MatConf: Generate a confusion matrix for evaluating thematic maps

This program creates a 'Confusion Matrix', a two dimensional table comparing the classified image data with reference to ground truth data. The output format may be a text file or a file in spreadsheet format (CSV).

The classification raster and the reference raster must correspond to the same geographic area, have the same resolution and have the same units related to the geographic reference system. Except for the background value, the range of values in the classified file must be within the range of the values in the ground truth file and they should have the same number of categories defined. This last restriction can be avoided using /CATEG_FIXAT parameter, when the number of columns will be fixed by ground truth raster categories. However they may be different types of files (excluding real type files that cannot represent category data) and the degree of compression is not important.

The output file contains the 'Confusion Matrix', in which the columns represent the values of the ground truth raster and the rows the values of the classified raster. The diagonal of the matrix shows the correctly classified pixels, and the off-diagonal elements of the table show the number of pixels in which the two rasters do not coincide for each combination of ground truth / classified image categories. The relations between the different elements of the table determine the types of error, the percent accuracies; omission errors and producer errors by column and commission errors and user accuracy by row.

At the end of the file there are various sums, indicating the global accuracy, the global accuracy weighted by surface area with different criteria, and the Kappa parameter to quantify the fit between the two rasters.

All this information can be saved in either of two formats: a text file format which can be paginated by choosing the maximum number of columns that fit in the report width, or the CSV format that has been designed to be read by a spreadsheet program. The CSV format is just a text format in which a specified character is used as a column separator. This separator can be specified in the advanced options. The default value is that value set as the list separator in the Windows Control Panel. As for the TXT format it is also possible to limit the report width when using CSV and so avoid going over the maximum number of columns supported by the spreadsheet application when the number of categories being considered is very large.