A method and system for analyzing visible light defects in a photovoltaic module

By embedding the GSCA module into the transfer learning network and employing a parallel design of channel and spatial attention, the problem of false detection caused by grid line texture interference in the visible light images of photovoltaic modules was solved, and high-precision defect identification was achieved.

CN121837285BActive Publication Date: 2026-07-03NANCHANG HANGKONG UNIVERSITY
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NANCHANG HANGKONG UNIVERSITY
Filing Date
2026-03-16
Publication Date
2026-07-03

Smart Images

  • Figure CN121837285B_ABST
    Figure CN121837285B_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for visible light defect analysis of photovoltaic modules. The method includes: acquiring at least one visible light image of a photovoltaic module and preprocessing the at least one visible light image to obtain an image dataset; training a pre-constructed improved transfer learning network based on the image dataset to obtain a defect recognition model, wherein a GSCA module is embedded in the residual network of the improved transfer learning network; inputting the acquired real-time visible light image into the defect recognition model, and the defect recognition model outputting a defect recognition result corresponding to the real-time visible light image. This method accurately suppresses the strong response generated by regular grid lines on the photovoltaic panel surface, while significantly enhancing the expression of subtle features of real defects such as bird droppings and dust, greatly reducing the probability of misjudgment caused by texture interference.
Need to check novelty before this filing date? Find Prior Art