This paper presents a method for performing traffic light detection using computer vision. Reliable traffic light detection and classification is crucial for automated driving in urban environments. By using big data and artificial intelligence, a complex dataset belonging to an urban area in China is preprocessed to determine the level of vehicular congestion, and then different machine learning algorithms are applied to a dataset of traffic light images in order to validate them in the urban environment to be studied, this process is explained step by step. The models obtained in this work can be applied in optical camera communication (OCC) systems, and also in intelligent transportation systems (ITS), using tracking channels for visible light communication (VLC). The two optical channels, VLC and OCC, are simulated in terms of the quality of information received in order to apply the previously generated datasets. In this work, a traffic light feature dataset has been generated from images and two traffic light classification models present in images and video frames have been generated from their features, obtaining a maximum accuracy of 94.52 %.