N14A-T008 TITLE: Object Cueing Using Biomimetic Approaches to Visual Information Processing

TECHNOLOGY AREAS: Air Platform, Battlespace, Human Systems

ACQUISITION PROGRAM: PMA 281

OBJECTIVE: Develop object recognition capabilities based upon biomimetic sensory, perceptual, and cognitive processes, which can effectively and efficiently process imagery data.

DESCRIPTION: In the data-rich operational world of our military forces, it is increasingly difficult to sift through the vast amount of information to determine that which is mission relevant. Efforts to develop automated object recognition, cueing, and assorted automated/semi-automated decision aids have been ongoing for many decades, but have yet to produce viable technologies able to reliably and accurately distinguish objects of interest from varieties of imagery data. As such, the burden of processing such data falls to laborious and expensive manual human processing. The complexity of such visual search and pattern recognition is certainly at the heart of the difficulties in attaining this capability, but some have argued that there may be a fundamental flaw in the 'brute force' algorithm approach that is traditionally used to discern patterns and identify objects. The simplest organism's ability to identify objects remains vastly superior to even the most sophisticated software, so efforts to model biological sensory, perceptual and cognitive processes have been underway in an effort to replicate the powerful functionality of simple biology.

It has been noted (e.g., Shoujue & Jianliang, 2005) that traditional pattern recognition methods of statistical modeling points in a high-dimensional space are generally inferior to biomimetic pattern recognition approaches; biomimetic approaches have proven successful in applied domains, such as with electro-optical (EO)/infra-red (IIR) imagery processing with high object classification (Pace & Sutherland, 2001). These biomimetic approaches have successfully modeled fundamental animal information processing (Lang et al., 2011) and general frameworks (Chikkerur and Poggio, 2011), as well as increasingly detailed biological representations (c.f., Hay et al., 2011) and present an opportunity for application in assorted remote sensing domains. These applications have been both highly specific (e.g., modeling highly specific areas of the brain such as retina, Tadross et al., 2000) as well as larger scale to include subcortical (Cecchi, Kozloski,Peck, and Rao, 2005) and cortical (e.g., see Lebiere & Anderson, 1993) structures.

PHASE I: Identify and research candidate biomimetic modeling approaches, functions, integrations, and concepts to support object identification and recognition, and demonstrate proof-of-concept.

PHASE II: Develop a prototype system capable of reliably and accurately discriminating land and sea objects from imagery data of quality comparable to typical imagery-type data that includes, but is not limited to full motion video (FMV), EO/IR data on moderate entity count scenarios.

PHASE III: Transition a final system capable of reliably and accurately discriminating land and sea objects from imagery-type data that includes, but is not limited to full motion video (FMV) and EO/IR data in high-entity count scenarios, to an appropriate Navy Unmanned Air System (UAS).

PRIVATE SECTOR COMMERCIAL POTENTIAL/DUAL-USE APPLICATIONS: Effective automated object recognition from imagery data for vehicles would be useful for the United States Coast Guard (USCG), Department of Homeland Security (DHS), Department of Energy (DOE), and other federal agencies for which protection from vehicle based threats is important. Commercial security entities could likewise benefit from automated processing of imagery data. Federal, state and commercial rescue organizations could also benefit from the ability to track objects. All organizations for which remote imagery is valuable could potentially benefit from this technology.

REFERENCES:
1. Cecchi GA, Kozloski J, Peck CC, Rao RA (2005) Mesoscopic modeling of thalamo-cortical circuitry: large-scale topology, oscillations and synchronization. Society for Neuroscience (SFN) abstract 617.12.

2. Chikkerur, Sharat; Serre, Thomas; Tan, Cheston; Poggio, Tomaso. (2010). What and where: A Bayesian inference theory of attention. Vision Research, 50(22), 2233-2247.

3. Chikkerur, S. and Poggio, T. (2011)."Approximations in the HMAX Model", MIT-CSAIL-TR-2011-021/CBCL-298, Massachusetts Institute of Technology, Cambridge, MA, April 14, 2011.

4. Hay E., Hill S., Schürmann F., Markram H, Segev I (2011). Models of Neocortical Layer 5b Pyramidal Cells Capturing a Wide Range of Dendritic and Perisomatic Active Properties. PLoS Computational Biology 7(7).

5. Hines, M. L., Davison, A. P., and Muller, E. (2009). NEURON and Python. Frontiers in Neuroinformatics, 3, 1-12.

6. King, J. G., Hines, M., Hill, S., Goodman, P. H., Markram, H., & Schurmann, F. (2009). A Component-Based Extension Framework for Large-Scale Parallel Simulations in NEURON. Frontiers in Neuroinformatics, 3, 10.

7. Kozloski, J., Sfyrakis, K., Hill, S., Schurmann, F., Peck, C., & Markram, H. (2008). Identifying, Tabulating, and Analyzing Contacts between Branched Neuron Morphologies. IBM Journal of Research and Development, 52(1/2), 43-55.

8. Lang, S., Dercksen, V. J., Sakmann, B., Oberlaender, M. (2011). Simulation of signal flow in 3D reconstructions of an anatomically realistic neural network in rat vibrissal cortex. Neural Networks, 24(9), 998-1011.

9. Lebiere, C., & Anderson, J. R. (1993). A connectionist Implementation of the ACT-R production system. In Proceedings of the Fifteenth Annual Conference of the Cognitive Science Society (pp. 635-640). Mahwah, NJ: Lawrence Erlbaum Associates.

10. Tadross, M., Whitehouse, C., Hornstein, M., Eng, V., and Micheli-Tzanakou, E. (2000). A Three-Dimensional Physiologically Realistic Model of the Retina, International Joint Conference on Neural Networks (1): 211-216.

11. Wang, S., Lai, J. (2005). Geometrical learning, descriptive geometry, and biomimetic pattern recognition. Neurocomputing, 67, 9-28.

12. Cao, W., Feng, H., Hu, L., He, T. (2009). Space Target Recognition Based on Biomimetic Pattern Recognition. First International Workshop on Database Technology and Applications, 25-26.