I'm working with a significant number of rasters (.tiff) as part of a landscape classification exercise. First I want to generate a dendrogram from the rasters (maximum of 7 different rasters) to assess the optimal number of classes for my classification. I would use Wards linkage and the square of the Euclidian distance. My preference is to use a hierarchy so that cluster membership (class) can be displayed as a series of 'maps' containing different levels of complexity (observational or phenon levels) that are still genetically related.
Can anyone recommend an R script (python, matlab) that allows a dendrogram to be generated from multiple rasters and then enables HCA to be performed on the rasters to produce a classified image?
Thank you.
I hope Matlab can do almost anything for you. See for example : https://au.mathworks.com/help/stats/dendrogram.html
Thanks Muhammad,
I was hoping to find a script that can handle rasters? They will have to be loaded into a dataframe before application of HCA. One of the challenges is that each raster is 2 GB in size and with 7 rasters this equates to 14 GB of data to load in the data frame.
Clint
Convert the rasters into feature data points by using feature vectors. That would enable them to be utilized by the dendrogram function and at the same time reduce the computational overhead. Feature characterization would depend on the content itself, for example, GLCM for texture modeling or RGB correlogram for color modeling.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification