Automatic Pore Typing Classification from 2D Images
- D. Floriello (Eni S.p.A) | A. Ortenzi (Eni S.p.A) | M. Idiomi (Eni S.p.A) | S. Ricci (Eni S.p.A) | A. Amendola (Eni S.p.A) | S. Carminati (Eni S.p.A) | E. Baralis (Politecnico di Torino) | P. Garza (Politecnico di Torino) | A. Pasini (Politecnico di Torino)
- Document ID
- Offshore Mediterranean Conference
- Offshore Mediterranean Conference and Exhibition, 27-29 March, Ravenna, Italy
- Publication Date
- Document Type
- Conference Paper
- 2019. Offshore Mediterranean Conference
- Pore types classification, Machine Learning, Pore Network Characterisation
- 6 in the last 30 days
- 29 since 2007
- Show more detail
Obtaining a detailed description of the reservoir evolution is an essential step for several reasons. Among the most important ones, there are the estimation the reservoir efficiency and the estimation of some of its dynamic properties.
The internal structure of reservoir rocks is traditionally investigated using thin sections that are prepared so that they represent the main geological units of the reservoir itself. At the optical and electron microscope, the thin sections (30 micron) allow for the recognition of the depositional and diagenetic features that are present in the sediments. Some of this information is conveyed through the pore types and their geometrical, morphological and appearance features. The type of the pores (e.g. primary or secondary) that can be found in the thin sections is related to their origin, the geological evolution of the rocks and of the reservoir itself.
However, several millions of pores can be present in an image and technicians have done the pore classification manually for more than 20 years. Such an activity, therefore, is extremely time consuming. Moreover, the technician can insert some of her own subjectivity in the interpretation of the pore types. Therefore, we aim at overcoming the outlined criticalities by an automatic classification of the pore types present in the images gathered with a scanning electron microscope. This will allow for a faster interpretation of the classes of the pores in the images and an improved objective classification. We want to achieve the automatic classification of the pore types by using only the geometrical, morphological and some simple appearance features of the pores as extracted from the images. We use an unsupervised algorithm to cluster the pores and from the result obtained, the type of the pores themselves is inferred. Several different algorithms have been tested on images coming from sandstone and carbonate reservoirs. The results of the proposed workflow are promising.
|File Size||746 KB||Number of Pages||6|