• Opto-Electronic Engineering
  • Vol. 49, Issue 4, 210307 (2022)
Zhuzhang Jin, Xuyuan Fang, Yanhui Huang, Caoqian Yin, and Wei Jin*
Author Affiliations
  • Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, Zhejiang 315211, China
  • show less
    DOI: 10.12086/oee.2022.210307 Cite this Article
    Zhuzhang Jin, Xuyuan Fang, Yanhui Huang, Caoqian Yin, Wei Jin. Satellite cloud image retrieval based on deep metric learning[J]. Opto-Electronic Engineering, 2022, 49(4): 210307 Copy Citation Text show less
    Visible band 1 cloud images of different weather (a) Cloudy; (b) Rainy; (c) Fair; (d) Typhoon
    Fig. 1. Visible band 1 cloud images of different weather (a) Cloudy; (b) Rainy; (c) Fair; (d) Typhoon
    Visible band 1 cloud images of different weather systems. (a) Snow; (b) Frontal surface; (c) Westerly jet; (d) Tropical cyclone
    Fig. 2. Visible band 1 cloud images of different weather systems. (a) Snow; (b) Frontal surface; (c) Westerly jet; (d) Tropical cyclone
    Overall algorithm flow chart
    Fig. 3. Overall algorithm flow chart
    Residual 3D-2D convolution neural network
    Fig. 4. Residual 3D-2D convolution neural network
    After training, the distance of the anchor-positive decreases and the distance of the anchor-negative increases
    Fig. 5. After training, the distance of the anchor-positive decreases and the distance of the anchor-negative increases
    The effects of the hash code length on model performance
    Fig. 6. The effects of the hash code length on model performance
    Retrieval results of cloudy weather image
    Fig. 7. Retrieval results of cloudy weather image
    Retrieval results of westerly jet cloud image
    Fig. 8. Retrieval results of westerly jet cloud image
    LossP@5/%P@20/%mAP/%
    TL83.6172.5364.21
    LTL85.9576.0271.64
    C-LTL90.9678.1475.14
    Table 1. The model retrieval performance of different loss functions in coastal cloud image dataset
    LossP@5/%P@20/%mAP/%
    TL80.5779.4771.23
    LTL87.7082.5573.07
    C-LTL85.2085.6380.14
    Table 2. The model retrieval performance of different loss functions in the North hemisphere cloud image dataset
    DatasetMethodsmAP/%P@5/%P@10/%P@20/%P@30/%
    沿海云图数据集KSH60.1073.6271.7367.7666.30
    DLBHS72.4476.1975.2874.3572.15
    MiLan68.0973.3872.5471.3770.92
    DSH68.6185.7681.1776.8973.24
    Proposed75.1490.9682.4178.1476.68
    北半球云图数据集KSH60.2768.5368.4367.4166.92
    DLBHS78.1384.6583.7083.0082.71
    MiLan74.9781.3380.6480.1679.85
    DSH70.2269.8672.8073.4073.90
    Proposed80.1485.2085.8485.6384.90
    Table 3. Comparison of retrieval performance between different retrieval methods
    Zhuzhang Jin, Xuyuan Fang, Yanhui Huang, Caoqian Yin, Wei Jin. Satellite cloud image retrieval based on deep metric learning[J]. Opto-Electronic Engineering, 2022, 49(4): 210307
    Download Citation