Hi,
This might be a very simple question but I am having some issues and was hoping someone could help. I have a multispectral (Red, Green, Blue, NIR) raster file of an urban area. I am trying to use supervised classification to sort out about 9 different land use types. For about 8 of these classes the classification works well but for the last (tin roof tops) it is a bit hit or miss. I think this is due to the NIR band as the multispectral image was not taken at solar nadir so there are shadows and due to sun angle some of the angled roof's are being mis-classified. I have tried selected new regions of interest but has not solved the problem. I also have a thermal image of this area and the roof type that is not classifying properly are normally cooler. Is there a way to use the thermal image to classify these cooler roof's and combine that with the supervised classification? Or is there a way of correcting the classification so I can use it to have average thermal temperature of each land use?
Thank you in advance.
Rick
