Pengenalan Karakter Plat Kendaraan Bermotor berbasis Citra dengan menggunakan Metode Canny dan Algoritma Backpropagation

Abstract

Transportation is important needs in daily life. However, there are so many problems in transportation system in our country. The one is public transportation. To overcome it, the government implement KIR. But this KIR has several weakness. One of them is manual data. As cosequnces, human error in listing can came out as the process goes on. Pattern recognition can be used to implement automatic number plate identification in this system. One of the method is canny filter. Canny filter is uses to obtain a good image in the character image acquisition. Characters based with 12X7 pixels are be converted into binary as input for Multi Layer Perceptron with 3 layers node number of each node 84, 50, 36. Artificial neural network is trained with back propagation algorithm with a learning rate parameter 0.3 and momentum 0.9. The training process will be terminated when the iteration reaches a maximum value of 10,000 or MSE (Mean Square Error) 0.0001. Recognition rate for numeral character is 100%, however recognition rate for letter character is little bit worser, 86,87%. So overall performance is 94,29% for the whole characters.