UMA METODOLOGIA PARA DETECÇÃO DE SINAIS DE TRÂNSITO EM VÍDEO

• Published in: Colloquium Exactarum (Online) Vol. 6; no. 3; pp. 26 - 44
• Main Authors: Nascimento, Antonio Carlos Paes, Silva, Francisco Assis da, Pazoti, Mário Augusto, Pereira, Danillo Roberto, Artero, Almir Olivette, Piteri, Marco Antonio
• Format: Journal Article
• Language: English; Portuguese
• Published: Universidade do Oeste Paulista 30.10.2014
• Subjects: Adaboost; Real Time; Traffic Signs Detection
• ISSN: 2178-8332; 2178-8332
• DOI: 10.5747/ce.2014.v06.n3.e086

In this work is performed the detection of road traffic signs using video frames of high resolution do not considering all the information printed on plate. Plates are searched considering only the information as edges, red circles with or without the diagonal lines representing prohibition, plates with yellow background, and plates with red background. The method used to perform the detection is proposed by Viola and Jones (2001), with a specific training method for traffic signals. To achieve good results, assessments of colors in the color space HSV were performed to eliminate regions that contain plates, and thus decrease the rate of false positives during the detection process. The results obtained from the experiments show that it is possible to detect traffic plates of various shapes in an acceptable processing time for a real time application.