Methodology - KGPS
Description of GPS Data Collection Campaigns for LANDMAP
Preamble / Data Collection Campaigns
In order to validate the various Landmap image and elevation products, two fieldwork campaigns were carried out to collect kinematic GPS data throughout the British Isles. When processed, this data yields accurate 3-D co-ordinates that can be used for quality assessment purposes, namely to validate the accuracy of the Landmap products.
Route Design
All of the routes followed were chosen for their optimal sky visibility, not only in terms of data collection from GPS satellites but also to permit easy identification from the SAR and SPOT satellite imagery. Where possible, routes followed included motorways and major trunk roads that generally afforded unobstructed satellite visibility. Inevitably due to the limited road network in some areas, for example the Lake District and the Scottish Highlands, it was necessary to follow routes that were less than ideal in terms of satellite visibility.
Equipment Used
Both campaigns used the technique of relative kinematic GPS positioning where the unknown mobile GPS receiver was positioned 'relative' to a second GPS receiver that occupied a point of known 3-D co-ordinates.
Each GPS surveying system used for each campaign comprised a high-quality geodetic GPS receiver, capable of recording dual-frequency GPS code pseudorange and carrier-phase measurements, and a geodetic quality choke-ring antenna capable of measuring these dual-frequency observations.
The mobile receiver was located inside the survey vehicle, a large estate car, for ease of operator interface and the antenna was mounted atop the vehicle on a roof-rack system. This ensured that the vehicle did not obstruct the antenna's sky visibility.
The receivers used were Leica Geosystem's System 500 and MC1000 unit as well as a Javad Eurocard receiver system. A Leica choke-ring antenna was selected specifically for both campaigns as it features some advanced multipath mitigation capabilities. This meant that it was able to greatly reduce the number of erroneous GPS signals entering the antenna that had been reflected off nearby structures (e.g. buildings, motorway furniture, high-sided vehicles and the survey vehicle roof) and thus reduce the chance of incorrect solutions.
Campaign 1
The first campaign, carried out in September 1999, required the kinematic GPS profiles for a number of pre-defined circular routes. This suited a 'Real-time Kinematic' (RTK-GPS) survey technique in which both GPS code pseudorange and carrier-phase measurements are recorded. This method is capable of yielding sub-decimetre accuracy over short baselines, generally less than 50 km.
The observing schedule was such that the reference receiver was established at a location deemed to be the centroid of the day's route so that the baseline distances from the 'local' reference receiver to mobile receiver would be kept to a minimum preventing the accumulation of distance-dependent errors. The mobile receiver would then be driven along the predefined route recording satellite observations at a rate of 5 Hz. Once the route was completed, the local reference station team was picked up and the entire team prepared to observe the next scheduled loop.
The mobile team covered almost 4000 miles during the 14 days of the first campaign with the predefined circular routes representing some 2800 miles of that total.
Campaign 2
The second campaign which took place during May and June of 2000 was geared to a different set of objectives and therefore had an observing schedule different to that of the first campaign. There was a requirement to observe some long GPS profiles that would essentially span a number of satellite-pass strips / several stereo-pair strips permitting some checking of the strip matching procedures using orthorectification techniques.
The establishment of a 'local' reference receiver station alongside each section of these proposed transects would have been too demanding in both time and logistics so an alternative processing approach was decided upon. The observing procedure was identical to that of the first campaign with the exception that the 'local' reference receiver remained in the same location for the duration of the campaign. A high-precision geodetic GPS receiver was established at a point of known co-ordinates at University College London where it collected GPS observations for the 9 days of this second campaign. The mobile receiver was driven along the required profiles recording data at a rate of 5Hz.
The routes followed for this second campaign contained a number of features as requested by the SPOT processing team that would aid them in their orthorectification tasks. One particular request was that a number of crossovers should be performed at major junctions whereby a mile or two of additional observations were taken on the feeder roads for the junction in question. Such manoeuvres provide the processing / imaging team with a greater number of features to identity and refer to as part of their orthorectification quality assessment routines. The nature of the road network in some areas meant that several long stretches of road were retraced or intersected which allowed some error checking.
Data processing
All GPS data was post-processed, eliminating the obvious difficulties of maintaining constant telecommunications between the reference and mobile receivers.
Leica's SKI-Pro GPS processing software was used to determine 3-D co-ordinates in the WGS-84 reference datum of the roving receiver antenna from dual-frequency GPS code pseudorange and carrier phase measurements.
Data from the first campaign was post-processed in an RTK-GPS mode that could make use of the more precise carrier-phase observations over shorter distance baselines. With the successful resolution of the carrier-phase ambiguities, this method is capable of yielding positions accurate to 2 cm in plain and 3-4 cm in height.
Results for the second campaign were derived from a technique called differential code GPS positioning (DGPS) that uses the code pseudorange measurements only. Use of this code-only technique is very common when working over longer distances as it becomes increasingly difficult to successfully resolve the carrier-phase ambiguities within the GPS position computation. This is due mainly to the ionospheric and tropospheric refraction effecting an increase in the levels of carrier-phase signal noise. DGPS positioning can provide positions with an accuracy of approximately 1.5 metres in plan and 2.5 metres in height.
Deliverables to LANDMAP SAR processing team
For the purposes of truthing and verifying the Landmap SAR DEM / imagery using GPS positions, it was necessary to perform the comparisons in the same datum. The GPS co-ordinates were converted from WGS-84 into the UK National Grid for local reference purposes and then reduced them to WGS-84 geoid heights using the EGM96 geoid model. The results for each day's route were provided in the following three formats:
1) WGS-84 position and WGS-84 ellipsoidal height
2) WGS-84 position and WGS-84 geoid heights using the EGM96 geoid
3) UK National Grid position and WGS-84 geoid heights using the EGM96 geoid
'WGS84 to UK National Grid Computation'
