Information retrieval of LED text on electronic road sign for driver-assistance system using spatial-based feature and Nearest Cluster Neighbor classifier

In traffic management, electronic road sign (ERS) plays an important role for providing a real-time traffic-related information. Electronic road sign generally displays discontinuous characters composed by a matrix of light-emitting-diode (LED) lamps, namely LED text. On the other hand, text recognition has been thoroughly studied for a long time, especially for vision-based intelligent transportation systems. However, LED text recognition still remains a challenge due to its different types and consisting of discontinuous character. Therefore, this work addresses a method to recognized LED text contained on electronic road sign, as an application to intelligent transportation systems (ITS).Our proposed method extracts supporting points representing as a center of LED segment on binary map of the input image. Character region generating is applied using density-based spatial region generation (DBSREN) method. The spatial information of each point contained on region is utilized for acquiring region feature. Nearest Cluster Neighbor (NCN) classifier is then used for training as well as recognizing stage. In experiment, our method achieves 87.4% and 61% of detection and recognition rate, respectively. The processing time is almost 12 frames per second. These results prove that the proposed method is an effective and fast to retrieve information from LED text contained on electronic road sign for robust ITS application.