Petrographic Cue-Conditioned Dual-Polarization Learning for Taxonomy-Consistent Sandstone Grain Recognition

Accurate sandstone-grain classification supports petrographic analysis and may provide grain-level information for subsequent reservoir-characterization workflows. Fine-grained recognition remains complicated by class imbalance, subtle optical differences, and the hierarchical relationships among grain categories. This study proposes the Petrographic Cue-Conditioned Taxonomy-Consistent Recognition Network (PC-TCRNet) for paired plane-polarized light and cross-polarized light micrographs. A Petrographic Cue Bank computes mask, edge, Gabor, color, gradient, and polarization-difference responses, which are used as image-derived proxies for petrographically relevant visual patterns. A Cue-Conditioned Polarization Fusion Encoder uses these cues to enhance dual-branch representations, and a Taxonomy-Consistent Prediction Head jointly models three parent categories and nine fine-grained classes. On the Sinopec Sandstone Grain Microscopic Dataset, PC-TCRNet achieved an accuracy of 0.807, a macro-F1 score of 0.730, and an area under the precision–recall curve of 0.792; a three-seed evaluation showed limited performance variation. Component, fusion, cue-branch, and hierarchy ablations indicated complementary contributions from prior-guided fusion and taxonomy consistency. An independent Yarlung Zangbo River cohort provided a preliminary cross-region evaluation. Implementation profiling further quantified the additional computational cost of the dual-branch design. These results support the use of petrographic priors and hierarchical supervision while motivating broader multi-region validation.

Authors

Institutions

Publication Details

Journal
Minerals
Published
2026-11-13
DOI
https://doi.org/10.3390/min16111188
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Petrographic Cue-Conditioned Dual-Polarization Learning for Taxonomy-Consistent Sandstone Grain Recognition

Yuan Liu, Xiaolu Yu, Jin Xu, Chen Mu et al.
2 links
Minerals
Advanced Neural Network Applications
article

Petrographic Cue-Conditioned Dual-Polarization Learning for Taxonomy-Consistent Sandstone Grain Recognition

Yuan Liu, Xiaolu Yu, Jin Xu, Chen Mu, Qianqian Du, Hong Jiang
article en

Abstract

Accurate sandstone-grain classification supports petrographic analysis and may provide grain-level information for subsequent reservoir-characterization workflows. Fine-grained recognition remains complicated by class imbalance, subtle optical differences, and the hierarchical relationships among grain categories. This study proposes the Petrographic Cue-Conditioned Taxonomy-Consistent Recognition Network (PC-TCRNet) for paired plane-polarized light and cross-polarized light micrographs. A Petrographic Cue Bank computes mask, edge, Gabor, color, gradient, and polarization-difference responses, which are used as image-derived proxies for petrographically relevant visual patterns. A Cue-Conditioned Polarization Fusion Encoder uses these cues to enhance dual-branch representations, and a Taxonomy-Consistent Prediction Head jointly models three parent categories and nine fine-grained classes. On the Sinopec Sandstone Grain Microscopic Dataset, PC-TCRNet achieved an accuracy of 0.807, a macro-F1 score of 0.730, and an area under the precision–recall curve of 0.792; a three-seed evaluation showed limited performance variation. Component, fusion, cue-branch, and hierarchy ablations indicated complementary contributions from prior-guided fusion and taxonomy consistency. An independent Yarlung Zangbo River cohort provided a preliminary cross-region evaluation. Implementation profiling further quantified the additional computational cost of the dual-branch design. These results support the use of petrographic priors and hierarchical supervision while motivating broader multi-region validation.

MineralsVol. 16(11)
Sinopec (China) (CN), Jiangnan University (CN)
Advanced Neural Network Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon