An Insulator Defect Detection Method Based on the Integration of RF-DETR and LLM

This paper presents a novel method for insulator defect detection through the integration of RF-DETR and large language models (LLMs). RF-DETR, a real-time object detection model based on the Transformer architecture, demonstrates high effectiveness in extracting visual features of insulators and potential defects. The combination with LLMs significantly enhances the accuracy and reliability of defect detection by leveraging their extensive knowledge and text generation capabilities to analyze and optimize RF-DETR's detection results. Experiments on the insulator detection dataset show that the proposed method achieves high precision and recall rates, outperforming traditional manual inspection methods. This study provides a promising direction for the application of deep learning and LLMs in power system inspection, contributing to the development of intelligent power grids.