为了应对光伏组件缺陷检测中人工检测速度缓慢以及使用YOLO等深度学习模型时速度较慢且硬件成本高的问题,提出一种基于SSDLite的多层特征融合轻量化目标检测方法。该方法采用MobileNetV2作为SSDLite模型的骨干网络,并从中提取3个不同层次的特征层进行特征融合。针对不同缺陷的尺寸特点,对模型中的先验框的大小也进行了重新设计。在MobileNetV2的瓶颈结构中引入CBAM注意力机制,以提高模型的检测精度。相比传统的SSDLite模型,该文模型平均精度从65.8%提高至72.4%,虽然速度略微下降,但已基本满足实际应用的需求。
Abstract
To address the issues of slow manual inspection and the high hardware costs and slower speeds associated with using deep learning models like YOLO in photovoltaic module defect detection, a lightweight object detection method based on SSDLite with multi-level feature fusion is proposed. This method employs MobileNetV2 as the backbone network of the SSDLite model and extracts three different feature layers for feature fusion. The sizes of the anchor boxes in the model are redesigned based on the size characteristics of different defects. Additionally, the CBAM attention mechanism is introduced into the bottleneck structure of MobileNetV2 to enhance the detection accuracy of the model. Compared to the traditional SSDLite model, the proposed model improves the mean average precision （mAP） from 65.8% to 72.4%. Although the speed slightly decreases, it still largely meets the requirements of practical applications.
关键词
光伏组件 /
目标检测 /
深度学习 /
SSDLite /
多层特征融合 /
MobileNetV2
Key words
photovoltaic modules /
object detection /
deep learning /
SSDLite /
multi-layer feature fusion /
MobileNetV2
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once: unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA, 2016: 779-788.
[2] REDMON J, FARHADI A.YOLO9000: better, faster, stronger[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA, 2017: 6517-6525.
[3] BOCHKOVSKIY A, WANG C Y, LIAO H M. YOLOv4: optimal speed and accuracy of object detection[EB/OL].2020: 2004. 10934. https://arxiv.org/abs/2004.10934v1.
[4] LIU W, ANGUELOV D, ERHAN D, et al.SSD: single shot MultiBox detector[M]//Lecture Notes in Computer Science. Cham: Springer International Publishing, 2016: 21-37.
[5] FU C Y, LIU W, RANGA A, et al. DSSD: deconvolutional single shot detector[EB/OL].2017: 1701.06659. https://arxiv.org/abs/1701.06659v1.
[6] JEONG J, PARK H, KWAK N. Enhancement of SSD by concatenating feature maps for object detection[EB/OL].2017: 1705.09587. https://arxiv.org/abs/1705.09587v1.
[7] GIRSHICK R.Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision (ICCV). Santiago, Chile, 2015: 1440-1448.
[8] REN S Q, HE K M, GIRSHICK R, et al.Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(6): 1137-1149.
[9] 王道累, 李明山, 姚勇, 等. 改进SSD的光伏组件热斑缺陷检测方法[J]. 太阳能学报, 2023, 44(4): 420-425.
WANG D L, LI M S, YAO Y, et al.Method of hotspot detection of photovoltaic panels modules on improved ssd[J]. Acta energiae solaris sinica, 2023, 44(4): 420-425.
[10] 孙建波, 王丽杰, 麻吉辉, 等. 基于改进YOLOv5s算法的光伏组件故障检测[J]. 红外技术, 2023, 45(2): 202-208.
SUN J B, WANG L J, MA J H, et al.Photovoltaic module fault detection based on improved YOLOv5s algorithm[J]. Infrared technology, 2023, 45(2): 202-208.
[11] 郭岚,刘正新.基于改进YOLOv5的光伏组件缺陷检测[J]. 激光与光电子学进展, 2023, 60(20): 148-156.
GUO L, LIU Z X.Detection of defects in photovoltaic modules by improved YOLOv5[J]. Laser & Optoelectronics Progress, 2023, 60(20): 148-156.
[12] SANDLER M, HOWARD A, ZHU M L, et al.MobileNetV2: inverted residuals and linear bottlenecks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA, 2018: 4510-4520.
[13] CHOLLET F.Xception: deep learning with depthwise separable convolutions[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA, 2017: 1800-1807.
[14] HOWARD A G, ZHU M L, CHEN B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[EB/OL].2017: 1704.04861. https://arxiv.org/abs/1704.04861v1
[15] WOO S, PARK J, LEE J Y, et al.CBAM: convolutional block attention module[M]. Lecture Notes in Computer Science. Cham: Springer International Publishing, 2018: 3-19.
[16] HU J, SHEN L, SUN G.Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA, 2018: 7132-7141.
PDF(1287 KB)
