Assessment of aggregate-bitumen coverage using entropy-based image segmentation

Adhesion between bitumen and aggregates has a major impact on the performance of asphalt mixes. In order to avoid stripping and consequent premature failure of a pavement, methods to assess the adhesion have been developed and standardised. These methods, such as the frequently applied rolling bottle test defined in EN 1269-11, are based on visual inspection of specimens and subjective evaluation. This shortcoming can be eliminated by employing image analysis that accurately and consistently identifies the portion of aggregate covered by bitumen. However, the image segmentation procedure cannot be based purely on intensity thresholding because the light reflections on the glossy surface of bitumen merge together with light aggregates, and dark aggregates coincide with diffuse bitumen reflections. In this article, a method based on local entropy calculation is proposed to tackle this issue in the form of an open-source software tool. Its algorithms are thoroughly described here, and their impact is illustrated using a model example. Software outcomes are compared with an advanced, yet more complicated, approach. How the software might be used in other fields is also outlined and discussed.