• Acta Optica Sinica
  • Vol. 38, Issue 12, 1215008 (2018)
Xing Liu*, Jian Chen, Dongfang Yang*, and Hao He
Author Affiliations
  • Missile Engineering College, Rocket Force University of Engineering, Xi'an, Shaanxi 710025, China
  • show less
    DOI: 10.3788/AOS201838.1215008 Cite this Article Set citation alerts
    Xing Liu, Jian Chen, Dongfang Yang, Hao He. Scene-Coupled Intelligent Multi-Task Detection Algorithm for Air-to-Ground Remote Sensing Image[J]. Acta Optica Sinica, 2018, 38(12): 1215008 Copy Citation Text show less
    Air-to-ground variable resolution scene. (a) High-altitude vision; (b) middle-altitude vision; (c) low-altitude vision
    Fig. 1. Air-to-ground variable resolution scene. (a) High-altitude vision; (b) middle-altitude vision; (c) low-altitude vision
    Models. (a) SSD model; (b) FSSD model
    Fig. 2. Models. (a) SSD model; (b) FSSD model
    Scene-coupled multi-task object detection model
    Fig. 3. Scene-coupled multi-task object detection model
    Information activation module. (a) Synchronous activation; (b) asynchronous activation
    Fig. 4. Information activation module. (a) Synchronous activation; (b) asynchronous activation
    Scene-assisted multi-task datasets. (a) Scene- object datasets; (b) remote sensing scene datasets
    Fig. 5. Scene-assisted multi-task datasets. (a) Scene- object datasets; (b) remote sensing scene datasets
    Visualization results of scene coupling multi-task model (VGG16) validation set
    Fig. 6. Visualization results of scene coupling multi-task model (VGG16) validation set
    Sequential scene change resolution target search. (a) Far view rasterization scene perception schematic; (b) high-altitude scene-aware guided object detection
    Fig. 7. Sequential scene change resolution target search. (a) Far view rasterization scene perception schematic; (b) high-altitude scene-aware guided object detection
    Base modelChannel addition mAP /%Channel concatenation mAP /%
    SSD82.1386.63
    FSSD86.7890.45
    Table 1. Two feature map channel fusion methods (synchronous activation, VGG16)
    Base modelObject detection mAP /%Scene classification mAP /%
    SSD-IA86.6398.21
    SSD-none83.1298.31
    FSSD-IA90.4598.69
    FSSD-none88.4498.56
    Table 2. Comparison of IA module on different framework models
    Base modelSynchronous activation mAP /%Asynchronous activation mAP /%
    SSD86.6384.31
    FSSD90.4588.36
    Table 3. Effect of synchronous and asynchronous activations on accuracy of object detection
    Feature extractionObject detection taskScene classification taskFrame rate
    AP /%precision /%
    CarTruckAirplaneBoatTownAirportWaters
    VGG1691.9777.5198.4293.9198.3298.7599.0130
    ResNet5093.1284.2399.1794.6799.3199.1299.5214
    MobileNetsv284.7679.3488.4586.5698.2297.4398.3246
    Darknetv283.1377.2185.3182.1497.5198.2198.7740
    Table 4. Scene-coupled multi-task model detection results based on different feature extractions
    AlgorithmAPmAP
    CarTruckAirplaneBoat
    SSD-VGG1684.2467.2298.3189.7784.89
    FSSD-VGG1688.8769.4597.6292.2887.05
    Proposed-VGG1691.9777.5198.4293.9190.45
    Table 5. Comparison of proposed algorithm with traditional object detection models %
    Feature extractionPrecision /%
    TownAirportWaters
    VGG1698.4499.7599.11
    ResNet5098.5198.9899.43
    MobileNetsv297.3297.9398.92
    Darknetv297.7397.6698.57
    Table 6. Classification results in remote sensing scenes under different feature extraction networks
    Xing Liu, Jian Chen, Dongfang Yang, Hao He. Scene-Coupled Intelligent Multi-Task Detection Algorithm for Air-to-Ground Remote Sensing Image[J]. Acta Optica Sinica, 2018, 38(12): 1215008
    Download Citation