
Pixel-by-Pixel
Classification for Zoning and Monitoring
Version
2.05 (September 2002)
Package
written by S. Griguolo, with the collaboration of M. Mazzanti
| The version 2.03.05 of ADDAPIX, partially funded
by FAO, project GCP/INT/578/NET , was released in 1996.
The rationale under the development of the package, that implemented some multivariate analyses (Principal Component Analysis, Classification, etc.) on time series of satellite images, was that: "...NDVI images are timely available for whole continents, and it would be a waste of resources to omit any effort to draw from them all possible useful information. ADDAPIX is an attempt to obtain useful synthetic results, at least from a qualitative viewpoint and in spite of the above mentioned unreliability, by processing large amounts of NDVI data." (ADDAPIX Manual, Introduction) Though initially aimed at the Early Warning community, and therefore developed with Cold Cloud Duration, NDVI time series and IDA/WINDISP image format in mind, ADDAPIX can easily handle any set of satellite images provided as bitmaps (RAW, BIN or BIL), and it has actually been applied so far for different purposes to a variety of images. In its executable form this package can be freely used by anybody for any type of application. The authors' only expectation is its use to be acknowledged in publications, as it is the custom in the scientific community. Now a Windows 32 version is being developed,
again with a partial contribution from FAO/ARTEMIS.
Changes in version 2.04.01 (released June 2000)
If the reconstruction of missing values is requested (in PREPDATA), but no code for missing values is defined for the concerned IDA type, the reconstruction cannot be carried out. Anyway, for de-clouding/de-misting purposes, each pixel's series is modified by filling local negative peaks caused by (even partial) cloudy or foggy conditions. This is ALWAYS done when the reconstruction or smoothing of a time series is requested. Be warned that this operation is inappropriate when working on radiometric bands, for which reconstruction/smoothing make usually no sense. To see how the procedure works, run PREPDATA twice on a time series of images: first do a simple merging without reconstructing/smoothing, then another run requesting reconstructing/smoothing. You will obtain two .DAT files that you can simultaneously visualize in DISPLAY to realize the way the procedure has operated. The picture below shows an example:
It was therefore necessary to have recourse to another DOS Extender (WDOSX by Michael Tippach), fixing some small incompatibilities existing with the ADDAPIX code. You can download separately the ADDAPIX
Manual from here and the Tutorial
documentation from here. They are
both in ACROBAT 4.0 PDF format, and are also included in the EXE files
listed above.
|
Download here
below the data for the Tutorial and the version 5.1 of the package.
|
|
|
|
|
|
|
|
|
|
|
describes the changes, presents some warnings and explains how to install the package |
Any remark, suggestion, request, feedback on analyses carried out using ADDAPIX, and bugs encountered (specially bugs!) will be highly appreciated. Please write to:
|
|
|
(page maintained by Silvio Griguolo - last update: September 30, 2002)