US 12,339,931 B2
Method for categorizing a rock on the basis of at least one image
Antoine Bouziat, Rueil-Malmaison (FR); Jean-Claude Lecomte, Rueil-Malmaison (FR); Renaud Divies, Rueil-Malmaison (FR); Sylvain Desroziers, Rueil-Malmaison (FR); and Arnaud Cayrol, Rueil-Malmaison (FR)
Assigned to IFP ENERGIES NOUVELLES, Rueil-Malmaison (FR)
Appl. No. 17/782,673
Filed by IFP Energies nouvelles, Rueil-Malmaison (FR)
PCT Filed Dec. 3, 2020, PCT No. PCT/EP2020/084389
§ 371(c)(1), (2) Date Jun. 6, 2022,
PCT Pub. No. WO2021/115903, PCT Pub. Date Jun. 17, 2021.
Claims priority of application No. 1914253 (FR), filed on Dec. 12, 2019.
Prior Publication US 2023/0008058 A1, Jan. 12, 2023
Int. Cl. G06N 5/01 (2023.01); G06F 18/2413 (2023.01); G06T 7/73 (2017.01); G06V 10/46 (2022.01)
CPC G06F 18/2413 (2023.01) [G06N 5/01 (2023.01); G06T 7/73 (2017.01); G06V 10/46 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30181 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A soil or subsoil exploitation method using a computer based method that classifies rocks based on rock images of the rocks to be classified which are stored in a rock image database comprising:
a) constructing a decision tree classifying the rock images using at least one descriptor that characterizes each of the rocks, the decision tree classifying lithologic facies of the rocks by using the descriptors to characterize the rocks which are selected from among an origin of the rocks, chemistry of the rocks, presence of foliation, presence of olivine, presence of bedding, and presence of mica;
b) associating descriptors characterizing the rocks to the rock images of the rock image database;
c) acquiring the rock images of the rock to be classified by at least one of photography, microscopy, and scanning of the rocks;
d) training a machine learning method including a neural network on the descriptors characterizing the rocks associated with the stored rock images of the rock image database and determining the descriptors of the acquired rock images to be classified by applying the trained machine learning method to the descriptors characterizing the rocks associated with each rock image of the rock image database;
e) determining the classification of the rocks to be classified by applying the decision tree to the determined descriptors of the rock images to be classified; and
f) exploiting the soil or subsoil by performing construction on the soil or in the subsoil according to the determined classifications of the rocks.