SBIR-STTR Award

InView and its close partner Rice University are world leaders in CS imaging algorithms and CS imaging sensor development. Over the last 8 years, there has been impressive progress on imaging architectures that seek to reduce the amount of data sensed by exploiting signal priors and task-specific imaging. We propose to leverage these latest advances in compressive sensing and computational imaging, many of which have been pioneered by the PIs and key personnel, to build imaging platforms tuned to change detection and the detection of fast moving objects such as rocket-propelled grenades. The three objectives of the Phase I efforts are to resolve the following CS change-detection challenges are to define reliable algorithms for rapid change detection, to define a small-form factor, low-cost hardware platform with sample speeds required to detect fast moving objects, and to perform system-level simulations to analyze the spectral, spatial and change-detection algorithms on a model of the platform. The Phase I work program that will meet the objectives listed above is focused on the preliminary design and feasibility demonstration of a multi-spectral infrared video sensor that uses advanced CS algorithms implemented on an innovative adaptive hardware platform.

Keywords:
Change Detection, Compressive Sensing, Area Monitoring, Sensor Network Data Reduction, Reduce Data,

The main goal of the Phase II program is to build and test an operational prototype compressive sensing (CS) camera capable of high-speed change detection. Based on the algorithmic development, hardware modeling and simulation achievements of Phase I, we have defined a real-time CS framework that will be implemented in Phase II in an operational prototype. This framework, similar to Phase I, begins with the high speed measurement capability of the CS platform and the programmability of the CS architecture that a digital micromirror array provides to implement the various measurement models. While data acquired in these models can be reconstructed into images to provide scene contextual information at selectable resolution, we concentrate in this proposal on the compressed domain data for high speed change detection based on the application of our Sum-to-One fast transform and Sparse-in-Time model statistics. We also investigate enhancement of these models using spectral domain measurement and unmixing techniques. In this proposal, we develop objectives for implementing the change detection computational framework in the CS spectral camera architecture developed in Phase I to achieve a Phase II high-speed change detection prototype that operates in the shortwave infrared.

Keywords:
Compressive Sensing, Shortwave Infrared, Change Detection, Spatial-Spectral Hybrid Subspace Sampling, Swir