Abstract:The small hot spot defect detection for photovoltaic (PV) farms is a challenging problem due to the feature vanishing as the network deepens. To solve this challenging problem, a novel residual channelwise attention gate network (RCAG-Net) is proposed by employing a novel RCAG module to achieve multiscale feature fusion, complex background suppression, and defect feature highlighting. In RCAG-Net, the novel RCAG module first realizes feature fusion by adding the features of different scale layers. Next, global average pooling (GAP) and multilayer perceptron (MLP) are used to dimension reduction and refinement of the fused features, then yielding an attention map for channelwise feature reweighting by gate mechanism, which employs selective transmission of the convolution neural network (CNN)-extracted features to achieve informative feature filtering. Moreover, residual connection from the fused features to the final output facilitates the insertion of the new RCAG into some classical pretrained models, without breaking its initial behavior. Finally, the proposed approach is validated through a real defect detection system, and the experimental result clearly verifies its effectiveness for small hot spot detection of PV farms.

What problem does this paper attempt to address?

This paper attempts to address the issue of small hotspot defect detection in photovoltaic power stations. Specifically, the problem of feature disappearance due to network deepening makes small hotspot defect detection very challenging. To solve this problem, the authors propose a novel Residual Channel Attention Gate Network (RCAG-Net), which achieves multi-scale feature fusion, complex background suppression, and defect feature enhancement by introducing a novel RCAG module. ### Main Issues 1. **Small Hotspot Defect Detection**: Small hotspot defects in photovoltaic power stations usually have a small area (≤32² pixels) and occupy a very small proportion of pixels in the image, making them difficult to detect accurately. 2. **Complex Background Interference**: The complex background of photovoltaic power stations can easily interfere with defect detection. ### Solutions 1. **RCAG Module**: - **Multi-Scale Feature Fusion**: By using upsampling, convolution, addition, and activation operations, features of different scales are fused. - **Global Average Pooling (GAP)**: Used to extract global features and perform dimensionality reduction. - **Multi-Layer Perceptron (MLP)**: Refines global features through fully connected layers and ReLU functions, generating channel attention maps. - **Gate Mechanism**: Filters information features by selectively transmitting features extracted by the convolutional neural network. - **Residual Connection**: Residual connections from fused features to the final output allow the new RCAG module to be inserted into classic pre-trained models without disrupting their initial behavior. 2. **RCAG-Net Architecture**: - **Input Image**: The original infrared image is resized and input into the network. - **Feature Extraction**: DarkNet53 is used as the backbone network to extract useful features from the infrared image. - **Feature Fusion**: Two RCAG modules are used to fuse multi-scale features, enhancing the feature representation capability for small hotspot defects. - **Prediction**: Multi-scale prediction layers output defect categories and locations. ### Experimental Validation The paper validates the effectiveness of RCAG-Net through an actual defect detection system. Experimental results show that this method performs excellently in detecting small hotspot defects in photovoltaic power stations, with high detection accuracy and fast speed. ### Conclusion This paper proposes a novel RCAG-Net, which effectively addresses the challenges of small hotspot defect detection in photovoltaic power stations by introducing the RCAG module, improving detection accuracy and robustness.

RCAG-Net: Residual Channelwise Attention Gate Network for Hot Spot Defect Detection of Photovoltaic Farms

RGR-Net: Refined Graph Reasoning Network for multi-height hotspot defect detection in photovoltaic farms

AEGLR-Net: Attention Enhanced Global-Local Refined Network for Accurate Detection of Car Body Surface Defects

AEGLR-Net

BAF-Detector: An Efficient CNN-Based Detector for Photovoltaic Cell Defect Detection

Surface Defect Detection of Solar Cells Based on Multiscale Region Proposal Fusion Network

Photovoltaic Panel Defect Detection Based on Ghost Convolution with BottleneckCSP and Tiny Target Prediction Head Incorporating YOLOv5

Deep Learning-based Method for PV Panels Segmentation and Defects Detection with Infrared Images

DCMF-AFNet: An anchor-free photovoltaic hot-spot fault detection network based on deformable context transformer and bi-branch multi-level feature fusion

Defect recognition of photovoltaic panels based on adaptive dimensional feature aggregation convolutional neural network

PGA-Net: Pyramid Feature Fusion and Global Context Attention Network for Automated Surface Defect Detection

A lightweight network for photovoltaic cell defect detection in electroluminescence images based on neural architecture search and knowledge distillation

Anomaly Detection Algorithm for Photovoltaic Cells Based on Lightweight Multi-Channel Spatial Attention Mechanism

GBH-YOLOv5: Ghost Convolution with BottleneckCSP and Tiny Target Prediction Head Incorporating YOLOv5 for PV Panel Defect Detection

Attention classification-and-segmentation network for micro-crack anomaly detection of photovoltaic module cells

PV-faultNet: Optimized CNN Architecture to detect defects resulting efficient PV production

A PV cell defect detector combined with transformer and attention mechanism

Detection of Defects in Power Grid Inspection Images Based on Multi-scale Fusion

Deep-Learning-Based Automatic Detection of Photovoltaic Cell Defects in Electroluminescence Images

CNN-based defect detection and classification of PV cells by infrared thermography method