class [ parameter=value ... ] [ inputfile outputfile ] class [ parameter=value ... ] [ inputfile ... directory ]
Parameters are: include_vars, training_sets, class_codes, unclass_code, reject_level.
class takes one or more input datasets and classifies them using training datasets from the training_sets list. One output file is created for each input file.
First, the list of classes has to be determined. The examples of classes are Land, Ocean, Clouds, etc...
Second, a list of image channels to be used for classification has to be determined, e.g. avhrr_ch1, avhrr_ch3, avhrr_ch4 for AVHRR image.
Third, training sets have to be supplied for each class. A training set is a collection of points selected from any image that has the same channels as the input. Each point is represented by its pixel value for each channel used for classification. A training set can be created, for example, with xvu --> Select --> (with cursor) option.
Each class will be presented in the output image by its integer code specified in the class_codes array. The unclass_code value will be assigned to the pixels that could not be classified for whatever reason.
After data is assigned to a particular class, an additional test of classification reliability is performed. If this data falls outside the class statistical distribution tail, then it is claimed unclassified. The reject_level value defines the percentage point for each distribution's tail.
Specifies variables to use for classification. They must be present in all training sets and input datafiles. All variables should be different. No wild characters are allowed.
There is no default.
Specifies names of files containing training statistics - one file per each class. All training sets should be different.
There is no default.
Specifies integer code for each class.
The default is an array of integers from 1 to the number of training sets.
Specifies integer code for unclassified data.
The default is 0.
Specifies in percents the tail part of distribution from where all the data will be claimed unclassified.
The default is an array of 1s.
[1] % class avin.cdf avin.class include_vars : char(255) ? [] avhrr_ch2 avhrr_ch4 training_sets : char(255) ? land.cls ocean.cls clouds.cls class_codes : int ? [1 2 3] unclass_code : int ? [0] reject_level : int ? [1 1 1] 3 3 3
Classifies one Terascan dataset into three groups (land, ocean and clouds) using two variables (avhrr_ch2 and avhrr_ch4). The output image will have four levels, from 0 to 3 (land=1, ocean=2, clouds=3, unclassified=0). All values falling into the 3% of distribution tail of each class are claimed unclassified.
xvu.
Last Update: $Date: 1998/05/28 19:43:30 $