Realtime Holographic Sensor Systems Using a Cluster of Heterogeneous CPU/GPU Processors
Digital in-line holographic imaging (DIHI) has thrived recently as an effective tool to obtain 3D information of objects with up to submicron spatial resolution. Combined with object tracking techniques, it also allows the study of moving objects, providing their 3D trajectories that represent the ﬂow field in microﬂuidics, including the motion of plankton in water, the motion of bacteria in cells, and much more. Due to the simplicity and low cost of DIHI instrumentation, DIHI-based optical sensors can potentially be produced in large quantities and deployed to monitor pollutants, detect carcinogenic and pathogenic cells, etc. in a number of environmental, medical, and military applications. However, the implementation of DIHI-based optical sensing is greatly hindered by the time-consuming process of numerically reconstructing and analyzing holograms.
The goal of this project is to overcome this obstacle and extend the applications of DIHI to optical sensing by exploiting the massively parallel computing capabilities of graphics processing units (GPUs). Preliminary results show that GPU-based processing has the potential to accelerate holographic applications on the order of 1,000X, enabling real-time reconstruction, analysis, and 3D visualization of holographic data.
Return to this PI's main page.