Development of a Cross-platform Algorithm for Application of Digital Holography in 3D Particle Detection

2019-08-16T16:27:36Z (GMT) by Yijie Wang
Digital holography (DH) has a variety of applications on measuring the 3D position of different kinds of particles, including the droplets created in drop breakups, seeding particles for flow velocity measurements, characterizations of the behavior of the microorganisms, etc. A particle detection method is required to extract the 3D information encoded in the interference patterns of the holograms, which is desired to be accurate and fast. As the accuracy of the particle detection method improves, the time efficiency of the method decreases. In this study, an optimization process is developed based on an existing method to shorten the processing time. The optimization process includes reducing the complexity of the method and introducing the parallel processing algorithm that can be implemented on cluster machines. The existing particle detection method is separated into several steps and analyzed. The most time consuming step, refining the threshold to separate overlapping particles, is the focus of complexity reduction optimization. A Python code is developed, based on object oriented programming, to implement the optimization. Message Passing Interface (MPI) is applied for parallel processing with a 24-core remote workstation. The optimized Python code is compared with the existing Matlab code in both time consumption and accuracy aspects with synthetic holograms. It is found that the optimization process is able to reduce the time consumption by about four times with an acceptable sacrifice in accuracy. Finally, a DIH system with the optimized method, is applied to characterize different kinds of solid particles. One is noted that the previous studies focus on measuring artificial particles or droplets which are both spherical particles, while most natural solid particles usually have irregular shapes. Equivalent diameter, circularity and aspect ratio are introduced to quantify the dimension and morphology of the irregular shapes. The statistics of the parameters are generated to characterize different kinds of the particles. The accuracy of the characterization of the particles are verified with the observation of the microscopic images of the particles, which can further prove the improvement of the optimized method for particle detection.