UGS-1m: Fine-grained urban green space mapping of 31 major cities in China based on the deep learning framework

Urban green space (UGS) is an important component in the urban ecosystem and has great significance to the urban ecological environment. The UGS-1m product provides the fine-grained UGS maps of 31 major cities in China, which is generated based on a deep learning (DL) framework.The DL framework consists of a generator and a discriminator. The generator is a fully convolutional network designed for UGS extraction (UGSNet), which integrates attention mechanisms to improve the discrimination to UGS, and employs a point rending strategy for edge recovery. The discriminator is a fully connected network aiming to deal with the domain shift between images. To support the model training, an urban green space dataset (UGSet) with a total number of 4,554 samples of size 512×512 is provided. Code for the UGSet and the UGSet will be available at: https://liumency.github.io/UGS-1m/. The main steps to obtain UGS-1m can be summarized as follows: a) Firstly, the UGSNet will be pre-trained on the UGSet in order to get a good starting training point for the generator; b) After pre-training on the UGSet, the discriminator is responsible to adapt the pre-trained UGSNet to different cities through adversarial training; c) Finally, the UGS results of the 31 major cities in China (UGS-1m) are obtained using 2,179 Google Earth images with a data frame of 7'30" in longitude and 5'00" in latitude, and a spatial resolution of nearly 1.1 meters. Evaluating the performance of the proposed framework on samples from five sample cities shows the validity of the UGS-1m products, with an average overall accuracy (OA) of 87.56% and an F1 score of 74.86%.Dataset Descriptions1) UGS-1m.zip: the fine-grained UGS map product of 31 major cities in China2) UGSet.zip: the large benchmark dataset to support and foster the UGS research3) GUB_Data.zip: the Global Urban Boundary data of each city4) GE_Imagery_DataFrame.zip: the grid data of the Google Earth images in “.shp” format, providing the image composition of each city5) other Zip files named by city names: the Google Earth images of each city

城市绿地（Urban green space, UGS）是城市生态系统的重要组成部分，对城市生态环境具有重要价值。本次发布的UGS-1m数据集产品提供了中国31个重点城市的精细化城市绿地地图，该数据集基于深度学习（Deep Learning, DL）框架生成。该深度学习框架由生成器与判别器两部分构成：生成器为专为城市绿地提取设计的全卷积网络（命名为UGSNet），其集成了注意力机制以提升对城市绿地的辨别能力，并采用点渲染（point rending）策略完成边缘修复；判别器则为全连接网络，旨在解决图像间的域偏移问题。为支撑模型训练，本研究同步发布了包含4554张512×512分辨率样本的城市绿地基准数据集UGSet。UGSet与本框架的相关代码将公开于：https://liumency.github.io/UGS-1m/。

UGS-1m数据集的生成流程可概括为以下三步：
a) 首先，依托UGSet对UGSNet进行预训练，为生成器获取优质的初始训练起点；
b) 完成UGSet上的预训练后，通过对抗训练，由判别器适配预训练后的UGSNet至不同城市场景；
c) 最终，利用2179张谷歌地球（Google Earth）影像完成中国31个重点城市的城市绿地提取，得到UGS-1m产品。所用影像的经纬度覆盖范围为经度7分30秒、纬度5分0秒，空间分辨率近1.1米。

在5个样本城市的测试样本上对所提框架的性能进行评估，结果验证了UGS-1m产品的有效性，其平均总体精度（Overall Accuracy, OA）达87.56%，F1分数为74.86%。

数据集详情如下：
1) UGS-1m.zip：包含中国31个重点城市的精细化城市绿地地图产品；
2) UGSet.zip：用于支撑与推动城市绿地相关研究的大型基准数据集；
3) GUB_Data.zip：各城市的全球城市边界（Global Urban Boundary, GUB）数据；
4) GE_Imagery_DataFrame.zip：格式为.shp的谷歌地球影像栅格数据，提供各城市的影像覆盖范围信息；
5) 其余以城市名称命名的压缩文件：各城市对应的谷歌地球原始影像。

搜集汇总

背景与挑战

背景概述

UGS-1m是一个基于深度学习框架生成的中国31个主要城市精细绿地地图数据集，包含绿地地图、基准数据集和全球城市边界数据，平均准确率87.56%。