The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
886 
Figure 3. The measurement residual of the Filter in model 1. 
Figure 4. The estimated position error in the model 1. 
Figure 8. The Std. Dev. of attitude in the Filter of model 1. 
The différence of positions from GPS end from Throstle 
Figure 5. The estimated drift of the gyroscope in the model 1. 
Figure 6. The estimated bias of the acce. in the model 1. 
Figure 7. The Std. Dev. of position in the Filter of model 1. 
Figure 9. The position difference between the results of Throstle 
in the model 1 and of Caravel PP. 
The attitude difference between Throstle and POSPac 
Figure 10. The attitude difference between the results of 
Throstle in the model 1 and of POSPac. 
In Figures 2 - 10, it can be found that the predicted residual and 
measurement residual of the KF with model 1 are about the 0.5 
- 1.0 meter. The spikes in the figures correspond to the aircraft 
banking where there were vehicle maneuvers. The estimated 
error of the position in the KF is about 10-20 centimeters. The 
estimated drifts of the gyroscopes for three axes are about 4, 5 
and -6 deg/h respectively. The estimated biases of the 
accelerometers are 0.3, -0.3 and 0 mg for the three axes 
respectively, and the biases of the axes corresponding to the east 
and north directions vibrate intensely with the different flight 
strips. That means the model 1 is not good enough to describe 
the stochastic error in the system. The std. deviations are about 
6cm for position, 4cm/s for velocity (not shown in figure), 
0.002 degrees for pitch and roll, 0.01 degrees for heading. The 
position difference between the results of Throstle in the model 
1 and of GPS is about 0.5 meter, which is consistent with the 
innovation of the filter. Because the GPS positioning accuracy 
is about 10 cm, the absolute position accuracy of the integrated 
system can be considered to be accurate at the 0.5-meter level. 
The attitude difference between the results of Throstle and of 
POSPac is about 0.1 degree. This result shows the 3- 
dimentional attitude accuracy in model 1 is at the 0.1-degree 
level as the attitude resolution in POS AV 510 system is