• Optoelectronic Technology
  • Vol. 43, Issue 2, 142 (2023)
Zhiyang XIAO1, Jianpu LIN1,2, Yongai ZHANG1,2, and Zhixian LIN1,2
Author Affiliations
  • 1School of Advanced Manufacturing, Fuzhou University, Quanzhou Fujian 362200, CHN
  • 2Fujian Science & Technology Innovation Laboratory for Optoelectronic Information of China, Fuzhou 350116, CHN
  • show less
    DOI: 10.19453/j.cnki.1005-488x.2023.02.006 Cite this Article
    Zhiyang XIAO, Jianpu LIN, Yongai ZHANG, Zhixian LIN. Detection of Beach Small Object Based on Multi‑layer Feature Map Information Fusion[J]. Optoelectronic Technology, 2023, 43(2): 142 Copy Citation Text show less
    The structure of FMIF-YOLO network
    Fig. 1. The structure of FMIF-YOLO network
    Structural schematic of GCSAM module
    Fig. 2. Structural schematic of GCSAM module
    Structural comparison between PANet and BIFPN
    Fig. 3. Structural comparison between PANet and BIFPN
    Schematic of CIOU_Loss
    Fig. 4. Schematic of CIOU_Loss
    Some labeling processes and beach monitoring scenes
    Fig. 5. Some labeling processes and beach monitoring scenes
    Experimental PR curves of different methods
    Fig. 6. Experimental PR curves of different methods
    Detection results of different networks
    Fig. 7. Detection results of different networks
    网络结构查准率/(%)查全率/(%)平均精度均值/(%)每秒传输帧数
    YOLOv589.4976.7083.2041.49
    YOLOv5+Alpha‑CIOULoss86.8578.0983.8342.37
    YOLOv5+BIFPN88.3876.7682.7040.00
    YOLOv5+GCSAM90.8978.9086.0628.57
    YOLOv5+BIFPN+Alpha‑CIOULoss87.5778.1484.3042.55
    YOLOv5+GCSAM +Alpha‑CIOULoss90.5281.2887.2228.49
    YOLOv5+GCSAM+BIFPN91.0278.9886.1228.24
    YOLOv5+GCSAM+BIFPN+Alpha‑CIOULoss91.4982.0387.5628.09
    Table 1. Experimental results of network model with different improvement strategies and their combinations
    算法模型

    查准率/

    (%)

    查全率/

    (%)

    平均精度

    均值/(%)

    每秒传输帧数
    SSD62.5668.3365.2332.01
    原始YOLOv589.4976.7083.2041.49
    YOLOv5+SElayer87.2871.9180.0232.78
    YOLOv5+CBAM89.1875.5982.3434.84
    文章方法91.4982.0387.5628.09
    Table 2. The results of different methods
    算法模型平均精度均值/(%)
    SSD15.24
    原始YOLOv531.17
    YOLOv5+SElayer30.70
    YOLOv5+CBAM33.12
    文章方法34.34
    Table 3. The results of different methods on the Visdrone dataset
    Zhiyang XIAO, Jianpu LIN, Yongai ZHANG, Zhixian LIN. Detection of Beach Small Object Based on Multi‑layer Feature Map Information Fusion[J]. Optoelectronic Technology, 2023, 43(2): 142
    Download Citation