Traffic Light Detection and Recognition Using Image Processing and Convolution Neural Networks

Automatic traffic light detection and mapping is an open research problem. In this paper, a method for detecting the position and recognizing the state of the traffic lights in video sequences is presented and evaluated using LISA Traffic Light Dataset which contains annotated traffic light video data. The first stage is the detection, which is accomplished through image processing technics such as image cropping, Gaussian low-pass filtering, color transformation, segmentation, morphological dilation, Canny edge detection, and Circle Hough transform to estimate the position and radius of possible traffic lights. The second stage is the recognition, whose purpose is to identify the color of the traffic light and is accomplished through deep learning, using a Convolutional Neural Network. Day and night images were used in both training and evaluation, giving excellent location rates in all conditions.