A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Automatic Building Extraction Based on High Resolution Aerial Images
Building extraction is a hotspot of research. Buildings in remote sensing images are numerous, and buildings have different distribution patterns in different regions, which means it is difficult to extract buildings by one traditional method for all areas. Deep learning plays more and more important part of computer version and image processing, which has made some progress in target detection and identification, especially in medical graphic recognition. In this paper, we compared three common datasets, and selected the most suitable one. Mask R- CNN is applied to this dataset. We provided a better method and this method can be used in different areas when dataset is rich enough for training. The method in this paper just need high resolution aerial images to extract buildings automatically without any other steps, because identification and extraction are achieved in one network. The research showed: the dataset we selected is suitable; Mask R-CNN can be used in high resolution remotely images; Mask R-CNN is good at building extraction, and greatly reduce the loss of classification and leakage. Therefore, this paper provided a suitable method for automatic building extraction.
Automatic Building Extraction Based on High Resolution Aerial Images
Building extraction is a hotspot of research. Buildings in remote sensing images are numerous, and buildings have different distribution patterns in different regions, which means it is difficult to extract buildings by one traditional method for all areas. Deep learning plays more and more important part of computer version and image processing, which has made some progress in target detection and identification, especially in medical graphic recognition. In this paper, we compared three common datasets, and selected the most suitable one. Mask R- CNN is applied to this dataset. We provided a better method and this method can be used in different areas when dataset is rich enough for training. The method in this paper just need high resolution aerial images to extract buildings automatically without any other steps, because identification and extraction are achieved in one network. The research showed: the dataset we selected is suitable; Mask R-CNN can be used in high resolution remotely images; Mask R-CNN is good at building extraction, and greatly reduce the loss of classification and leakage. Therefore, this paper provided a suitable method for automatic building extraction.
Automatic Building Extraction Based on High Resolution Aerial Images
Hu, Yiwen (author) / Guo, Fenglin (author)
2019-10-01
2161577 byte
Conference paper
Electronic Resource
English
Automatic Georeferencing of Aerial Images Using Stereo High-Resolution Satellite Images
| Online Contents | 2011
Automatic Extraction of Buildings from Aerial Images
| British Library Conference Proceedings | 1997
Automatic 3D building extraction from aerial and space images for earthquake risk management
| Taylor & Francis Verlag | 2011
| British Library Conference Proceedings | 2008