy and reduce calculation time at the same time.(3)Regularly and randomly distributed particles in the reconstruction space are respectively simulated by 3D position reconstruction using MART algorithm and improved MART algorithm.The absolute error of particle position inversion using MART algorithm is ± 2.9mm,and the depth error is slightly less than ± 2.9mm.The absolute error of particle position inversion using improved MART algorithm is± 0.59 mm,and the depth error is slightly less than ± 0.59 mm.Simulation results prove that improved MART algorithm has a good particle position reconstruction effect.(4)Regular and randomly distributed particles in the reconstruction space are respectively subjected to 3D position reconstruction experiments using MART algorithm and improved MART algorithm.The relative error of particle position inversion using MART algorithm is 5.16%,and the depth error is ± 9.45 mm.The relative error of particle position inversion using improved MART algorithm is1.09%,and the depth error is ± 2.01 mm.Experimental results prove that improved MART algorithm does have a good effect of particle position reconstruction.