This data set contains vector data of building roofs in 90 cities in China (selected comprehensively based on city administrative levels and regional distribution, See Appendix 1 for city list). It is mainly produced based on deep learning semantic segmentation models and multi-source remote sensing images. First, the original images are preprocessed, and stratified sampling and visual interpretation are performed according to the city level and its regional distribution to produce training and test data. Then the training data is input into the deep learning semantic segmentation model for training, making it suitable for the building roof extraction task. Based on the test data, the performance of the building roof extraction model is evaluated using general indicators for result evaluation in the deep learning field. Finally, this model was applied to the task of extracting building roofs in 90 cities in China, and the building roofs were automatically extracted and vectorized. This data set can provide important data support for related research based on building roofs (such as rooftop solar potential assessment, urban planning, etc.) at the urban and national scales.

 CnOpenData obtained permission from the Smart City Sensing and Simulation Laboratory of Nanjing Normal University to include the data in the public data area for scholars to browse and read. Sample data can be downloaded below. For complete data download, please go to National Tibetan Plateau Science Data Center (http ://data.tpdc.ac.cn/zh-hans/news/c1b0ee71-2ceb-4276-8f5c-b53d57caa88d).

Data description

• Dataset year: This data set automatically identifies and extracts urban building roofs based on the deep learning semantic segmentation model and multi-source remote sensing images. The original image was obtained in 2020. The raster pixel value resolution of the building roofs extracted by the deep learning semantic segmentation model The rate is 1m, and the actual year of the recognized building roof vector elements depends on the shooting time of the remote sensing image.

• City level description: The 90 cities are divided into 4 levels according to their administrative levels: deputy national level and full provincial level, deputy provincial level, quasi-deputy provincial level and main department level. For details, please see the data download below.

Data details

1. Scale: None
2. Projection: Albers
3. File size: 143360.0MB
4. Data format: Esri Shapefile
5. Spatial range: North: 53.55; South: 3.86; West: 73.66; East: 135.05

Field description

Sample data

Beijing (the source file is in Esri Shapefile format, the table is just a preview of the data attribute table)

Quotation

• Data reference:
  Smart City Perception and Simulation Laboratory of Nanjing Normal University. Rooftop vector dataset of 90 urban buildings in China (2020). National Tibetan Plateau Scientific Data Center, DOI: 10.11888/Geogra.tpdc.271702, CSTR: 18406.11.Geogra.tpdc. 271702, 2021. [NANJING NORMAL UNIVERSITY Lab of smart city sensing and simulation. Vectorized rooftop area data for 90 cities in China (2020). National Tibetan Plateau Data Center, DOI:10.11888/Geogra.tpdc.271702, CSTR:18406.11.Geogra .tpdc.271702, 2021]

• Citation of the article:
  Zhang, Z., Qian, Z., Zhong, T., et al. (2022). Vectorized rooftop area data for 90 cities in China. Scientific Data, 9(1): 1-12.

Data update frequency

Updated from time to time

文件下载