An algorithm for text detection in images from street view, basing on Haar-like features and AdaBoost classification is proposed in this paper. The idea is intended for searching a wanted place in a foreign city with the business name, the scene text and the street address. There are two contributions in this paper. First the difficulty of locating the specified buildings exactly in an unfamiliar city is analyzed, and then a novel application associated with a mapping application to the real view of target place is presented in this paper. Another is training a cascade AdaBoost classifier using several weak classifiers with Haar-like features, which examined our approach. The feasibility of proposed application has been evaluated using 216 training samples and 128 testing samples captured by ourselves in street view. The overall performance of our experiment is improved, a precision of 72.6% and a recall of 79.9%.
Reading text in street views using Adaboost: Towards a system for searching target places
2009 IEEE Intelligent Vehicles Symposium ; 227-232
2009-06-01
1940324 byte
Conference paper
Electronic Resource
English
Reading Text in Street Views Using Adaboost: Towards a System for Searching Target Places
British Library Conference Proceedings | 2009
|Target Track System Design Based on SVM and AdaBoost
IEEE | 2008
|TARGET SEARCHING DEVICE, TARGET SEARCHING METHOD AND TARGET SEARCHING PROGRAM
European Patent Office | 2018
|Transportation Research Record | 2006
|