Granulometry using Mathematical Morphology and Motion

摘要

This paper describes an efficient framework for image-based granulometry, which provides three main contributions. First, unlike most of the existing work on particle delineation, which is based on the use of detected edges, our method-a region-based approach-relies on the use of mathematical morphology techniques. Second, we have explored the use of motion in order to recover the visible volume of the detected lumps through the recovery of dense disparity maps. Third, we have applied the proposed framework to the measurement of oil sand lumps, a problem not addressed in the literature of granulometry. The on-line analysis provided by the developed framework has been evaluated successfully on video streams. From the experimental results, we believe that the proposed framework can be applied efficiently to various real applications that are not limited to granulometry systems.