Neural network system of traffic signs recognition

This paper describes an approach for detection and recognition of traffic signs in real time with account for illumination and distance changes. A small single-board computer Raspberry Pi 2 and a webcam Hama AC-150 were used to implement the proposed algorithm. A scheme for determination traffic sign location uses color filter with morphological operators and Canny edge detector, identification of sign type is based on multilayer perceptron neural network. Variations of five traffic signs were used to train and test an algorithm. As a result experiments were successfully performed. Developed system is robust to light changes and is able to recognize traffic signs 20 cm in diameter from 1.5–2 m distance.