Détection et Reconnaissance Des Panneaux de signalisation Routière

Abstract

In order to facilitate the driving process and to avoid road accidents, which have become incremental in recent times, and which have become dangerous for human life in general and for the lives of children in particular, so a human being does great attention in finding an appropriate solution to get out of this problem or at least to minimize it. So the solution is to cross the field of computer vision that we study and use to identify this type of images. Our job is to implement a desktop application to extract signs from an image. After a pretreatment step, explained in detail in the "Conception and development" chapter, we use the top-down approach which requires to select, first, areas that may be areas of interest, then to focus our treatment to these extracted areas to validate if it is really a road sign. The top-down approach avoids us to treat a large area of the image, and consequently a saving of time of execution. Which implies better performance of the application. Another phase takes place after these treat- ments to identify a panel by considering a base of road signs. The identification of a sign is translated by the display of the corresponding image of the base, or to dis- play a text explaining the plate or even to emit a sound indicating which panel it is