High-rate Modeling and Computing for Optical Systems---Gigapixel Image Formation and X-ray Imaging Physics

Abstract

With the rapid development of computational sensing technologies, the volume of available sensing data has been increasing daily as sensor systems grow in scale. This is sometimes referred to as the "data deluge". Many physical computing applications have to spend great effort on meeting the challenges of this environment, which has prompted a need for rapid and efficient processing of massive datasets. Fortunately, many algorithms used in these applications can be decomposed and partially or fully cast into a parallel computing framework. This dissertation discusses three sensing models---gigapixel image formation, X-ray transmission and X-ray scattering---and proposes methods to formulate each task as a scalable and distributed problem which is adapted to the massively parallel architecture of Graphics Processing Units (GPUs). For the gigapixel images, this dissertation presents a scalable and flexible image formation pipeline based on the MapReduce framework. The presented implementation was developed to operate on the AWARE multiscale cameras, which consist of microcamera arrays imaging through a shared hemispherical objective. The microcamera field-of-views slightly overlap and are capable of generating high-resolution and high dynamic range panoramic images and videos. The proposed GPU implementation takes advantage of the prior knowledge regarding the alignment between microcameras and exploits the multiscale nature of the AWARE image acquisition, enabling the rapid composition of panoramas ranging from display-scale views to gigapixel-scale full resolution images. On a desktop computer, a 1.6-gigapixel color panorama captured by the AWARE-10 can be delivered in less than a minute, while 720p and 1080p panoramas can be stitched at the video frame rate. We next present a pipeline that rapidly simulates X-ray transmission imaging via ray-tracing on GPU. This pipeline was initially designed for statistical analysis of X-ray threat detection in the context of aviation baggage inspection, but it could also be applied in the modeling of other non-destructive X-ray detection systems. X-ray transmission measurements are simulated based on Beer's law. The highly-optimized OptiX API is used to implement ray-tracing, greatly speeding code execution. Moreover, we use a hierarchical representation structure to determine the interaction path length of rays traversing heterogeneous media described by layered polygons. The validity of the pipeline was verified by comparing simulated data with experimental data collected using a Delrin phantom and a laboratory X-ray imaging system. On a single computer, 400 transmission projections (125 by 125 pixels per frame) of a bag packed with hundreds of everyday objects can be generated via our simulation tool in an hour, compared to thousands of hours needed by CPU-based MC approaches. Further speed improvements have been achieved by moving the computations to a cloud-based GPU computing platform. Finally, we describe a high-throughput simulation algorithm for X-ray scatter based on a deterministic but sampled approach built upon the previously described GPU-centric ray-tracing framework. Compared to Monte Carlo and Monte Carlo-based hybrid methods our approach is orders of magnitude faster and (in contrast to the deterministic method) allowing for modeling of scatter radiation in arbitrary imaging configurations and to any order. Qualitative and semi-quantitative validation have been conducted by comparing data obtained with the simulated pipeline and a laboratory X-ray scattering system. As for the speed of execution, on a single computer, a scatter image (125 by 125 pixels) of a simple 3D shape collected in a pencil beam geometry can be generated in minutes, while a realistic bag model collected in a fan-beam geometry takes about an hour.