围岩分类方法、训练方法及设备
By employing a semi-supervised learning method and utilizing a collaborative training framework of random forest and support vector machine models, confidence pseudo-label samples are generated and multiple rounds of iterative training are performed. This solves the problem of label scarcity in surrounding rock classification, achieves high-precision and stable surrounding rock classification, reduces the cost of manual annotation, and improves the model's generalization ability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NAT ENG LAB FOR HIGH SPEED RAILWAY CONSTR
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing rock classification methods struggle to fully utilize unlabeled data in scenarios where labels are scarce, leading to decreased classification accuracy, class imbalance, and insufficient generalization ability. Furthermore, existing semi-supervised methods suffer from iterative accumulation of pseudo-label errors and unstable recognition performance under class imbalance.
A semi-supervised learning method is adopted, which uses a collaborative training framework of random forest model and support vector machine model. Multiple rounds of iterative training are carried out using labeled and unlabeled samples to generate confidence pseudo-label samples and gradually add them to the training set. Combined with data preprocessing and cross-validation techniques, an efficient surrounding rock classification model is constructed.
It significantly reduces the cost of manual annotation, improves the accuracy of surrounding rock classification, enhances the model's generalization ability and real-time recognition ability, suppresses the propagation of false label errors, and improves the model's classification performance under label-scarce conditions.
Smart Images
Figure CN122413093A_ABST