Title: Solar panel defect segmentation method based on improved SegFormer

Authors: Lili Jin; Lisheng Wei

Addresses: School of Electrical Engineering, Anhui Polytechnic University, Wuhu, 241000, China ' School of Electrical Engineering, Anhui Polytechnic University, Wuhu, 241000, China

Abstract: In order to solve the problem of insufficient feature extraction ability and low segmentation accuracy in solar panel defect image segmentation, an improved Segformer segmentation model Segformer-EFDF is proposed. First of all, the output embeddings at each stage of the encoder are composed of efficient channel attention (ECA) and depthwise separable convolution (DSC). Secondly, a lightweight feature fusion module (FFM) is designed to gradually fuse multi-scale feature information and improve the ability of the network to locate and distinguish different defects. Finally, a loss function combining dice loss and focal loss is used to optimise the imbalance of samples and alleviate the impact of the large proportion of background pixels on model learning. The research shows that the proposed method improves the segmentation performance of solar panel defect images and has a certain application prospect.

Keywords: solar panel; defect segmentation; SegFormer.

DOI: 10.1504/IJISTA.2026.153220

International Journal of Intelligent Systems Technologies and Applications, 2026 Vol.24 No.2, pp.201 - 214

Received: 13 Mar 2025
Accepted: 19 Dec 2025
Published online: 29 Apr 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article