ASSESSING THE WEATHERING OF BASALTIC AGGREGATES USING IMAGE TEXTURE ANALYSIS
Abstract
Mineral aggregates are the most mined materials in the world. Aggregates are a component of composite materials such as asphalt concrete, used mainly in highway´s paving. The importance of aggregate is because it serves as reinforcement to add strength to the overall composite material. Aggregates should be obtained from non-weathered rocks. Aggregates from weathered rocks commonly present low resistance to abrasion and compression, especially under conditions of frequently repeated vehicle's load, such as on highways. This paper presents an evaluation of image texture analysis when applied to detect the presence of weathering in aggregates. An evaluation of the texture analysis using histograms, entropy, local binary patterns and local phase quantization as descriptors aiming to identify weathering in basaltic aggregates is presented. These texture descriptors and the Naive Bayes, IBK, J48 and Multilayer Perceptron machine learning algorithms were used to detect the presence of weathering in aggregates. The results obtained indicated that red channel stands out as descriptor to differentiate weathered from non-weathered aggregates with an average error around 4%.