Bangladeshi road sign recognition based on DtBs vector and artificial neural network

Road sign recognition from road image is a crucial area of intelligent transportation system which role is to raise awareness of drivers, pedestrians and automated driving system. In this regard, a framework has been proposed in this paper for recognizing Bangladeshi road sign (BRS). For detecting the road sign (RS), two natural properties of a BRS is utilized, they are - border color rim and shape of the RS. Secondly distance to borders (DtBs) vector is used for shape verification of BRS. For recognizing those candidate regions artificial neural network (ANN) has performed. Based on these ideas initially, YCbCr color model is used to eliminate the illumination sensitiveness and statistical threshold value is used for color segmentation. Next, labeling and filtering is used to extract the shapes of candidate region. DtBs vector is used to verify the region of interest (ROI) of BRS. Affine transformation is used for triangular and quadrangular region to avoid sheared road sign (RS) blob. Finally, ANN is used to recognize the BRS blob. Various road signs are utilized to test the proposed framework under various conditions and results are presented to verify its efficiency.