Outdoor scenes interpretation suitable for blind people navigation

One important problem for blind people during their navigation in the outdoor environment, is the need for external help while reaching a specific place. This dependence could be reduced supplying the blind person with a smart mobile system capable of giving simple scene descriptions and useful landmarks. In this paper a system which gives an interpretation of real scenes through single (monocular) image analysis is proposed. Indoor and outdoor scenes in natural situations, without any expedient to make image interpretation easier, have been used to validate the system. The main goal of the proposed architecture is to provide help for blind people navigation, but the techniques developed are suitable for robotics applications too. In order to guarantee the maximum flexibility, it is assumed an absolute lack of “a priori” knowledge about the scene contents. Scenes are described using simple primitives such as surfaces, zones and their spatial relations. A new vanishing point detection method has been developed