Classification#
Description of Modules#
Cluster analysis and segmentation are useful machine learning or multivariate statistical analysis techniques which can perform an automatic interpretation of multi-band raster data.
A multiple band raster dataset, imported through the raster menu is input. If more than one dataset is imported, they must first be merged/stacked into a multiband dataset. It is also important to normalise the data prior to classification to ensure that the data values of the different datasets fall in the same range. If this is not done, the results will be biased towards datasets with large values, e.g. magnetic data.
The Classification menu provides several classification/clustering tools.
Context Menu#
Context menus are available for classification modules after they have been executed. These are accessed by right-clicking on the green classification module. If the module has not been run it will be blue and the context menu will give you the option to select the input raster bands. Cluster Analysis, Crisp Clustering and Supervised Classification have the same context menu while Fuzzy Clustering has a slightly different context menu. The context menu for Image Segmentation is identical to the raster data context menu.
Classification context menus contain the raster data context menu that acts on the input raster datasets. Only the items applicable to the classification results will be discussed here.
Context menus for the Classification module.#