井筒稳定成像表征与风险智能识别数据集

井筒稳定成像表征与风险智能识别数据集是一个面向深度学习与储层精细表征的多模态、像素级标注专业数据集。本数据集针对火成岩储层非均质性强、流体识别难度大的挑战，系统处理并集成了电成像测井（FMI）与常规测井曲线数据。核心内容为1,200张经过地质专家精确解释的“电测井成像填充图”，12731张单颗粒岩屑图，即对原始电成像图像中的裂缝、孔洞、层理等关键地质结构、返出岩屑颗粒进行了像素级语义分割标注。数据集以高分辨率图像（TIF, PNG）和结构化标签（JSON）的形式提供，完整保留了测井深度信息。本数据集为开发基于深度学习的火成岩储层参数预测、裂缝自动识别、储集空间定量评价及井筒稳定风险智能预警模型提供了无可替代的训练与测试数据，对推动复杂储层测井解释的智能化进程具有重要价值。

The Dataset for Wellbore Stability Imaging Characterization and Intelligent Risk Identification is a professional multimodal, pixel-level annotated dataset targeting deep learning and fine reservoir characterization. Aiming at the challenges of strong heterogeneity of igneous rock reservoirs and difficulty in fluid identification, this dataset systematically processes and integrates fullbore formation microimager (FMI) logging and conventional logging curve data. Its core content includes 1,200 precisely interpreted "electrical logging imaging fill maps" by geological experts, and 12,731 single-grain cuttings images, which have undergone pixel-level semantic segmentation annotations for key geological structures such as fractures, vugs, and bedding in the original electrical imaging images, as well as the discharged cuttings particles. The dataset is provided in the form of high-resolution images (TIF, PNG) and structured labels (JSON), and fully retains the logging depth information. This dataset provides irreplaceable training and test data for developing deep learning-based models for igneous rock reservoir parameter prediction, automatic fracture identification, quantitative evaluation of reservoir space, and intelligent early warning of wellbore stability risks, and holds significant value for promoting the intelligent advancement of complex reservoir logging interpretation.

搜集汇总

背景与挑战

背景概述

该数据集是一个面向深度学习与储层精细表征的多模态专业数据集，针对火成岩储层挑战，集成了电成像测井和常规测井数据，包含1,200张经过精确标注的电测井成像填充图和12,731张单颗粒岩屑图，提供了像素级语义分割标注。它以高分辨率图像和结构化标签形式存储，用于支持火成岩储层参数预测、裂缝自动识别和井筒稳定风险智能预警模型的开发，推动测井解释智能化。