Integrating Attention and Categorization Behaviors in Robotics

This work presents an active vision system for control tasks involving attention and pattern categorization based on visual sensory information. The system is currently implemented in a robot consisting of an articulated stereo-head with four degrees of freedom (pan, tilt, left and right verge). As a practical result, the robot is able to analyze all regions of its environment, selected according to salience maps. This involves performing attentional shifts and an eecient feature extraction and pattern recognition. Also the system incrementally construct an attentional map, and keep the map consistent with a current perception of the environment. Experimental result shows succesful applications in inspection, monitoring, or surveillance tasks.