Algorithms for IR data

Period of Performance: 06/20/2014 - 03/30/2015


Phase 1 SBIR

Recipient Firm

Numerica Corp.
5024 Technology Parkway Array
Fort Collins, CO 80528
Principal Investigator

Research Topics


ABSTRACT: Detection of dim phenomena in images generated by short-wave infrared (SWIR) systems is complicated by the fact that objects of interest generate very faint trails on the raw imagery due to image artifacts; e.g., background clutter, sensor noise, internal reflections, atmospheric effects, etc. Track-before-detect approaches typically used to solve the dim object detection problem must strike a balance between detection thresholds and false alarm rates. In this proposal, we present a novel TBD approach for detecting exceptionally dim targets while minimizing false alarms. Our proposed method for solving the problem uses recent advances in the detection of anomalies in noisy and corrupt data to formulate the task as a convex optimization problem for which fast efficient solvers are available. We propose to achieve additional improvements in efficiency by combining the new techniques with multi-scale methods. The value of our novel approach for detecting dim phenomena in cluttered IR data will be demonstrated during Phase I using a library of unclassified simulation data. BENEFIT: The convex optimization based approach for dim object detection and clutter mitigation described in this proposal will result in increased detection performance while minimizing false alarms. The ability to leverage previous experience in tracking and IR data processing provides substantial added value. The most promising near-term transition paths for this technology are Air Force programs of record that have strong potential for the new technology. These include the Sustainment and Modernization of Optical and Radar Sensors (SMORS) program, the Ground-based Electro Optical Deep Space Surveillance System (GEODSS), the Advanced Space Superiority Technology and Engineering Requirements (RASTER), and the Beam Control Engineering and Analysis (BCEA) program, of which the last two are currently in the acquisition stage. Beyond this primary transition path, Numerica's team has identified Army CERDEC programs of interest within the Night Vision and Electronic Sensors Directorate that would benefit by incorporating our solutions. Additionally, Numerica plans to validate our proposal concept with Boeing Directed Energy and Strategic Systems. In a potential Phase II program, Numerica intends to engage with Boeing to prepare plans to integrate our Track Before Detect algorithm into one of Boeing's electro-optical systems.