基于电力知识驱动的输电隐患样本可控生成方法、装置、设备、介质和产品
By acquiring scene instruction templates and large-scale text-image synthesis models from transmission line inspection images, and combining dynamic weight functions and rule violation coefficients, transmission hazard samples that conform to power business rules were generated. This solved the problems of insufficient coverage of multiple types of hazards and poor scene adaptability in existing technologies, and achieved high-precision hazard sample generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA SOUTHERN POWER GRID ARTIFICIAL INTELLIGENCE TECHNOLOGY CO LTD
- Filing Date
- 2026-04-14
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies lack the ability to cover multiple types and complex hazards when generating power transmission hazard samples, and the generated samples have poor scenario adaptability, which cannot meet the needs of multimodal question answering models.
By acquiring scene instruction templates from transmission line inspection images, an initial hidden danger sample is generated using a large image-text composite model. The model parameters are then updated using a dynamic weight function and a rule violation coefficient to generate a target hidden danger sample that conforms to power business rules.
It improves the generation accuracy and scenario adaptability of power transmission hazard samples, realizes deep integration of power business rules and sample generation, and meets the needs of multimodal question answering models.
Smart Images
Figure CN122047233B_ABST