A photovoltaic image segmentation method based on an improved SegFormer
By improving the SegFormer architecture and introducing a multi-scale fusion spatiotemporal attention mechanism and a boundary-assisted supervision branch, the problems of insufficient segmentation accuracy and unclear edge recognition in photovoltaic image segmentation are solved, realizing high-precision photovoltaic module detection and boundary recognition, which is suitable for intelligent detection and automated operation and maintenance of photovoltaic power plants.
CN122435256APending Publication Date: 2026-07-21HUANENG CLEAN ENERGY RES INST +2
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUANENG CLEAN ENERGY RES INST
- Filing Date
- 2026-03-27
- Publication Date
- 2026-07-21
Smart Images
Figure CN122435256A_ABST
Abstract
The application provides a photovoltaic image segmentation method based on an improved SegFormer, and the application comprises the following steps: acquiring a flight or satellite remote sensing image of a photovoltaic module, and constructing a data set containing an image sample and a labeled mask; performing data enhancement on the image sample, and generating a boundary supervision label based on the labeled mask; constructing a semantic segmentation model introducing a multi-scale feature fusion attention mechanism, a boundary auxiliary supervision branch and a context semantic enhancement module; using a joint loss function containing a region segmentation and boundary prediction error to optimize and train the model, and using the trained model to segment the photovoltaic image and output the component boundary information. The application improves the photovoltaic module boundary recognition and global semantic understanding ability, the segmented contour is clearer and more continuous, and high-precision region segmentation and boundary discrimination of the photovoltaic image are realized.
Need to check novelty before this filing date? Find Prior Art