Video processing for V2V communications: a case study with traffic lights and plate recognition

In recent years there have been changes in the way cars are designed. Car manufactures put a lot of effort on safety and systems that provide information to the driver with the long-term objective of achieve a complete self-driving car. Nowadays the most effective approach relies on data fusion of information, coming from a plethora of different sensors like RADARs and videocameras. While some of these sensors are already available on commercial cars, others will be introduced step-by-step. Therefore, data fusion algorithms should address the possibility to manage duplicated information and using this redundant information to validate the received data. In this work we describe two near-real time algorithms which exploit the video stream acquired by on-board camera. One allows for the identification of traffic light status and the second one addresses vehicles tracking and plate recognition.