danaroth/jasper_ridge
收藏资源简介:
Jasper Ridge是一个常用的高光谱数据集,用于多种高光谱分析任务。该数据集包含512 x 614像素,每个像素在224个通道上记录,波长范围从380 nm到2500 nm,光谱分辨率高达9.46 nm。由于原始高光谱图像过于复杂,难以获取真实地面数据,因此选择了一个100 x 100像素的子图像,起始像素位于原始图像的第(105,269)个像素。在去除第1-3、108-112、154-166和220-224通道(由于密集的水蒸气和大气效应)后,剩下198个通道(这是高光谱分析中常见的预处理步骤)。数据中潜藏着四个端元:“#1 Road”、“#2 Soil”、“#3 Water”和“#4 Tree”。
Jasper Ridge is a widely used hyperspectral dataset for various hyperspectral analysis tasks. It has a size of 512 × 614 pixels, with each pixel recorded across 224 channels, covering a wavelength range from 380 nm to 2500 nm and featuring a spectral resolution as high as 9.46 nm. Given the high complexity of the original hyperspectral image and the difficulty in acquiring ground-truth data, a 100 × 100 pixel sub-image is selected, whose starting pixel is located at the (105, 269)th position of the original image. After removing channels 1–3, 108–112, 154–166 and 220–224 due to dense water vapor and atmospheric interference, 198 channels remain, which is a common preprocessing step in hyperspectral analysis. Four endmembers are latent in this dataset: "#1 Road", "#2 Soil", "#3 Water" and "#4 Tree".
数据集描述
Jasper Ridge 是一个流行的超光谱数据集,广泛用于多个研究领域。该数据集包含 512 x 614 像素,每个像素在 224 个通道上记录,波长范围从 380 nm 到 2500 nm,光谱分辨率高达 9.46 nm。由于原始超光谱图像过于复杂,难以获取地面真实数据,因此考虑使用一个 100 x 100 像素的子图像,该子图像从原始图像的 (105,269) 像素开始。在预处理过程中,去除了通道 1-3、108-112、154-166 和 220-224(由于密集的水蒸气和大气效应),最终保留了 198 个通道。该数据集包含四种端元:“#1 Road”、“#2 Soil”、“#3 Water”和“#4 Tree”。
快速浏览

数据集来源
该数据集最初由 Feiyun Zhu 收集,原始数据可从以下链接获取:http://www.escience.cn/people/feiyunZHU/Dataset_GT.html
使用此数据集时,请引用相关论文:
@misc{zhu2017hyperspectral, title={Hyperspectral Unmixing: Ground Truth Labeling, Datasets, Benchmark Performances and Survey}, author={Feiyun Zhu}, year={2017}, eprint={1708.05125}, archivePrefix={arXiv}, primaryClass={cs.CV} }




