A dual-task auxiliary method and system for rock painting dating and a storage medium
By constructing a dual-task model based on DINOv2, integrating HSV color histograms and HOG manual descriptors, and performing feature recalibration, the reliability problem under various degradation conditions in rock art dating was solved. This achieved stability and metadata integrity in rock art dating and regional interpretation, and improved the efficiency and accuracy of rock art surveys.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies are insufficient to reliably improve the reliability of dating and regional interpretation in rock art under various degradation conditions. Furthermore, existing strategies are inadequate when long-tail imbalance, degradation noise, and subtle temporal differences exist, and cannot meet the cataloging requirements for metadata integrity.
A dual-task model is constructed using DINOv2 transfer learning, which integrates HSV color histograms and HOG handmade descriptors. Feature recalibration is performed through the SE module to construct a time/region dual-task model. Weighted cross-entropy optimization is then used to achieve multi-task classification.
It improves the stability and reliability of rock art dating, provides verifiable structured clues, supports rock art survey cataloging and expert review, and enhances discovery efficiency and record matching accuracy.
Smart Images
Figure CN122416129A_ABST