Extraction de panneaux de signalisation routière dans des images couleurs

Bahman Soheilian (1) , Aurore Arlicot (1) , Nicolas Paparoditis (1)

1 MATIS - Méthodes d'Analyses pour le Traitement d'Images et la Stéréorestitution

In this paper we present a robust algorithm for detection and recognition of road signs from color images. Our approach is based on matching models with geometric and radiometric panels. The general strategy consists of three major steps. The first of them detects the pixels of a given color and provides regions of interest in the image. The second step is to detect signs of known geometric shapes (circles, triangles, squares) by a RANSAC technique in regions of interest previously obtained. This step provides potential candidate panels that will be evaluated in the final stage. This step attempts to match image correlation, each candidate with the correct sign from the group reference panels of the same shape. We present results on images of streets acquired by a mobile vehicle scanning in urban areas. Finally we discuss the system performance in terms of detection and robustness.

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Bahman Soheilian, Aurore Arlicot, Nicolas Paparoditis. Extraction de panneaux de signalisation routière dans des images couleurs. Reconnaissance des Formes et Intelligence Artificielle, Jan 2010, France. pp.1-8. ⟨hal-00717032⟩