Unlocking the Potential for IRS-1C Data
Geometric correction and data fusion processes offer the key to new applications
and users
By Dr. Phillip Cheng and Dr. Thierry Toutin

The introduction of the IRS-1C satellite, launched two years ago, and, more recently, the IRS-1D satellite, launched in December 1997, has increased the highest resolution available for commercial optical imagery from 10m, offered by SPOT, to 5.8m. The improvement in the resolution of satellite images has broadened the applications for satellite images to areas such as urban planning, data fusion with aerial photos, and the integration of cartographic features with GIS data. This trend will continue as new satellites capable of imaging at even higher resolutions are launched in the coming years. For these reasons, there is now a growing interest in using the higher resolution IRS-1C data.
    The greatest advantage in using a satellite image is that it covers a much larger area than an aerial photo. For example, a standard IRS-1C panchromatic scene contains 3 arrays of 4096 pixel detectors. The 3 arrays of detectors combined together cover approximately 6100 square km. This is almost equivalent to 33, 1:60,000 scale aerial photos! More importantly, the number of ground control points (GCPs) required is less, and the cost is reduced.
    Most applications of satellite images or aerial photos require a geometric correction process called orthorectification to correct the data according to the user's ground reference system. Because they are planimetrically correct, the resulting orthoimages can then be used as maps.
    Several questions arise regarding the orthorectification of IRS-1C data, the first of which deals with the accuracy of the geometric modeling method used for orthorectification. The two most common methods are the polynomial method and the collinearity condition method. This choice leaves open the question of which is the best method to correct the IRS-1C data in order to obtain the highest accuracy?
    The second question involves the data format. The IRS-1C data are distributed in two different formats, i.e., the EOSAT fast format and the super structure format. The EOSAT fast format is a very simple format and contains no ephemeris and attitude data. The lowest correction level available is systematic correction. The super structure format contains ephemeris and attitude data, and the lowest correction level is raw data with or without radiometric correction. Which format gives the best accuracy in orthorectification?
    The third question is how to generate the best geometric and radiometric corrected mosaic from the IRS-1C data. This is an issue because the IRS-1C data are generated in 3 different arrays with small overlaps.
    The final question addresses how to fuse the orthorectified IRS-1C data with the aerial photo or with the digital elevation model (DEM). Fusion with the aerial photo is particularly useful for areas that are not covered by the aerial photo where the IRS-1C data can be used instead.
    To find the answers to these matters, IRS-1C panchromatic (5.8m) data with 3 arrays distributed in the EOSAT fast format and the super structure format, and a 1:60,000 scale aerial photo scanned at 300 dots per inch (approximately 5.0m ground resolution) of Irvine, California were used. The area has an elevation range between 0 and 2500m.

Which Geometric Modeling Method?
The polynomial method is a very simple method for correcting images. However, the method does not correct distortions introduced during image acquisition. This deficiency limits the polynomial method's use to small areas with flat terrain. It requires many GCPs and only corrects locally around the GCPs.
    The collinearity condition method is a mathematical representation of the physical law of the transformation between the image and ground space. It corrects the entire image globally and also takes into consideration the distortions due to terrain.
    To correct the IRS-1C satellite image, the collinearity condition method - developed by the author at the Canada Centre for Remote Sensing (CCRS), Natural Resources Canada - were used. The method is based on principles related to orbitography, photogrammetry, geodesy and cartography. This method reflects the physical reality of the complete viewing geometry and the following distortions that may occur during the image formation:
(1) distortions due to the platform (position, velocity, and attitude);
(2) distortions due to the sensor (orientation, integration time, and field of view);
(3) distortions due to the Earth (geoid, ellipsoid, and relief); and
(4) distortions due to the cartographic projection (ellipsoid, and cartographic reference).
    The greatest advantage of this collinearity condition method is that it has been applied to VIR data (Landsat, SPOT, IRS, MOS), as well as SAR satellite data (ERS, JERS-1, SIR-C and RADARSAT) and can easily be modified to support other satellite and airborne sensors. Since the method can adjust simultaneously, more than 1 input image can be used to improve the relative accuracy of the positioning of superimposed images. Based on good quality GCPs, the accuracy of this method was proven to be within 1/3 of a pixel for VIR satellite images and 1 resolution cell for SAR images. Since the method was developed, numerous tests have been performed with different data sets and study sites to demonstrate that it is precise, robust, economic, and consistent. The method has been integrated into PCI's OrthoEngine Satellite Edition software.

Which Data Format?
Tests with the collinearity condition method were performed using the IRS-1C data in both EOSAT fast format and super structure format. Ten GCPs and 10 independent check points (ICPs) were collected from each array using 1:24000 scale maps. The ICPs were collected inside the area bounded by the GCPs and were not used in determining the geometric model and its parameters. Table 1 shows a summary of the results. The errors of both GCPs and ICPs were about 1 pixel for the super structure format. The results using EOSAT fast format were much worse. The maximum error of the ICPs for EOSAT fast format was up to 32m (about 6 pixels). Since an irreversible systematic correction was applied to the EOSAT fast format data, the geometry of the image related to the satellite and sensor was consequently destroyed. The superstructure format contains raw images with ephemeris data. A radiometric correction was applied to the data, preserving the geometry and integrity of the image related to the satellite and sensor. It was then possible to orthorectify directly from the raw data to the final image with high accuracy. Only one resampling was required, and, hence, the radiometric quality of the image was preserved.

How to Mosaic IRS-1C Data?
Because IRS-1C data in super structure format are distributed in 3 different arrays with small overlaps, it is always necessary to orthorectify and to mosaic the 3 arrays together. When GCPs were collected from each image and the geometric model was applied to each image separately, there were small discontinuities between the orthoimages in mosaicking. This was due to the inaccuracy or the distribution of GCPs. The problem was solved by adding tie points or altimetric points between the overlap areas to improve the relative accuracy of the geometric model and the mosaicked images. In addition, tie points or altimetric points were used for areas where GCPs could not be collected. The geometric model adjusted simultaneously for all the images such that the resulting orthoimages superimposed with each other had minimal discontinuity on the overlap areas.
    Since the 3 arrays have common characteristics such as the orbit, the resolution, and the field of view, it may be possible to combine the 3 arrays together mathematically to reduce the number of GCPs and to improve the mosaicking results. Research in this methodology is currently being carried out at CCRS.

How to Fuse IRS-1C Data with Aerial Photo?
In order to fuse the IRS-1C data with an aerial photo, it is necessary to orthorectify the aerial photo. The geometric modeling method based on the space resection by collinearity is the best method to orthorectify an aerial photo. This is a purely numerical method that simultaneously yields six independent parameters expressing the space position and angular orientation of a tilted aerial photo. To utilize this method, the calibrated focal length of the camera lens and a minimum of three GCPs with X, Y, and Z ground coordinates must be known. This method permits the use of a redundant number of ground control points; hence, least squares computational techniques can be used to determine most probable values for the 6 parameters. In addition, the method can be extended to multiple blocks of photos by simultaneously using a bundle adjustment method.
    To test data fusion of the IRS-1C data with an aerial photo, PCI's OrthoEngine Airphoto Edition software was used. The software uses the space resection by collinearity for the geometric modeling. Ten GCPs and 8 ICPs were collected from the aerial photo. Table 2 shows a summary of the results. The fused results using the IRS-1C data and the aerial photo were very satisfactory.

How to Fuse IRS-1C Data with DEM?
CCRS has developed an interesting method for perceiving depth from integrated remote sensing and geoscientific data. This method is based on the effect called chromo-stereoscopy: the depth (such as relief, gravimetry, magnetism, etc.) is encoded in colors, and then decoded by simple refraction prisms, contained in the ChromaDepthª glasses. The "normal" flat color images can be viewed in 2-D, but when viewed with the ChromaDepth' glasses, the images become 3-D. An example of an orthorectified IRS-1C image was generated using this method.

Summary
High-resolution IRS-1C data can be used in many mapping and GIS applications. To precisely orthorectify the data, the super structure format and the collinearity condition method should be used. With these guidelines, pixel or sub-pixel accuracy can be obtained in the orthorectification. To mosaic three IRC-1C images of the same area, tie points or altimetric points can be collected in addition to GCPs. Furthermore, it is possible to fuse the IRS-1C data with the aerial photo or with the DEM. The software with the geometric model has been successfully applied in Switzerland and Germany to use IRS-1C data for mapping and classification. Since the data can now be acquired from both IRS-1C and IRS-1D, a growing interest for different applications is expected in the near future.

About the Authors:
Dr. Philip Cheng is a senior software engineer at PCI Enterprises, Richmond Hill, Ontario, Canada. He may be reached at 905-764-0614. Dr. Thierry Toutin is a senior research scientist at the Canada Centre for Remote Sensing, Natural Resources Canada. He may be reached at 613-947-1293.

Back