Identificação de placas de trânsito através da classificação de imagens usando redes neurais

Abstract

Recently, automotive vehicles have included driver's aid systems, which analyze the environment through the acquisition and digital processing of images, including modules for identification and recognition of transit boards. For this process, the initial step is to identify the transit boards present in the image. This work shows the steps developed, such as converting to grayscale, thresholding and segmentation of the image, to isolate transit boards in images acquired by a camera coupled to a Raspberry Pi B +. Shortly after, it was accomplished the artificial neural network learning, and this showed promising results with the recognition of a particular plaque. Thus, the Raspberry Pi B + sends commands to an Arduino Uno, which controls a self-follower line, performing actions on the engines to perform maneuvers according to the recognized plaque.