• Acta Optica Sinica
  • Vol. 38, Issue 11, 1128001 (2018)
Xiaonan Zhang1、2、*, Xing Zhong1、3、*, Ruifei Zhu1、3, Fang Gao3, Zuoxing Zhang1、2, Songze Bao1、2, and Zhuqiang Li3
Author Affiliations
  • 1 Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences, Changchun, Jilin 130033, China
  • 2 University of Chinese Academy of Sciences, Beijing 100049, China
  • 3 Key Laboratory of Satellite Remote Sensing Application Technology of Jilin Province, Chang Guang Satellite Technology Co., Ltd, Changchun, Jilin 130102, China
  • show less
    DOI: 10.3788/AOS201838.1128001 Cite this Article Set citation alerts
    Xiaonan Zhang, Xing Zhong, Ruifei Zhu, Fang Gao, Zuoxing Zhang, Songze Bao, Zhuqiang Li. Scene Classification of Remote Sensing Images Based on Integrated Convolutional Neural Networks[J]. Acta Optica Sinica, 2018, 38(11): 1128001 Copy Citation Text show less
    Architecture of integrated neural network
    Fig. 1. Architecture of integrated neural network
    Flow chart of integrated network construction
    Fig. 2. Flow chart of integrated network construction
    Scene images of NWPU-RESISC45 dataset
    Fig. 3. Scene images of NWPU-RESISC45 dataset
    Accuracy, loss value and learning rate versus number of cycles in training process of ResNet-50.(a) Accuracy; (b) loss value; (c) learning rate
    Fig. 4. Accuracy, loss value and learning rate versus number of cycles in training process of ResNet-50.(a) Accuracy; (b) loss value; (c) learning rate
    Classification results based on CNN
    Fig. 5. Classification results based on CNN
    Accuracy and loss value versus number of cycles in training process of BP network. (a) Accuracy; (b) loss value
    Fig. 6. Accuracy and loss value versus number of cycles in training process of BP network. (a) Accuracy; (b) loss value
    Confusion matrix obtained after classification prediction of dataset by integrated model
    Fig. 7. Confusion matrix obtained after classification prediction of dataset by integrated model
    Single accuracy comparison with those of other algorithms
    Fig. 8. Single accuracy comparison with those of other algorithms
    Classification accuracies and prediction time of various models
    Fig. 9. Classification accuracies and prediction time of various models
    Impact of number of scene categories classified by AlexNet on performance of integrated model
    Fig. 10. Impact of number of scene categories classified by AlexNet on performance of integrated model
    ModelInput size /(pixel×pixel)Batch size /frameNumber of cyclesTraining accuracy /%
    Experiment IExperiment II
    AlexNet224×22425630081.2285.46
    ResNet-50224×22425630086.5290.52
    ResNet-152224×22412860085.1190.11
    DenseNet-169224×22412860082.4487.44
    VGG-16[2]---87.1590.36
    Proposed model---88.4792.53
    Table 1. Training parameters and results of each network
    MethodColor-histogramBoVWVGG-16ResNet-50ProposedCompetition
    Accuracy /%27.5244.9790.3690.5992.5393.41
    Standard deviation0.21840.20510.06730.06570.05930.0451
    Prediction time /s--0.620.470.412.26
    Table 2. Performance comparison among several algorithms
    MethodAccuracy /%(experiment I)Accuracy /%(experiment II)
    GIST[2]15.9017.88
    LBP[2]19.2021.74
    Color histograms[2]24.8427.52
    BoVW+SPM[2]27.8332.96
    LLC[2]38.8140.03
    BoVW[2]41.7244.97
    GoogLeNet[2]82.5786.02
    VGG-16[2]87.1590.36
    AlexNet[2]81.2285.16
    Two-streamDFF[13]80.2283.16
    ResNet-5087.6990.59
    Proposed model89.3492.53
    Table 3. Average accuracy comparison with those of other algorithms
    Xiaonan Zhang, Xing Zhong, Ruifei Zhu, Fang Gao, Zuoxing Zhang, Songze Bao, Zhuqiang Li. Scene Classification of Remote Sensing Images Based on Integrated Convolutional Neural Networks[J]. Acta Optica Sinica, 2018, 38(11): 1128001
    Download Citation