Sistema computacional para identificar níveis de degradação em barras de aço

Abstract

Reinforcement corrosion in concrete structures is one of the most important pathological manifestations that leads to degradation and structural damage of buildings, while represents risks to people lives. Analysis and identification of degradation levels of these structures is characterized as a repetitive, monotonous and time-consuming process. In this scenario, computer systems based on computer vision have been applied to similar problems to replace or assist human specialists. This paper presents a proposal to classify different degradation levels in steel bars through computer vision. A widely known set of tools in the machine learning area was used and validated the proposal. The best results achieved recognition rates of 75.6% by using LPQ descriptor and SVM machine learning algorithm.