Experimental
Fuzzy k-NN classification that yields soft class membership and optional probability surfaces.
remote_sensing classification knn fuzzy legacy-port
| Name | Description | Required | Default |
|---|---|---|---|
inputs | Array of single-band input rasters. | Required | ['band1.tif', 'band2.tif', 'band3.tif'] |
training_data | Point/polygon vector training data path. | Required | training.shp |
class_field | Class field in training_data attributes. | Required | class |
scaling | Feature scaling mode: none (default), normalize, standardize. | Optional | none |
k | Number of neighbors (default 5). | Optional | 5 |
m | Fuzzy exponent parameter (> 1; default 2.0). | Optional | 2.0 |
output | Optional output classified raster path. | Optional | — |
probability_output | Optional membership-probability raster path. | Optional | — |
Run fuzzy kNN and output both class and confidence rasters.
wbe.fuzzy_knn_classification(class_field='class', inputs=['band1.tif', 'band2.tif', 'band3.tif'], k=7, m=2.0, output='fuzzy_knn_classified.tif', probability_output='fuzzy_knn_probability.tif', training_data='training.shp')