抄録
The earth observation satellite, Landsat-5 launched in 1984, carried a high resolution imaging instrument called the Thematic Mapper. It is possible that the satellite occurs a high frequency attitude fluctuation caused by scanning mirror and high gain antenna drive, et al.. Fortunately, the magnitude of a high frequency attitude fluctuation was less than 30 meters of sensor spatial resolution of the TM imagery and didn't occur conspicuous geometric distortions in imagery. But, an Enhanced TM imagery of 15 meters spatial resolution of Landsat-6 scheduled to be launched in 1993, may be apt to cause attitude fluctuations due to the satellite bus. Therefore, it is necessary to develop an accurate attitude determination in regard to these problems.
There are two types of attitude sensors onboard the Landsat spacecraft: Inertial Measurement Unit and Angular Displacement Sensors. The IMU measures angular rate from DC to about 2 Hz. The ADS measures angular deviations from 2 Hz to 125 Hz. The geometric correction of TM imagery requires a time history from DC to the highest frequency at which a significant motion occurs about 100 Hz. To estimate the attitude is necessary to use an inverse filter with a compensated transfer function of these two sensors. However, some troubles would occur if an inverse filter is applied in straight forward manner. This paper discribes a high accurate estimation algorithm for attitude determination based on the crossover network model proposed by Sehn et al., and evaluates estimation accuracy using a simulated data. The estimation accuracy of 10-2μ rad order at 0.01 Hz to 100 Hz is obtained. Our proposed algorithm may be applied to the attitude determination for the platforms of ADEOS, EOS, scheduled to be launched in 1996 and 1998, respectively.
