Rock Thin Section Mineral And Grain Boundary Detection With One DNN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying minerals and grain boundaries in rock samples are inefficient and subjective, often requiring manual interpretation and failing to utilize the full information from polarized light microscopy images, with separate algorithms typically used for each task.
Innovation Solution
A single deep neural network (DNN) is trained using a combined dataset of optical and electron microscopy images, leveraging features from XPL images and FFT transforms to simultaneously detect grain boundaries and mineral types in rock samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection by geologists is used to identify minerals and grain boundaries, then expertise and interpretative judgment are applied, but subjectivity and inconsistency arise due to human bias
Solution Approach 1:
The patent replaces manual geological inspection with an automated deep neural network system that processes polarized light microscopy images. The DNN algorithm automatically identifies minerals and grain boundaries without human intervention, eliminating subjectivity and ensuring consistent, reproducible results across different samples and analysts.
Solution Approach 2:
The system creates a digital copy of the physical thin section through polarized light microscopy imaging, then processes this digital representation through neural network algorithms. This allows repeated analysis of the same sample without physical manipulation, ensuring consistent results while maintaining the integrity of the original sample.
2Measurement precision
If separate algorithms are used for mineral identification and grain boundary detection, then each task can be optimized independently, but overall system complexity increases and processing time increases
Solution Approach 1:
The patent merges mineral identification and grain boundary detection into a single integrated deep neural network model. This unified approach simultaneously performs both functions using one algorithm, reducing overall system complexity and processing time while maintaining high accuracy for both tasks through shared computational resources and coordinated processing.
Solution Approach 2:
The deep neural network is designed as a universal system that performs multiple functions (mineral identification, grain boundary detection, and mineral composition analysis) within a single framework. This multi-functional approach eliminates the need for separate specialized algorithms while maintaining optimization for each specific task through the model's architecture and training processes.
3Ease of manufacture
If only simple features from XPL images are used for analysis, then processing is simplified, but valuable information recorded in the image arrays is discarded
Solution Approach 1:
The patent replaces simple feature extraction methods with a deep neural network that automatically learns and extracts complex patterns from polarized light microscopy images. The DNN processes the full complexity of XPL image data, capturing latent information related to mineral composition, crystal orientation, and grain boundary characteristics without requiring manual feature selection or simplification.
Solution Approach 2:
The system transforms the input XPL images through multiple processing stages, converting pixel intensity data into meaningful geological parameters such as mineral identification, grain boundary detection, and composition analysis. This parameter transformation process extracts valuable information that would be lost in simpler analysis methods while maintaining the computational efficiency of automated processing.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides a consistent and efficient method for identifying grain boundaries and minerals, reducing human bias and improving accuracy by utilizing the full information from XPL images, eliminating the need for separate models and enhancing the speed and reliability of mineral analysis.
Implementation Method 1
PPL and XPL images can be acquired as hyperstacks (arrays) of images, each image from the stack potentially containing important information regarding the underlying mineralogy. Also note that different images from the same stack may reveal different features of the same mineral.
Data Source
AI summary
A method for generating a training dataset for determining grain boundaries and minerals in a thin section of a rock sample, includes receiving the thin section of the rock sample, generating optical images of the thin section with an optical tool, generating mineral phase images of the thin section with an electron microscopy tool, computing first and second pseudo-images based on different features extracted from the optical images, generating the training dataset based on (1) the optical images, (2) the mineral phase images, and (3) the pseudo-images, and training a single deep neural network, DNN, based on the training dataset to simultaneously determine a mineral type and grain boundaries in the thin section of the rock sample.