Ramin Rahimi Dgafari
National Cartographic Centre of Iran (NCC),
P.O. Box 13185 – 1684 Tehran, Iran
phone + 9821 6000031-9, fax :+9821 6001971
e-mail: ncc [email protected]
Nowadays, There is much concern to use SPT Satellite data in topographic mapping specially for those areas where these data are the best solution (like flat desert area with a few features). To do so we should seriously consider to those factors which play some roles in effective use of SPOT data in topographic mapping. In this paper using SPOT images for making topographic maps at scale 1/50000 will be studied from different point of views. Including completeness (Image content) and interpretability; Moreover some accuracy standards, accuracy of SPOT data in topographic tasks, Stereo-measurements and methods to select necessary control points will also be evaluated.
Spot I was launched on Feb. 1986 and since then it has been transmitting high quality imagery of the earth back to ground stationsand it has been in a near polar circular orbit, it has 10m. panchromatic pixel size. off nadir viewing capacity, stereo-capability and ability to commission imagery for specific areas and therefore we can say SPOT has characteristics which should be a boon to the cartographic community. There are different cartographic applications for SPOT images for example topographic map inspection, map revision and compilation and generation of digital elevation data. Image quality and geometric accuracy of SPOT data are essential elements in cartographic applications. In this paper applications of ASPOT imagery for topographic mapping at scale 1:50000 will be evaluated.
SPOT data has been available since 1986. We should consider that mapping from SPOT is not simply a matter of tracing off detail from the imagery into the map and the geometry of the new image is both complex and dynamic. The SPOT HRV sensor has an array of 600 charged couple device (CCD) sensors which sample reflected radiation every few milli-seconds to create a digital image swath 10 m. long and 60 Km. wide. Each swath has a perspective but off nadir view and a complete image is built up swath by swath while the stellite rotates around its three axes. SPOT data are recorded by two push-broom scanners abroad the satellite. These scanners can operate in two separate modes: multi spectral (ms) or panchromatic (pan) with 20m. and 10 m. spatial resolution respectively and it should be noted that MS images provide relatively high spectral contrasts while the PAN images have much more subtle contrasts and trade off between them should be taken. Also SPOT sensors can record either vertical or oblique image data and the oblique data can be corrected to produce rectified images. SPOT has different processing levels for different applications. The most basic level is ISA, corrected for sensor radiometric calibration but with no geometric correction. Level 1B data is additionally corrected for known geometric distortions including earth rotation, earth curvature, sensor viewing angle and satellite attitude variations. Level 2 is rectified for a map projection system using ground control points and it does not take into account terrain relief distortions and level S is rectified to a reference scene for multidate studies. There are also some auxiliary data including scene information, sensor data and ephemeris data. There are limiting factors in SPOT data acquisition e.g. clouds, mist, dust… make it difficult to obtain data when and where we want and the lack of ease in identification requires more ground effort.
For making true decisions about using SPOT for topographic maps at scale 1:50000 we should consider to interpretability of SPOT images. Although all of the geographic information required for a 1:50000 topo map are not provided by SPOT but however most high priority map features can be accurately mapped from SPOT images and also field completion and verification are very important for classification and identification of features. Man features are not directly distinguishable on the imagery but their presence is discovered b the context of their surroundings. In general most features especially roads, railways and built up areas are considered easily detected (THIRL WALL, 89) and uniqueness of tone, color, size and/or shape of features are important controlling factors for identification and also other factors are locational setting of features and their interrelationship. In general we can say that line features are very easily detected roads, railways and rivers (THRIL WALL, 89).
The two types of SPOT image are complementary in qualities while MS images have a higher spectral content, PAN images have better spatial information and a balance between these two kinds of images should be taken for each specific project and many interpretation problems can be solved b using MS and PAN images together and it should also be noted that stereo-analysis is not necessary for extraction of geographic features but it assists in clarifying interpretation. SPOT images should be selected carefully for each specific mapping project and is preferable to use images from different seasons and if we can only use one image for mapping, a summer image is recommended, it should be noted that two important factors for interpreting details from SPOT imagery are the quality of imagery and the experience of the operator.
Under this title we are considering about image content which defined as the information on a specific set of features. Image content is dependent on many factors particularly it depends on spatial resolution of the image in relation to the size, shape and pattern of the features and also the spectral response of the features as they are expressed through tone or color and contrasts on the image. An experiment has been do in a test area in QUEBEC, CANADA (THIRLWALL, 89) and it has resulted that for both two kinds of SPOT image the information content is very good but the PAN images provide much details and higher spatial resolution which displayed better representation of size and shape than MS images but it does not mean that mapping results use both PAN images and we should use both PAN and MS images with suitable balance together. It should also be said that another test has resulted that the percentage of detail on the 1:10000 scale map plotted from the POT imagery is 92.1 percent and the same for the 1:50000 scale map is 82.3 percent and these figures don’t take account of errors of commission which would also need to be edited during field verification (GUGAN & DOWMAN, 88). IT should also be noted that SPOT imagery contains a lot of useful information which can be utilised for map revision purposes when image interpretation is less important than the recognition of areas of changes.
Now we want to evaluate the accuracy of SPOT data for mapping of scale 1:50000. When we are selecting a source of imagery for a particular mapping task, we should consider to several important points: what data are available, will they give sufficient accuracy and required information and is it an economic procedure? There are different sources of errors for SPOT data e.g. the satellite has perturbations i.e. rocks about it’s axis and varies in its’ attitude, height and velocity: Therefor a line which is really straight on the ground does not image as such and it appears as a wavy skewed line. Now we should consider to accuracy standards in more details. for example NATO standards require that 90% of plasmatic errors fall within 25m. for a 1:50000 scale map and 12.5 m. for 1:25000 map if they are to be considered as accurate (class A ) (FOX, 91). Using this criteria certainly SPOT panchromatic data could be used to produce positionally accurate planimetric maps at both these Scales although line maps from these cultural information. However none of the current SPOT sensors is able to produce images with sufficient resolution of quality to the revision of 1:50000 scale maps without the need for a considerable amount of filed completion work. In general we should say although SPOT images can be used as a source for the generation of contours and spot heights compatible with a 1:50000 scale map, so that 90% of heighting errors fall within 10 m., the method of acquiring stereo imagery using oblique across-track viewing on multiple passes is far from ideal. The main system influencing the heighting accuracy of stereo imagery is the mirror viewing angle which determines the base: height ratio. It has been indicated a test (GUGAN, DOWMAN 88) that the heighting accuracy of a stereo pair with the maximum B/H ratio would be better than 4m. and it is better than planimetric accuracy. Also it should be noted that control data can be taken from small sale maps with little degradation in the SPOT model accuracy; And also the detection of gross errors can be difficult if we consider to large number of parameters but if the stable four parameter solution is used (just orbital parameters), gross errors are easily detected. The accuracy and completeness evaluation indicates that SPOT is a potential source of imagery for 1:50000 and smaller scale mapping. Planimetric accuracy of >=m. and heitht accuracy of >=8m. are suitable for these scales. In general we can are that the geometric accuracy of SPOT data is adequate for map at scales 1:50000 and smaller; Therefore SPOT can be considered as a realistic alternative to very high altitude aerial photography for mapping at medium and small scales.
Geometric rectification of stellite images to a reference coordinate system demands control points with known coordinates in both coordinate systems and we can say that 30D positioning of SPOT imager can be divided into two different methods; using the analytical plotters and applying the bundle adjustment method or using the digital image in the CCT and applying image processing methods and both methods need precise ground coordinate and stereo image coordinate of CCPs. Therefore questions about quantity and distribution of control points must be resolved. The main criteria for a ground control points is first that it shall be easily and uniquely defined and second it shall be located with high precision. The choice of method for measurement of control points depends on the demands for economy, geometric accuracy and availability of existing ground measurements. An important fact for achieving correct ground controls with accurate coordinates is that their definition in the satellite image and on the ground sould be identical. The selection of control points depends on the planned accuracy distribution of points and the number of points and it also has to be decided by the availability of image coordinates. There are different methods for acquiring 3-D coordinates of GCPs : ground surveying., aerial photogrammetry and using maps. Ground surveying produces accurate coordinates but it is expensive and it is not possible to be applied to inaccessible areas and in photogrammetry method the adjustment is time consuming and it is also expensive and can not be used for inaccessible areas. Using maps is applicable to inaccessible areas and it is inexpensive and effective but in case of small scale maps problems can be anticipated when the image resolution is smaller than thee map accuracy and difficult in identifying objects from updated maps. It should also be noted that it is difficult to get image coordinates of selected GCPs form the small selected SPOT images. As result of analysis of acquisition method of GCPs ground surveying method proved to be the best and in case of using maps as sources of 3-D coordinate of GCP accuracy fell for height. We can also say that with better satellite sensor system performance the number of control points for one image can be reduced without loss of accuracy. In general it can be said that a good method for acquiring control points is ground surveying and usually the most cost – effective network is created from existing data and supplemented with new control.
In the SPOT stereo scene the morphology is visible in more detail that in a map at scale 1:100000 or 1:50000 so the terrain forms are clearly visible but other topographic features such as roads or settlements are not clearly visible. A test has shown [GRABMAIER, TULADHAR, VERSTAPPEN 1988] that contours from SPOT are appreciably consistent and much more detailed than those from the map and the morphologic fidelity is obviously better. The resulting isometric view has also shown much more realistically the roughness of the terrain. For the sum of all mentioned error influences a standard deviation of 40m. has been found in that test which indicates that the SPOT stereo measurements are more accurate than we can expect the topographic map to be. In general we can measurement from SPOT is so high but it is not a realistic source of information for other topographic details.
There are two primary problems in the application of stellite remote sensing as an operational method of map compilation namely a lack of continuous stereo cover and relatively low resolution will be solved b future satelitte systems and the third problem which exist due to clouds will be solved by radar sensors. The launch of several new radar and mapping sensors before the year 2000 should dvelop the potential of stellite imagery into a true mapping capabilit providing images which can be employed operationally for thematic and topographic mapping at small and medium scales. At the present SPOT images should be considered in relation to other image data; for example SPOT M image provide about the same spectral information as LANDSAT TM images while all SPOT images provide greater spatial detail. Trade-off must be made between the increased data obtainable from SPOT images and the lower cost of other satellite images. We saw that we can obtain a planimetric and height accuracy about m. and 8m. respectively which are suitable for topographic maps at scales 1:50000 and smaller: Therefore in general we can say that spot can be considered as a realistic alternative to very high altitude aerial photography for mapping at medium and small scales and there are also several viable cartographic applications which are map inspection and revision and digital elevation model generation.
- Canada centre of geomatics (CCG). 1988. “Topographic features/ data capture criteria” (Project notes).
- DOYLE. F.J. 1984 “Surveying and mapping with SPOT data” ITC journal 1984-4.
- FOX D.J. “Mapping from satellite remote sensing: past. present and future” The cartographic journal (1991) Vol. 28(1) pp. 55-58
- GRAMAIER K., TULADHAR A. M., VERSTAPPEN H. T. “Stereo mapping with SPOT” ITC journal (1988) Vol. 2pp. 149-154
- GUGAN. D.J., 1987 “Practical aspects of topographic mapping from SPOT imagery” Photogrammetric record. 12(69) : 349-355
- GIGAN D.J. DOWMAN L.J. ” Accuracy and completeness of topographic mapping from SPOT imager” Photogrammetric record (1988) 12(72) pp. 787-796
- HARTLEY, W.S. “Topographic mapping with SPOT-1 data : A practical approach by the ordnance survey, Photogrammetric record, 12(720, 1988, PP. 833-846
- MIKHAIL, E.M., F.C. PADERES. Aand J. A. FAGERMAN 1989″ Batch and on line evaluation. of stereo SPOT imagery : Proceedings ASPRS/ ACSM 1989 convention, Baltimore, Maryland, April 2-7, 1989, vil. 3, p. 31-40.
- RAGGAM, J., BUCHROITHNER, M.F. AND MANSBERGER, R.” Relief mapping using nonphotographic spaceborne imagery.” ISPRS journal of photogrammetry and remote sensing, 44(19890), PP. 21-36
- THIRLWALL S. L. “CISM journal ACSGC (1998) Vol. 43 (2) PP. 133-140
- Turner, .M., D.R. STAFFORD. 1987 “Operational revision of national topographic maps in Canada using landsat images” in ITC journal 1987-2.
- YEO, BOCK-MO, CHO, GI-SUNG, YOM, JAE-HONG, KWON, HYUN” Accuracy evaluation by GCP acquisition method in bundle adjustment for SPOT imagery” ASPRS Washington, 1991.