SAR Pre-Processing: Flight Track Reconstruction using GPS and INS
Contact Persons
Dipl. El.-Ing. ETH Arnold Barmettler |
Keywords
SAR pre-processing, motion compensation, DGPS, IMU, INS, GPS/IMU merging
Quick Reference
The task of SAR processing requires a very accurate knowledge of the geographical position of the radar sensor. Besides for airborne synthetic aperture radar such a system is of interest for e.g. air flight traffic, and photogrammetry.
In the airborne case, two distinct navigation systems are usually available:
- an acceleration based inertial navigation system (INS), that assesses positional data with high dynamic but poor accuracy and
- a satellite based (differential) global positioning system (DGPS) with typical high positional accuracy but low temporal bandwidth.
Within this project, an exploitation of the favourable characteristics of both systems has been performed. Several methods were investigated in order to achieve a positional data set of the flight track, which is redundant, reliable, highly dynamic and absolutely accurate (temporal and spatial).
Both, off- and on-line systems were considered during the study.
Some Results
For the analysis of the characteristics of the DGPS and INS navigation systems, various data sets were available from flight campaigns. The following figure shows the original and compensated positional signals from the DOSAR flight campaign held in 1994 in Solothurn, Switzerland.
Example of original and enhanced positional data
from the DOSAR flight campaign in 1994.
The raw data used to merge the different navigational systems are dipicted in the graph above: The inertial data (IMU) and DGPS (crosses at 1 Hz intervalls). The resulting signal is shown as continous signal close to the DGPS samples.
Methods
Off-Line Correction
A first task is to temporal synchronize the various data sources. Since the sampling rate is quite different and the available INS data are only patches of several seconds duration out of several hours of flight time and may even be missing a reference time indication (cf. following figure), rigorous correlation procedures were developped.
Ground track of the DOSAR sensor during several hours of
data aquisition. In red the track while aquiring
hires data of our reference test site.
Among several correlation algorithms applied, the method of Sum of Mean Normalized Absolute Difference showed the best signal to noise ratio in assessing the correct time offset of the INS data patch: (R and S are the two time sequences sampled to the same data rate).
The compensation of the positional data was performed using an algorithm which uses the DGPS signal as positional reference:
Block diagram of the algorithm used to compensate
the low-order drift of the inertial measurement unit.
On-Line Correction
Operational navigational equipment requires a near-real-time correction of the drifts inherent to inertial measurement units. We used adaptiv controllers to continously estimate the current error signal in the DGPS and INS positional data. One of the simplest controllers used is shown in the following figur.
