Door crossing and state identification using robotic vision

Abstract

This project is aimed at the implementation of an autonomous vehicle guide (AGV) to navigate in internal environments, such as hospitals, offices, residences and so on. The work presents a new 3D algorithm for robotic vision, used for automatic identification of the state of doors, without any previous information about the environment, such as maps, location of obstacles, illumination and the like. The characteristics of the doors that would influence the decision of the robot on the identification of its state, also not ignored, such as texture, color, dimensions. Published in the 8th IFAC International Sysmposium on Robot Control (SYROCO 2006), this work presented the application of the Hougth Transform in images captured by the robotic vision. The coordinates of Hougth then presented the location of the edges of the doors and the analysis of cross-corelation of the found ones, gave us the configuration of the state of the door (open, closed, half open). With the results obtained, the navigation of the autonomous robot became more secure, since the risks of collision were minimized. Navigation using 3D vision, with real-time processing, also reduced the risk of collision with objects dynamically positioned in the environment. In the first part of this document, we present the article published in (SYROCO 2006). In the second part, we present some of the experimental results obtained during the project.