Method of detecting at least one geological constituent of a rock sample
The present invention relates to a method of detecting a geological constituent (cge) of a rock sample, by use of a machine learning method (ALG) trained with a limited number of training images (IAP), through segmentation (SEG) and partitioning (DEC) of training images (IAP) into multiple patches.
1. A method of detecting at least one geological constituent of a rock sample from training images, comprising steps of:
a) segmenting the training images by digital processing into a segmented image of at least two colors with a first color of each area of the training images comprising no geological constituent, and with at least a second color of each area of the training images comprising a geological constituent;
b) partitioning the segmented training images of at least two colors and the segmented images into patches;
c) training a machine learning classification algorithm by using the patches of the segmented training images of the at least two colors and by using the patches of the segmented training images for classifying the patches of the training images according to colors of the patches of corresponding segmented training images;
d) acquiring an image of the rock sample;
e) partitioning the acquired image of the rock sample into the patches; and
f) detecting the at least one geological constituent in each of the patches of the acquired image of the rock sample by application of a model formed by the trained machine learning application of the classification algorithm to the patches of the acquired image of the rock sample.
2. A detection method as claimed in claim 1 , wherein the geological constituent is selected from one of microfossils, nanofossils, plant debris, minerals, and pollen spores.
3. A detection method as claimed in claim 1 , wherein each one of the at least second colors corresponds to a type of geological constituent.
4. A detection method as claimed in claim 1 , wherein a number of training images ranges between 3 and 20.
5. A detection method in accordance with claim 4 , wherein the number of training images ranges between 3 and 10.
6. A detection method as claimed in claim 1 , wherein the machine learning algorithm uses an artificial neural network, selected from one of a convolutional neural network and a fully convolutional neural network.
7. A detection method as claimed in claim 1 , wherein the image of the rock sample is acquired from a section of the rock sample.
8. A detection method as claimed in claim 1 , wherein the image of the rock sample is acquired by use of an optical or electron microscope, using light, photography, synchrotron radiation tomography scanning or X-ray imaging.
9. A detection method as claimed in claim 1 , wherein the machine learning classification algorithm is trained to classify the patches of the training images, by analyzing the color of a central pixel of the patches of the segmented images.
10. A detection method as claimed in claim 9 , wherein the acquired image of the rock sample is partitioned into patches by partitioning a patch which surrounds each pixel, and the model formed by the trained machine learning classification algorithm determines whether the surrounded pixel belongs to the geological constituent.
11. A detection method as claimed in claim 1 , wherein the machine learning classification algorithm is trained to segment an entire surface of the patches of the training images to delimit the geological constituent.
12. A detection method as claimed in claim 11 , wherein the geological constituents are detected by delimiting each patch of the acquired image of the rock sample of the geological constituent.
13. A detection method as claimed in claim 1 , wherein the detection method comprises at least one of:
counting a number of geological constituents of the acquired image of the rock sample from detection of the geological constituents;
determining from detection of the geological constituents a proportion of a rock sample volume occupied by geological constituents;
estimating at least one of morphological, textural characteristics of the geological constituents; and
applying a supervised classification method for categorizing the geological constituents from the detection of the geological constituent.
14. A method of exploiting a soil or a subsoil comprising:
a) detecting at least one geological constituent of a rock sample by using the method of claim 1 ; and
b) exploiting the soil or the subsoil according to the detection of the geological constituent of the rock sample.
15. A soil or subsoil exploitation method as claimed in claim 14 , wherein the soil or subsoil exploitation comprises constructing a structure on the soil or in the subsoil, constructing gas storage in the subsoil, exploiting raw materials of the soil or the subsoil, and wherein raw materials contained in the soil or the subsoil is one of rock, a material or a fluid.
16. A method of determining climate in a geographical area through geologic ages, comprising:
a) taking at least two rock samples at different depths of an underground formation;
b) detecting at least one geological constituent for each rock sample by using the detection method of claim 1 ; and
c) determining the climate and geologic age in the geographical area according to the detected at least one geological constituent.