We developed several machine learning techniques for growth stages classification based on MODIS (or later incorporating LANDSAT). Currently, we obtained results on our own annotated dataset based on field survey. Are there some open annotated dataset that we can compare our algorithms to this data?
You can find all europe land cover database in CORINE classification Scheme in lambertian conformal conic projection in fallowing web page in 100m and 250m spatial resolution:
http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2006-raster-2
You can use the global land cover facility to get archived Landsat data. Follow this link http://glcfapp.glcf.umd.edu:8080/esdi/
You can check the following site for MODIS and Landsat:
http://earthexplorer.usgs.gov/
http://ladsweb.nascom.nasa.gov/data/search.html
http://daac.ornl.gov/cgi-bin/MODIS/GLBVIZ_1_Glb/modis_subset_order_global_col5.pl
https://mrtweb.cr.usgs.gov/
Dear Wisam adn Abeyou,
Thank you for your reply. Are those data already been annotated or assigned object/observation classes for each pixel? We are looking for the 'annotated' data to test our algorithms, not just the open dataset.
You can try the ZY-3 satellite data, which is high resolution satellite with 2.1m Panchoromatic and 5.8m Multispectral. http://www.sasmac.cn/portal_space/ Link to the "English" on the right-upper of the website. Or just send mail to me.
If US can be your test area, you should consider using WELD Landsat data (http://globalmonitoring.sdstate.edu/faculty/roy/roy.html) as the data source, and use CDL (cropland data layer, http://www.nass.usda.gov/research/Cropland/SARS1a.htm) and NLCD data for validation. They have the same resolution (30m) and map projection, so they can be easily lined up. I recently did an algorithm test using WELD and CDL so I know it works.
Specifically, WELD have the processed Landsat 7 weekly, monthly, seasonal and annual data for over 10 years, and have the processed Landsat 5 & 7 monthly data for three years. CDL has 30m resolution data for conterminous US after 2009, and has 85% - 95% classification accuracy for main crops.
You can find all europe land cover database in CORINE classification Scheme in lambertian conformal conic projection in fallowing web page in 100m and 250m spatial resolution:
http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2006-raster-2
Dear Mohamad Ivan Fanany,
you can find some global land use / land cover produts here
http://www.globallandcover.com/GLC30Download/index.aspx
http://maps.elie.ucl.ac.be/CCI/viewer/index.php
You can download Landsat or Modis images in http://glovis.usgs.gov/, you only have to register
Dear all,
Thank you for your kindly reply. Actually, we have been developed generic model for paddy growth stages detection using remote sensing data. Firstly, we built the model using low spatial resolution (MODIS 250m), and now we extend our model by using higher resolution RS (such as Landsat, SPOT, etc). What we are going to do now is to validate our model using annotated RS data that labeling each growth stage of paddy crop, i.e. vegetative stage, reproductive stage, ripening stage, harvesting stage, and plowing stage.
- Badges
- Science topic