IRSAMap:高分辨率土地覆盖矢量数据集

数据来源： 结合 Jilin-1 卫星影像（0.5 m）、Google Earth 与 Open Cities AI Challenge，保证分辨率统一至 0.5 m。覆盖范围： 共 79 个典型区域，横跨六大洲，总面积约 1000 km²。其中，中国境内 16 区（419 km²），Google Earth 数据 33 区（436 km²），Open Cities AI 数据 30 区（155 km²）。数据规模： 约 5000 幅切片（1024×1024 像素），共计 180 万实例，涵盖建筑、道路、水体、植被等 10 类自然与人造要素。采样策略： 中国区按九大生态分区挑选代表性区域，境外按大洲划分，确保地理多样性与场景均衡。方法：通过包含人工标注、模型推理和人工校正三个步骤的迭代标注过程，生成了覆盖全球 79 个区域的地表覆盖矢量数据集。

Data Sources: This dataset combines Jilin-1 satellite imagery (0.5 m resolution), Google Earth data, and the Open Cities AI Challenge dataset, with all data standardized to a uniform resolution of 0.5 m. Coverage: A total of 79 typical regions spanning six continents are included, with a combined total area of approximately 1000 km². Specifically, 16 regions (419 km²) are located within China, 33 regions (436 km²) are sourced from Google Earth, and 30 regions (155 km²) are derived from the Open Cities AI Challenge dataset. Dataset Scale: The dataset contains approximately 5000 image slices (1024×1024 pixels), totaling 1.8 million instances, covering 10 categories of natural and man-made features including buildings, roads, water bodies, and vegetation. Sampling Strategy: For domestic Chinese regions, representative areas are selected based on China's nine major ecological regional divisions; for overseas regions, sampling is conducted by continent to ensure geographic diversity and balanced scene distribution. Methodology: An iterative annotation workflow consisting of three stages—manual annotation, model inference, and manual correction—was utilized to generate the global land cover vector dataset covering all 79 regions.

搜集汇总

背景与挑战

背景概述

IRSAMap是一个高分辨率土地覆盖矢量数据集，基于吉林一号卫星影像（0.5米分辨率）、Google Earth和Open Cities AI Challenge数据源构建，覆盖全球六大洲79个典型区域，总面积约1000平方公里，包含约5000个图块和180万个实例，涵盖建筑物、道路、水体和植被等10个类别。该数据集通过迭代标注方法确保数据质量，并采用地理多样性采样策略，适用于大规模土地覆盖制图和分析。