Boston University | Center for Computational Science
HomeNews and EventsResearchEducationPeopleSeminarsFacilitiesContact Us

CCS Seminar Proposal - for presentation on 10/10/2003
Applying Neural Networks to Remote Sensing and Robotic Sensor Fusion

Sigfried Martens
Cognitive and Neural Systems
Boston University
October 10, 2003
Remote sensing of forests and robotic sensor fusion can be viewed, in part, as supervised learning problems, mapping from sensory inputs to labeled outputs. This talk will introduce supervised learning from an applied perspective, using ARTMAP neural networks. Applications will be presented, each teaching the neural network a mapping from a set of input values to a set of class labels. In particular, two main applications will be discussed: satellite remote sensing of forests, and fusion of sonar and camera-based sensors from a mobile robot. In the remote sensing study, the neural network is used to evaluate the relative importance of spectral and terrain data for map-making. Maps produced by the neural network are compared with maps produced by human experts. In the second study, the neural network is used to combine sensory information from two different modalities, producing a more accurate distance percept than is afforded by either sensory input separately. This improved distance percept produces occupancy grid visualizations of the robot’s environment. These visualizations point to specific problems of raw sensory information processing and demonstrate the benefits of using a neural network system for sensor fusion.

copyright © 2006, Center for Computational Science | Boston University , MA, 02215