Abstract:
A plate number recognition system is a computer based system, integrated with the red light system, that is designed for the identification of the plate number of a vehicle violating the beating the red light rule.
The system has been developed to augment the current manual apprehension system of the Metro Manila Development Authority (MMDA) in apprehending erring motorists. The system addresses issues on potential bribery between motorists and traffic enforcers and situation where in no traffic enforcers could monitor a particular intersection.
It possesses functionalities that are under the fields of computer vision and neural networks. This includes Edge Detection, Connected Components, Morphological processes, Feed-forward neural network and the Backpropagation method of training. The edge detection is used in the process of localizing plate number candidates which is helped by the different morphological methods. Connected Components segment the localized plate number candidates to separate the potential characters of a plate number. To recognize the segmented characters, they are then used as inputs for the feed-forward neural network trained by the backpropagation method.
