Image Processing and Machine Learning Approaches for Petrographic Thin Section Analysis (Russian)

Interactive Machine Learning Improves Accuracy of Coal Porosity Segmentation in Focused Ion Beam–Scanning Electron Microscopy Images

“…The confusion matrix was generated for each segmentation method using MATLAB to evaluate the applicability of the segmentation method. , In addition, 15 slices that were not part of the training set were used to test the overall comprehensive segmentation quality. We calculated accuracy ( A ) according to the following equation A = normalT normalP + normalT normalN normalT normalP + normalT normalN + normalF normalP + normalF normalN × 100 % where TP, TN, FP, and FN are evaluation metrics for classification, representing true positive (TP), true negative (TN), false positive (FP), and false negative (FN), respectively.…”

“…For instance, if the actual porosity of a sample is 5%, but the algorithm identifies all the pixels of the image as pores, the FN value would be 0, and R would be 100%. While also suffering from some weaknesses, 46 the F 1 score is designed to balance these effects and provide a more comprehensive measure of segmentation quality, with a higher value of F 1 score representing a higher quality of segmentation.…”

“…The confusion matrix was generated for each segmentation method using MATLAB to evaluate the applicability of the segmentation method. 46,47 In addition, 15 slices that were not part of the training set were used to test the overall comprehensive segmentation quality. We calculated accuracy (A) according to the following equation…”

Automatic identification of fossils and abiotic grains during carbonate microfacies analysis using deep convolutional neural networks

“…With the rise of artificial intelligence and computer vision in the last decade, petrographic analysis studies have diversified into various areas, such as the sorting level of particles (Shu et al, 2018), rock classification (Cheng and Guo, 2017;Shu et al, 2017), and microfacies classification (Pires de Lima et al, 2020). Budennyy et al (2017) employed machine learning to evaluate the properties of structural objects in thin sections, such as grain, cement, voids, and cleavage. Their models achieved up to 80% accuracy and proved a way to conduct an automatic quantitative and qualitative analysis of thin sections by applying image processing and statistical learning methods.…”

Petrographic Features Prediction of Carbonate Reservoir Based on Thin Sections Using Deep-Learning

“…Moreover, the resulting interpretations are often biased and/or inconsistent since they are highly subjective to the experience and expertise of the individual geoscientist, imaging technique, and scale of visualization (de Lima et al, 2020;Lokier and Al Junaibi, 2016). These challenges, compounded by advancements in computer-aided analysis methods have motivated a large number of applied research studies that aimed at automation of petrographic thin section analysis using machine learning (Budennyy et al, 2017;Jobe et al, 2018;Koeshidayatullah et al, 2020a;Marmo et al, 2005). In this paper, an effective method for petrographic analysis of micro-textures classification and microfossils recognition using deep-learning is proposed.…”