The Image and Signal Processing (ISP) group at the Universitat de València has harmonized a big database of labeled multi- and hyperspectral images for testing classification algorithms. We think that, like in other related fields of science, data sharing and reproducibility are the only ways for fostering true advance in remote sensing data processing. So far we have harmonized 43 image datasets, both multi- and hyperspectral. We want to expand this database as much as possible in order to objectively evaluate algorithms and submitted papers. We provide training pairs (spectra and their labels) and test spectra. Researchers are able to train their algorithms off-line, and then evaluate their accuracy over an independent, fixed, spectra test set per image. The system returns accuracy and robustness measures of your algorithm in that test set, as well as a ranked list of the best methods. The datasets and the automatic testing system will be available as soon as no data copyright conflict is identified.
The database contains 7 million multispectral images from the MSG sensor. Images correspond to 200 landmark locations in the globe during 2010. They are fully labeled with cloud vs cloudfree classes. A landmark is essentially a ground Control Point or geometric feature on the Earth surface with known location (typically a small coastline area or island). Landmarks are essential in image registration and geometric quality assessment. Matching the landmark accurately is of paramount relevance, and the process can be strongly impacted by the cloud contamination of a landmark. This a challenging problem for classification, in which the main goal is the automatic detection of clouds over landmarks. Spatial and temporal information are typically used, as well as the need for illumination compensation and feature extraction are a must.
Vegetation-related parameters (LAI, fCover, Chlorophyll content) from hyperspectral spectroradiometer measurements or from airborne hyperspectral images, atmospheric parameters (temperature, moisture, emissivity and ozone) from infrared sounders, carbon, heat and and water fluxes upscaling from eddy-covariance flux towers, etc.
This is a 21 class land use image dataset meant for research purposes. There are 100 images for each of the following classes. Each image measures 256x256 pixels. The images were manually extracted from large images from the USGS National Map Urban Area Imagery collection for various urban areas around the country. The pixel resolution of this public domain imagery is 1 foot. Please cite the paper by Yi Yang and Shawn Newsam, "Bag-Of-Visual-Words and Spatial Extensions for Land-Use Classification, "ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (ACM GIS), 2010, if you find this database useful.
This database contains the familiar USGS mineral database for spectral unmixing. The interested user may find useful to look at the MATLAB demos in the Spectral Unmixing tool