Method for automated classification of porosity in microscopic slides and computer-readable storage media

Description

/ 27 “METHOD FOR AUTOMATED CLASSIFICATION OF POROSITY IN MICROSCOPIC SLIDES AND READABLE STORAGE MEDIA BY COMPUTER" TECHNICAL FIELD

[001] The present invention is situated within the scope of oil exploration and production, specifically in the processes of production, modeling, simulation, and evaluation of reservoirs. Thus, the present invention focuses on presenting a new methodology that standardizes the analysis of porosity through petrographic images and evaluates, at the pixel scale, the relationships of porous elements with their adjacent elements, enabling more precise classification and quantification, favoring studies and data analysis. In this sense, the new methodology allows for the massive acquisition of data, potentially supporting decisions regarding the production of oil reservoirs and making information from these studies available in a timely manner within the oil exploration and production chain, in addition to minimizing user / generator error in this type of information. FUNDAMENTALS OF THE INVENTION

[002] Petrographic characterization is a crucial step in oil field studies, being used to acquire knowledge about source rocks, seal rocks and, mainly, reservoir rocks. Microscopic studies aim at the qualification and quantification of solids (mineral phases) and voids (popularly known as porosity).

[003] In petrography, the rock data analyzed are reduced to a volume of infinitesimal thickness, considered two-dimensional and randomly oriented (in most cases). The constituents found there (among them, porosity) are considered two-dimensional sections, representative or not, of the sampled volume. The most technical definition for porosity in this context is that of PorEl (from Petition 870250081304, dated 10 / 09 / 2025, page 8 / 84 / 27 (English Porous Element), which refers to a two-dimensional section that captures a more or less complex arrangement of elements of the porous system.

[004] In any case, the characterization of PorEls involves both the acquisition of quantifiable parameters, such as area and geometric attributes (axis sizes, sphericity, porosity area, etc.), and the classification of the pore type. This classification can vary according to the objective of the study, encompassing different criteria. In general, this classification suggests a genetic origin of the pore, often materialized in its relationship with adjacent constituents, that is, its position and interaction within the rock system.

[005] Porosity presents itself as a diversity of forms, with varying geometric aspects and dimensions, resulting from the sum of processes (geological or otherwise) to which the rock volume has been subjected. The acquisition of this type of information is, even today, carried out manually, based on the specialist's evaluation, who obtains as a main product a relative quantification of the main types of pores classified. This activity is highly imprecise, as it is susceptible to user biases, as well as fatigue and inaccuracies related to a lack of expertise. Since this type of information has so many points of attention, its use is compromised, and it ends up having restricted use in more detailed characterizations.

[006] One of the biggest concerns associated with petrographic classification in the context of oil exploration and production is the accurate and timely identification of reservoir rock characteristics, such as porosity, permeability, and mineralogical composition. These parameters are fundamental for evaluating the quality and productive potential of reservoirs. The geological variability of the formations can make consistent analyses difficult, making the use of advanced methodologies indispensable. Petition 870250081304, dated 10 / 09 / 2025, page 9 / 84 / 27

[007] In this context, the need arises for the development of automated techniques, such as the use of artificial intelligence and machine learning, aimed at optimizing the process of classification and complete quantification of porosity in microscopic slides, reducing analysis time and increasing the reliability and uniformity of results. STATE OF THE ART

[008] Document US2022325613 A1 discloses a computer-implemented method that includes: accessing a first input dataset that encodes a plurality of petrophysical properties of a first set of wells in a reservoir; performing one or more petrochemical rock type (PRT) labels at least in part based on the first input dataset; at least in part based on one or more petrochemical rock type (PRT) labels, training one or more models for the reservoir using one or more machine learning algorithms; accessing a second input dataset that encodes the plurality of petrophysical properties of a second set of wells in the reservoir and applying one or more models to a second input dataset to determine a reservoir feature where the second set of wells differs from the first set of wells.

[009] The first set of input data may include one or more measurement records that encode petrophysical properties of rocks. Petrophysical properties may include: a porosity, a permeability, a pore geometry, a capillary pressure, and a saturation height function. One or more machine learning algorithms may include: a support vector machine (SVM), a self-organizing map, a random forest, an artificial neural network, a convolutional neural network (CNN), a UNet, and a ResNet. One or more models may be configured to run at least one of the Petition 870250081304, dated 10 / 09 / 2025, p. 10 / 84 / 27 following: a regression, a classification, a clustering, or a segmentation. Porosity is intercorrelated with permeability or relative permeability. Saturation height is intercorrelated with capillary pressure. Capillary pressure is intercorrelated with permeability or relative permeability, and pore geometry is intercorrelated with rock type. The workflow can access data that includes input core data and log data. Core data may refer to measurements taken on core samples taken from a well site. SCA data may include capillary pressure measurements, nuclear magnetic resonance spectroscopic data, relative permeability data, and X-ray diffraction (XRD) data.

[0010] Core petrography data may include scanning electron microscopy (SEM) data, thin sections, and core photographs. A deep neural network-based regression model can be used to establish a mapping between the input measurements listed above and the output properties targeted by the petrophysical modeling process. Once the model has been trained on the input training data and then validated based on, for example, validation data, machine learning models can then be applied to predict porosity / permeability (phi-k), saturation height function (SHF), petrophysical rock type (PRT), simultaneously or individually, from new measurement data.

[0011] The method described in the document includes: accessing an initial input dataset that encodes a plurality of petrophysical properties of a set of wells in a reservoir; performing one or more petrochemical rock type (PRT) labeling operations; training one or more models for the reservoir using one or more machine learning algorithms. Petrophysical properties may include: a porosity, a permeability, a pore geometry, a Petition 870250081304, dated 10 / 09 / 2025, p. 11 / 84 / 27 capillary pressure and a saturation height function. On the other hand, the document is silent regarding the possibility of extracting statistical distributions of properties (area, axis size, roughness, perimeter, etc.), frequency, and spatial distribution of each type of porous element, in order to detail the characterization of the rock.

[0012] Thus, document US2022325613 A1 does not provide a solution that directly and reliably correlates the geometric characteristics of pores, such as size, perimeter, and area, with their genetic classification. Although the state of the art presents classification systems that use geometric attributes, such as division by size, these systems do not establish a precise relationship with the genetic type and geometric properties of the pores.

[0013] Document US2024193427 A1 discloses a method for predicting the occurrence of a geological feature in a thin-section image, the method comprising the steps of: (a) providing a trained backpropagation-enabled classification process, the backpropagation-enabled classification process having been trained by (i) providing a training set of thin-section images; (ii) determining the scale of each of the thin-section images in the training set; (iii) extracting training image fractions from the training set of thin-section images; (iv) defining a set of geological features of interest, wherein the set of geological features comprises a plurality of classes; (v) selecting a class to label each of the training image fractions;(vi) insert the extracted training image fractions with associated labeled classes into the backpropagation-enabled classification process; and (vii) iteratively compute a prediction of the probability of occurrence of the class in the extracted training image fractions and adjust the parameters in the backpropagation-enabled classification model accordingly; Petition 870250081304, dated 10 / 09 / 2025, page 12 / 84 / 27 appropriate, thus producing the trained backpropagation-enabled classification process; and (b) use the trained backpropagation-enabled classification process to predict the occurrence of the class in an untrained thin-section image by (i) providing an untrained thin-section image; (ii) determining the scale of the untrained thin-section image; (iii) extracting an untrained image fraction from the untrained thin-section image, the extracted untrained image fraction having the same absolute horizontal and vertical length as the extracted training image fraction used to train the networks; and (iv) inserting the extracted untrained image fractions into the trained backpropagation-enabled classification process; (v) predicting a probability of occurrence of the class in the extracted untrained image fraction;and (vi) combine the probabilities for the extracted untrained image fractions to produce an inference for the occurrence of the class in the untrained thin-section image.

[0014] The method predicts the occurrence of each of the classes that characterize a geological feature in a geological thin-section image. A geological feature of interest may include: texture type, grain size, grain type, cement type, mineral type, rock type, pore size, and porosity type. The scale of each image in the training set is determined by the number of pixels corresponding to the length of the graphic scale bar, which is superimposed on the top of the image, subdivided by the absolute dimensions that the graphic scale bar represents, which is located next to the graphic scale bar at the top of the image, having a characteristic resolution or number of pixels in the horizontal or vertical direction.Untrained image fractions are extracted from an untrained thin-section image, exhibiting the same absolute dimensions as the extracted training image fractions, and fed into a classification-enabled process. Petition 870250081304, dated 10 / 09 / 2025, page 13 / 84 / 27 trained backpropagation. Predictions are produced by showing the probability of occurrence of classes A, B, C, D in extracted untrained image fractions, which are combined to produce an inference of classes A, B, C, D for the untrained thin-section image, helping to identify vertical trends in the frequency or distribution of classes in relation to depth, and to reveal clues about the capacity of rocks to store and produce hydrocarbons and the best way to extract these resources.

[0015] The method described in the document aims to predict the occurrence of a geological feature in a geological thin-section image using a backpropagation-enabled classification process trained by inserting extracted training image fractions that have the same absolute horizontal and vertical length and associated markers for classes from a predetermined set of geological features and iteratively computing a prediction of the probability of occurrence of each of the classes for the extracted training image fractions. On the other hand, the document is silent regarding the possibility of extracting statistical distributions of properties (area, axis size, roughness, perimeter, etc.), frequency, and spatial distribution of each type of porous element in order to further characterize the rock.

[0016] Thus, as can be observed, document US2024193427 A1 does not present an approach to pixel-level classification, although it mentions the pixel as the basic unit for image evaluation; that is, there is no reference to classification considering adjacent pixels or belonging metrics. The method described in the document uses pixels for advanced image processing, but without mentioning the use of double classification, distinguishing between pores and non-pores and evaluating the pixel's belonging to the area of ​​a given constituent. In this sense, in prior art documents there is no reference to the use of scaled images. Petition 870250081304, dated 10 / 09 / 2025, page 14 / 84 / 27 pixel, performing multiple classifications before reaching the final stage, with the aim of defining genetic classes of pores.

[0017] Document CN117454261 A discloses a method for characterizing the pore structure of petroleum reservoir material with the aim of providing a pore structure evaluation method and its application that combines feature screening algorithms and support vector machine. The pore structure evaluation method that combines feature screening and SVM algorithm, including the following steps: Identify and characterize the pore structure types of core samples and classify the pore structure types of core samples; Screen the model input features based on the Fscore algorithm and build a sample dataset through core calibration recording; Build a pore structure type profiling identification model based on the SVM algorithm and select test samples to verify the model's generalization performance;Combining linear discriminant analysis and backpropagation neural network algorithm to verify model performance; The pore structure type prediction model based on the collaborative F-score SVM algorithm is used to identify and apply to wells drilled throughout the area; Based on the identification of single-well pore structure types from well logging, the sequential Gaussian simulation method is used to establish a three-dimensional geological model of the pore structure types of the main oil-bearing strata and predict favorable zones for porosity and permeability development.

[0018] The pore structure evaluation method described in the document combines feature screening and SVM algorithm, including identifying and characterizing pore structure types and classifying pore structure types, screening model input features and Petition 870250081304, dated 10 / 09 / 2025, page 15 / 84 / 27, proposes to construct a sample dataset, build a pore structure type profiling identification model, combine linear discriminant analysis and backpropagation neural network algorithm to verify the model's performance, and establish a three-dimensional geological model of pore structure types and predict favorable zones for porosity and permeability development. On the other hand, the document is also silent regarding the possibility of extracting statistical distributions of properties (area, axis size, roughness, perimeter, etc.), frequency, and spatial distribution of each type of porous element in order to detail the rock characterization.

[0019] Thus, as can be observed, document CN117454261 A does not deal with the classification of porosity, but rather with quantitative and geometric parameters of void spaces, since the methodology presented in this document addresses porosity parameters. Many of these parameters are not acquired from petrographic images, but from calculations derived from petrographic analyses that relate porosity and permeability. Furthermore, despite mentions of the use of algorithms, the document presents little conceptual development on the petrographic and petrological attributes necessary for the classification of porosity.

[0020] As can be observed, despite technological advances and the development of new tools, the state of the art does not reveal a solution capable of performing the automated analysis and classification of porosity with precision, speed, and uniformity to meet the demands of petrographic characterization of porous reservoir rock systems, using petrographic images, and evaluating, at the pixel scale, the relationships of porous elements with their adjacencies.

[0021] The invention described below stems from ongoing research in this segment, the focus of which is to develop a Petition 870250081304, dated 10 / 09 / 2025, page 16 / 84 / 27: a solution capable of standardizing porosity analysis through petrographic images, and evaluating, on a pixel scale, the relationships of porous elements with their adjacencies, using artificial intelligence resources for complete classification and quantification, favoring studies and data analysis, as well as increasing the accuracy, reliability, and uniformity of results. OBJECTIVES

[0022] The objective of the present invention is to offer an automated method for analyzing and classifying porosity in microscopic slides, using advanced technological tools such as artificial intelligence, aiming to provide greater precision, speed, and uniformity of results. This method seeks to overcome the limitations of traditional methods, offering an effective solution for detailed characterization of porous systems in reservoir rocks, optimizing the data acquisition process and reducing the imprecision of analyses performed by specialists.

[0023] Thus, the present invention describes an automated method for classifying porosity by obtaining geometric data and analyzing photomicrographs, i.e., images of petrographic thin sections on a microscopic scale. This method uses advanced techniques, such as artificial intelligence, and allows for the precise identification of the characteristics of porous systems, ensuring uniformity and reliability in the results.

[0024] Furthermore, the present invention standardizes porosity analysis through petrographic images, at pixel scale, evaluating the relationships of porous elements with their adjacencies and enabling the massive acquisition of data, which can support decisions regarding the production of oil reservoirs and make information from these studies available in a timely manner within the oil exploration and production chain. SUMMARY OF THE INVENTION Petition 870250081304, dated 10 / 09 / 2025, page 17 / 84 / 27

[0025] The present invention relates to a method for automated classification of porosity in microscope slides, comprising the steps of: Define S100 as a porosity classification system, in which a reference system is selected and adapted to a model based on digital images, using fundamental image units, where the fundamental image units are represented by pixels and where each pixel is classified based on its individual properties and spatial relationships with adjacent pixels; To obtain an S200 image of microscopic slides, where the pixel size of the image is smaller than the smallest porosity structure to be identified; The S300 binarizes the image, performing segmentation to distinguish between pore and non-pore regions based on color properties, where pore regions, corresponding to the pore spaces of the sample, are identified by means of predefined color thresholds; Identify S400 porous elements (PorEls), delineating each of the porous elements and extracting their geometric properties; The S500 segmentation analyzes the constituents of the non-porous phase, dividing the solid areas into discrete elements such as grains, particles, and cement, with individual labels assigned to each of these constituents; To correlate S600 porous and solid elements, performing a spatial analysis of the neighborhood of porosity pixels in relation to adjacent solid phase pixels to determine the porosity type classification based on predefined spatial interactions; Apply S700 geometric parameters, including additional properties of porous elements (PorEls), where the additional properties include aspect ratio and area distribution; Petition 870250081304, dated 10 / 09 / 2025, page 18 / 84 / 27 classify S800 the porous elements (PorEls), assigning to each porous element (PorEl) a final classification according to the reference system; and extract S900 the statistical distributions to generate data, computing and storing said quantitative data related to the distribution, frequency and geometric properties of the porous elements (PorEls) for detailed analysis.

[0026] Additionally, the present invention also relates to a computer-readable storage medium which, upon receiving a set of instructions, will perform the steps of the method of the present invention. BRIEF DESCRIPTION OF THE FIGURES

[0027] The present invention will be better understood from the detailed description and figures described below that refer to it.

[0028] Figure 1 illustrates an image representation with sufficient resolution to faithfully represent the rock constituents.

[0029] Figure 2 illustrates a representation of the binarization of the image.

[0030] Figure 3a illustrates a porous element (PorEl).

[0031] Figure 3b illustrates geometric properties to be considered in the porous element (PorEl).

[0032] Figure 3c illustrates the limits of the porous element (PorEl).

[0033] Figure 4a illustrates a representative diagram of a photomicrograph with examples of classifications of solid constituents.

[0034] Figure 4b illustrates a representative diagram of a photomicrograph with examples of classifications of solid constituents.

[0035] Figure 4c illustrates a labeling on the photomicrograph.

[0036] Figure 5a illustrates a segmentation of Petition 870250081304, dated 10 / 09 / 2025, page 19 / 84 / 27 solid constituents in binarized photomicrograph.

[0037] Figure 5b illustrates a pixel-level representation of pixel classification.

[0038] Figure 6a illustrates a representation of the segmentation of 'non-Pore' and 'Pore' phase elements for identification.

[0039] Figure 6b illustrates a detailed representation of the segmentation of 'non-Pore' and 'Pore' phase elements for identification of internal porosity and dissolved edges.

[0040] Figure 7a illustrates a spatial correlation of porous elements (PorEls) with solid constituents.

[0041] Figure 7b illustrates in detail a spatial correlation of porous elements (PorEls) with solid constituents.

[0042] Figure 7c illustrates in detail a spatial correlation of porous elements (PorEls) with solid constituents.

[0043] Figure 7d illustrates in detail a spatial correlation of porous elements (PorEls) with solid constituents.

[0044] Figure 7e illustrates in detail a spatial correlation of porous elements (PorEls) with solid constituents.

[0045] Figure 8a illustrates a spatial correlation of porous elements (PorEls) with solid constituents.

[0046] Figure 8b illustrates in detail a spatial correlation of porous elements (PorEls) with solid constituents.

[0047] Figure 8c illustrates in detail a spatial correlation of porous elements (PorEls) with solid constituents.

[0048] Figure 9a illustrates a classification of porous elements (PorEls) based on belonging labels.

[0049] Figure 9b illustrates in detail a classification of porous elements (PorEls) based on the labels of Petition 870250081304, dated 10 / 09 / 2025, page 20 / 84 / 27 belonging.

[0050] Figure 9c illustrates in detail a classification of porous elements (PorEls) based on membership labels.

[0051] Figure 9d illustrates in detail a classification of porous elements (PorEls) based on belonging labels.

[0052] Figure 10a illustrates a classification of porous elements (PorEls) based on geometric properties.

[0053] Figure 10b illustrates in detail a classification of porous elements (PorEls) based on geometric properties.

[0054] Figure 10c illustrates in detail a classification of porous elements (PorEls) based on geometric properties.

[0055] Figure 10d illustrates in detail a classification of porous elements (PorEls) based on geometric properties.

[0056] Figure 11 illustrates a flowchart of the sequencing of steps required to carry out the present invention.

[0057] Figure 12a illustrates a representative image of the interparticle / interelement pore.

[0058] Figure 12b illustrates a representative image of the intraparticle / intraelement pore.

[0059] Figure 12c illustrates a representative image of the porus vugularis.

[0060] Figure 12d illustrates a representative image of a fracture pore / channel. DETAILED DESCRIPTION OF THE INVENTION

[0061] The following description constitutes only a preferred embodiment within the scope of the present invention.

[0062] The present invention aims to offer a Petition 870250081304, dated 10 / 09 / 2025, page 21 / 84 / 27 automated method for analysis and classification of porosity in microscopic slides, using advanced technological tools, such as artificial intelligence, aiming to provide greater precision, speed and uniformity of results.

[0063] The methodology addressed by the present invention provides a complete step-by-step process for classifying porosity through the analysis of photomicrographs (microscopic-scale images of petrographic thin sections), based on the main current porosity classification systems published in the literature. Porosity classification, although time-consuming, is of great value for the genetic interpretation of rocks, being essential for understanding the distribution of voids in the rock and its complete characterization. The possibility of automation, to be achieved through the development of algorithms based on the sequence of steps described herein, tends to promote an effective gain in productivity and quality of information acquired.

[0064] In technical terms, what is generically called porosity in petrographic thin sections is defined as PorEl (Porous Element), and it is an aggregate with real continuity (for the two-dimensional section represented in the studied thin section) of elements of a porous system (porous chambers and pore throats). Such structures, fundamental for the storage and transmission of fluids in a reservoir rock, are consequences of the geological history imposed on that volume of rock, and present characteristics that can aid in this interpretation (geometric properties, dimensions, relationships with adjacent mineral phases, etc.).

[0065] It is also worth mentioning that, for cases of automation of the proposed methodology, it is essential that the resolution of the images to be used must be sufficient to accurately represent the most representative PorEls modes of the analyzed texture (in other words, that Petition 870250081304, dated 10 / 09 / 2025, page 22 / 84 / 27 (the pixels that make up the image have dimensions smaller than the mode of the porosity existing in the slide / photomicrograph)

[0066] In general, classification systems are used that can condense information regarding the genesis and spatial relationship of the PorEls with their surroundings. This classification system must be translated to the reality of a digital image, where the fundamental element (pixel) is the unit of reference.

[0067] Therefore, the first step is to define S100, a porosity classification system, in which a reference system is selected and adapted to a model based on digital images, using fundamental image units, where the fundamental image units are represented by pixels and where each pixel is classified based on its individual properties and spatial relationships with adjacent pixels.

[0068] Below, in Table 1, there are examples of porosity classification (PorEls), using the pixels of this porous element as a reference. Table 1 exemplifies this step, through two columns: the pore type of the classification system used, and the pixel distribution pattern of the PorEl that meets this classification. The porosity classification system is an intergranular porosity classification system, an intragranular porosity classification system, a vugular porosity classification system, and a fracture porosity classification system. Table 1 Pore ​​Classification (based on Choquette & Pray, 1970) Representative image Pore characteristic Pore characteristic in image pixels Interparticle / Interelement Figure 12a Selective pore. Empty space between particles The outermost pixels of PorEl make contact, for the most part, with Petition 870250081304, dated 10 / 09 / 2025, page 23 / 84 / 27 Pixels belonging to the perimeter of particles” Intraparticle / Intraelement Figure 12b Selective pore. Empty space within particles (original or diagenetic product). The outermost pixels of the PorEl make contact, for the most part, with pixels belonging to the perimeter of a single particle. Vulgar Figure 12c Non-selective pore. Empty space resulting from the dissolution of more than one mineral phase, generally larger than the mode of the main constituent. The outermost pixels of the PorEl make contact, for the most part, with pixels belonging to the perimeters of various constituents. The area of ​​this PorEl is larger than the mode of the areas of the most common constituents in the image. Fracture / Channel Figure 12d Non-selective pore. Empty space, generated by fracturing (with or without associated dissolution), of one or more mineral phases. It has a very low aspect ratio (less than 0.1). The outermost pixels of the PorEl make contact with pixels belonging to the perimeters of various constituents.Geometric properties such as Aspect Ratio. Petition 870250081304, dated 10 / 09 / 2025, page 24 / 84 / 27 They show extremely elongated PorEls.

[0069] It is worth noting that this step can be customized by the user, who can use, or even create, a classification system that suits their needs. Additionally, the classification system creates or adapts classification tables that relate porosity types to pixel distribution patterns.

[0070] As can be seen in Figure 1, there is a definition of the material to be used (photomicrographs - microscopic images). It is necessary to obtain images with sufficient resolution to faithfully represent the rock constituents.

[0071] Figure 1 shows the pixel size small enough to accurately represent all PorEls modes present in the image field. Figure 1 is a representation of a photomicrograph, not to scale, of a granular rock texture, where different grains are in contact and empty spaces (porosity, in blue) occur between them. In detail (red rectangle), a view of the regular, gray-colored square mesh would represent the image pixels, where it is possible to obtain S200 an image of microscopic slides, in which the pixel size of the image is smaller than the smallest porosity structure to be identified, that is, the smaller the pixel dimensions, the more defined the image.

[0072] As can be seen in figure 2, the next step is to binarize the S300 image, performing segmentation to distinguish between pore and non-pore regions based on color properties.

[0073] This S300 step aims to separate what is porous (void / non-solid) from what is non-porous (mineral phases). Generally, petrographic thin sections of reservoir rocks are impregnated Petition 870250081304, dated 10 / 09 / 2025, page 25 / 84 / 27 by colored resin (preferably blue), which causes the rock's porosity to exhibit characteristic coloration (with varying shades, according to crystallinity, granulometry, impregnation effectiveness, etc.). As a general trend, pore regions, corresponding to the porous spaces of the sample, are identified by means of predefined color thresholds. This process is carried out by a series of programs and algorithms available online, or in free applications (e.g., defining RGB thresholds - Red, Green, Blue - to encompass the shades related to porosity). In practice, this binarization is nothing more than assigning a Pore or Non-Pore label to each pixel of the image, according to its color properties.Figure 2 presents an example of binarization, in which all mineral constituents are colored with a specific color (black), and porosity is identified as another segment (in white). In the detail of Figure 2, there is a view, in a square pixel mesh, of this binarization.

[0074] Next, the next step is to identify S400 porous elements (PorEls), delineating each of the porous elements and extracting their geometric properties, whose geometric properties of the porous elements (PorEls) include area, perimeter, shape parameters, perimeter-to-area ratio, and aspect ratio.

[0075] Once the image is binarized, there are two phases (segments) in the image: one related to solids (non-Poro) and the other related to voids (Poro), which are distributed as areas of various shapes and sizes. It is fundamental that each individual perimeter of the Poro segment be considered as an element (by definition, a Porous Element, or PorEl). For each PorEl, extract its typical geometric properties (area, perimeter, perimeter / area ratio (PoA), aspect ratio, gamma) in order to foster a quantification database (to be used in later steps). In Petition 870250081304, dated 10 / 09 / 2025, p. 26 / 84 / 27 In practice, each pixel assigned as Poro receives a specific label for each PorEl (i.e., in addition to the Poro label, the pixel would receive a designation belonging to PorEl x, where x is the number of the identified PorEl). Figures 3a, 3b, and 3c demonstrate, in an identified PorEl, some geometric properties to be considered. In terms of pixels, the detailed representation shows that those identified as Poro receive the label belonging to PorEl x, detailed in Figures 3b and 3c. The blue perimeter, marked in both Figures 3b and 3c, reveals the PorEl boundaries (poro wall), and are given by the edges of the outermost pixels.

[0076] As can be seen in figures 4a, 4b and 4c, the next step is to segment the S500 non-porous phase constituents, dividing the solid areas into discrete elements, such as grains, particles, elements, cement, etc., where individual labels are assigned to each of these constituents, said segmentation is performed by means of artificial intelligence algorithms trained to recognize mineralogical and textural patterns.

[0077] It is important to highlight that the boundaries are well established so that the classification of these constituents is accurate. Figures 4a, 4b, and 4c show a representative diagram of a photomicrograph, with examples of classifications of solid constituents, as highlighted in Figures 4a and 4b. There are several ways to perform this step, from manual assignment by a user to the application of machine learning algorithms for pattern recognition. In the latter case, extensive training of the algorithms is recommended to encompass as many variations as possible (in size, nuances of internal structures, coloration, etc.) associated with the geological context, artifacts from the preparation of the petrographic slide, and limitations of photodocumentation, and which do not promote the modification of the imposed label. It is also interesting that each identified class Petition 870250081304, dated 10 / 09 / 2025, page 27 / 84 / 27 (e.g.: grain) have its elements individualized: within the class of grains (A, B, C), each closed perimeter would represent a specific grain (e.g.: grain A1, grain A2, grain A3, etc., representing several individualized grains of the same nature).

[0078] In practice, within the non-Pore phase, all constituent types would be discretized. Each pixel assigned to non-Pore would receive another label, associated with its belonging to a specific non-Pore segment, for example: the pixels that, in the photomicrograph, make up grain A1 receive the label that relates them to grain A1, as illustrated in figure 4c.

[0079] Just like PorEls, each element of each non-Pore segment must have its geometric properties quantified. The S500 step is crucial for the success of the proposed methodology, and has two main approaches, as illustrated in the sequence of figures 5a, 5b, 6a and 6b.

[0080] In figure 5a, the segmentation of non-Pore constituents - once the photomicrograph is binarized, only the pixels identified as non-Pore participate in the segmentation into elements / structures. This approach is more easily reproduced through automation algorithms, as it links the identified element to characteristics of the solid material that composes it. It is possible to note, from figure 5a, that the perimeter of the solids, identified as diverse constituents: A2, A4, B2, D3) comprise only pixels identified as non-pore.This is evident in the case of constituents A4 and B2, where, in the case of A2, the perimeter of the constituent acquires a re-entrant profile resulting from the porosity (blue coloration phase) that appears to penetrate the constituent. In B2, the pore is isolated within the constituent, and its pixels are not considered part of it. In the detail of Figure 5b, the pixel-level representation of how pixel classification would work following this approach is shown: each solid constituent composed of its... Petition 870250081304, dated 10 / 09 / 2025, page 28 / 84 / 27 respective pixels identified as non-porous, while the different PorEls are individualized according to their areal continuity.

[0081] Segmentation into elements / structures occurs on both pixels identified as non-Porous and Porous, as shown in Figures 6a and 6b. This approach attempts to account for any elements / structures that have undergone fragmentation and / or dissolution, generating internal porosity or dissolved edges. In terms of automation, it is up to the user to establish algorithms that, in addition to identifying the solid element based on a smaller volume of data, are able to infer the original boundaries of that element, an activity that is not always easy.

[0082] If you choose to perform this activity manually, it would add considerable time to the activity (depending on the number of such cases in the photomicrograph studied). It is worth noting that, if you choose this approach, the pixels identified as belonging to the pore segment, which already have a label related to the respective PorEl to which they belong, receive a new label that expresses a relationship of belonging (or not) to a certain constituent of the solid phase, e.g.: contained in the perimeter of grain A, not contained in the perimeter of solids, detailed in Figure 6b. It is possible to note that, from the example in the figure, the perimeter of the solid constituents (A2, A4, B2 and D3) are distinct from the approach exemplified by Figures 5a and 5b, especially constituents A4 and B2, which incorporate, within the area limited by the perimeter of the constituent, a sub-area occupied by porosity - represented as some pixels in the example.In Figures 6a and 6b, two examples are shown: one related to internal porosity within a constituent connected to the porosity between constituents (A4), where the pixels identified as porosity (blue color) located in a concave area of ​​the perimeter of constituent A4 were labeled as belonging to PorEl1 (given its areal continuity with... Petition 870250081304, dated 10 / 09 / 2025, page 29 / 84 / 27 this) as well as “belonging to the area identified as “Grain A4. This interpretation of the constituent area composed of both solids and porosity (with the necessary inference of the original perimeter of the constituent) is an arduous task, especially in attempts at automation.

[0083] In this case, it is concluded that the “PorEl” is composed of subareas with different contexts (in this case, a context “between the constituents, and another “within a constituent. The second case is typified by the constituent “B2, which has an isolated pore within its perimeter. This pore is defined as a specific PorEl, which has the same characteristic for each pixel: “belonging to PorEl2 and belonging to “Grain B2.

[0084] The next step involves correlating S600 porous and solid elements, performing a spatial analysis of the neighborhood of porosity pixels in relation to adjacent solid phase pixels to determine the porosity type classification based on predefined spatial interactions, in which S601 pixels adjacent to porous elements (PorEls) are evaluated to determine their spatial allocation, identifying whether they are intergranular, intragranular, or isolated.

[0085] In this S600 stage, the two levels of information obtained from the binarization and segmentation of the binarized phases will be articulated. Each pixel of the image, labeled as “Poro” and the respective “PorEl” of which it is a part, or as some solid phase and its respective “constituent,” will be spatially correlated in two ways, which may be articulated, as illustrated in the sequencing of figures 7a, 7b, 7c, 7d, 7e and in the sequencing of figures 8a, 8b, and 8c.

[0086] Following the aforementioned figures, there is a survey, for each PorEl, of its outermost pixels (the pixels that represent contact with the “pore wall”): what are the labels of the adjacent “non-Pore” pixels? Petition 870250081304, dated 10 / 09 / 2025, page 30 / 84 / 27 analyze, for each PorEl boundary pixel (in yellow in figures 7b, 7c, 7d, 7e and 8b and 8c), its eight adjacent neighbor pixels (four edge neighbors, and four vertex neighbors - highlighted in dashed lines in the examples of the aforementioned figures).

[0087] Each of these neighbors is evaluated according to its “belonging label” (e.g., “belongs to PorEl x”, “within the perimeter of constituent A”). It is evaluated, out of the total number of neighbors, which neighbor(s) are identified by the “non-Poro” label. Depending on the result, the classification of that PorEl type is assigned, according to the rule established in step S100. Figures 7a, 7b, 7c, 7d, and 7e exemplify this process, considering five cases (c1, c2, d1, e1, e2) of pixels classified as “poro” in contact with pixels of another nature.

[0088] In dotted line, the perimeter of the immediately neighboring pixels, for each case (reference pixel is named with (A) , (B) , (C) , (D) , (E) ). Note, for the cited example, the various labels of the adjacent pixels in the five cases shown, which shows the “interconstituent” nature of this PorEl (second example of the classification table mentioned in step I).

[0089] Figures 8a, 8b and 8c provide another example, with a simpler case, where the outermost pixels of PorEl have the same classification (“belonging to B2”), which suggests a PorEl entirely contained within the perimeter of constituent B2, that is, an intraconstituent PorEl, as per the classification exemplified in Table 1.

[0090] Additionally, each PorEl will have an internal evaluation of the labels assigned to its pixels. If the labels related to "pixel ownership" are all the same, the PorEl is selective, and its classification will be according to the table loaded in step S100; if there is variation in these ownership relationships, the PorEl is non-selective, and a different classification will be assigned, also through the Petition 870250081304, dated 10 / 09 / 2025, page 31 / 84 / 27 rules established in the table loaded in step S100.

[0091] In the example illustrated in the sequence of figures 9a, 9b, 9c and 9d, two cases are presented: In figure 9c, PorEl is made up of pixels classified as pores belonging only to the perimeter of PorEl 1, and pixels classified as pores also belonging to the perimeter of constituent A4. This indicates that there are two porosity allocation groups for the same PorEl, highlighting its non-selective nature. On the other hand, figure 9d exemplifies a simple case, in which PorEl is made up of pixels with the same belonging relationships (belonging to PorEl2 and belonging to constituent B2), indicating the selective nature of this PorEl.

[0092] Next, S700 geometric parameters are applied, including additional properties of porous elements (PorEls), where additional properties include aspect ratio and area distribution, where additional geometric parameters, including circularity, elongation and fractal dimension, can be used to enhance the classification of porous elements (PorEls).

[0093] If the relationship of the PorEl pixels to their adjacencies is not sufficient for classifying that element, other information gathered can be used, such as the size (area) of the PorEl, or its aspect ratio. This is because certain types of pores are also defined by their geometric characteristics, as exemplified in the sequence of figures 10a, 10b, 10c and 10d and in Table 1. In figures 10a, 10b, 10c and 10d, there is an example of a case where the complete classification of the PorEls is possible only through the evaluation of a geometric property (in this case, Aspect Ratio - short axis / long axis). Both identified PorEls have pixels with the same belonging relationships (belonging to the respective PorEls and to constituent A1), but their geometric characteristics are essential for classification. In this case, the ratio between the axes of these PorEls was indicative. Petition 870250081304, dated 10 / 09 / 2025, page 32 / 84 / 27 sufficient for the classification of porosity, according to the classification table inserted in step S100.

[0094] Next, it is possible to classify the porous elements (PorEls) using S800, assigning each porous element (PorEl) a final classification according to the reference system.

[0095] Once the PorEls are classified, it is possible to extract S900 statistical distributions of properties (area, axis size, roughness, perimeter, etc.), frequency and spatial distribution of each type of PorEl, in order to detail the characterization of the rock, wherein the data extracted in the S900 step can be stored in a database to perform S901 automated statistical analyses and visualization of results, wherein the statistical analysis includes a spatial distribution of porosity, frequency histograms and mapping of geometric trends in the photomicrograph.

[0096] Figure 11 illustrates a flowchart representing the sequencing of steps for carrying out the present invention.

[0097] Based on the same concept as the previous embodiment of the method, an embodiment of this application further provides a computer-readable storage medium. The computer-readable storage medium stores a set of instructions. When the computer program instructions are loaded and executed on a computer, the procedures or functions according to embodiments of the present invention are wholly or partially generated.

[0098] A computer can be a general-purpose computer, a dedicated computer, a computer network, or any other programmable device. Computer instructions can be stored on a computer-readable storage medium or can be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the Petition 870250081304, dated 10 / 09 / 2025, page 33 / 84 / 27 Computer instructions can be transmitted from one site, computer, server, or data center to another site, computer, server, or data center in a wired manner (e.g., a coaxial cable, an optical fiber, or a digital subscriber line (DSL)) or wirelessly (e.g., infrared, radio, or microwave). The computer-readable storage medium can be any usable medium accessible by the computer, or a data storage device, such as a server or data center, integrating one or more usable media. The usable medium can be a magnetic medium, an optical medium, a semiconductor medium, or the like. Petition 870250081304, dated 10 / 09 / 2025, page 34 / 84

Claims

1 / 3 CLAIMS 1. A method for automated classification of porosity in microscopic slides, characterized by comprising the steps of: defining (S100) a porosity classification system, in which a reference system is selected and adapted to a model based on digital images, using fundamental image units, wherein the fundamental image units are represented by pixels and wherein each pixel is classified based on its individual properties and spatial relationships with adjacent pixels; obtaining (S200) an image of microscopic slides, wherein the pixel size of the image is smaller than the smallest porosity structure to be identified; binarizing (S300) the image, performing segmentation to distinguish between pore and non-pore regions based on color properties; wherein the pore regions, corresponding to the pore spaces of the sample, are identified by means of predefined color thresholds;Identify (S400) porous elements (PorEls), delineating each of the porous elements and extracting their geometric properties; segment (S500) the constituents of the non-porous phase, dividing the solid areas into discrete elements, such as grains, particles, and cement, where individual labels are assigned to each of these constituents; correlate (S600) the porous and solid elements, performing a spatial analysis of the neighborhood of porosity pixels in relation to the adjacent pixels of the solid phase to determine the classification of the porosity type based on predefined spatial interactions; apply (S700) geometric parameters, including additional properties of the porous elements (PorEls), where the additional properties include aspect ratio and area distribution;classify (S800) the porous elements (PorEls), assigning to each porous element (PorEl) a final classification according to the reference system; and extract (S900) the statistical distributions to generate data, computing and storing said quantitative data related to the distribution, frequency and geometric properties of the porous elements (PorEls) for detailed analysis.

2. Method, according to claim 1, characterized in that the porosity classification system is an intergranular porosity classification system, an intragranular porosity system, a vugular porosity system, a fracture porosity system, or a system customized by a user.

3. A method according to claim 1 or 2, characterized in that the classification system creates or adapts classification tables that relate porosity types to pixel distribution patterns.

4. A method, according to any one of claims 1 to 3, characterized in that the geometric properties of the porous elements (PorEls) comprise area, perimeter, shape parameters, perimeter-to-area ratio, and aspect ratio.

5. Method, according to any one of claims 1 to 4, characterized in that the segmentation step (S500) Petition 870250013755, dated 20 / 02 / 2025, page 65 / 83 3 / 3 of the non-porous phase constituents is performed by means of artificial intelligence algorithms trained to recognize mineralogical and textural patterns.

6. Method, according to any one of claims 1 to 5, characterized in that the step of correlating (S600) the porous and solid elements further comprises the step of evaluating (S601) pixels adjacent to the porous elements (PorEls) to determine their spatial allocation, identifying whether they are intergranular, intragranular or isolated.

7. Method, according to any one of claims 1 to 6, characterized in that in the step of applying (S700) geometric parameters, additional geometric parameters are used, including circularity, elongation and fractal dimension, to enhance the classification of porous elements (PorEls).

8. Method, according to any one of claims 1 to 7, characterized in that the data extracted in step (S900) are stored in a database to perform (S901) automated statistical analyses and visualization of results, wherein the statistical analysis contemplates a spatial distribution of porosity, frequency histograms and mapping of geometric trends in the photomicrograph.

9. Computer-readable storage medium, characterized in that it comprises a set of instructions, wherein when the set of instructions is executed on a computer, it causes the computer to perform the method as defined in any one of claims 1 to 8. Petition 870250013755, dated 20 / 02 / 2025, pp. 66 / 83

Images