Algorithm Improvement And Software Development Of Particle Field Digital Holographic Reconstruction

Digital holography is a three-dimensional imaging technology.It uses an image sensor to record information about an object in the form of interference fringes,and restores the intensity and phase of the light field during reconstruction to restore the three-dimensional information of the object.Digital holography technology is widely used in particle field,velocity field and other fields for its advantages such as simple light path,flexible recording and reproduction.The particle field hologram records the information of each particle in the field.Through a series of operations such as pre-processing,holographic reconstruction,depth of field expansion,particle identification,and particle positioning,the 3D coordinates and particle size information of the particles can be obtained.Because the particles in the particle field are distributed in different axial positions,in order to obtain a focused image of all particles,it is necessary to reconstruct the cross section of the hologram at a certain interval in the axial direction.Therefore,a major problem of holographic reconstruction is that the speed is too slow,especially when a large hologram or a large axial range of the particle field needs to reconstruct multiple sections,the calculation amount is too large and the time is too long.At the same time,the existing particle field holographic processing software lacks the hardware control function of the holographic system,and it cannot play the maximum performance of the computer during the reconstruction calculation.In view of the above problems,the research contents of this article are as follows:(1)For the problem that the calculation amount of hologram reconstruction is too large,the holographic algorithm process is optimized in a targeted manner,the algorithm process structure and intermediate variables are improved,and the reconstruction process is accelerated at the algorithm process level.At the same time,OpenMP multi-thread technology and CUDA technology are used to realize the reconstruction algorithm based on the two-level parallel architecture.The algorithm allocates image-level tasks to different threads and pixel-level tasks to different CUDA cores,thereby maximizing the parallelism of the computer and greatly speeding up the reconstruction.For 5000×5000 hologram reconstruction,the calculation speed of this algorithm is 48.3 times faster than the original single-thread reconstruction algorithm.This technology has great value for the application of digital holography in a wide range of particle fields and real-time particle field detection.(2)The holographic reconstruction processing software was written based on QT,and developed with a modular software architecture,which realized the hardware control of hologram shooting and the system process of reconstructing and processing the entire holographic measurement.It can maximize the computing performance according to the hardware of different computers.At the same time,it meets the requirements of academic research and practical production applications.This software has been applied to the coal fineness meter of the power plant.After taking the hologram,the software can complete the reconstruction process in 1.5 seconds,and calculate the distribution of fineness of coal powder in the current pipeline in real time,which is of great significance for timely guidance of the operation of the power plant's milling system.