Sistema embarcado para identificação de pistas de rolamento

Abstract

The ground on which a motor vehicle is traveling is one of the main factors that affect its handling and, consequently, its safety. In order to contribute to the evolution of the automotive industry, this work presents the development of an embedded system that uses a multilayer perceptron neural network to identify driving lanes, classifying them in asphalt, stone paving or sand. Neural networks are widely used for pattern recognition, because similarly to natural intelligence, they are able to learn from examples and become able to distinguish data from different classes. However, developing such an application in an embedded system becomes a challenge due to the limited resources. Using an Arduino and an RGB sensor, the designed hardware collects informations about the color of the surface on which the sensor moves and identifies the type of driving lane. Therefore, the device can provide important information to existing systems, allowing them to make the most appropriate decisions to adapt to the situation they are in.