Clustering-based particle detection method for digital holography to detect the three-dimensional location and in-plane size of particles

Digital holography (DH) has been extensively applied in particle field measurements due to its promising ability to simultaneously provide the three-dimensional location and in-plane size of particles. Particle detection methods are crucial in hologram data processing to determine particle size and particle in-focus depth, which directly affect the measurement accuracy and robustness of DH. In this work, inspired by clustering algorithms, a new clustering-based particle detection (CBPD) method was proposed for DH. To the best of our knowledge this is the first time that clustering algorithms have been applied in processing holograms for particle detection. The results of both simulations and experiments confirmed the feasibility of our proposed method. This data-driven method features automatic recognition of particles, particle edges and background, and accurate separation of overlapping particles. Compared with seven conventional particle detection methods, the CBPD method has improved accuracy in measuring particle positions and displacements.