You are here

Developing a Seamless Integration Between Machine Learning Techniques and Rule-Based Classification of Remotely Sensed Imagery

Award Information
Agency: Department of Defense
Branch: Army
Contract: DACA42-03-C-001
Agency Tracking Number: A022-2026
Amount: $120,000.00
Phase: Phase I
Program: SBIR
Solicitation Topic Code: N/A
Solicitation Number: N/A
Solicitation Year: N/A
Award Year: 2003
Award Start Date (Proposal Award Date): N/A
Award End Date (Contract End Date): N/A
Small Business Information
P.O. Box 8226
Missoula, MT 59807
United States
HUBZone Owned: No
Woman Owned: No
Socially and Economically Disadvantaged: No
Principal Investigator
 Stuart Blundell
 Chief Operating Officer
 (406) 829-1384
Business Contact
 David Opitz
Title: Chief Executive Officer
Phone: (406) 829-1384
Research Institution

The Army has a critical need to accelerate and improve terrain analyses from remotely sensed imagery to support the increasingly mobile requirements of the Army Warfighter. Existing techniques for terrain analysis, topographic, and reproduction supportare slow, labor intensive processes that do not meet the needs of the Force XXI digital battlefield. Previous research has shown that incorporating ancillary data, such as GIS thematic data layers or DEM derived rasters, into rule-based classificationscan increase the accuracy and precision of land-cover and land-use classification. However, the process of rule generation from these data is a significant challenge without expert knowledge or sophisticated computer science programming skills.Artificial intelligence techniques, including machine learning and rule-based expert systems, are now emerging in COTS geoprocessing software for tasks such as feature extraction and image classification. Visual Learning Systems, Inc. (VLS) introduced theFeature AnalystT extension for ESRI's ArcGIS software in 2001. Widely recognized in the industry as the first viable machine learning application incorporating spatial context in the feature extraction process, the underlying Feature Analyst architecturewill be used to automatically generate and refine a rule base using a novel approach called theory refinement. The proposed Phase I research strongly supports the Army's SBIR program goals of developing a seamless integration between machine learningtechniques and rule-based classification of remotely sensed imagery. In addition to supporting the Army's terrain analysis needs the proposed system will also support scientific research, environmental modeling, local government planning, and federalgovernment security programs. The connectionist theory refinement system proposed here has strong commercial potential for the GIS software industry as a means of lowering the costs of geospatial database maintenance using remotely sensed imagery.

* Information listed above is at the time of submission. *

US Flag An Official Website of the United States Government