Designing Automatic Number Plate Recognition (ANPR) Systems Based on K-NN Machine Learning on the Raspberry Pi Embedded System

Abstract

Research on vehicle number plate recognition or Automatic Number Plate Recognition (ANPR) is mostly done by researchers to produce an introduction that has high accuracy. Several methods of introduction are carried out such as introduction to edge detection and morphology, relationship analysis between objects, machine learning and deep learning. In this research a K-NN machine learning ANPR system was developed in character recognition. The method of analyzing relationships between objects is used to localize number plates. The system that was developed also added an artificial intelligence to be able to find out the fault of the number plate recognition and fix it based on the position of the character group in the number plate. The ANPR system is designed to be an Embedded system so that it can be implemented to be able to carry out the identification of two-wheeled and four-wheeled vehicle license plates. The ANPR system was also developed to be used in the parking management system. In this research the recognized number plates are limited to private number plates in Indonesia. In testing, the system is made capable of recognizing the number plates of two-wheeled vehicles and four-wheeled vehicles on vehicles that have a standard license plate according to Polri regulations, both in the font type and the number plate writing format. The results of vehicle number plate recognition reached an accuracy of 98%.