You are here

Automatic Target Recognition from Motion Video

Award Information
Agency: Department of Defense
Branch: Defense Advanced Research Projects Agency
Contract: DAAH0103CR270
Agency Tracking Number: 03SB1-0213
Amount: $99,000.00
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
6 New England Executive Park, Burlington, MA, 01803
DUNS: 094841665
HUBZone Owned: N
Woman Owned: N
Socially and Economically Disadvantaged: N
Principal Investigator
 Matthew Antone
 Sr. Research Engineer
 (781) 273-3388
Business Contact
 John Barry
Title: Contracts Manager
Phone: (781) 273-3388
Research Institution
We propose a new automatic target recognition (ATR) system that operates on video data, exploiting the inherent redundancy and spatio-temporal coherence of motion imagery to create three-dimensional target signatures. Unlike previous approaches that useshape alone, our method will additionally model and predict object surface appearance properties, in the form of bi-directional reflectance distributions (BRDFs), to compensate for environmental effects such as varying illumination.We will employ parameterized reflectance models and pre-defined shape and material characteristics to create and refine object models on-the-fly, both for the formation of target templates and for recognition of these templates in subsequently observedmotion video. Tight integration between tracking, prediction, modeling, and registration reflects the tight coupling of the underlying recognition problem and facilitates optimal signature estimation by incorporating all available information. We willimplement a prediction engine that guides refinement of geometry and appearance by accounting for environmental lighting, shadows, and material properties as well as object shape and reflectance. We will also apply robust correlation techniques to estimatetarget match probabilities across multiple video frames. Finally, we will evaluate the performance of our model generation and ATR techniques, and compare with existing approaches. Two major technologies will emerge from this effort. Automatic creation and refinement of 3D object models from video, incorporating both surface shape and appearance properties, will have significant impact on time-critical construction of targetsignatures, site modeling, and photo-realistic image based rendering. The larger program objective of robust video-based ATR has direct application to military scenarios that require high-precision confirmatory identification and tracking, and tocommercial sectors such as law enforcement, security, and video monitoring and surveillance.

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

US Flag An Official Website of the United States Government