Overview

This work is part of a project which aims to enable a team of robots Koala with rotating camera to keep a watch on industrial areas. Our objective in this project is to develop a visual cortical map. This map will allow the robot to extract visual primitives featuring its environment. We focus our purpose on the visual task, inspired by the biological strategy of ocular movements for scene analyses. In order to interpret its environment, the human eye has two sorts of vision :

	a central vision  analyses accurate visual details,
	a peripheral vision gives contextual information  for the next movement of the eye.

Our model of cortical map is inspired by self organizing maps (Kohonen maps). We add retinotopic property in the behavior of the map.

Our system is composed of two elements :

	a retina with a centered area has an accurate vision of details and the peripheral area gives contextual information about the environment of the robot,
	a cortical map gathers the robustness property of self organizing map and retinotopic property.

Computer Science D.E.A (Post-graduate diploma) report (word document - in french)