I5PR5
UNISPACE III - ISPRS Workshop on
“Resource Mapping from Space”
9:00 am-12:00 pm, 22 July 1999, VIC RoomB
Vienna, Austria
I5PR5
width and positioning accuracy - differ significantly what makes
a software designed to control the ScanER station considerably
more complex (and the station cost - perceptibly higher). Despite
of these essential internal differences, the stations very little
differ in weight and size of the hardware and practically do not
differ in a user’s interface of applications which control them. As
to ScanViewer and Catalogue Manager, they work as well with
any type of data created with ScanEx stations. Correspondingly,
a sequence of actions a user has to carry out to get an image -
from obtaining the satellite orbital elements to registering the
image in the data base - is the same, and the actions are described
m the same terms and executed with identical program controls.
In other words, a user familiar with the ScanEx station caa
master the ScanER station in half of an hour, and this half of an
hour is needed mostly to leam features in behavior of the
satellites - not the applications. This all gives us still more
grounds to define the ScanER as a personal station.
The whole process of obtaining images and preparing them for
the following thematic processing includes the following steps:
1. Approximately once in a week, a user must get fresh orbital
elements and calculate a schedule - also for a week - in the
station control application (the Resurs Receiver for ScanER
station). Since Resurs-Ol satellites transmit data not
permanently, a user must select lines in the shcedule accordingly
to the so-called "Resurs 1-4 Measurement Program" (a schedule
of downloads distributed by the Center of Program Research of
Russian Space Agency (CPR of RSA)) and save the schedule
into a file.
2. After the schedule is loaded into the control applicatioa it
receives data automatically. The user must only observe an image
and stop the reception when he find a level of errors in the image
too high, or restart the reception in a case of accidental loss of
the signal.
3. The user looks through the received image with the
ScanViewer, check the image georeference and correct it in the
first-order approximation, manually selects fragments for the
long-term storage and saves them in new files. Taken into acount
in the fragmentation are a noise level, surface illumination and
the image contents itself (say. long pieces of open sea surface are
usually cut out). But the main is cloudness: if no clouds an image
may need no fragmentation as well as a dense cloudness may
make the whole image invalid for any further usage. Usually an
image is cut into 2 - 4 fragments.
4. Selected fragments are written to CD-ROMs. When a CD is
full, the Catalogue Manager is called to scan the CD and write a
description of images into the data base.
An example of concrete thematic application
As an example, involving multitemporal features in the field of
decoding characteristic gives a great opportunity for mapping
infrastructure of forest massifs. To this must be added that
factors of forest growth became more recognizable after
multitemporal analyze. In our work we had found five principle
time periods substantial for such research, especially in northern
and middle taiga regions.
But in spite of the best opportunities that remote sensing
investigations can give to monitoring forested lands there are
some problems. Main part of them are associated with wide
spectrum of decoding characteristics different forest ecosystems.
Information excessive of the remote sensing data increase the
quality of the local experimental results, but strongly worse the
results of regional and global estimations. The main difficulty on
our view is incompatible and unreproducible methods image
processing the remote sensing data.
The neuro-based decoding technology can give a chance to avoid
most part of that problems. Using algorithms ANN SOM (Self-
Organizing maps. T. Kohonen, Springer Verlag. 1997) allow
cvantifay remote sensing image on homogenous areas. Also, it
allow to connect this rated regions with thematically constants
feature of current mapping object.
So, analysis of remote sensing data in suggested technology
begins from measurement image quality and checking up its
suitability for the studying current decoding task. After that tire
process of teaching neuron-net begins. Various sizes of teaching
massifs and its stmcture or topology in the field of remote sensed
image allow to obtain very high selectivity of decoding
knowledge
On the basis of this technology during the period 1997-1999
different thematic maps (1:50 000 - 1:1000 000) were created.
The most part of them obtained on the basis of Resurs-Ol N3
satellite images and deal with the problems of the forest
exploitation, crave dynamic and land use in forested lands.
Region of application includes different territories of northern,
middle and southern taiga in the board of Russian Federation.
The thematic application of the Resurs-0 images, besides
forestry, is wide: multitemporal environmental monitoring,
agriculture, control for the hazardous territories, control for the
seashore zones, etc.
A network of ScanER stations
The first ScanER station was put on work in Moscow in May
1996 for receiving information from Resurs-Ol #3. It was
followed by stations in Kurgan (December, 1996), Salekhard
(May, 1997), Krasnoyarsk (August, 1997), Khanty-Mansiysk
(October, 1997), N. Novgorod (November, 1997), Tomsk (2
stations. December, 1997). Ufa (January, 1998) and S-
Petersburg (end of April, 1998), Yuzhno-Sakhalinsk (June.
1998), Irkutsk (August, 1998) installed and exploited by local
environment protection services. Now there are three stations are
operating in Moscow, Irkutsk and Khanty-Mansiskfor receiving
information from Resurs 0-1 #4.
At present two stations operate in Moscow and Yuzhno-
Sakhalinsk for reception information from Resurs-Ol #3 on
International Archives of Photogrammetiy and Remote Sensing. Vol. XXXII Part 7C2, UNISPACE III, Vienna. 1999
75
