Raster_SupervisedClassificationService

Raster_SupervisedClassificationService

This service serves to reclassify multispectral satellite data using supervised classification method. The classification process is supported by a training map, which is used to identify the spectral signatures for the classification process.

Operation:

supervised - Two supervised classification methods : the maximum-likelihood classifier (MLC) algorithm or the sequential maximum a posteriori - estimation (SMAP) algorithm.

With a sufficient number of training areas, the results of the SMAP are superior to the MLC. A comparison of SMAP and MLC can be found in McCauley and Engel (1995).

Request Parameters:

sourceURLArray: URL Array of a group of imagery files to be classified; Default format: GeoTIFF
trainingImageURL: URL of ground truth training map; Default format: GeoTIFF
classificationMethod: Type of classification method; Parameter format: string; Options: MLC, SMAP; Default: MLC; MLC (the maximum-likelihood classifier algorithm): It is important that the training areas are homogeneous samples. Since training areas cover several pixels, small local variations are included for the definition of the classes. The training areas can be prepared from auxiliary maps such as already available land use maps. It is also important not to generate very small areas since these will be ignored for statistical resons.[Markus Neteler and Helena Mitasova, 2008]; SMAP ( the sequential maximum a posteriori - estimation algorithm): A combined radiometric/geometric reclassification method uses an image pyramid approach which also takes neighborhood similarities into account (Schowengerdt, 1997:107, see also Ripley, 1996:167-168). This combination leads to a significnt improvement of the classification results(Redslob, 1998). A second advantage is that the module also accepts a single channel data. Note that the number of training areas defines the number of classes. The training areas should cover several pixels; otherwise they will be ignored if the pixel number is too small.[Markus Neteler and Helena Mitasova, 2008]
outputGeoTiffType: Type of output GeoTIFF file. The output Format Type is specified as GeoTIFF; Parameter format: string; Options: Byte,Int16,UInt16,UInt32,Int32,Float32,Float64,CInt16,CInt32,CFloat32,CFloat64

Response Parameters:

returnURL: URL of output raster map to hold classification results
returnFormat: Format of output raster map; Default: GeoTIFF

For further informations

ServiceLocation
WSDL
JavaInterface