中国城市足迹级建筑物高度数据集(2018年)

该数据为中国城市足迹级建筑高度数据集合，利用DSM、DEM和建筑物轮廓形态特征，以随机森林回归的机器学习算法为核心算法，实现了中国十个大城市超过百万栋建筑的高度估算。相较于传统栅格格式的建筑高度数据，本数据集提供了以建筑底面为单元的矢量建筑高度数据，展现了更加细节的城市三维形态，可以更好地作为基底数据应用于城市环境模拟、城市分析等相关研究。本数据生产流程包含了实地测量、跨区域泛化检验等质量控制步骤，具有较好的精度和鲁棒性。目前发布数据包含北京、长沙、成都、杭州、南京、上海、沈阳、天津、武汉、西安十个大城市的核心建成区数据，每个城市为一个独立的数据压缩包，数据格式为ESRI Shapefile。后续会逐步推进其他城市的数据建设。

This is a dataset of footprint-level building height data for Chinese cities. Leveraging Digital Surface Model (DSM), Digital Elevation Model (DEM), and building contour morphological features, with Random Forest Regression as the core machine learning algorithm, it has achieved height estimation for over one million buildings across ten major Chinese cities. Compared with traditional raster-format building height data, this dataset provides vector building height data in units of building footprints, presenting more detailed urban 3D morphology, and can better serve as baseline data for studies such as urban environment simulation and urban analysis. The data production process includes quality control steps such as field surveys and cross-region generalization validation, endowing the dataset with good accuracy and robustness. Currently, the released data covers core built-up area data of the ten major cities: Beijing, Changsha, Chengdu, Hangzhou, Nanjing, Shanghai, Shenyang, Tianjin, Wuhan, and Xi'an. Each city is provided as an independent compressed data package, with the data format being ESRI Shapefile. Subsequent data construction for other cities will be gradually promoted.

搜集汇总

背景与挑战

背景概述

该数据集是2018年中国十个大城市（包括北京、上海等）的足迹级建筑物高度数据，通过随机森林回归机器学习算法结合多源遥感数据生成，以矢量格式提供超过百万栋建筑的详细高度信息。其特点在于以建筑底面为单元，能更精确地展现城市三维形态，适用于城市环境模拟和分析研究，且经过质量控制，平均绝对误差为4.68米，具有较好的精度和鲁棒性。