We have developed an airborne multi-spectral imaging system specifically designed to detect and monitor wildfires. The system was developed using commercial off-the-shelf (COTS) hardware and software where possible. There are six spectral bands recorded covering the standard red, green and blue visible photographic channels and short wave infrared, mid-wave infrared and long-wave infrared channels. The spectral regions recorded by the cameras were chosen to optimize detection probability and minimize false alarms for wildland fires and to produce concurrent high-resolution visible mapping of the fire area. The airborne imager is flown on a Piper Aztec aircraft specially modified for this purpose.

The camera sub-system is gimbaled in the aircraft airframe so that the field of view may be swept across the flight path ('scanned') to increase ground coverage area. A single 'scan' consists of four overlapping images with a resolution in the infrared of 1600 pixels across track (with a nadir resolution of 3m) and 10,000 pixels across track (with a nadir resolution of 0.5m) in the visible bands. The three infrared channels have 640 X 512 pixel sensors while the visible camera has a 4096 X 4096 pixel sensor. A commercial inertial measurement unit allows orthographic correction of the images. Three MS Windows computers record camera and flight data and control the scanning motor and instrument peripherals. Operator interface and 'quick-look' views of acquired imagery are available during the flight via a convenient touch-panel input/display device.

We will discuss initial fire experiments flown in the Rochester, NY area and show fire detection results obtained from these experiments using several algorithms.

This work was supported by the National Aeronautics and Space Administration under Grant NAG13-02051. This financial support has been greatly appreciated