Summary

Direct georeferencing, i.e. the direct measurement of position and attitude of imaging sensors by a combined GPS/inertial system, increases the efficiency and reduces costs for data capture with airborne sensors. The direct measurement of exterior orientation parameters for example

• enables a faster acquisition of the exterior orientation.
• removes limitations to the flight path during image acquisition.
• avoids additional problems of image matching required for automatic aerial triangulation.

Additionally, direct georeferencing using GPS/inertial sensors is an essential pre-requisite for the operational processing of airborne scanner databased on the pushbroom principle.

The different aspects mentioned above have shown the potential applications of direct georeferencing. Using integrated GPS/Inertial systems many applications can be realized more efficiently and economically. In its final stage even traditional aerial triangulation (AT) might become obsolete if

• the exterior orientation is obtained with sufficient accuracy from integrated GPS/inertial systems.
• there are no remaining errors in the calibration of the multi-sensor system (GPS, IMU and camera).

Direct georeferencing is based on the combination of GPS and inertial sensor components. GPS offers the possibility to determine world-wide position and velocity information at a very high absolute accuracy. The principle of inertial navigation is based on the measurements of linear accelerations and rotational rate increments of a body relative to an inertial coordinate frame. The actual position, velocity and attitude information is obtained from an integration process. INS are self contained systems but due to the integration process the accuracy is not constant but time dependent. The relatively high short term accuracy deteriorates with increasing time. Therefore, both sensors are of complementary error behaviour. The integration of GPS/inertial components will

• improve the system reliability
• provide higher accuracy and higher data rate
• determine the complete set of exterior orientation parameters

Using an integrated GPS/inertial system to determine the position and attitude of an imaging sensor (e.g. aerial camera) directly, the standard approach of indirect georeferencing has to be modified. Since normally the positioning sensor (GPS antenna) and the attitude device (IMU) are physically shifted from the imaging sensor a constant displacement vector has to be added to obtain the position of the camera perspective centre in the mapping reference frame. Similarly, a constant misorientation exists between the IMU body frame defined from the axes of the sensor and the imaging coordinate frame (so-called misalignment or boresight alignment angles). The obtained attitudes from GPS/inertial are related to the IMU sensor axes and not to the desired image coordinate frame at first. Therefore, this attitude offset has to be taken into account to obtain the "correct" orientation parameters of the camera perspective centres. Finally, the measurements of the integrated system have to be interpolated on the exposure times of the imaging sensor to overcome the time offset between the different sensors. Additional important aspects like definition of rotation angles (navigation angles versus photogrammetric angles) and the use of different reference frames should only be mentioned in this context.

Sensor configuration for direct georeferencing

Integrated sensor orientation (ISO) is the favoured approach to determine the overall system calibration. Besides consideration of correlations between calibration parameters this integrated approach allows the correct handling of accuracy of directly measured GPS/inertial exterior orientation elements and includes GPS/inertial components and the imaging sensor itself. The functional model is based on the well-known collinearity equations. The GPS/inertial data are used as direct observations of exterior orientation parameters. This functional model is extended by additional parameters to model systematic offsets or linear drifts (if present) of directly measured positions and attitudes, where the attitude offsets compensate for the inherent boresight alignment angles also.

Principle of integrated sensor orientation (ISO) for system calibration

Besides overall system calibration ISO is used to compensate remaining errors even with very reduced number of ground control in the mission site itself, which is essential for highest accuracy and reliability in cases where pure direct georeferencing is not able to fulfil the aspired accuracy demands.

Results

Right now only very few integrated GPS/inertial systems are available providing such a high accuracy potential that can fulfil even highest photogrammetric accuracy requirements. Nevertheless, it has still to be proven, if the specified accuracy and especially reliability could be reached in the standard and very high kinematic airborne environment of photogrammetric applications. This was the motivation for several test flights performed at Institute for Photogrammetry (ifp) to estimate the accuracy potential of commercially available high-end integrated GPS/inertial systems in combination with several airborne digital or analogue sensor systems. Almost all of the tests were flown over the well controlled photogrammetric test site Vaihingen/Enz maintained by ifp. In addition to this, very extensive tests were done in the OEEPE http://www.oeepe.org) framework (now renamed to EuroSDR (http://www.eurosdr.net/)), where a special test was dedicated exclusively on the performance of direct georeferencing and integrated sensor orientation.

History of ifp test flights with different GPS/inertial system installations

Within the ifp tests two commercially available high-quality integrated systems were investigated, namely the

• " POS/AV 510 DG from Applanix Corp. Markham Ontario, Canada ( http://www.applanix.com) and the
• " AEROcontrol-IId system from IGI, Kreuztal, Deutschland ( http://www.igi-ccns.com).

Both GPS/inertial systems typically are combined with an classical Z/I RMK-Top airborne camera. Due to this combination with a standard photogrammetric airborne camera, independent data for the GPS/inertial exterior orientations were provided from photogrammetry (traditional aerial triangulation) for first accuracy estimations. The overall accuracy control is obtained from independent check points on the ground.

The general integration and sensor concept of both GPS/inertial systems is quite similar. The GPS/inertial data integration is based on a loosely coupled decentralized Kalman filter approach, where the differential GPS phase data processing is done first and the obtained DGPS positions and velocities serve as update information for the later inertial data processing. The only difference from a hardware point of view is the IMU: The Litton LR86 dry-tuned gyro (1998) inertial unit installed in the POS/AV 510 DG is meanwhile replaced by other newer units i.e. Applanix AIMU (dry-tuned gyro system) or Litton LN200 (fibre-optical system), where the AEROcontrol unit (IGI IMU-IId) utilizes fibre-optical gyros.

Inside view of LR86 IMU	Inside view of IMU-IId
System installation POS/AV test LR86 IMU mounted at RMK-Top camera	System installation AEROcontrol test IMU-IId mounted at RMK-Top camera
System installation POS/AV test LN200 IMU mounted at RC30 camera

Flight trajectory ifp POS/AV test, December 1998	Flight trajectory ifp AEROcontrol test, June 2000

In our case the ifp test field Vaihingen/Enz about 25km north-west of Stuttgart/Germany was prepared for such accuracy investigations. This test site covers an area of about 30kmē with more than 100 signalized points determined from static GPS surveys. Since the accuracy of these object points is about 5cm and better, they are used as references for the overall system accuracy check. Besides this a DTM from LIDAR measurements is available for the test site providing reference data to estimate the quality of photogrammetric DTM generation. The location of the reference points is optimized for two different image flights with different flying height and image scale. The medium scale flight covers the whole test site and is flown in cross-pattern at a flying height of 2000m above ground. During the three long (east-west) and three cross (north-south) strips altogether 36 images are captured. The large image scale block is flown in the eastern part of the area and consists of two north-south strips (hg=1000m) only, providing 16 independent values for camera air stations. Since a wide-angle camera was used in most test campaigns the different flying heights result in image scales of 1:13000 and 1:6000, respectively. Within the most recent signalization campaign the number of control points was significantly densified in the western part of the test area. Hence future large scale flights will be done in that part of the test site.

The individual test results from these different ifp test series can be found in several publications, i.e. the results from the first GPS/inertial test together with University of Calgary, Department of Geomatics are given in Škaloud et al. (1996), the results from the pushbroom line scanner test flights with DPA and HRSC-A systems are published in Fritsch (1997) and Haala et al. (2000) whereas the investigations on commercial GPS/inertial system performance can be found in Cramer (1999) and Cramer (2001).

From the well controlled experimental flight tests at ifp an accuracy of object point determination about 5-20cm (RMS) for the horizontal and 10-25cm (RMS) for the vertical component was obtained after direct georeferencing of RMK imagery and with optimal system calibration. The accuracy variations are most likely due to the tested different block geometries - large image overlap providing strong block geometry with several multi-ray points positively influences the object point accuracy. Multiple image rays compensate for remaining errors in the orientation elements. This quality is quite remarkable and allows for lots of efficient applications.

Accuracy of direct georeferencing of RMK images (scale 1:13000, hg=2000m)

In general, from the results of these different test campaigns - focussing on the role of GPS/inertial components as one part of the sensor systems - the following conclusions are drawn:

• Direct sensor orientation based on GPS/inertial systems provides high flexibility since this method can be used with any type of sensor (frame/line, analogue/digital). There are no longer restrictions on flying regular block structures. Since there is no need for tie point matching this method of image orientation will succeed even in applications where traditional transfer of tie points and processing is problematic like coastlines and dense forest regions. The use of GPS/inertial sensors is essential for later efficient data processing of push-broom line sensor data, like ADS40.
• Direct determination of exterior orientation parameters using high-quality integrated GPS/inertial systems can reach high accuracy fairly close to standard photogrammetry - but only if a correct GPS/inertial data processing (including efficient GPS/inertial error control, correct transformation between different coordinate and mapping frames, datum problems) and an appropriate overall system calibration (including GPS/inertial components and camera self-calibration) is guaranteed for the specific mission site.
• Direct georeferencing without any ground control is possible but highly unreliable since non corrected systematic or gross errors remain undetected and directly deteriorate the quality of sensor georeferencing. Hence, the use of a certain number of check point data in the mission area itself is recommended to provide redundant data for quality assessment and quality control. If errors are present these check points in the test site may serve as control data for integrated sensor orientation to compensate the error effects.

Although it is explicitly admitted that direct georeferencing is a powerful technology with wide influence on the georeferencing process the results from such academic performance tests typically suffer from the following points: Due to the lack of separated calibration and mission sites the calibration is performed in the mission area itself, which is different from later practical use. In order to estimate the maximum possible accuracy of GPS/inertial systems these test flights are carefully planned, most often the flight turns are flown with low banking angles to preserve good satellite constellation. In addition to that, highly skilled experts (sometimes the system manufacturers themselves) spend a lot of time for the subsequent data processing and the overall system evaluation. Again, such test conditions are non-typical for the later use of GPS/inertial systems in operational production environments, showing quite clearly that the results from these performance tests might be too optimistic compared to the results from later practical use. From these points the transfer of these results on truly operational projects has to verified by analysing the experiences using GPS/inertial systems in production process.

First results from operational practice are already published showing that in principle the accuracy potential of performance tests is verified even in production process. Nonetheless, to reach this quality a proper overall system calibration has to realized. In ideal cases such calibration should be valid for several mission flights, which seems to be possible concerning the stability of boresight alignment angles as long as no change in the system installation is done. A long-term a priori calibration of position offsets seems to be more critical. Experiences from operational calibration flights have shown significant and variable position offsets in the vertical component for several mission days with which will affect the quality of direct georeferencing if their a priori calibration is not done properly. In addition to this, sometimes a significantly worse performance was obtained from operational use of GPS/inertial systems, since the maximum accuracy performance of GPS/inertial integration was not fully exploited during operational GPS/inertial data integration. This definitely is not a deficiency of this technology in general but clearly shows the need for careful data analysis and processing before using the data for system calibration and direct georeferencing. Although operator skills with such data processing will increase pushed by the increased use of this technology in future practice, still some projects will remain where pure direct georeferencing is not able to fulfil the aspired accuracy demands. In that cases GPS/inertial data should be used for integrated sensor orientation (ISO).

Integrated sensor orientation offers the most effective way to calibrate the overall system and to compensate any remaining errors even with very reduced number of ground control. The large impact on object point quality is depicted in the following figures, where the object space residuals from independent check points are given for direct georeferencing (DG) and integrated sensor orientation (ISO). The block shown here consists of one single flight line with 7 images (RMK, normal-angle, large-scale flight). After application of integrated sensor orientation the quality of object point determination increases significantly. This result is based on the use of one additional ground control point in the centre of the strip which is sufficient for the refinement of system calibration parameters in the mission area itself.

Object point quality of operational DG	Object point quality after using ISO

Publications:

• Cramer, M.: Direct geocoding - is aerial triangulation obsolete?, in: D. Fritsch & R. Spiller, eds, ‘Photogrammetric Week '99', Herbert Wichmann Verlag, Heidelberg, pp. 59–70. 1999.
• Cramer, M., D. Stallmann & N. Haala: Direct georeferencing using GPS/inertial exterior orientation for photogrammetric applications, IAPRS, Volume XXXII/B, Amsterdam, The Netherlands, July, pp. 198-205. 2000.
• Cramer, M.: Performance of GPS/Inertial Solutions in Photogrammetry, in: D. Fritsch & R. Spiller, eds. 'Photogrammetric Week '01', Herbert Wichmann Verlag, Heidelberg, pp. 49–62. 2001.
• Cramer, M.: Experiences on operational GPS/inertial system calibration in aiborne photogrammetry, Journal GIS - Geoinformationssysteme 14(6), June 2002, Wichmann Verlag, Heidelberg, Germany, pp. 37-42.
• Cramer, M. & Stallmann, D.: System Calibration for Direct Georeferencing. International Archives on Photogrammetry and Remote Sensing IAPRS, Volume XXXIV, Com. III, Part A, pages 79-84, ISPRS Commission III Symposium, Graz, September 2002.
• Cramer, M.: Integrated GPS/inertial and digital aerial triangulation – recent test results, in: D. Fritsch, ed., 'Photogrammetric Week '03, Herbert Wichmann Verlag, Heidelberg, pp. 161–172, 2003.

for further information contact Michael Cramer