AUTOMATED MAPPING FROM LINEAR ARRAY SENSOR DATA
Shunji Murai and Ryosuke Shibasaki
Institute of Industrial Science
University of Tokyo
7-22, Roppongi, Minatoku
Tokyo, Japan
ABSTRACT
The objective of the study is to propose an automated mapping
from linear array sensor data in stereo mode.
The study deals with the following items:
1) Production of simulated stereo images which will be
obtained from linear array sensor, by using the digital
terrain model (DTM) and the corresponding scene corrected
LANDSAT MSS data.
2) Determination of exterior orientation parameters given as
the function of time, by using ground control points.
3) Automated mapping by applying auto-correlation to the
stereo imagery.
Computer simulation was done for the test area where the
national elevation data bank and scene corrected LANDSAT
data are available. The test area was selected in the
mountaneous area including high mountaines covered with snow,
flat plane, rivers, reservoirs and so on. Simulated stereo
images were generated by computer with a given exterior
orientation parameters. Two different stereo angles of 45°
and 60° were compared.
A mathematical model to determine the orientation parameters
was developed with use of ground control points which were
selected in the imagery, and/or attitude and tracking data
of the satellite. Accuracy of exterior orientation was
compared for two different stereo angles. Standard deviation
of residuals ranges 0.3 to 0.5 pixel for the control points,
while 0.3 to 1.0 pixel for unobserved points.
Three auto-correlation methods to recognize the conjugated
points in the stereo imagery, were compared with respect to
the accuracy and computing time. As the result of comparison,
minimum absolute Siam method were selected as the optimum
method. Accuracy of height measurement was about one pixel
RMS.
SIMULATED STEREO IMAGERY
Computer simulation was done for the following variables.
. stereo angle
. satellite orbit
. satellite attitude
483