Article April/May 2000

Warming Up for ISPRS Amsterdam

A Look at Current and Future Imagers, Imagery & Systems

By Professor Gordon Petrie 

To begin with, we can certainly expect a great deal of attention at the forthcoming ISPRS Congress to be focused on the new generation of optical airborne and spaceborne imagers. Of course, optical imagers have always been important, indeed vital parts of the mapping process, yet at the same time, they have been something of a side-show at past ISPRS Congresses. On the airborne side, over the last 30 years, the very high performance frame-type film cameras from the two major European suppliers - Zeiss (now Z/I Imaging) and Wild (now LH Systems) - have totally dominated the scene. Over this long period, the only really major change was the adoption of image movement compensation (IMC) in combination with gyro-controlled mounts and fine-grain high-resolution film to give markedly improved ground resolution. By contrast, on the spaceborne side, over the same period, there has been a steady progression in the form of scanners producing digital image data with ever smaller ground pixel sizes - from 80m (MSS), through 30m (TM), 10m (SPOT) to 6m (IRS-1C/D). During this time, optical space imaging technology has been dominated by the use of scanners equipped either with rotating optical-mechanical elements, as used on Landsat (since 1972), or linear array sensors, as used on MOMS (since 1983), SPOT (since 1986) and IRS (since 1995).

Frame-Type Imagery & Linescan Imagery

Thus there has been a very clear distinction from the geometric, technological and the physical product points of view between (i) airborne analogue film data consisting of discrete frame-type images with a very high geometric resolution; and (ii) spaceborne digital data comprising continuous strips of linescan imagery having a much lower geometric resolution. At the Amsterdam Congress, one may confidently expect imagers to take centre stage with the advent of a completely new generation of devices equipped with all-digital sensors. Furthermore the Congress should reveal a complete cross-over and mixture of the previously distinctive technologies - including frame-type digital cameras being operated from space and pushbroom linear array scanners being operated from aircraft.

Kodak's Airborne Digital Frame Cameras

In recent years, small-format digital frame cameras with areal arrays of CCD detectors have appeared that are suitable both for airborne and spaceborne imaging operations. Thus Kodak has produced its relatively inexpensive Megaplus range of monochrome cameras, typically with 2k x 2k = 4 Megapixel arrays. These have been used to take both single-shot pan images and sequential exposures in conjunction with the use of a rotating filter wheel placed in front of the camera to produce multi-band images - as in the case of the Sensys Technologies AA497 Airborne Multi-spectral Digital Camera (AMDC). Kodak's own DCS 460 CIR cameras employ 2k x 3k = 6 Megapixel arrays with integral filters (giving 18 Megapixels for three bands) to produce colour infra-red (CIR) false-colour images. Over the last two or three years, these have proven to be very popular - especially those developed with special mounts for use in small aircraft and integrated with GPS sets by Positive Systems in the USA (with its ADAR system) and GeoTechnologies (with its ADPS) in the UK. These have given a number of commercial mapping companies, environmental organisations and universities experience of using digital cameras in light aircraft with low operating costs, especially for applications such as crop, environmental or disaster monitoring - where rapid local response is a key issue. Notwithstanding the limited ground coverage of each frame image produced by such cameras and the very large numbers of these images needed to cover any substantial area of terrain, the users are enthusiastic about this development and can be expected to share their enthusiasm at the Congress.

Other Airborne Digital Frame Cameras

Going up the scale in terms of CCD array size, 4k x 4k = 16 Megapixel areal arrays have been utilized in the experimental airborne digital cameras developed by IGN (France) and Ohio State University (USA). Currently at the top of the resolution and format range, Philips have produced  a 7k x 9k = 63 Megapixel array and Lockheed-Martin-Fairchild an 8k x 8k = 64 megapixel array. However the manufacture of such large arrays lies at the very edge of current chip fabrication technology. Furthermore, to get such a large number of individual imaging elements to work properly, each with the same response and avoiding dead pixels, is very difficult - as is the radiometric calibration of these sensors. With low chip yields, this makes large-format CCD areal arrays very expensive to produce. In this context, there has never been any question about the adequacy of the geometric resolution of digital cameras - e.g. the Kodak cameras use areal arrays with a 9.2 mm pixel size. However the small array size gives a limited ground coverage, especially when compared with the 25k x 25k = 625 Megapixels of a aerial film camera image digitized at the same pixel size of 9.2mm. However, of the two major manufacturers, Z/I Imaging is now taking the plunge into this area with its new Digital Modular Camera (DMC) concept involving the use of multiple cameras (i) to get over the ground coverage limitations, and (ii) to produce multi-band, multi-spectral images. If the actual hardware DMC camera is shown at the Amsterdam Congress, then it is certain to be a centre of attention.

SSTL's Digital Space Cameras

At the same time, in parallel with these airborne developments, digital cameras are starting to be mounted in satellites. Thus, for example, low-cost digital cameras equipped with off-the-shelf 1k x 1k CCD areal arrays from Kodak and lenses from Leica have been installed and used in the experimental UoSAT-12 mini-satellite produced by SSTL in the UK to validate key mini-satellite bus and payload technologies. Even these inexpensively produced cameras are producing pan images with ground pixel sizes of 10m (equivalent to that of the SPOT Pan sensor) and multi-spectral images with a 30m ground pixel (equivalent to that of Landsat TM) - albeit with limitations in their ground coverage. The UoSAT-12 cameras also employ the approach of sequential exposure of the constituent band images to produce multi-spectral images - like that adopted on the Sensys Technologies AA 497 airborne camera mentioned above. Whereas, in the earlier SSTL TM-Sat, three separate cameras are being used to produce the component band images simultaneously to create multi-spectral images - in a similar manner to that proposed with Z/I Imaging's DMC camera.  

Other Spaceborne Cameras

Digital cameras using areal arrays were also installed in EarthWatch's EarlyBird satellite. Unfortunately, although the satellite was launched successfully in December 1997, the on-board power supply failed four days later. A similar camera was to have been mounted in NASA's Clark satellite. However, in February 1998, the project was terminated due to cost overruns and the delays associated with the non-availability of the launcher. In summary, regarding future prospects in this field, the new digital cameras with areal arrays that will be discussed at the Amsterdam Congress are just the beginning of this development. There is still a long way to go before these digital cameras can compete directly with current large-format film cameras: in this respect, everything is dependent on the successful development of larger areal arrays and their availability in quantity at a reasonable cost.  

Airborne Pushbroom Scanners

At Amsterdam, we shall also see the entry of the airborne pushbroom scanner based on the use of linear CCD arrays into the mainstream of photogrammetry. The technology has undergone a long gestation period. The original concept of the three-line scanner with fore/nadir/aft pointing allowing along-track stereo-imagery to be acquired both from the air and from space is that devised by Hofmann in 1972 and has been nurtured ever since by the German Aerospace Centre (DLR). Under its sponsorship, in parallel with the development of the technology for use in the MOMS, MEOSS, Mars96 and Mars Express space missions, a series of airborne versions of the three-line scanner have also been built. These have included the EOS (in 1978), the Digital Photogrammetric Assembly (DPA), the Wide-Angle Airborne Camera (WAAC) and, most recently, the High-Resolution Stereo Camera (HRSC). The use of the last of these (the HRSC-A) by DLR and the French ISTAR company has resulted in a series of most impressive mapping products, including high-resolution multi-spectral orthoimages and DEMs. Now the technology has been taken up by the second of the two major aerial film camera manufacturers, LH Systems. The results achieved with the engineering version of the company's new scanner installed in a gyro-controlled mount and utilizing a 12,000 pixel linear array with a pixel size of 6.5mm were shown publicly at the beginning of 1999. A further prototype model was flown in January 2000. If, as promised, the production version of the imager featuring a multi-spectral capability with four lines recording images simultaneously in the blue, green, red and near-IR bands and the use of 20,000 pixel arrays in each line does appear, then undoubtedly it will be another star attraction in the Technical Exhibition at the Congress.

Space Pushbroom Scanners

Here the emphasis will almost certainly be on the products from the new high-resolution space imagers. After the protracted development of the technology and several disappointments over failed launches, at last, Space Imaging's IKONOS with its Kodak-built pushbroom scanner - whose pan sensor is equipped with a 13,500 pixel linear array with a 12 mm pixel size - has been placed successfully in orbit and has come into commercial operation. Certainly we should expect to see and hear a great deal about the products and the applications of the IKONOS imagery at the Congress. This will be reinforced by the first images (if all goes well!!) from the competing QuickBird, EROS and OrbView satellites, all of which are scheduled to be launched during the next few months before the Congress takes place towards the end of July. Although the resulting Pan imagery is being labelled as "high-resolution", some sense of perspective needs to be kept about the use of the term in this particular context. Thus the 1m ground pixel of the new space imagery is equivalent to that obtainable from modern 1:40,000 scale aerial photography. Whereas a 20 to 25cm ground pixel can fairly readily be obtained from 1:10,000 scale aerial photography and still larger scale photography - in the scale range 1:3,000 to 1:6,000 - with a 5 to 10cm ground pixel size is in regular use for the large-scale mapping of urban areas. Thus the biggest value of the "high-resolution" space images could well be that of allowing images to be acquired for remote areas and over countries that have severe restrictions regarding the taking and dissemination of aerial photography of their territory. But the pricing of the new imagery as compared with that of comparable aerial photography will also be a decisive factor in its take-up. Again this whole matter should become clearer at the Congress and it will be very interesting to see how the issue of the Space Imaging company refusing to release the sensor model of IKONOS to the system suppliers will be resolved.  

Imaging Spectroscopy

During the last few years, much of the attention of the remote sensing community has been given to the development of imaging spectroscopy. With this technology, the imaging of the ground takes place using a scanner that provides images in a large number of contiguous, narrow, but discrete spectral bands so that a complete spectrum is obtained over a wide range of visible and infra-red wavelengths for the area being imaged. Usually this technique is termed hyperspectral imaging with the term "hyper" replacing "multi" to convey the idea of the much large number of individual bands or channels being covered as compared with the small number of much broader bands used with multi-spectral imagery. To achieve this, suitable prisms or gratings are used to refract the incoming radiation differentially on to an array of detectors that can capture the full range of up to several hundred narrow spectral bands. Much of the impetus for this development has come from NASA, which has funded the development and construction of a number of alternative hyperspectral scanner designs both in-house (e.g. those built by JPL and GSFC) and by outside contractors (e.g. TRW).

Airborne Hyperspectral Scanners

Although the eventual deployment of these hyperspectral scanners will be in space vehicles, up till now, almost all of the existing imagers have been operated from airborne platforms to prove the design, operation, performance and reliability of the new systems. Prominent among these is the Advanced Visible Infra Red Imaging Spectrometer (AVIRIS) constructed by JPL and operated from high-flying NASA aircraft. Besides the many NASA sponsored developments, a number of commercial suppliers - e.g. GER (USA), ITRES Research (Canada) and Integrated Spectronics (Australia) - have entered this field and have sold airborne systems to various mining exploration companies and to government organisations involved in environmental monitoring. One can expect the results from this development and its applications to be presented at the Congress. They are eagerly awaited and sought by many field and environmental scientists.

Spaceborne Hyperspectral Scanners

The story regarding spaceborne hyperspectral devices has been punctuated by failures and disappointments - as has so much of optical remote sensing from space in recent years. In particular, NASA's Lewis satellite with its two alternative hyperspectral imagers built by TRW and GSFC respectively was lost shortly after its launch in August 1997. But a determined effort is now under way to retrieve this rather dire situation. Thus NASA's newly launched Terra satellite has various sensors with multiple band imaging capabilities in the form of its ASTER, MODIS and MISR scanners. In two or three months' time, NASA will also launch its EO-1 developmental satellite with its advanced ALI multi-spectral linear array scanner and its Hyperion hyperspectral imager with 220 spectral bands - the latter instrument being derived from that lost on the Lewis satellite. EO-1 will be orbited in formation with both Landsat-7 and Terra for comparative purposes. Again, if indeed all goes well. then one can expect the images and preliminary results from all three satellites to be presented and discussed before a large audience at the Amsterdam Congress. In particular, there has been a big revival of interest with the advent of this latest satellite (L-7) in the Landsat series. Furthermore, the availability of its multi-spectral imagery with its wide ground coverage at a medium resolution and at a low cost seems certain to be reflected in papers given in the appropriate technical sessions and in the images that will be displayed on the stands in the Technical Exhibition.  

Radar Imagery

Dealing with microwave radar imagery is not easy - in this respect, your reviewer still bears the scars of his own considerable involvement with this type of imagery during the 1980s. And there is still no sign of solutions to some of the fundamental difficulties - including the occurrence of speckle or clutter; foreshortening; layover; dead areas due to radar shadow; etc. - that are experienced with this type of imagery. Notwithstanding your reviewer's previous (poor) experience, it is obvious that currently there is a big revival of interest in this field. Much of this has been fuelled by the recent developments in interferometric SAR (InSAR or IfSAR) for DEM generation. The basic idea is quite an old one - having been introduced originally by the Goodyear company in the mid-1970s. However, since then, the technology and the subsequent processing of the data have slowly been developed to a much more mature state. This has resulted in much activity taking place recently using data acquired both from airborne and spaceborne platforms. Indeed current interest is literally sky-high - it really is a hot topic!!  

Airborne SAR Imagery

Once again, much of the basic research and development in this field has been carried out by NASA with JPL to the fore. This work has resulted in the development of systems such as the TOPSAR/AIRSAR dual-frequency SAR and the IFSARE InSAR system (in cooperation with ERIM). The latter has formed the basis of the STAR-3i system now being operated on a commercial basis for DEM and orthoimage generation by the Canadian Intermap company. Another Canadian company, Atlantis, is also operating the venerable CCRS SAR-580 system on a commercial basis. As a result, some really large contracts have been completed in North and Central America (e.g. in Puerto Rico, Panama and Colorado) using these two systems. Besides the InSAR developments, work still continues at CCRS and Vexcel using stereo-radar for height determination and mapping. If the North Americans do indeed come in force and present their work at the ISPRS Congress, there will be plenty of interest - especially in terms of the accuracy and completeness of the DEMs and orthoimages produced by airborne InSAR methods in comparison with those generated from aerial photography and airborne laser scanning. The upsurge of interest in both airborne and spaceborne radar has also been reflected in the positioning of some of the system suppliers - who have allied themselves with specialist radar software companies. Examples include Z/I Imaging offering Atlantis's EarthView InSAR processing software and ERDAS's new InSAR, StereoSAR and OrthoRadar packages which have been developed in cooperation with Vexcel (USA) and NPO Mashinostroenia (Russia). PCI 's RadarSoft is another (home-brewed) software suite for use in this area of SAR imagery. Ask about all of these on the exhibition stands!

Spaceborne SAR Imagers

There has been something of a lull in this field with regard to InSAR data collection activities after the Tandem Mission of ERS-1 and -2 in 1995 allowing two-pass InSAR operations. However processing of the data still continues - e.g. DEM and image data covering a large area (130,000 sq. km.) of Labrador was produced from 23 ERS-1/-2 tandem-mode pairs by Atlantis Scientific and completed at the end of 1999. Since the end of the Tandem Mission, ERS-2 has continued to collect data on its own for those areas that are covered by suitable ground stations and this has been supplemented by the similar widespread activities of the Canadian RADARSAT. So there has been plenty of space SAR imagery available for those who find benefit from its application. In this respect, the RADARSAT orthoimage mapping mission covering the whole of Antarctica, carried out in cooperation with NASA, is particularly outstanding. Experimental work using repeat-pass RADARSAT InSAR data has also been carried out by CCRS and Atlantis. Furthermore RADARSAT stereo-pairs using images with same-side and opposite-side configurations obtained from different orbits have been used by Vexcel and by CCRS for DEM generation. But in this particular area, the Congress limelight will surely shine most brightly on the NASA-JPL and DLR Shuttle Radar Topography Mission (SRTM) with its aim of generating a DEM of the whole of the Earth's land mass lying between latitudes 60 N and S. This mission is currently under way from the Space Shuttle Endeavour using the single-pass InSAR technique made possible through the innovative use of a 60m telescopic mast to carry the second antenna. Although the processing of all the data collected during its ten day mission will take at least two years to complete, one would expect some preliminary results to be given at the Amsterdam Congress. If so, they will be of great interest to many participants.

Airborne Laser Scanning

Like so many of the current "new" technologies, airborne laser scanning has, in fact, had to undergo a long, slow and difficult development period since it was first devised in the 1970s. But now it is mature, operational and exciting, with a large number of systems having been built and put into service, both in Europe (e.g. TopoSys, TopScan and TopEye) and in North America (Optech, Nortech, EagleScan, etc.). Almost all of the devices in current use employ cross-track scanning using a downward pointing laser and time (and therefore distance) measurements of the returns from the ground objects in conjunction with an integrated DGPS/INS system to determine continuously the position and attitude of the sensor. After processing these measurements, dense elevation data in the form of a DEM is produced along a narrow swath of the terrain. Part of the attraction of the method is that the rapid pulsing rate and dense sampling allows penetration of the vegetation canopy to give both the height of the vegetation and of the terrain surface (the so-called "bald Earth" !!) below. Building roof elevations are another result from such surveys. Another important point is that a modern laser scanning system can readily be fitted into a small plane or helicopter. However, at present, since the effective operational flying height (and therefore the swath width) over which most airborne scanning lasers can be operated is limited, the method has been applied mainly to "corridor" surveys, e.g. in the Netherlands - where water management is so important - along coasts, rivers, canals, dikes and polders. Similar surveys have been carried out along power transmission lines, pipelines, railway networks and roads in other countries. Since laser scanning only produces DEM data, not image data, it frequently needs to be supplemented by imagery taken with a digital or video camera. Whether the laser scanning technique is cost effective over large areas of terrain in competition with aerial photogrammetric mapping or airborne InSAR surveys is a matter that will no doubt be discussed and debated at the Congress. In this respect, the new AeroScan laser scanner developed for use by the EarthData and Spencer B. Gross mapping companies in the U.S.A. can reputedly be operated at flying heights up to 20,000 ft. (6,000m). If this is correct, then it could change the situation entirely.

Integrated DGPS/INS Systems

What has also become very clear over the last few years, especially with airborne scanning devices - whether pushbroom linear array scanners, InSAR radar imagers or laser scanners - is the ever growing importance of integrated DGPS/INS systems. This has come about since the very accurate DGPS measurements can only be made at a comparatively wide time interval (typically once per second) whereas the INS gives measurements at a much smaller time interval (typically 200 times per second). Thus the DGPS data gives very accurate in-flight positions, but only at well spaced intervals. By contrast, the INS data has a lower absolute accuracy but provides frequent measurements with a high relative accuracy between successive measurements. This helps to determine the short-term changes in the platform position and attitude - which is especially important when considering the high speed of operation of a scanning device. Thus the 3D coordinates of the actual point in the air from which each line scan originates can be determined more accurately by using the INS data to help carry out the interpolation between the DGPS positions and to handle the rapid changes in the sensor's tilts arising from atmospheric turbulence. Since the individual lines of the scanned data are being acquired at intervals of a few milliseconds, the INS data is essential in providing the positional and attitude data required for each line if the scanner images or data are to be used for photogrammetric purposes. But these integrated DGPS/INS systems are still very expensive (up to $200,000 per unit).

Aerial Film Cameras

Which brings one back finally to the classical "old fashioned" frame-type film cameras. Notwithstanding all the new or recently developed all-digital imaging technologies that command so much of the current attention of the photogrammetric and remote sensing community and of the discussion conducted above, these film cameras and the products derived from them still set the standard against which everything else is judged. In this context, it is worth remembering that, at the present time, 99% of all topographic mapping is being carried out using images acquired by frame-type film cameras. Their combination of large-format, wide coverage, high resolution and low geometric distortion is still unrivalled. Which means that, notwithstanding the inconvenience and expense of first having to chemically process and then scan the films for use in digital processing systems, they will still be with us and serving us well for quite some time to come. Nonetheless, it is still important to consider the matter of the impact of integrated DGPS/INS systems on these film cameras as well. Positional accuracies of 20 to 30cm and attitude accuracies of 20 arc-seconds are being claimed for systems such as the Applanix Position & Orientation System (POS). Indeed, it is claimed that the accuracies of these values are such as to eliminate the need for aerial triangulation. This remains to be proven, but, at the very least, they should certainly reduce the control point requirements for such triangulation operations. Whether the absolute orientation of each individual stereo-model can indeed be achieved using the DGPS/INS data resulting in the elimination of the aerial triangulation process is another matter that needs to be proven through extensive and rigorous testing. No doubt, we can expect this type of research work to be reported at the Congress too. If it is successful, then it has considerable implications for the future of all types of digital photogrammetric systems and operations.

Conclusion

Given the rich menu that will be offered, I am really looking forward to the sumptuous meal that can be consumed at this Congress!!

Professor G. Petri, Department of Geography & Topographic Science, University of Glasgow, Glasgow, G12 8QQ,

Scotland, U.K.