In response to the severe lighting conditions and complex background in the production site of power plant, an improved YOLOv5s abnormal state detection method for slag extractor was proposed to ensure the safe and efficient operation of slag extractor in complex power plant environments. On the basis of YOLOv5s network, the ShuffleNet was introduced to replace the original backbone network and to achieve network lightweight by reducing network parameters. At the same time, an improved convolutional attention module was added to the ShuffleNet, and more attention was paid to the target features of the slag extractor scraper by concatenating space and channel attention mechanisms. The weighted bidirectional feature pyramid Bi-FPN (Bilateral Feature Pyramid Network) and bounding box regression loss SIoU (Scaled IoU) function were introduced to obtain more effective feature maps for feature information to improve target detection accuracy. The research results show that the improved model significantly reduces the number of parameters, reduces the model volume by 15.2%, improves the average accuracy of mAP (mean Average Precision) by 2.2%, and reduces detection time by 58.0%. While ensuring detection accuracy, real-time and accurate detection of abnormal states of the slag extractor is achieved.
为有效评估改进模型YOLOv5s-SCB的性能,改进网络模型的IoU阈值默认为0.5,选取精确率P(precision)、召回率R(recall)、平均精度均值mAP(mean average precision)和综合评价指标F1作为模型的识别精度评价指标,选取模型体积和检测时间作为模型性能的评价指标[17]。选取目标检测模型YOLOv3-tiny[18]、YOLOv4-tiny[19]、YOLOv5s、YOLOv7x与YOLOv5s-SCB进行比较,其中改进模型与其他模型mAP训练曲线变化如图8所示。
MaN N, ZhangX Y, ZhengH T,et al.ShuffleNet V2:practical guidelines for efficient CNN architecture design[C]//European Conference on Computer Vision.Cham:Springer,2018:122-138.
[6]
ZhengZ H, WangP, LiuW,et al.Distance-IoU loss:faster and better learning for bounding box regression[J].Proceedings of the AAAI Conference on Artificial Intelligence,2020,34(7):12993-13000.
HuY C, TianS F, GeJ A.Hybrid convolutional network combining multiscale 3D depthwise separable convolution and CBAM residual dilated convolution for hyperspectral image classification[J].Remote Sensing,2023,15(19):4796.
[12]
HouQ B, ZhangL, ChengM M,et al.Strip pooling:rethinking spatial pooling for scene parsing[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Seattle:IEEE,2020:4002-4011.
[13]
RedmonJ, DivvalaS, GirshickR,et al.You only look once:unified,real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas:IEEE,2016:779-788.
[14]
LinT Y, DollárP, GirshickR,et al.Feature pyramid networks for object detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Honolulu:IEEE,2017:936-944.
[15]
ZhangY F, RenW Q, ZhangZ,et al.Focal and efficient IOU loss for accurate bounding box regression[J].Neurocomputing,2022,506(11):146-157.
[16]
GuZ C, ZhuK, YouS T.YOLO-SSFS:a method combining SPD-conv/STDL/IM-FPN/SIoU for outdoor small target vehicle detection[J].Electronics,2023,12(18):3744.
[17]
PanS J, YangQ.A survey on transfer learning[J].IEEE Transactions on Knowledge and Data Engineering,2010,22(10):1345-1359.
[18]
TanL, LvX Y, LianX F,et al .YOLOv4_Drone:UAV image target detection based on an improved YOLOv4 algorithm[J].Computers & Electrical Engineering,2021,93(3):107261.
[19]
WangC Y, BochkovskiyA, LiaoH Y M.YOLOv7:Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[EB/OL].(2022-07-06)[2023-12-04].