Détection et reconnaissance de texte dans les documents vidéos. Et leurs apports à la reconnaissance de personnes

This article presents the different steps used to recognize characters for multi-modal person recognition systems in video (ANR REPRE challenge). Text detection is achieved by a technique based on the text features (texture, color, contrast, geometry, temporal information, measure of accumulated gradients). The text recognition is then performed by the free software Google Tesseract. The method was evaluated on a TV news corpus containing 59 videos from the France 2 French TV channel. MOTS-CLÉS : Détection des textes, Vidéo Optical Character Recognition (OCR), reconnaissance