ibm-nasa-geospatial/hls_burn_scars
收藏资源简介:
--- size_categories: - n<1K license: cc-by-4.0 language: - en --- # Dataset Card for HLS Burn Scar Scenes ## Dataset Description - **Homepage: https://huggingface.co/datasets/nasa-impact/hls_burn_scars** - **Point of Contact: Dr. Christopher Phillips (cep0013@uah.edu)** ### Dataset Summary This dataset contains Harmonized Landsat and Sentinel-2 imagery of burn scars and the associated masks for the years 2018-2021 over the contiguous United States. There are 804 512x512 scenes. Its primary purpose is for training geospatial machine learning models. ## Dataset Structure ## TIFF Metadata Each tiff file contains a 512x512 pixel tiff file. Scenes contain six bands, and masks have one band. For satellite scenes, each band has already been converted to reflectance. ## Band Order For scenes: Channel, Name, HLS S30 Band number 1, Blue, B02 2, Green, B03 3, Red, B04 4, NIR, B8A 5, SW 1, B11 6, SW 2, B12 Masks are a single band with values: 1 = Burn scar 0 = Not burned -1 = Missing data ## Class Distribution Burn Scar - 11% Not burned - 88% No Data - 1% ## Data Splits The 804 files have been randomly split into training (2/3) and validation (1/3) directories, each containing the masks, scenes, and index files. ## Dataset Creation After co-locating the shapefile and HLS scene, the 512x512 chip was formed by taking a window with the burn scar in the center. Burn scars near the edges of HLS tiles are offset from the center. Images were manually filtered for cloud cover and missing data to provide as clean a scene as possible, and burn scar presence was also manually verified. ## Source Data Imagery are from V1.4 of HLS. A full description and access to HLS may be found at https://hls.gsfc.nasa.gov/ The data were from shapefiles maintained by the Monitoring Trends in Burn Severity (MTBS) group. The original data may be found at: https://mtbs.gov/ ## Citation If this dataset helped your research, please cite `HLS Burn Scars` in your publications. Here is an example BibTeX entry: ``` @software{HLS_Foundation_2023, author = {Phillips, Christopher and Roy, Sujit and Ankur, Kumar and Ramachandran, Rahul}, doi = {10.57967/hf/0956}, month = aug, title = {{HLS Foundation Burnscars Dataset}}, url = {https://huggingface.co/ibm-nasa-geospatial/hls_burn_scars}, year = {2023} } ```
样本规模分类: - 样本量小于1000 许可证:CC BY 4.0 语言: - 英语 # HLS烧伤疤痕场景数据集卡片 ## 数据集说明 - **数据集主页**:https://huggingface.co/datasets/nasa-impact/hls_burn_scars - **联系方式**:Christopher Phillips 博士(cep0013@uah.edu) ### 数据集概述 本数据集包含2018-2021年美国毗邻本土地区的烧伤疤痕区域的协调Landsat与Sentinel-2(Harmonized Landsat and Sentinel-2, HLS)影像及配套掩膜文件,共计804张512×512像素的影像场景。本数据集的核心用途为训练地理空间机器学习模型。 ## 数据集结构 ## TIFF文件元数据 每张TIFF文件均为512×512像素的影像。影像场景包含6个波段,掩膜文件则仅含1个波段。对于卫星影像场景,所有波段已转换为反射率值。 ## 波段顺序 影像场景波段顺序如下: 通道序号 | 波段名称 | HLS S30 波段编号 1 | 蓝光波段(Blue) | B02 2 | 绿光波段(Green) | B03 3 | 红光波段(Red) | B04 4 | 近红外波段(NIR) | B8A 5 | 短波红外1(SW 1) | B11 6 | 短波红外2(SW 2) | B12 掩膜文件为单波段影像,各像素值含义如下: 1 = 烧伤疤痕区域 0 = 未燃烧区域 -1 = 数据缺失 ## 类别分布 烧伤疤痕区域:11% 未燃烧区域:88% 无有效数据:1% ## 数据划分 804份文件已被随机划分为训练集(占比2/3)与验证集(占比1/3),两个目录下均包含影像、掩膜文件及索引文件。 ## 数据集构建流程 在将形状文件与HLS影像场景匹配对齐后,以烧伤疤痕区域为中心截取512×512像素的影像块。对于位于HLS影像瓦片边缘附近的烧伤疤痕区域,其中心位置会相应偏移以适配。 研究人员已手动筛选掉存在云覆盖与数据缺失的影像,并人工验证了烧伤疤痕区域的存在性,以确保影像场景尽可能干净无干扰。 ## 源数据说明 影像数据来自HLS V1.4版本。关于HLS的完整说明与获取方式,请访问:https://hls.gsfc.nasa.gov/ 本次数据集所用的烧伤疤痕形状文件由监测烧伤严重度趋势(Monitoring Trends in Burn Severity, MTBS)团队维护,原始数据可通过以下链接获取:https://mtbs.gov/ ## 引用说明 若本数据集对您的研究有所帮助,请在出版物中引用`HLS Burn Scars`。以下为示例BibTeX引用条目: bibtex @software{HLS_Foundation_2023, author = {Phillips, Christopher and Roy, Sujit and Ankur, Kumar and Ramachandran, Rahul}, doi = {10.57967/hf/0956}, month = aug, title = {{HLS Foundation Burnscars Dataset}}, url = {https://huggingface.co/ibm-nasa-geospatial/hls_burn_scars}, year = {2023} }
数据集概述
数据集名称
HLS Burn Scar Scenes
数据集描述
该数据集包含2018-2021年美国连续地区的Harmonized Landsat和Sentinel-2烧伤疤痕影像及其相关掩码。共有804个512x512场景,主要用于训练地理空间机器学习模型。
数据集结构
- 文件类型:TIFF
- 分辨率:512x512像素
- 场景包含:六个波段(已转换为反射率)
- 掩码包含:一个波段
波段顺序
- 场景波段:
- 1: Blue (B02)
- 2: Green (B03)
- 3: Red (B04)
- 4: NIR (B8A)
- 5: SW 1 (B11)
- 6: SW 2 (B12)
- 掩码波段:
- 1: Burn scar
- 0: Not burned
- -1: Missing data
类别分布
- Burn Scar: 11%
- Not burned: 88%
- No Data: 1%
数据分割
数据集随机分为训练集(2/3)和验证集(1/3),每个部分包含掩码、场景和索引文件。
数据来源
- 影像来源:HLS V1.4
- 烧伤疤痕数据来源:Monitoring Trends in Burn Severity (MTBS) 维护的Shapefile
引用信息
@software{HLS_Foundation_2023, author = {Phillips, Christopher and Roy, Sujit and Ankur, Kumar and Ramachandran, Rahul}, doi = {10.57967/hf/0956}, month = aug, title = {{HLS Foundation Burnscars Dataset}}, url = {https://huggingface.co/ibm-nasa-geospatial/hls_burn_scars}, year = {2023} }




