Análise Comparativa de métodos de segmentação e técnicas de aprendizado de máquina com aplicação no reconhecimento automático de placas de identificação de veículos

Abstract

The present study deals with the problem of Automatic Number Plate Recognition (ANPR), which involves the use of different techniques, including Optical Character Recognition (OCR) and machine learning, to perform the process of character recognition of a vehicle license plate in an image. As this process is done under real conditions, there are difficul ties with regard to various types of noise that may be present in the acquisition of images, such as irregular lighting, occlusion of characters due to some kind of object attached to the vehicle license plate and positioning of the characters, among others. Since the ap plied segmentation method has a direct influence in the attenuation of the negative effects of these noise, one of the objectives of this study is to evaluate which of these methods optimizes the performance of both the OCR developed and the Tesseract OCR engine module | a software considered to be the state of the art in this area. Lastly, using the precision and recall values as evaluation measures, we analyze which machine learning technique obtain the best results in character recognition and compare these results with Tesseract in order to verify the performance of the proposed approach