中国90座城市建筑物屋顶矢量数据集（2020）

该数据集包含中国90座城市（根据城市行政等级及区域分布综合选取，城市名录详见附件1）建筑物屋顶矢量数据。主要基于深度学习语义分割模型和多源遥感影像进行制作。首先，对原始影像进行预处理，并根据城市等级及其区域分布情况进行分层采样以及目视解译，制作训练和测试数据。然后将训练数据输入深度学习语义分割模型进行训练，使其适用于建筑物屋顶提取任务，并基于测试数据，采用深度学习领域结果评价一般性指标对建筑物屋顶提取模型性能进行评价。最后，将此模型应用于中国90座城市建筑屋顶提取任务中，自动提取建筑物屋顶并进行矢量化。该数据集可以为城市乃至全国尺度以建筑物屋顶为基础的相关研究（如屋顶太阳能潜力评估、城市规划等）提供重要数据支撑。

This dataset contains vector data of building rooftops from 90 cities in China, which were comprehensively selected based on their administrative hierarchy and regional distribution (the complete list of cities is provided in Appendix 1). This dataset was primarily developed using deep learning semantic segmentation models and multi-source remote sensing imagery. The workflow is as follows: First, raw remote sensing imagery was preprocessed, and training and test datasets were constructed via stratified sampling and visual interpretation based on the administrative levels and regional distribution of the selected cities. Next, the training dataset was input into the deep learning semantic segmentation model for training, to tailor the model to the building rooftop extraction task. The performance of the trained model was then evaluated using general-purpose evaluation metrics standard in the deep learning field, based on the test dataset. Finally, the trained model was applied to the rooftop extraction task for all 90 Chinese cities, enabling automatic extraction and vectorization of building rooftops. This dataset can provide valuable data support for urban and even national-scale studies centered on building rooftops, including rooftop solar potential assessment, urban planning, and other relevant research.

搜集汇总

背景与挑战

背景概述

该数据集包含中国90座城市2020年的建筑物屋顶矢量数据，基于深度学习语义分割模型和多源遥感影像制作，空间分辨率≤1米，适用于屋顶太阳能潜力评估和城市规划等研究。数据格式为ESRI Shapefile，开放获取，大小为24.70 GB。