Detection and recognition of Bangladeshi road sign based on maximally stable extremal region

Abstract

Road sign detection and recognition systems (RSDRS) offers an autonomous driver support system, which mainly concerns of safety for drivers, road users as well as passengers as a portion of Advanced Driver Assistance System (ADAS). RSDRS are mainly used to help the drivers (especially those who have a disability to drive) by warning them about the existence of road signs to lessen the risks in a situation of driving disruption, tiredness and in rough weather. Though many RSDRS has been proposed in the literature as well as many research areas. But detecting and recognizing of road sign is still challenging matter because of uneven illumination, complex background, different shape, various distance and different angle. This research aims to detect and recognize road signs of Bangladesh. A system has been proposed in this paper for automatic detection and recognition of Bangladeshi road sign which generates a chromatic normalized image that has two channels and detects the candidate regions as maximally stable extremal regions (MSERs), which offers physique in a variation of illumination conditions of Bangladeshi road sign. The recognition process is done by a cascade of support vector machine (SVM) classifiers in which the images are trained by using histogram of oriented gradient (HOG) features.