|
Hello,
A basic linear spectral unmixing model can be created with the following nodes. We don’t have the ability to add screen captures in the forum, so I will try to explain:
1. Add a File node, Dataset node, ROI Statistics node, Linear Spectral Unmixing node, View node, and Data Manager node to a new, empty model. For the File node, select the “Raster” type and choose an input raster file. For the Dataset node, select the “Regions of Interest” dataset type and choose an input ROI file.
2. Connect the File and Dataset nodes to the front of the ROI Statistics node. The raster file should connect to the Input Raster field, and the ROI file should connect to the Input ROIs field.
3. Connect the end of the ROI Statistics node to the front of the Linear Spectral Unmixing node. In the Edit Connection Parameters dialog, connect “Mean” under ROI Statistics to “Endmembers” under Linear Spectral Unmixing. This lets you use the ROI means as the spectral endmembers.
4. Connect the File node to the Input Raster field of the Linear Spectral Unmixing mode.
5. Connect the output of the Linear Spectral Unmixing node to the View node, and again to the Data Manager node.
When you run the model, the spectral unmixing raster will appear as a RGB image in the ENVI view. Linear Spectral Unmixing creates an output raster that consists of separate “Abundance” bands for each input endmember (showing the relative abundance of the feature of interest, 0 to 1), plus a separate RMS Error band. If you look at the Data Manager, the band names will not show “abundance” or “error”, which can cause some confusion as to what band corresponds to which endmember.
You can add some logic to create more informative band names in the output, but it is too complex to explain here. One of our developers created a variation of this model that provides better output band names so that you can tell which are the abundance bands and which is the RMS Error band. Please contact Tech Support so we can send you a sample model file.
|