stella28 Posted January 10, 2014 Report Share Posted January 10, 2014 (edited) Hi Everyone, As far as I know, this is impossible and the only way we can do this is by visual interpretation. But out of desperate need of data, the satellite images from Google Earth are all we have to assess damaged sites. I got the images as JPEG from Google Earth and made a mosaic and saved it as TIFF. The mosaic is already true color, but by any chance, does anyone know if there are algorithms/methods how to extract/classify features using a single band or true image? Thank you and I would really appreciate the help. Best, Stella Edited January 10, 2014 by stella28 Quote Link to comment Share on other sites More sharing options...
3dbu Posted January 10, 2014 Report Share Posted January 10, 2014 if you can have a indicative class is posible make a fake classification using this tiff image in arcgis you digitalise sample area and give class name and then make a iterative classify.... Quote Link to comment Share on other sites More sharing options...
Lurker Posted January 11, 2014 Report Share Posted January 11, 2014 there is a function called feature extraction, using many software, you can extract feature from images Quote Link to comment Share on other sites More sharing options...
stella28 Posted January 12, 2014 Author Report Share Posted January 12, 2014 Thank you very much for the responses. In regards with the feature extraction, the image should have a high spatial/spectral resolution right before you can do this, right? Quote Link to comment Share on other sites More sharing options...
peralta2794 Posted January 12, 2014 Report Share Posted January 12, 2014 If it is possible, but not recommended. Quote Link to comment Share on other sites More sharing options...
meodensi Posted January 14, 2014 Report Share Posted January 14, 2014 Why not? It depends on the algorithm. As far as your algorithm is good enough which can simulate human's brain (like what you're thinking and doing with visual interpretation), you can use Google Earth images for classification. An important thing of this data source is it's been changed in color (spectral) but kept its texture. You know already damage site is mostly related to texture. You have to use two dates of data (even from Google Earth) to compare. The damage sites look "different" from the previous date. There are two possibilities: - The "difference" is a newly built-up area - The "difference" is the damage site After detecting the difference, you can check the shape of them. If the shape is round/square/rectangle... lets say it's a newly built-up. Otherwise, it's a damage. In this case, you have to think of Feature Extraction function as Lurker already referred to. Quote Link to comment Share on other sites More sharing options...
kingamir Posted January 20, 2014 Report Share Posted January 20, 2014 Brother, If u download usgs landsat 8 data. You can choose geotiff data. Geotiff data can be easily classified using simple supervised classification in ArcGIS or ENVI. HTH Quote Link to comment Share on other sites More sharing options...
kingamir Posted January 20, 2014 Report Share Posted January 20, 2014 The other choice is makesure ur data download from google earth is already georeferenced. Open GIS program such as ArcGIS and manually create shapefile for polygon / point or line feature extraction.HTH Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.