DIGRP

Description

DIGRP groups input data layers according to class and unique-conditioned regions. This serves as a preprocessing step for the spatial data integration algorithms in DIEST, used to predict geological events. DIGRP also estimates a metric for each input layer, indicating how relevant a layer is to the prediction task.

Parameters

digrp(file, dbic, kosn, sasn, unicon, datfil)

Name	Type	Caption	Length	Value range
FILE *	str	Input file name	1 -
DBIC *	List[int]	Input raster layer	1 - 16
KOSN *	List[int]	Known occurrences bitmap segment	1 - 48
SASN	List[int]	Surveyed area bitmap segment	0 - 48
UNICON *	List[int]	Unique condition region IDs channel	1 - 1
DATFIL *	str	ASCII data file name	1 -

* Required parameter

Parameter descriptions

FILE

Specifies the name of the PCIDSK file containing input layers to be read, and to which the UNICON channel will be written.

DBIC

Specifies the input image channels for which the data layers will be integrated and used to predict geological events. Each channel contains a categorized data layer. Up to 16 layers may be specified.

KOSN

Specifies the number of the segment containing the known occurrence bitmap. ON pixels in the bitmap correspond to occurrence. This bitmap represents the 'a priori' knowledge of the occurrence of a geological event whose future occurrence is to be predicted. It serves as a training data set. If several bitmaps are specified, a pixel is set to ON if any of the bitmaps has an ON pixel.

SASN

Specifies the number of the segment containing the surveyed area bitmap mask.

If the SASN parameter is missing, the entire image is assumed to have been surveyed. If more than one bitmap segment is specified, these are OR'ed together to form the mask (a pixel is set to ON if any of the bitmaps has an ON pixel).

UNICON

Specifies the channel to receive the uni-condition region IDs. Because there may be a large number of uni-condition regions, it is strongly recommended that a 16-bit channel be used to hold the IDs. This channel is required by DIEST.

DATFIL

Specifies the name of the ASCII file in which to save the grouped data layer and uni-condition region statistics. This data file is required by DIEST.

Details

The DIGRP and DIEST set of functions integrates multiple layers of spatial data and produces a prediction model for a specific geological event. The prediction model is in the form of a favorability map where, at each pixel, a higher value indicates that the geological event is more likely to occur at this location.

DIGRP groups and summarizes the input data layers and saves the statistics in the specified ASCII data file. This serves as a preprocessing step for the spatial data integration algorithms in DIEST, used to predict geological events. DIGRP also estimates a metric for each input layer, indicating how relevant a layer is to the prediction task. Regions with unique conditions are grouped together and their IDs are saved in the unique condition (UNICON) output channel. Typically, a user only needs to apply DIGRP once for a set of input channels. DIEST may then be applied many times to produce favorabilities of occurrence using various algorithms. The data file from DIGRP and the UNICON channel are used by the estimation algorithms in DIEST.

It is important that each input channel contain only a small number of classes, rather than a raw gray-level image. Any continuous raw data should be classified or thresholded to obtain a theme map consisting of a few class labels.

DIGRP also computes a metric for each input layer, which indicates how relevant the layer is to the prediction of the geological event, based on a "correlation score" of the layer with the known occurrence map. The metric is in the range of 0 to 1, with 0 being least relevant and 1 being most relevant. The metric for the input layers is printed to the standard output.

Example

DIGRP groups the input channels and generates an image of uni-condition regions. This image is to be saved in channel 15 (must be at least a 16-bit channel). Some statistics of the uni-condition regions are written to an ASCII file (DATFIL="land.dat"). Later, the information in the UNICON image and in the data file may be used by DIEST to generate favorability maps.

from pci.digrp import digrp

file='irvine.pix'
dbic=[1,2,3]
kosn=[10,11]
sasn=[11]
unicon=[15]
datfil='land.dat'

digrp(file, dbic, kosn, sasn, unicon, datfil)