首页--天文学、地球科学论文--测绘学论文--摄影测量学与测绘遥感论文--测绘遥感技术论文

Building Footprint Generation Using Deep Learning Methods

Abstract第5页
1 Introduction第7-11页
    1.1 State of the Art第7-8页
    1.2 Motivations and Objectives第8-9页
    1.3 Outline of the Thesis第9-10页
    1.4 Summary第10-11页
2 Methodology第11-43页
    2.1 Urban area Delineation第11-18页
    2.2 Preprocessing第18-22页
        2.2.1 Normalization第19页
        2.2.2 Coregistration第19-20页
        2.2.3 Refinement第20页
        2.2.4 Output Representation第20-22页
    2.3 Convolution Neural Network第22-33页
        2.3.1 Fully Convolution Network第22-23页
        2.3.2 U-Net第23页
        2.3.3 Seg Net第23页
        2.3.4 Res Net-DUC第23-24页
        2.3.5 ENet第24-26页
        2.3.6 Generative Adversarial Network第26-33页
        2.3.7 FC-Dense Net第33页
    2.4 Graph Model第33-40页
        2.4.1 Conditional Random Field第34-38页
        2.4.2 Graph Convolution Network第38-40页
    2.5 Proposed Approach第40-42页
        2.5.1 Network Architecture第40-42页
    2.6 Summary第42-43页
3 Experiments第43-61页
    3.1 Study Sites and Datasets第43-44页
    3.2 Accuracy Metrics第44页
    3.3 Results of Different Convolution Neural Network第44-50页
    3.4 Results of Convolution Neural Network Combined with Different Graph Models第50-60页
        3.4.1 Comparison between Condition Random Field as Recurrent Neural Network (CR-Fas RNN) and Convolution Condition Random Field (Conv CRF)第50-51页
        3.4.2 Results of Convolution Neural Network (CNN) Combined with Convolution Con-ditional Random Field (Conv CRF)第51-57页
        3.4.3 Results of Convolution Neural Network Combined with Graph Convolution Network第57-60页
    3.5 Summary第60-61页
4 Concluding Remarks第61-63页
    4.1 Disscussion and Conclusion第61-62页
    4.2 Outlook第62-63页
Acknowledgements第63-64页
Bibliography第64-68页

论文共68页,点击 下载论文
上一篇:福州建设海峡西岸旅游中心城市研究
下一篇:我国高新技术企业竞争力提升问题研究--以厦门市为例