ANVIL - Automated Target Recognition on Multi-Spectral Imagery
Project Anvil (Applied Analysis Spectral Analytical Program) is a software program that is currently part of the ERDAS imagery processing package. Anvil takes MultiSpectral imagery, using ERDAS, and goes inside of a pixel to determine the content of the pixel, rather then classifying the whole pixel as current techniques do. Current technology can classify down to as little as 20% of a pixel. This means that if the object you are looking for takes up only 20% of the area of the pixel, you should still be able to identify it. A quick example of the process is as follows. You have a SPOT or Landsat Multispectral image of an area and you are looking for something specific, i.e., concrete roads. By going into the image and already knowing spectral properties of many items in the scene, i.e. grass, trees, etc, you can remove all of these and if the remaining spectral properties match that of the type of road you are looking for, that basically means you have road there.
- Provided operational capability to USAFE
- Automated process to accomodate user needs (shortcut keys/userguide to the basics etc.)
- 1995 MERIT projects
Current projects include:
- Bathymetry for NAVSPACECOM. Provide capability to determine water depth for possible landings, etc. On going. Completion sometime in later half of 1996.
- USAFE project. Provide USAFE with an initial Multispectral imagery processing capability, project to have follow-on options to go through June/July 96 time frame.
- MERIT/AD follow-on. This is a follow-on project sponsored by Defense Dissemination Program Office (DDPO) to take this process and create a Hyperspectral processing capability. Make Hyperspectral imagery for new sensors and make it look like Multispectral imagery.
|Join the GlobalSecurity.org mailing list|