An image analysis technique for automatic traffic sign detection and classification is proposed. It makes it possible, after proper training, to detect, recognize and classify vertical road signs from video frames acquired on a moving vehicle equipped with cameras, as well as to identify anomalies with respect to road sign regulations (positioning and visibility). The experimental results show that this technique allows one to correctly detect and classify almost all vertical signs and, mainly in extra-urban environment, it can be considered as highly reliable, apart from being really versatile and user-friendly for road inventory and road maintenance purposes.