Rock Particle Color And Texture Extraction With Partial Convolution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing feature extraction techniques struggle with extracting texture features from rock cutting instance images due to arbitrary shapes and sizes, background pixels, and imaging errors, leading to inefficiencies and inaccuracies in analyzing rock particles.

Innovation Solution

Implementing global average pooling (GAP) and partial convolution in an autoencoder (AE) to analyze rock particle images, which ignores invalid pixels and makes feature extraction invariant to rotation, translation, and scale, while using a variational autoencoder (VAE) to control latent space distribution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional feature extraction techniques are used on rock cutting instance images, then the extraction process can be performed, but the accuracy is reduced due to background pixels and arbitrary shapes affecting the features

Engineering Contradiction:

Improvefeature extraction accuracyVSAvoidbackground pixels interference

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent extracts and removes background pixels from rock cutting instance images before feature extraction. By separating the background from the rock particle content, the harmful interference is eliminated, allowing accurate extraction of color and texture features from only the relevant rock particle regions.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the rock cutting instance images into valid rock particle regions and background regions. This segmentation allows the feature extraction process to focus only on the rock particle content while excluding background pixels, thereby improving measurement precision.

Inventive Principle:

Principle #1Segmentation

2Adaptability or versatility

If feature extraction is performed on rock particles with arbitrary shapes and sizes, then all rock particles can be analyzed, but the features become inconsistent due to shape similarity being affected by background

Engineering Contradiction:

Improveability to handle arbitrary shapesVSAvoidfeature consistency

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent extracts only the rock particle content from arbitrary-shaped instances by removing background pixels. This ensures that feature extraction operates on consistent rock particle regions regardless of the original instance shape or size, maintaining feature consistency across diverse inputs.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies feature extraction specifically to the local rock particle regions after background removal, rather than processing entire arbitrary-shaped instances. This localized approach ensures that features are extracted from consistent rock particle content, achieving both adaptability to arbitrary shapes and stability of feature composition.

Inventive Principle:

Principle #3Local quality

3Ease of manufacture

If pixel-based reconstruction loss is used in autoencoder for texture feature extraction, then the model can be trained, but the reconstructed images become blurry which degrades texture feature quality

Engineering Contradiction:

Improvemodel training simplicityVSAvoidtexture feature accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent changes the loss function parameter from pixel-based reconstruction loss to perceptual loss based on pre-trained network features. This parameter change maintains the simplicity of autoencoder training while significantly improving texture feature accuracy by comparing high-level features rather than raw pixel values.

Inventive Principle:

Principle #35Parameter changes

4Measurement precision

If manual analysis of rock cuttings is performed, then detailed inspection can be conducted, but the process is time-consuming and labor-intensive

Engineering Contradiction:

Improveanalysis accuracyVSAvoidturnaround time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical analysis with an automated computer-based system that uses autoencoders for feature extraction and machine learning models for classification. This substitution maintains high analysis accuracy while dramatically reducing the time required, eliminating labor-intensive manual inspection.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent implements self-service automated analysis where the system independently performs feature extraction, processing, and classification of rock particle images without human intervention. This automated self-service approach achieves both high accuracy and fast turnaround time by eliminating manual analysis steps.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20250285406A1Method to extract color and texture information from rock particle instance images

Publication Date: 2025.09.11 SCHLUMBERGER TECH CORP

AI summary

Systems and methods are provided to extract features (e.g., colors, textual features) from arbitrary shaped and sized images by implementing global average pooling (GAP) and partial convolution in an autoencoder (AE) for analysis of the images. A global average pooling (GAP) layer may be used at the last layer of the encoder of the AE to make the feature rotation and translation invariant and scale equivariant. In addition, partial convolution may be used in the encoder to logically ignore the invalid pixels (e.g., background, image error, other object) or any pixel (e.g., in any area) in the images.