As one of the main routes for the mobility of people and goods around the world, paved streets and roads must undergo regular maintenance in order to have suitable conditions of use. However, with extensive paved meshes and budget constraints, the optimization of these resources is essential. In this way, it is necessary a frequent monitoring of these paved roads to identify priority sites and to create a cost-effective maintenance plan. However, this monitoring is often performed by visual inspection using trained humans, which is highly costly. In this sense, researchers have tried to develop automatic computer vision tools to perform this monitoring, classifying images of paved stretches in an attempt to identify deterioration in them. On the other hand, Deep Neural Networks are currently the state of the art in general texture classification. Therefore, this work proposes to investigate the use of Deep Neural Networks for automatic detection and classification of degradation in asphalt pavement images. With this, it is expected that deep neural networks will also become the state of the art for this problem, with a greater recognition accuracy of degradations, as a basis for the future development of geographic information systems with information about the asphaltic pavements of cities, allowing its managers to reduce preventive and/or corrective maintenance costs.
News published in Agência FAPESP Newsletter about the scholarship: