Identity verification using computer vision for automatic garage door opening

We present a novel system for automatic identification of vehicles as part of an intelligent access control system for a garage entrance. Using a camera in the door, cars are detected and matched to the database of authenticated cars. Once a car is detected, License Plate Recognition (LPR) is applied using character detection and recognition. The found license plate number is matched with the database of authenticated plates. If the car is allowed access, the door will open automatically. The recognition of both cars and characters (LPR) is performed using state-ofthe- art shape descriptors and a linear classifier. Experiments have revealed that 90% of all cars are correctly authenticated from a single image only. Analysis of the computational complexity shows that an embedded implementation allows user authentication within approximately 300ms, which is well within the application constraints.