Análisis de métodos para reconocimiento de estado de la luz de los semáforos aplicada para carros de conducción autónoma.

Abstract

This study about how to recognize the light of the traffic light touch different algorithms of computer vision and machine learning with the objective to be applied to self-driving cars. Also, the search about the legal regulations of the self-driving cars national and internationals that now are few or nothing regulations in different countries. The methodology was the experimental, trying different algorithms and parameters tuning to obtain the best approach in the test dataset. The dataset was of 1484 images, split in 80% for training and 20% for test, also finally was used locals traffics lights images for validate the model. In the study shows that implement machine learning can improved the approach of the recognizer the light a different of using only computer vision algorithms. The result was a computer model that can process and predict the state of traffic light of images of never see before, for a future application in self driving.