IMPLEMENTASI METODE JARINGAN SARAF TIRUAN BACKPROPAGATION PADA SISTEM PENGENALAN NOMOR GERBONG KERETA API BATUBARA RANGKAIAN PANJANG (BABARANJANG); IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK BACKPROPAGATION METHOD ON BABARANJANG WAGON NUMBER RECOGNITION SYSTEM

Khorunisa, Rilla, R. Sumiharto

2016 | Skripsi | FMIPA

Batubara rangkaian panjang (Babaranjang) is a train owned by PT. KAI used for hauling coal in South Sumatra. In each of wagon for hauling coal there are wagon number that should be noted by an operator. It is used for reporting a database of coal weight for the next distributor. A system of recording wagon number done by an operator TLS manually by looking at the wagon number directly and typed on datalog. It is valued uneffective , because of error possibility in recording wagon number is very large. In this study , done the implementation of artificial neural network backpropagation on Babaranjang wagon number recognition system. The wagon number recorded and saved in video format, then processed by digital image processing using MATLAB. The first step is pre-processing used to extract the wagon number automatically and improve the quality of image. The next stage is image segmentation, this stage consists of two parts, those are wagon number image segmentation and every character wagon number segmentation. Segmentation step uses the vertical projection and horizontal projection. After obtaining image character wagon number, done the introduction of characters use artificial neural network backpropagation. Training step use multilayer perceptron architecture that consist some parameters, such as learning rate, momentum constant and transfer functions. The results of the study show the system can recognize wagon number well. Training step is done with 295 characters and obtained value accuracy is 95.93%. While, the testing step performed by 294 characters and obtained value accuracy is 96.26%.

Kata Kunci : Babaranjang, artificial neural network, backpropagation