Quantification par analyse d'images de la granulométrie des roches fragmentées : amélioration de l'extraction morphologique des surfaces, amélioration de la reconstruction stéréologique

Abstract : This research deals with a more general context of the quality control for some important operations in mining production, predominantly blasting, crushing and grinding. It is interested more in the granulometry quantification of the fragmented rocks by using image processing. In general, the development work is mainly related to the improvement of two most important stages of measurement : the contours extraction of the fragments at the image (2D) and the rebuilding of the 3D size-distribution-curve (stage of stereology, 3D). Nowadays, screening (or more general the sifting) is the most used and important process for measuring the granulometry of the fragmented rocks ; It is therefore considered as the current reference of measurement. Consequently, any measurement established by image processing will then be validated on the basis of screening's result.

With regard to the stage of contours extraction of fragments, the major problem is how to determine the correct filtering of the noise at the image. First of all,new tools for filtering, based on the residual morphological transformations, and rather powerful and adapted to the case of fragment images of rocks, are introduced and evaluated. The study, undertaken on grey level images, approaches the various aspects of the noise problems from a topographic point of view. Surfaces of the fragments are then extracted carefully with the help of a segmentation of the image. The extraction process is principally based on the construction of a watershed controlled by the markers, which are extracted by filtering. The robustness and the automation of the step are afterward validated for several cases of images of rocks fragments.

The second problem carried out in this research is the rebuilding of 3D size-distributioncurve starting from the surfaces extracted by image processing. In order to avoid the interference of many skews related to this stage, we recommend to separate the problem into two subproblems, according to the origin of analysed materials (cut down Heap, truck, belt conveyor) :
- in the case of weak covering and overlapping : an experimental model of volumic rebuilding bringing the data of the refusal materials (surface and volumes), and based on the results of reference (screening), is developed. The two laws of the sizes and the thickness (laws of rebuilding), which make possible to determine precisely the size and volume of surfaces, are then validated in experiments.
- for the complex case of the appearance of masking between fragments : we commence by putting some essential assumptions for the adaptation of the theory of the random structures model into our fragmented rocks case. The analysis of the limitation of a direct application of the random model of dead leaves (assumption of the Booleandiagram) may lead us to develop a new method of acquisition based on an incidence with variable of illumination. This new method inspired from the technique "shape-fromshading", makes us possible to obtain an additional information on covering (solid drop shadow between fragments) for the application of the statistic of the intact grains.

Finally, the results obtained in this research for the case of the fragments of rocks as well as for other type of images of granular elements, enable us presently to consider a valorisation not only for mineral industry but also for other fields such as the agro industry, the pharmaceutical industry and cement factories.
Liste complète des métadonnées

https://pastel.archives-ouvertes.fr/pastel-00579926
Contributor : Bibliothèque Mines Paristech <>
Submitted on : Friday, March 25, 2011 - 2:16:26 PM
Last modification on : Monday, November 12, 2018 - 10:53:55 AM
Document(s) archivé(s) le : Sunday, June 26, 2011 - 2:44:47 AM

Identifiers

  • HAL Id : pastel-00579926, version 1

Citation

Souhaïl Outal. Quantification par analyse d'images de la granulométrie des roches fragmentées : amélioration de l'extraction morphologique des surfaces, amélioration de la reconstruction stéréologique. Géologie appliquée. École Nationale Supérieure des Mines de Paris, 2006. Français. ⟨NNT : 2006ENMP1355⟩. ⟨pastel-00579926⟩

Share

Metrics

Record views

1203

Files downloads

11992