Open-Vocabulary Object Detection for Reliable Transformer Stand Installation Inspection

Ensuring the safety and proper placement of various components on transformer stands is critical in the installation inspection process within power grids. Traditional object detection methods often face challenges when dealing with the evolving requirements of installation processes and the potential introduction of unknown objects. This paper proposes a multimodal approach based on open-vocabulary object detection to enhance the reliability of transformer stand installation inspection. By integrating visual and textual information, the model effectively identifies known components and makes informed decisions when encountering unseen elements. Experimental results demonstrate the model’s robustness in complex scenarios, including background clutter and occlusions, as well as its ability to adapt to new equipment or signage. This approach not only improves detection accuracy and stability but also offers a flexible and adaptive solution for ensuring the safety and integrity of transformer stand installations. The findings of this study provide valuable insights into the application of open-vocabulary object detection in industrial settings, highlighting its potential for advancing intelligent inspection processes within power systems.