IMAGE PROCESSING FOR ROAD SIGNS INVENTORY Bruno L. Associate professor – University of Palermo – bruno@ing.unipa.it Parla G. PhD student – University of Palermo – parla@ing.unipa.it Rizzo A. PhD – University of Palermo – rizzo@ing.unipa.it ABSTRACT In this paper a detection road sign system, based on the analysis of colour images, is presented. This system aims at recovering the images regions that have colours similar to those present in road signs (red, white, black, blue). For each colour and inside adequate colour space, heuristic threshold decision are identified. Dynamic thresholds don’t suffer from the illumination change effects of the whole recorded sequence. Actually the recording carried out in this procedure preparation phase were performed during the day and during good visibility conditions. The binary images (afterwards thresholding) are associated so as to derive colour combination that are present inside the different road signs class. The check of regions size is carried out. The procedure disregards signs shape. Even if the system suggested doesn’t work in real time, we optimize performance velocity for every image size and for every camcorder now available. Road authorities, who often do not have enough reliable information about their actual road sign inventory, could use the results of images analysis for the purpose of inventory, inspection and maintenance. Keywords: Road sign, image analysis, road signs inventory