Identifikasi Plat Nomor Kendaraan Dengan Zone Based Feature Extraction Menggunakan Metode Klasifikasi Bacpropagation

Abstract

Automatic license plate identification is one of the important features required for vehicle data recording system to be use in applications such as parking system, automatic highway gate system, etc. Several research has been conducted to devise a reliable method to identify vehicle license plate. In this research, we aim to implement an automatic license plate identification system using the zone based feature extraction method and artificial neural network for classification. The data is obtained from 100 units of vehicle using a 5 MP mobile phone camera. The preprocessing step consists of converting the images to grayscale, followed by noise reduction using median filter, edge detection using Canny with a threshold of 0,2 and 0,5. Afterwards, we perform segmentation using 8-connected labelling component to obtain the characters. The zone based feature extraction used is image centroid and zone using the most efficient zone. The fastest and the highest accuracy will be choosen as the most efficient zone. In this research 14 zone extraction have an efficient result. We used the backpropagation neural network with 25 input neurons, 30 hidden neurons, and 36 output neurons representing each characthers (alphabet and numerals). The best result for indvidual character recognation is 85,32% while the best recognition rate for the whole plate is 40,61%