|
GPS for geometric correction of remotely sensed imagery: possibilities after termination of SA
Background
The higher accuracy (sub-meter/centimeter level) GCP coordinate determination with the help of GPS requires DGPS, because of GPS signal degradation from various sources like SA, atmospheric errors etc. After the decision of US department of defense (DOD) on 1st May 2000 to turn off SA of GPS signals, accuracy range of hand held stand-alone GPS receivers has improved considerably as compared to earlier range of about 100 meter (Chalermchon et al., 2001). In this changed scenario, even a hand held GPS receiver in stand-alone mode may also fulfill the medium accuracy requirements, which may be sufficient for the geometric correction of IRS-1C/D imagery. In India, conventional source of GCP coordinates is Survey of India (SOI) topographic maps on scales 1:50,000 and 1:25,000. These maps are based on very old survey work, and for some of the areas even 1:25,000 maps are not available. The GCP accuracy requirement is higher for the geometric correction of IRS-1C/D, PAN images as compared to the accuracy of available large-scale maps.
GPS is available in various configurations depending on the accuracy requirements and accordingly its cost may vary from about 20 thousands to 20 lakhs. The DGPS configuration for higher accuracy is quite expensive and in the developing countries like India cost is one of the major factor in the adaptability of any new technology like GPS. In such circumstances, hand-held GPS may prove to be a cost effective solution when accuracy requirements are not very demanding. At present, different category hand-held GPS receivers are available, depending on the number of GPS signal channels and their frequency. In the geometric correction of remotely sensed data, GCP collection cost can be reduced considerably if we are able to know about achievable accuracy of hand-held GPS.
Objective
The objectives of this research paper are mentioned below:
- Accuracy analysis of GCP coordinates observed by hand-held GPS in stand alone mode and its comparisons with the total station measurements.
- Investigation of geometric correction accuracy of IRS-1C/D PAN sensor imagery, using GCPs derived from hand-held GPS in stand-alone mode.
Study Area, Data Set and Equipments used
The present research work was carried out for the Kanpur city and its adjoining area in India. Hand-held GPS receiver of 12 channels and L1 frequency (model GS5 of Leica Geosystems) and DGPS receiver of dual frequency having L1 and L2 channels (model SR530 of Leica Geosystems) were used for the GPS observations. The Trimble make total station equipment was used for the measurement of distances between GPS observation stations, in order to verify the distances computed from GPS observations. The IRS-1C/D, PAN sensor images as per details given in table 1, were used for the geometric correction in ILWIS 3.0 software.
Methodology
Research work was carried out as per the steps mentioned below:
- Selection of eight different well distributed GPS observation stations in the I.I.T. Kanpur (IITK) campus, which are separated by distance in the range of 200 to 500 m. Out of the above stations, some stations are obstructed by trees (station no. 3, 5 and 7) and one station is completely free from obstructions (station no. 8) as it is on the roof top of high rise building.
- Measurement of true distance and height difference between above stations with the help of total station equipment of millimeter level precision.
- Selection of permanent features like intersections of highways, roads, canals, railway lines as GCPs in the Kanpur city and its adjoining area for the geometric correction of satellite imagery.
- GPS observations on the above mentioned stations and GCPs, with the help of hand-held GPS (GS5 receiver) in stand-alone mode. These observations were taken on WGS 84 ellipsoid in two different coordinate systems namely: geodetic (latitude, longitude and height above ellipsoid) and projected coordinates (eastings, northings) in Universal Transverse Mercator (UTM) projection system. In case of GPS observation stations of IITK campus, repeated observations were taken at different times and dates spread over a period of about one month.
- Three different GPS observations for longer duration (about 8 to 14 hrs.) on station no. 8 of IITK, using DGPS receiver.
- Computation of mean and standard deviations of observations taken on different dates for IITK observation stations.
- Computation of distances between stations from mean values of eastings and northings observed by hand-held GPS.
- GCP-based geometric correction of IRS-1D satellite PAN sensor imagery in ILWIS 3.0 software, with the help of GCPs derived from hand-held GPS in WGS 84 and UTM system. In the image rectification process affine transformation and nearest neighbor resampling methods were used.
- Determination of geometric correction accuracy at test GCPs, by computing discrepancy between coordinates read from rectified images and GPS observed one.
Results
The results of present work are summarized below:
- Spatial distribution of selected GPS observation stations, in terms of eastings and northings are shown in fig. 1.
- Hand-held GPS observations on different dates for one of the IITK area station are shown in table 5. The deviations of different date observations from mean values are shown in fig. 2
- The standard deviations of GPS measurements (eastings, northings and height above ellipsoid) at various stations are shown in fig. 3.
- The distances between different stations were computed from the mean values of coordinates measured by GPS and these are listed in the table 2, along with the total station measured distances. The difference between GPS and total station measured distances were computed and the same is plotted in the fig. 4.
- For station no.8, GPS observations were also taken with the help of dual frequency receiver for three different dates and mean values of coordinates were computed by single point positioning (SPP) of collected data. Also mean values of the same station coordinates were computed from hand-held GPS observations of different dates. These results are given in table 6.
- For GCP-based geometric correction of remotely sensed images, the root mean square (RMS) errors of affine transformation and number of GCPs used is given in table 3.
- Discrepancies between coordinates read from geometrically corrected images and their hand-held GPS observed values, for various distributed test GCPs are given in table 4.
|
|
