2.1.5 Design of an Intelligent Vehicle License Plate Recognition System
Yap Keem Siah [14] proposed different types of Neural Networks that can be
used to classify the characters. For example, multi-layer preceptors network with back
propagation learning algorithm (BP Network), Fuzzy ARTMAP neural network, learning
vector quantization (MTLVQ) and also probalistic neural network (PNN). In this
experiment, the BP network has the highest recognition rate among all neural network
methods with 99.67%. However, there are number of factors must be considered strictly
for the neural network comparisons such as learning rate, initial weight and etc.
Therefore, second experiment was conducted to test the neural networks on their
recognition ability with training sets. From this second experiment, the Fuzzy ARTMAP
can memorize 100% of trained samples. It is the best networks used in memorizing
trained data.