• Acta Optica Sinica
  • Vol. 38, Issue 4, 0428001 (2018)
Xinfang Xie, Xin Xu*, Hao Dong, Han Wu, and Luoru Li
Author Affiliations
  • School of Electronic Information, Wuhan University, Wuhan, Hubei 430072, China
  • show less
    DOI: 10.3788/AOS201838.0428001 Cite this Article Set citation alerts
    Xinfang Xie, Xin Xu, Hao Dong, Han Wu, Luoru Li. A Semi-Supervised Dimension Reduction Method for Polarimetric SAR Image Classification[J]. Acta Optica Sinica, 2018, 38(4): 0428001 Copy Citation Text show less
    Spatial scatter distribution of polarimetric features of Flevoland. (a) σHH0, σHV0, σVV0; (b) H, α, A; (c) ρHH-VV, ρHH-HV, ρVV-HV
    Fig. 1. Spatial scatter distribution of polarimetric features of Flevoland. (a) σHH0, σHV0, σVV0; (b) H, α, A; (c) ρHH-VV, ρHH-HV, ρVV-HV
    Flowchart of polarimetric SAR image classification based on SLDA
    Fig. 2. Flowchart of polarimetric SAR image classification based on SLDA
    (a) Pseudocolor image; (b) labeled truth in classification of RADARSAT-2 Flevoland
    Fig. 3. (a) Pseudocolor image; (b) labeled truth in classification of RADARSAT-2 Flevoland
    (a) Pseudocolor image; (b) labeled truth in classification of AIRSAR Flevoland 1
    Fig. 4. (a) Pseudocolor image; (b) labeled truth in classification of AIRSAR Flevoland 1
    (a) Pseudocolor image; (b) labeled truth in classification of AIRSAR Flevoland 2
    Fig. 5. (a) Pseudocolor image; (b) labeled truth in classification of AIRSAR Flevoland 2
    Relationship between training data number and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Fig. 6. Relationship between training data number and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Relationship between dimension number and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Fig. 7. Relationship between dimension number and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Relationship between labeled sample proportion and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Fig. 8. Relationship between labeled sample proportion and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Relationship between balance factor and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Fig. 9. Relationship between balance factor and classification accuracy. (a) RADARSAT-2 Flevoland; (b) AIRSAR Flevoland 1; (c) AIRSAR Flevoland 2
    Relationship between feature number and classification accuracy
    Fig. 10. Relationship between feature number and classification accuracy
    Classification maps of RADARSAT-2 Flevoland. (a) SLDA+KNN; (b) KNN; (c) PCA+KNN; (d) LPP+KNN; (e) LDA+KNN; (f) local detail maps
    Fig. 11. Classification maps of RADARSAT-2 Flevoland. (a) SLDA+KNN; (b) KNN; (c) PCA+KNN; (d) LPP+KNN; (e) LDA+KNN; (f) local detail maps
    Classification maps of AIRSAR Flevoland 1. (a) SLDA+KNN; (b) KNN; (c) PCA+KNN; (d) LPP+KNN; (e) LDA+KNN
    Fig. 12. Classification maps of AIRSAR Flevoland 1. (a) SLDA+KNN; (b) KNN; (c) PCA+KNN; (d) LPP+KNN; (e) LDA+KNN
    3D scatter distribution diagram of AIRSAR Flevoland 2. (a) SLDA; (b) PCA; (c) LPP; (d) LDA
    Fig. 13. 3D scatter distribution diagram of AIRSAR Flevoland 2. (a) SLDA; (b) PCA; (c) LPP; (d) LDA
    Classification maps of AIRSAR Flevoland 2. (a) SLDA+KNN; (b) KNN; (c) PCA+KNN; (d) LPP+KNN; (e) LDA+KNN
    Fig. 14. Classification maps of AIRSAR Flevoland 2. (a) SLDA+KNN; (b) KNN; (c) PCA+KNN; (d) LPP+KNN; (e) LDA+KNN
    CategorySLDAPCALPPKNNLDA
    Water93.0793.7793.5793.8194.18
    Farmland91.9488.1586.5088.9491.13
    Forest88.3981.5981.6883.0181.05
    Building81.0268.6571.3367.4269.10
    OA88.8883.2483.2383.6784.25
    Table 1. Classification accuracy of RADARSAT-2 Flevoland%
    CategorySLDAPCALPPKNNLDA
    Steam bean96.8894.999.7094.3098.37
    Forest98.5995.0347.0195.0895.66
    Potato89.7784.9023.0385.2388.04
    Lucerne92.7490.707.3490.8292.99
    Wheat95.0288.4892.3188.9793.54
    Bare soil89.2692.075.9592.0893.61
    Beet88.0881.511.1282.1380.51
    Rapeseed92.7992.0714.3792.2885.13
    Pea66.9172.321.8673.1376.83
    Grass85.4673.711.4273.7175.36
    Water98.5593.8748.2095.5299.46
    OA91.0187.4434.7087.7489.11
    Table 2. Classification accuracy of AIRSAR Flevoland 1%
    CategorySLDAPCALPPKNNLDA
    Wheat97.9397.2695.5097.3597.44
    Rapeseed98.8198.4452.5898.5199.52
    Beet87.7483.3819.3583.6873.61
    Onion53.6450.790.9351.2933.75
    Corn47.3844.124.7844.0643.76
    Lucerne97.7496.575.5496.6091.34
    Barley92.5789.4736.2489.4473.62
    Flax97.3296.411.7896.5192.26
    Fruit96.4392.2010.4592.1793.03
    Grass70.1665.2115.2265.3053.29
    Pea68.6873.10073.0153.74
    Steam bean41.0439.9214.5039.5620.75
    Potato97.9497.9236.3598.0397.59
    OA92.5390.8147.8490.8985.59
    Table 3. Classification accuracy of AIRSAR Flevoland 2%