MambaVSS-YOLOv11n: State Space Model-Enhanced Multi-Defect Detection in Photovoltaic Module Electroluminescence...

Area of Science:

• Renewable Energy Systems
• Materials Science and Engineering
• Artificial Intelligence and Machine Learning

Background:

• Growing global demand for sustainable energy necessitates advancements in solar photovoltaic (PV) power generation.
• Defective PV modules reduce power conversion efficiency and operational lifespan, impacting overall system reliability.
• Accurate and efficient defect detection during PV module manufacturing is crucial for ensuring product quality.

Purpose of the Study:

• To develop an advanced electroluminescence (EL) image-based multi-defect detection method for PV modules.
• To address limitations in existing models regarding small object detection and complex backgrounds in EL images.
• To provide a lightweight yet accurate solution for real-time defect inspection in PV production lines.

Main Methods:

• Proposed MambaVSS-YOLOv11n model integrating Mamba's Vision State Space (VSS) module and optimized C3k2 Bottleneck structure.
• Utilized a dataset of 692 labeled EL images covering six defect types: Broken Gate, Cold Solder Joint, Black Spot, Scratch, Microcrack, and Suction Mark.
• Employed Space-to-Depth Convolutional (SPD-Conv) Layer for efficient downsampling and Inner Mask Distance Penalized Intersection over the Union (Inner-MDPIoU) loss function for enhanced accuracy.

Main Results:

• MambaVSS-YOLOv11n achieved a 18.1% reduction in parameters compared to YOLOv11n, demonstrating model lightweighting.
• Improved detection performance with mAP@0.5 reaching 0.869 and mAP@0.5:0.95 reaching 0.637.
• The model effectively mitigated the impact of small objects and complex backgrounds, enhancing detection accuracy.

Conclusions:

• MambaVSS-YOLOv11n offers a lightweight and high-performance solution for multi-defect detection in PV module EL images.
• The proposed method significantly enhances detection accuracy and computational efficiency for PV manufacturing quality control.
• This technology supports PV manufacturers in achieving real-time, reliable defect inspection, contributing to the energy transition.