InFuse-an integrated expert neural network for intelligent sensor fusion

A discussion is presented of an architecture called InFuse (Intelligent Fusion) for the multiple sensor fusion problem. InFuse exploits the notion of combining expert systems and neural networks to capture the advantages of both technologies. The application involves the extraction of natural terrain features from imagery provided by multiple sensors. In addition to the imagery, terrain knowledge in geographical databases needs to be integrated. The uniqueness of the approach lies in its ability to combine symbolic data of spatial databases and domain-specific knowledge about sensor behavior in expert systems with example-induced learning from neural networks to achieve a high classification rate. Results are presented to demonstrate the power of the approach