A Butterfly Valve Opening Indicator Analysis Method Based on Image Synthesis and Keypoint Detection

To address the engineering challenge of analyzing butterfly valve opening indicators in the intelligent inspection of pumped storage power stations, this study proposes a fusion analysis method based on image synthesis and Keypoint Detection. First, a layer mask is used to separate the valve opening indicator region, which is then rotated to a specified angle and combined with the background layer to generate new images, thereby enhancing data diversity. Subsequently, a keypoint dataset is constructed using both the synthesized and original data. Then, the YOLOv11-pose model is employed for keypoint detection, enabling precise identification of critical components such as pointers and barcodes. Finally, based on the detected keypoints, barcode segmentation and indicator information analysis are performed. Experimental results demonstrate that this method significantly improves analysis accuracy and efficiency, providing a new approach to enhancing the intelligence level of power station inspections and showing promising application prospects.