Traffic sign detection — A new approach and recognition using convolution neural network

Traffic Sign Recognition (TSR) system is a component of Driving Assistance System (ADAS). The TSR system assists the drivers in safe driving as road signs provide important information of the road. This research focuses to design and develop a TSR system by using color cues and Convolution Neural Network (CNN) as both features extractor and classifier for Bangladeshi traffic signs. In the first step, after image acquisition, some pre-processing task is performed. Then the image is segmented using color information of HSV color model. After that, morphological closing is executed to fine the segmented image. Consequently, after filtering the image by using region properties and shape signature, the desired region is cropped. Finally, the extracted sign area is classified by means of automatic features extraction with deep CNN. The experimental results illustrate that the proposed algorithm shows comparable performance with good recognition accuracy.