Use of this function requires a license for Whitebox Workflows for Python Professional (WbW-Pro). Please visit www.whiteboxgeo.com to purchase a license.
This tool performs an evaluation of the reflectance properties of multi-spectral image dataset for a group of digitized class polygons. This is often viewed as the first step in a supervised classification procedure, such as those performed using the min_dist_classification or parallelepiped_classification tools. The analysis is based on a series of one or more input images (inputs
) and an input polygon vector file (polys
). The user must also specify the attribute name (field
), within the attribute table, containing the class ID associated with each feature in input the polygon vector. A single class may be designated by multiple polygon features in the test site polygon vector. Note that the input polygon file is generally created by digitizing training areas of exemplar reflectance properties for each class type. The input polygon vector should be in the same coordinate system as the input multi-spectral images. The input images must represent a multi-spectral data set made up of individual bands. Do not input colour composite images. Lastly, the user must specify the name of the output HTML file. This file will contain a series of box-and-whisker plots, one for each band in the multi-spectral data set, that visualize the distribution of each class in the associated bands. This can be helpful in determining the overlap between spectral properties for the classes, which may be useful if further class or test site refinement is necessary. For a subsequent supervised classification to be successful, each class should not overlap significantly with the other classes in at least one of the input bands. If this is not the case, the user may need to refine the class system.
min_dist_classification, parallelepiped_classification
def evaluate_training_sites(self, input_rasters: List[Raster], training_polygons: Vector, class_field_name: str, output_html_file: str) -> None: ...