A smart sensor based visual landmarks detection for indoor robot navigation

Abstract

In this article we present a multisensor system for usual landmarks detection in indoor autonomous robot navigation context. This smart sensor combines a CCD camera with a 3D laser camera. The landmarks detection is based on the collaboration between intensity image processing which ensures the planar quadrangle landmarks detection itself and ad hoc 3D laser camera which copes with the specific constraints of our problematic. The process consists in grouping horizontally/vertically oriented segments extracted from the intensity image. The constraints are applied thanks to a relaxation scheme. The next step consists in confirming or rejecting the detected quadrangles regarding their planarity by focusing the 3D laser camera in the environment area defined by these visual targets.