42 January/February 2003 • GI News

IMAGERY

Eyes in

SPOT-5 HRS spaceborne along-track stereo imager using forward- and backward-pointing CCD linear arrays. (Source: Spot Image)

GI News • January/February 2003 43

IMAGERY

the SkyImagery

from spaceplatforms

by ProfessorGordon Petrie

IMAGERY FROM SPACEBORNEPLATFORMS

Spaceborne frame camerasAs with the airborne platforms, both film anddigital frame cameras have been used fromspace, though in much smaller numbers.

Photographic frame camerasSince the Space Shuttle flights with the ESAMetric Camera and the NASA Large FormatCamera (LFC) in 1983, no Western filmcameras have been flown in civilianspacecraft, though this is not to say that filmcameras may not have been used since thenin American military reconnaissancesatellites.

However, the main known operators offilm frame cameras from space have beenRussian agencies. These cameras have beenused on numerous short duration missions

from comparatively low altitudes (220 to 270kilometres) using film return capsules. Bycontrast, the corresponding Americanagencies appear to have discontinued theiruse of this particular technique in the mid tolate 1980s.

Extensive coverage has been acquired fortopographic mapping purposes using theRussian TK-350 large format film camerahaving a ground resolution of 7 to 10 metres.A similar performance (5 to 10 metre groundresolution) has been achieved using the KFA-1000 reconnaissance camera, a pair of whichhave often been operated together within a

Over the last few years, the terrain imagingfield has expanded into an almost bewilderingarray of imagers, from frame cameras andline scanners to microwave imagers. In thefirst of these two articles, we focused onimagers that are used mainly on airborneplatforms. This month, we look at the imagersmounted on spaceborne platforms.

For spaceborne platforms, the operatingaltitude may vary from an absolute minimumof 160 kilometres to a practical maximum of1,000 kilometres. All spaceborne platformsoperating within this altitude range travel atvery high speeds (up to seven kilometres persecond) over the ground. This means thatrapid image motion may need to beconsidered in the design or operation of theimager. Furthermore, on the one hand, thehigh operating altitude of a spacecraft allowsa synoptic view of the ground to be imagedthat cannot be obtained from any airborneplatform. On the other hand, it alsomeans that very powerful andphysically large telescopes areneeded if high-resolution opticalimages are to be acquired fromspace. In turn, these pose size andweight problems in the spacecraftand need powerful launchers to liftthem into a suitable orbit.

The provision of adequate powerto allow active radar (SAR: syntheticaperture radar) imagers to beoperated over long distances fromorbiting space platforms is a specialproblem for the designers andoperators of these devices. The size,weight and cost of the largesatellites that can accommodatehigh performance imagers hasgrown enormously. For example, theEuropean Space Agency’s Envisat,launched in March 2002 with itsoptical and microwave imagingsensors, is the size of a large bus,weighs eight tons and cost $2.3 billion. It alsorequired the use of the most powerful andmost costly European launcher (Ariane-5) toplace it in orbit.

Therefore, there is a strong move towardssmaller, cheaper satellites as exemplified bythe SSTL (Surrey Space Technology Ltd)micro- and mini-satellites. These can carrysmall yet quite powerful optical imagers.However, one certainly can’t mount the largeultra-high-performance optical and SARimagers used by the American NationalReconnaissance Office on such smallsatellites.

44 January/February 2003 • GI News

single satellite using a split-vertical (lowoblique) configuration to increase the cross-track coverage. However, the KFA-3000 (ahigher-resolution model in the KFA series)and the KVR-1000 panoramic film camerahave produced images with still better groundresolutions of two to three metres. These havebeen sold quite widely to a variety of users inthe Western world, besides their use byRussian agencies. Since 2000, the RussianSovinformsputnik organisation has beenmaking available imagery with one metre (DK-1) and 1.5 metre (DK-2) ground resolution. Itis not clear whether these images have beenproduced using film cameras or linescanners, or perhaps even both.

Digital frame camerasFor spaceborne digital frame cameras therehas been considerably less developmentactivity than in the airborne field. Both the

commercial EarlyBird and theNASA Clark satellites were tohave orbited the same types ofdigital frame cameras in 1997.These would have provided bothpanchromatic (three-metreground pixel) and multispectral(15-metre ground pixel) images.However, EarlyBird suffered apower supply failure a few daysafter its launch, and the Clarksatellite was cancelled due tocost overruns. In each case, a 2kx 2k pixel CCD areal array was tohave been used as the imagingsensor in the camera.

Apart from these non-operational examples, the mainbuilder and actual user of digitalframe cameras has been SSTL,based in Guildford. Its camerashave been deployed as bothpanchromatic and multispectralimagers. In the case of the

UoSAT-12 satellite launched in 1999, a pair ofmultispectral frame cameras have beenmounted side-by-side in the satellite witheach camera being tilted in oppositedirections relative to the flight line (or groundtrack) to provide a wider coverage in thecross-track direction, as has been done withthe KFA-1000 film camera mentioned above.The additional pan camera mounted on theUoSAT-12 satellite points in a vertical (nadir)direction to produce a 10 x 10 kilometreimage with a 10-metre ground pixel size. TheCCD areal arrays used in both the pan andmultispectral frame cameras of the UoSAT-12

IMAGERY

The area of high-resolution pushbroomscanners has been the one where

commercial activity has been greatest, but ithas also been one of considerable

disappointment and large financial loss. Inturn, the EarlyBird (in December 1997),

EROS-A (in January 1998), IKONOS-1 (inApril 1999), QuickBird-1 (in November

2000) and OrbView-4 (in September 2001)satellites have all been lost either at launch

or immediately afterwards. The onlysuccessful launches have been those of the

second IKONOS (in September 1999),EROS-A1 (in December 2000) and

QuickBird-2 (in October 2001).

are only 1k x 1k pixels in size. The TMSat builtby SSTL for Thailand uses three verticallypointing cameras, each equipped with adifferent colour filter, mounted side-by-sidewith parallel optical axes in the samearrangement as that used on the Z/I Imagingand the various American airborne framecameras discussed earlier. Again simple andinexpensive 1k x 1k pixel CCD areal arraysfrom Kodak have been used as the imagingsensors in each of these cameras.

Spaceborne line scannersLine scanners are the classical type ofimaging device used for remote sensing fromspace. Over the last 30 years, there has been asteady progression in line scannersproducing digital image data with ever smallerground pixel sizes – from 80 metres (MSS),through 30 metres (TM), 10 metres (SPOT),six metres (IRS-1C/D) and one metre(IKONOS) to 0.6 metre (QuickBird) – in thepan imagery available for civilian use.

Very low resolution scannersHowever, it is interesting to note thecontinuing strong interest by the scientificcommunity in quite low-resolution linescanners that provide synoptic views andwide-area coverage, principally forenvironmental monitoring and the mapping ofland cover and vegetation on a regional orglobal scale. Thus there is a surprisingly largeuse of the low-cost AVHRR (Advanced VeryHigh Resolution Radiometer) imagery with itscoarse 1.1 kilometre ground pixel sizecombined with a swath width of 2,400kilometres. It comprises 4-band images in thevisible (VIS), near infrared (NIR), medium-wave infrared (MWIR) and long-wave infrared(LWIR) regions respectively. Often mapsshowing Normalised Difference VegetationIndex (NDVI) values over extensive(continent-wide) areas are derived from theseimages. Into the same category falls the datafrom the so-called Vegetation imagermounted on SPOT-4, again having a one-kilometre ground pixel size and a 2,250-kilometre swath, but giving coverage in theblue, red, NIR and short-wave infrared(SWIR) spectral bands. This imager has alsoformed part of the instrumental package onthe SPOT-5 satellite that was launched in May.

Falling into the same general category isthe SeaWIFS imagery with 1.1 kilometreground pixel size and 2,800-kilometre swathprovided by the OrbView-2 satellite. Finally,there is the Along Track Scanning Radiometer(ATSR) that has been operated on board the

ERS-1 and ERS-2 satellites. A more advancedversion (AATSR) has been mounted on thenew Envisat (figure 1). This produces fourimages in the SWIR, MWIR and LWIR (2)bands respectively having a one-kilometreground pixel size over a 500-kilometre swath.These infrared images are supplemented bythree other images recorded in the green, redand NIR parts of the spectrum, principally forvegetation studies.

Low resolution scannersNext we come to the group of linescanimagers that give an improved groundresolution, but with a narrower swath width.These include the Indian WiFS (Wide FieldScanner) with its 188-metre pixel/804-kilometre swath combination and the RussianRESURS-01 with its 160-metre pixel/600-kilometre swath.

Into the same group falls the MediumResolution Imaging Spectrometer (MERIS)that is now being operated from the newEnvisat. This provides images in 15 spectralbands with a 300-metre ground pixel over aswath width of 1,150 kilometres. However,much of the current interest in this sector isfocused on the imagery being acquired by theMODIS (Moderate Resolution ImagingSpectrometer) scanners that are mounted onNASA’s Terra and Aqua satellites. These give36 spectral bands ranging over the whole ofthe optical spectrum (λ = 0.4 to 14µm) withground pixel values varying from 250 metresto one kilometre over a swath width of 2,330kilometres. The second MODIS instrument ismounted on the Aqua satellite, which waslaunched in May. The MODIS image data,

together with that from AVHRR and SeaWIFS,is being taken down routinely for much ofEurope by the ground receiving station at theUniversity of Dundee in Scotland, which issupported by NERC (Natural EnvironmentResearch Council).

Medium resolution scanners This is the original sector of civilian satelliteremote sensing. With the recent withdrawalfrom service of Landsat-5, only Landsat-7 ofthis classic series remains in operation. ItsETM+ line scanner provides continuous stripsof pan imagery with a 15-metre ground pixelsize. This is combined with its multispectralTM scanner providing 30-metre ground pixelimages in seven spectral bands; and it stillretains its original wide 185 x 185 kilometresareal coverage. This attractive specification,combined with its very low pricing, meansthat it retains a very prominent, if not theleading, position in this sector.

Certainly the SPOT-1 to SPOT-4 satelliteswith their 10-metre pan and 20-metremultispectral imagery (but with only fourbands) do have an edge in terms of theirsomewhat higher ground resolution.However, the more limited ground coverage ofa single SPOT scene of 60 x 60 kilometresmeans that a minimum of nine SPOT scenesis needed to cover the area of a single L-7scene – which gives an enormous priceadvantage to the L-7 imagery.

The IRS-1C/D imagery continues to beacquired, but apparently only in smallquantities. For the future, both Spot Imagewith its newly launched SPOT-5 satellite andISRO with its forthcoming Cartosat-1 – which

Figure 1: The Advanced Along Track Scanning Radiometer (AATSR) makes use of an unusual conical scanning system whichhas its rotation axis pointing ahead of the satellite in the along-track direction. This allows a double scan to be made acrossthe earth’s surface – the one at an angle of 47 degrees to the nadir ahead of the satellite, the other in the nadir direction.(Drawn by Mike Shand).

GI News • January/February 2003 45

IMAGERY

46 January/February 2003 • GI News

is scheduled to be orbited later this year –have decided to move upwards in theresolution stakes through the installation ofline scanners producing pan imagery in the2.5 to 5 metre ground pixel range.

Slightly further along the line, with ascheduled launch in November 2003, is theUK’s TOPSAT (figure 2). This enhancedmicro-satellite is being built by SSTL, while itspushbroom linescan imager is beingdesigned and built by the Rutherford AppletonLaboratory. TOPSAT will generate imagerywith ground pixel values of 2.5 metres (pan)and 5 metres (multispectral) over a 10 x 10-kilometre area. The project is being funded bythe British National Space Centre and theMinistry of Defence, though apparently moreas a technology demonstrator thanthe semi-commercial SPOT and IRSseries. DERA (now QinetiQ) andInfoterra are also partners in thisproject.

High resolution scannersThe area of high-resolutionpushbroom scanners has been theone where commercial activity hasbeen greatest, but it has also beenone of considerable disappointmentand large financial loss. In turn,EROS-A (in January 1998),IKONOS-1 (in April 1999),QuickBird-1 (in November 2000)and OrbView-4 (in September 2001)satellites have all been lost either atlaunch or immediately afterwards.

The only successful launcheshave been those of the secondIKONOS (in September 1999),EROS-A1 (in December 2000) andQuickBird-2 (in October 2001).Although they all fall within this group and allfeature off-nadir pointing and stereo-imagingcapabilities, they are otherwise quite differentin terms of their imaging performance.IKONOS generates pan and multispectralimagery with ground pixel sizes of one metreand four metres respectively. EROS-A1produces pan imagery only, having a 1.8-metre ground pixel size. QuickBird – now fullyoperational – produces 0.6-metre pan and2.5-metre multispectral images.

The EROS-A1 satellite continually changesthe orientation, or pointing direction, of itsscanner imager as it travels forward over thetarget area. This gives a longer dwell time forits linear array sensor, thus allowing morelight to reach it and so improving the qualityof the resulting image. The QuickBird images

have a greater swath width, in spite of theirhigher resolution. This is achieved by usingmultiple linear arrays butted together in thefocal plane. While OrbView-3 is scheduled forlaunch early in 2003 (if ORBIMAGE’s presentfinancial problems can be overcome), thefollow-on satellites to the present IKONOS,EROS and QuickBird satellites are notscheduled to appear until 2003 or 2004 at theearliest. The Indian Cartosat-2 with its one-metre ground pixel imagery is also due to belaunched around the same time.

Satellite stereo-imageryAlthough it has been possible with the SPOTand IRS satellites to generate stereo-imageryfor mapping and DEM (digital elevation

model) production from overlapping imagestaken cross-track from different orbits usingthe satellites’ off-nadir pointing facility, thisoften poses difficulties since the images maybe taken several months apart. In this case,due to seasonal changes in the terrainvegetation and hydrology, the ground area willlook quite different on the two overlappingimages, so making stereo-viewing andautomatic image matching impossible. TheGerman MOMS-2 imager, flown on both theSpace Shuttle and the MIR space stationduring the 1990s, demonstrated theadvantages of along-track stereo-imagerytaken near-simultaneously using the three-line imaging technique with fixed forward-,nadir- and backward-pointing linear arrays.

This success has given rise to further

imagers of this type. The first of these is theASTER imager built in Japan and mounted onNASA’s Terra satellite. This operationalsatellite features nadir- and forward-pointinglinear arrays with a 27.6-degree anglebetween them. The ground pixel size is 15metres and the swath width 60 kilometres.The users are very enthusiastic about thisimagery. The second instrument is the HRSimager that has been mounted on the newlylaunched SPOT-5 satellite (figure 3 – beingthe main, introductory picture on page 42).This features forward- and backward-pointingarrays with a ±20-degree angle between them and generates a rectangular 5 x 10metre ground pixel over a 120-kilometreswath width. The payload of the Indian

Cartosat-1 is also reported toinclude a similar forward/backward pointing stereo-imager, though not too much isknown about it. It is nowscheduled for launch in 2003. A fourth stereo-imager is theJapanese PRISM device with thefull three-line implementation ofthe technology. It will produceoverlapping images with a 2.5-metre ground pixel size overa 35-kilometre swath. This willoperate from the ALOS satellitewhich, after several delays and postponements, is nowscheduled to be launched in thesummer of 2004.

It will be most interesting tosee the results of the wide-areaDEMs that can be generatedfrom these dedicated opticalstereo-imagers in comparisonwith those derived from the

SRTM radar interferometric (InSAR) missionin terms of their respective geometricaccuracies.

Hyperspectral imageryThis area, like that of high-resolution opticalimaging satellites, has been one ofconsiderable disappointment and loss.NASA’s Lewis satellite, with two alternativetypes of imaging spectrometers on board,was lost shortly after its launch in August1997. OrbView-4, carrying a similar device,suffered a failed launch in September 2001.

But now there has been some success.The experimental Hyperion hyperspectralimager producing images with 220 bands anda 30-metre ground pixel size over a narrow7.5-kilometre wide swath has been

IMAGERY

If the whole spaceborne SAR scene has beenfull of delays, it is also full of innovative

proposals or projects, especially with regardto the acquisition of InSAR imagery. Part ofthe rescheduling of Radarsat-2 is due to theredesign and upgrading of the satellite so

that it can carry out the proposal for atandem mission with a Radarsat-3 satellite to

be produced some years later. In thismission, the two satellites would fly side-by-side in a closely controlled formation with anoptimised base line to allow the generationof InSAR data for the production of DEMs

(digital elevation models).

Figure 2: above (a): Artist’s impression of the TOPSAT satellite currently being constructed at the Surrey Space Centre in Guildford by SSTL. (Source: SSTL)

below (b): A cross-section diagram showing the folded optics of the pushbroom scanner being constructed by the Rutherford Appleton Laboratory for use on TOPSAT. (Source: RAL)

GI News • January/February 2003 47

IMAGERY

IMAGERY

48 January/February 2003 • GI News

operational on board NASA’s EO-1 satellitethat was launched in November 2000.Obviously MODIS with its 36 bands (thoughwith a 1.1 kilometre ground pixel size) canalso be considered to fall within this class.Following on from this, ESA’s small PROBAexperimental satellite built in Belgium wasorbited successfully in October 2001 using anIndian launcher. It features the CHRISimaging spectrometer built in the UK by SiraElectro-Optics (figure 4). The resultinghyperspectral images produce 19 differentspectral bands with a ground pixel size of 17metres over an area of 13.5 kilometres

square. The results of the hyperspectralimages being generated by all of theseimaging spectrometers are eagerly awaited bythe scientific community.

It is worth noting that Sira has also beennominated to build the multispectral imagerswith six spectral bands for the proposedGerman RapidEye satellites for which SSTLhas also been selected to build the mini-satellite platforms. The RapidEye images willhave a 6.5-metre ground pixel size over a 160-kilometre swath width, produced by a pair ofpushbroom scanners mounted side-by-sidein the satellite. The projected dates for the

launch of the four RapidEye satellites, whichare to be launched in two pairs, are 2003 and2004.

Satellite microwave imagersThe routine SAR imaging of the Earth fromspace has continued via the ERS-2 andRadarsat-1 satellites, both of which werelaunched and came into operation in 1995.

ERS-2 remains operational despite theloss of both its original and its back-up gyros.The satellite uses its horizon sensors toprovide signals to its reaction wheels tomaintain the correct pointing of its SAR forimage data collection. Radarsat-1 alsocontinues in operation. Although SARimagery is not to everyone’s taste, theCanadian Space Agency remains indefatigablein terms of promoting the use of Radarsatimagery and finding fresh applications for it.In this respect, the SAR radar mosaicscovering the whole of Australia and the wholeof Antarctica have been particularlynoteworthy.

Now attention is being focused on theirsuccessors. After several postponements, thelaunch of ESA’s Envisat with its AdvancedSAR (ASAR) took place on 1 March 2002. Thefirst images (with 30-metre ground pixel size)were released at the end of March. In the caseof Radarsat-2, there was a very strongdifference of opinion between the US andCanadian governments concerning thesecurity aspects of its higher resolution SARimagery – three-metre ground pixel in finebeam mode v. the eight-metre of Radarsat-1.After a good deal of public acrimony, thematter now seems to have been settled withcertain restrictions placed on thedissemination of the data. Radarsat-2 is nowscheduled to be launched in November 2003,instead of in 2002 as originally planned.

Yet another spaceborne SAR that hassuffered delays is the PALSAR, which is to bemounted on the Japanese ALOS satellite. Thisis an L-band SAR with its fine beam modeproducing imagery with a 10-metre groundpixel size. Again, as noted above, the launch ofALOS has been postponed from 2002 to2004.

The TerraSAR-X is a newly announcedGerman project to orbit an X-band SAR, beingfunded jointly by DLR – the German SpaceAgency – and Astrium under a public-privatepartnership. It is scheduled for launch in 2005.

Interferometric SARIf the whole spaceborne SAR scene has beenfull of delays, it is also full of innovative

Figure 4:

above (a): The CHRIS hyperspectral imager thatis mounted on the recently launched ESAPROBA satellite.

left (b): Diagram of the imaging spectroscopyprinciple used in the CHRIS imager. (Source:SIRA Electro-Optics)

While OrbView-3 is scheduled for launch early in 2003 (if ORBIMAGE’s present financial problems can be overcome),

the follow-on satellites to the present IKONOS, EROS andQuickBird satellites are not scheduled to appear until 2003

or 2004 at the earliest. The Indian Cartosat-2 with its one-metre ground pixel imagery is also due to be

launched around the same time.

Figure 5: The Radarsat-2/-3 Tandem Mission. (Source: Radarsat)

Receive-onlymicro-satellitesin "cartwheel"configuration

Radar illuminatore.g. ENVISAT, ALOS

TerraSTAR, Radarsat II

Radar footprinton Earth's surface

��������������������

Figure 6: The concept of the ‘interferometric cartwheel’ to be implemented as part of the Franco-Italian Pleiades project.(Drawn by M. Shand)

proposals or projects, especially with regardto the acquisition of InSAR imagery. Part ofthe rescheduling of Radarsat-2 is due to theredesign and upgrading of the satellite so thatit can carry out the proposal for a tandemmission with a Radarsat-3 satellite to beproduced some years later. In this mission,the two satellites would fly side-by-side in aclosely controlled formation with anoptimised base line to allow the generation ofInSAR data for the production of DEMs(figure 5).

Besides the Radarsat-2/3 tandem mission,another interesting proposal is that of an‘interferometric cartwheel’ which forms partof the Franco-Italian Pleiades project. Thebasic concept is based on the use of threepassive micro-satellites flying in a preciseformation behind a single active SAR satelliteand carrying receivers that would pick up thereflected signals from the ground illuminatedby the active SAR transmitter (figure 6). It hasbeen suggested that Envisat could in fact actas the active ‘illuminator’ satellite for theCartwheel – which is quite a thought! In themeantime, intensive processing workcontinues on the InSAR data collected by theShuttle Radar Topography Mission (SRTM).NASA-JPL (Jet Propulsion Laboratory), withthe help of Boeing Autometric, Intermap andthe Vexcel Corporation, is processing the C-band data, while DLR in Germany is mainlyprocessing the X-band data.

CONCLUSIONAs this article has tried to show, we are in anexciting and hectic period of development ofnew digital imagers, both in the optical andmicrowave areas. Already numerousexamples have been built or are being builtand will be coming into operational use soon.Still more imagers are being developed.

So the future is indeed bright, and endusers can look forward to acquiring a greatvariety of imagery from these many differentdevices.

However, some tough decisions aboutpricing will have to be made by the operatorsof these new airborne and spaceborneimagers. They certainly do not come cheaply. ...we are in an exciting and hectic period of development of

new digital imagers, both in the optical and microwave areas.Already numerous examples have been built or are being built

and will be coming into operational use soon. Still moreimagers are being developed. So the future is indeed bright,and end users can look forward to acquiring a great variety

of imagery from these many different devices.

Professor Gordon Petrie is with theDepartment of Geography & Topographic Science, University of Glasgow. Email: g.petrie@geog.gla.ac.uk

GI News • January/February 2003 49

IMAGERY