TechMatch: SBIR Opportunity View

Site Search:

Licensable Patents

Success Stories

Technology Needs

(Register Here)

DOD TechMatch Sponsors

SBIR Opportunity View

Print This Page

Back to Listing

A08-059

Click For Printer Friendly Verson




Go To DOD SBIR Site
Basic Information:

Title:	A psychologically inspired object recognition system
Program:	SBIR
Technology Area:	Information Systems, Ground/Sea Vehicles, Electronics, Human Systems
Open Date:	4/21/2008
Close Date:	6/18/2008

Description:

Recognizing and identifying an object from a video input turns out to be a very difficult problem. The problem stems from the fact that a single object can be viewed from an infinite number of ways. By rotating, obscuring, or scaling a single object, one can create multiple representations of an object - which makes the problem of matching the object to a database of objects very difficult. The problem expands exponentially when objects that need to be identified have never been viewed before. Combine these limitations with the wide variety of objects which might be identified, and the problem becomes intractable. One solution is to study and understand how human beings recognize objects in the real world and duplicate that functionality in a series of algorithms. Recent research (Tarr and Bulthoff, 1995) has indicated that humans use not one algorithm, but multiple algorithms for the task of object recognition - depending on the object being recognized and the situation at hand. Specifically, research has shown that people use template based algorithms (i.e. similar to the database matching algorithms described earlier) in addition to Geon based (Beiderman, 1995) algorithms and feature based algorithms. These three algorithms are used in conjunction with a fourth algorithm, a contextual cueing algorithm, which limits the overall search space. Finally, human spatial memory is able to mentally rotate objects in order to match the object to different representations (Shepard & Cooper). Therefore, a rotation algorithm would complete the set of algorithms. Geon based algorithms use specific shapes represented as simple forms in a template matching schema. For example, a cup could be represented in memory as a circular tube with a tubular handle extending from its side. Or a tree could be represented as a triangle for the leaves and branches, and a single tube extending from the bottom of the triangle to represent the trunk. This is similar to a template based matching algorithm, but the representations are more abstract. The abstract nature of the representations allows for a more general matching algorithm to be used. In addition to this algorithm, feature based algorithms would be used in conjunction with the Geon based algorithms. Feature based algorithms use features as rules to determine the identity of an object. For example, a tiger has stripes and a leopard has spots. The advantage of feature based algorithms is that they are especially immune to problems of rotation, scaling and obscurance. In other words, the key features of a leopard are still visible irrespective of the orientation or scaling of the object. In summary, new research suggests that a system based on the Geon models used in combination with template and feature based matching systems and context specific filtering algorithms which can rotate objects offer a better solution to the object recognition problem than the use of simple template based systems.

Objective:

To create an object recognition system based on the newer psychological models of object recognition by using a series of different algorithms to identify a variety of objects in different orientations. Such a system would be extremely beneficial for robotic control/intelligence and would allow for an exponential expansion of robotic capabilities and intelligence

Phase I:

Implementations of the system should be specific to the theories mentioned in this call, with an emphasis on the latest cognitive psychological theories of object recognition. The Phase I output should demonstrate a conceptual integration of the different stated algorithms, with details of the integration expounded. The Phase I process should show that, with further implementation, the system could scale-up and be able to recognize a wide variety of objects from a wide variety of viewpoints. Phase I documentation should include all aspects of hardware as well as software integration as well as the theoretical aspects of integration. Training of the system needs to be addressed in detail. An ideal system should be able to be trained by someone of limited experience or expertise with the system; however, this is not a necessary requirement.

Phase II:

Phase II will include full implementation of a prototype. Recognition rates should be above 95 percent for trained objects and be rotationally and scale invariant. Additionally, the prototype should be able to recognize objects and symbolically label objects that the system has never been previously exposed to (i.e. objects outside the training set). The prototype should use video input (VGA) into the object recognition system. The prototype should be further trained on a wide variety of objects to show the systems scalability to a database of a large number of items. The objectives outlined in the Phase I of the program still apply.

Phase III:

Produce a fully integrated system that can serve as a stand-alone system and can be tested in experiments to confirm the object recognition capabilities of the system. Recognition rates around 98 percent would be the preferred target. The system should be designed to be modular enough to allow for hosting on a variety of hardware platforms.

References:

1. Biederman, I. 1995. Visual object recognition. In Kosslyn and Osherson, 1995. 2. Shepard, R. N., & Cooper, L. A.. 1982. Mental images and their transformations. Cambridge, Mass.: MIT Press. 3. Tarr, M. J., & Bulthoff, H. H. 1995. Is human object recognition better described by geon-structural descriptions or by multiple views Journal of Experimental Psychology: Human Perception and Performance, 21, 1494-1505.

SBIR Keywords

Object recognition, computer vision, sensors, target recognition

TechMatch Keyword(s):

Computer Software

Applications

Data Fusion

Data Visualization and Graphics

Image Processing and Pattern Recognition

Mathematical Sciences

Robotics

Sensors

Optical, Visual

Modeling

DISCLAIMER:

The Solicitations listed on this site are copies from the various SBIR agency solicitations and are not necessarily the latest and most up-to-date. For this reason, you should use the agency link listed below which will take you directly to the appropriate agency server where you can read the official version of this solicitation and download the appropriate forms and rules. The Official Website can be found at: http://www.acq.osd.mil/sadbu/sbir/solicitations/index.htm

© 2002, 2009 by West Virginia High Technology Consortium Foundation.
All rights reserved. Disclaimer.