Architecture is an important product of human activities. There are many buildings in a place, which most likely means that the place has a large population and strong economic vitality. Therefore, building data is a highly valuable data that can be used in many directions such as population research and economic research.
At present, the extraction of large-scale building data, such as global or nationwide, is mainly carried out through remote sensing image recognition technology. For remote sensing images, you can simply understand that remote sensing images are images of the ground viewed from the sky. Since it is a bird's eye view, the building you see is the roof of the building, so the building data of the large area that we can currently access is the building roof data! For example, we have previously shared the city-wide building block data of 90 cities across the country in 2020 (you can check the previous article for details)! But this data does not cover the whole country, and the data is only for 2020! So how to obtain long-term building roof data covering the entire country?
This time we share with you the nationwide 2.5m resolution building roof data from 2016 to 2021! This data set was generated by the team of Professor Tang Hong of Beijing Normal University by using 2016-2021 Sentinel-2 images. This is also China's first full-coverage and multi-year building remote sensing recognition results data set. The data format is raster format (.tif), and the country is divided into 215 spatial grids. The pixel values of the raster are 0 and 255, where 0 represents the non-building area and 255 represents the building roof area. The data coordinates are GCS_WGS_1984.
The following is a detailed introduction of the data:
01Data preview
Let’s take the Shanghai building roof area in 2021 as an example to preview:
Let’s take a look at the details:
02Data details
Data introduction:
In their research, the Beijing Normal University team proposed a deep learning method called "Space-time Perception Super-Resolution Segmentation Framework (STSR-Seg)" to extract images from relatively low-resolution images in a large spatial range. Reliable high-resolution building roof areas. The team generated a long-term China Building Roof Area (CBRA) dataset with a resolution of 2.5m by using Sentinel-2 images from 2016-2021. This is the first comprehensive and multi-year building remote sensing recognition result data set in China.
Official website:
https://zenodo.org/record/7500612
Data naming:
CBRA is divided into 215 spatial grid tiles, named "CBRA_year_E**_N**.tif", where "year" is the sampling year, "E**" and "N**" refer to the upper left corner of the tile data longitude and latitude coordinates. Some data for 2021 are as follows:
Data year:
2016-2021
Raster values:
The cell values of the raster are 0 and 255, where 0 is the non-building area and 255 represents the building roof area.
Data coordinate system :
GCS_WGS_1984
Data Format:
tif
Spatial resolution:
2.5 m
Reference papers:
https://essd.copernicus.org/articles/15/3547/2023/essd-15-3547-2023.html
Citation format:
Liu, Zeping, Tang, Hong, Feng, Lin, & Lyu, Siqing. (2023). CBRA: The first multi-annual (2016-2021) and high-resolution (2.5 m) building rooftop area dataset in China derived with Super-resolution Segmentation from Sentinel-2 imagery (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7500612
If you have data usage requirements, please quote according to the requirements of the official platform. For more data details, please check the official website!
At the bottom of the article is our public account business card. We will regularly introduce various types of urban data and data visualization and analysis technologies. Regarding the nationwide 2.5m resolution building roof data from 2016 to 2021, everyone is welcome to pay more attention to us to learn more!