SBIR-STTR Award

Gitam Technologies Inc. (GTI) proposes to develop novel HSI/MSI algorithms for detection, recognition and tracking of dismounts, vehicles and other man-made objects (Fortifications, Ordnance, weapons, IEDs, etc.). Preliminary work on dismount/vehicle detection using PCA with VIS-HSI data indicated improved performance when compared to traditional EO processing. In this work, we will extend the work to incorporate NIR-HSI bands. For the dismount case, a major focus will be in the phenomenology analysis, and exploitation of NIR band facial/skin data for face recognition, as well as for dismount tracking in darkness using NIR-HSI sensors. Preliminary results on moving target tracking using change detection over successive HSI bands have been promising. In this project, we shall consider Active Contour Tracking of dismounts and vehicles using Level Sets where the contours are represented by energy-minimizing splines known as Snakes. Genetic algorithm will be used to harness large quantity of high-quality spatio-spectral HSI data for automated feature extraction. In addition to traditional HSI algorithms, such as PCA, LDA and ICA, we will also study the feasibility of a new information theoretical Spectral Information Measure algorithm, which is specially suited for exploiting spectral variability, similarity, and discrimination from hyperspectral images.

The primary goal of this effort is to develop advanced Hyperspectral Image (HSI) data and algorithms for early detection of plant degradation due to Chemical/Biological agents. During Phase I, proof-of-concept demonstrated that HSI algorithms are capable of detecting de-greening in arabidopsis plants infused with covert de-greening circuits. In Phase II, the major objectives are: (1) Extend genetic engineering towards more operational viability, i.e., subject larger and mature plants to a wider range of chemical and biological agents, (2) Develop advanced Detection/Classification algorithms: multiple-hypothesis detect/ID for multiple plant specimens affected by different chem-bio agents, signature-based temporal change/anomaly detection, kernelization of linear algorithms to account for nonlinearities, ICA-based unmixing of HSI data, genetic algorithm for automated feature extraction, (3) HSI aided Triage resource management for distinguishing live and deceased dismounts in urban calamity region using HSI thermal-IR bands, and (4) Generation of and experimentation with synthetic remote sensing data and analytical prediction models: Incorporate healthy/de-greened plant and human skin spectral reflectance/emissivity signatures within FASSP/DIRSIG modeling environment, add atmospheric/illumination/sensor effects to generate synthetic electro-optical imagery that an airborne sensor might observe from a distance, apply appropriate detect/ID algorithms on the synthetic images, and perform model based sensitivity analysis to explore detection bounds.

Keywords:
Biosensors, Plant-Sentinel, Chem.-Bio Detection, Automated Detection Of Plant Degradation, Hyperspectral Image Processing (Hsi), Eo/Ir, Sensor And Sig