monster-monash/Tiselac

TiSeLaC（时间序列土地覆盖分类）数据集是为了2017年欧洲机器学习与知识发现数据库原理与实践会议上的土地覆盖分类挑战而创建的。该数据集由2014年留尼汪岛的23个Landsat 8卫星图像时间序列生成，大约每16天采样一次。这是一个像素级别的数据集，每个时间序列代表单个像素的变化值。每个像素提供10个时间序列特征，包括7个覆盖可见和红外频率的表面反射率特征和3个由这些波段派生的指数——归一化差异植被指数、归一化差异水指数和亮度指数。数据集包含了99,687个时间序列，每个序列长度为23。

The TiSeLaC (Time Series Land Cover Classification) dataset was created for the land cover classification challenge held in conjunction with the 2017 European Conference on Machine Learning & Principles and Practice of Knowledge Discovery in Databases. It is generated from a time series of 23 Landsat 8 satellite images of Reunion Island, sampled approximately every 16 days in 2014. The dataset consists of 99,687 pixel-level time series, each with 10 features over 23 time points, representing about one year of data. Each time series includes seven surface reflectance features across visible and infrared frequencies, and three indices derived from these bands.