Year
2008
Abstract
Since 2006, Sandia National Laboratories has developed a sensor fusion application that combines raw security sensor data in an effort to reduce nuisance alarms for security systems. This Fusion Framework allows various algorithms to be \"plugged in\" to combine sensor information. The purpose of the software is to reduce alarm data rates as well as nuisance alarms caused by environmental conditions. The software is part of an effort by Sandia to create a security detection system that can operate in uncontrolled exterior environments beyond the perimeter of a standard security system for a facility. User confidence in the detection system depends on the application of sensor fusion algorithms. Too many false or nuisance alarms can overwhelm operators, making a security system unusable. The fusion framework team has developed a modular software framework that handles the input, output, and flow of sensor events through the fusion application. Various algorithms can then be attached to the framework to handle sensor fusion processing. Algorithm developers can concentrate on development of sophisticated algorithms without the need to spend time writing code to handle data input and output. This architecture allows new fusion algorithms to be created, configured, and tested quickly. The design also supports reconfiguration of algorithm parameters in real time to adapt algorithms for various external events such as time of day or changing weather conditions. The fusion framework and a select set of algorithms have been deployed for testing. This paper describes the sensor fusion framework and presents some results of its performance in the field over the last several months.