Parallel-Architecture-Based Feature Extraction and Sensor Fusion for Object-Oriented Image
Parallel-Architecture-Based Feature Extraction and
Sensor Fusion for Object-Oriented Image
LNK Corporation, Inc.
6811 Kenilworth Avenue, Suite 306
Riverdale, MD 20737
Srinivasan Raghavan (301-927-3223)
Imagery gathered by the earth observation systems (EOS)
places a heavy burden on computational resources. Retrieving and
archiving massive image databases generated using EOS requires
efficient and real-time algorithms. This project's approach to
solving this problem is to use the object-oriented design theme.
This theme allows image features, such as land cover, vegetation,
and other important image characteristics, to be used as an index
for the images in the database. To aid the process of feature
extraction, Phase I will develop a synergic framework of neural
networks and an expert system supported by fuzzy logic.
Specifically, parallel algorithms will be developed in conjunction
with this framework to achieve sensor fusion and feature
extraction. The company will show a proof-of-concept demonstration
of the parallel algorithms on a parallel machine, Zephyr
(Wavetracer Inc.), that is available in-house.
Potential Commercial Application:
Potential Commercial Applications: Applications include remote
sensing for agricultural purposes, EOS data management, coastal
feature extraction and habitat loss analysis, weather
understanding, and environmental monitoring.