This paper describes a working prototype for road traffic sign recognition, presenting experimental results and evaluation. The elaboration chain involves the acquisition of the videos provided by digital cameras fixed on a vehicle and a video processing system based on degenerate decision trees, made up of two classifiers adopting two different techniques: the detector, which operates on Haar-like features and detects the regions of interest (ROI) where candidate traffic signs can be found, and the recognizer, which makes use of standard MPEG-7 visual descriptors computed in a metric space enabling the similarity approach for selecting the correct pattern in each ROI. Here they present an implemented working prototype based on a novel approach that combines pattern detection and MPEG-7, resulting in a system able to reach a very high overall precision. The paper is organized as in the following: in Section 2 they describe the analysis model adopted for the prototype implementation which is presented in Section 3; Section 4 shows the experimental results achieved with a sample of more than 120000 images containing about 600 traffic signs.
Automatic video processing for traffic sign recognition
2009
4 Seiten, 2 Bilder, 1 Tabelle, 19 Quellen
Conference paper
English
TRAFFIC SIGN RECOGNITION DEVICE AND TRAFFIC SIGN RECOGNITION METHOD
European Patent Office | 2017
|TRAFFIC SIGN RECOGNITION DEVICE AND TRAFFIC SIGN RECOGNITION METHOD
European Patent Office | 2023
|