G. E. BORMANN Wild Heerl~ruggLtd. CH-943.5 Heerbrugg, S~~;itzerlatid E. VOZIKIS Techt~iculUtli~ersit!/ Vienna A-1040 Viet~nu,Austria Transformation with Digitally -Controlled Differential Rectifiers

An unconventional application of modern orthophoto equipment.

INT-RODUCTION The navigation charts of central Europe ancl HE CONTENT of existing maps is very often the elsewhere are, however, on the Lambert confor- T basis for new cartographic products. In mal (angle-true)conical projections of the Interna- many cases the extracted information will be used tional (Hayford) or Bessel Ellipsoids with either for various different purposes. It happens quite one or two length-true parallels at scales of 1 : 1 frequently, however, that a geometric projection million or 1 : 500 000. The meridians are straight other than the original one is desired or needed. lines converging at the (North) Pole and parallels Cartographers are then hced with a problem are imaged as concentric circles about the (North) which could in the past only be solved in either a Pole. In this particular case the 1 : 500 000 scale crude and not very accurate procedure or in a very Aeronautical Chart ICAO 2253-B of Switzerland costly way. Meshwise re-projection with a rectifier (one length-true parallel, Bessel Ellipsoid) had andlor re-drawing was necessary. consequently to be transformed into a (special) Today digitally controlled instruments and sys- rectangular Cassini projection of 1 : 1 million tems, which were basically developed for the pro- scale.

Assr~~c'r:The geometrically correct tr(lr~sforr?lcltiot~of the entire mu))content from one curtogruphic projection to another is tlr~oltl problon. It curl be soloetl economicully with digitally-controllet1 dfirerlticll rectifiers st~cll(1s the Wild AVIOPLAN OR1, using udditioilul softr~tlrernotlttles. Esmnz))les illt~strtltethe potential of this method and sltorc; ulso the gretrt udoc~tltclgesof the color cupu- bility of the OR1 system. duction of orthophotographs, offer a new eco- The work was successfully carried out on the nomic solution of the problem. The Wild AVIO- AVIOPLAN OR1 in Heerbrugg using additional PLAN OR1, a differential rectifier, belongs to this and modified software originally developed at the category of equipment. Institute of Photogrammetry of the Technical Uni- versity in Vienna, Austria. The converging merid- ians became parallel and the circles were rectified to straight lines with high accuracy. The A practical case in the field of aviation naviga- entire map content was transfor~nedsimultaneous- tion, approximately two years ago, led to a closer ly into the new projection in its original colors. investigation of the new possibilities. In connec- Although the differences of the two cartographic tion with some of the data supplied by an inertial projections are obvious and considerable at the navigation system, a special navigation chart was map scale, even over such a small area as Switzer- needed very quickly. The specifications asked for land, another more spectacular example was se- a rectangular grid formed by the geographic me- lected for presentation in a publication where the ridians and parallels (with additional constraints). scale of reproduction is limited by the page size.

PHOTOGRAMMETRICENGINEERING AND REMOTESENSING, 0099-1 112/83/4!309-1317$02.25/0 Vol. 49, No. 9, September 1983, pp. 1317-1323. O 1983 American Society of Photogrammetry The choice fell on the physical map of Europe CENTER OF PROJECTION taken from the Swiss Atlas for High Schools, a creation of the well known Swiss cartographer Prof. Ed. Imhof.

The principle of orthophoto production is well known and the operation of the AVIOPLAN OR1 has been described in detail in several publica- tions (Kraus, 1976; Stewardson; 1976). We will therefore restrict the discussion of these items to some background remarks which seem to be nec- essary for the appreciation of the map transforma- tion method. In general, orthophotos cannot be obtained di- rectly with a camera. A subsequent process is nec- essary to remove the perspective distortions from, for instance, an aerial photograph. The perspec- tive effects are caused by central projection when the photographed terrain surface is not flat and/or the camera axis was not perpendicular to the datum plane. //// For the elimination of these effects it is neces- ORTHOPHOTO sary to resolve the content of the original photo- t graph into small area or line elements. They have s-l to be changed in scale and rotated as well as shift- FIG.1. Basic relations between central and or- ed into the correct geometric position before re- thogonal projections. composition to a new photograph, the orthophoto, can take place. Figure 1 illustrates the principle of taking an The OR1 applies the principle of differential (aerial) photograph. A regular, rectangular grid in rectification of an infinite number of line elements the (horizontal) datum plane (x,y) is assumed to of finite (selectable) length which are arranged in undergo vertical parallel projection onto an undu- adjacent profiles. The orthophotograph or rec- lating terrain surface (or a model of that surface). tified image is composed on a rotating film drum There it forms a spatial network of irregularly which is shifted in the direction of its axis at the shaped quadrangles. By central projection this end of each profile scan. The amount of shift corre- grid net is consequently deformed by the shape of sponds to the length of the line element, i.e., the the terrain and by the spatial orientation of the length of the selected slit mask through which the camera axis. photographic image has to pass. The production of an orthophoto corresponds to The input data to the OR1 process-computer are the inverse process. The irregular grid (u,a) has to image coordinates (u,~)supplied from magnetic be transformed into the regular rectangular grid tape. With reference to Figure 1, this means that (x,y) or, multiplied with a scale factor, into (X,Y) as the projection scan width S can be chosen be- is shown in Figure 2. This transformation is, how- tween 3 and 15 mm. The Y-interval (AY) in the ever, only possible when the shape of the terrain, rectified image is always 1 mm; however, it is lin- the orientation of the camera axis, the spatial posi- tion of the projection center, and the principal dis- tance of the camera are known in some digital fox-in. Normally, this is not the case u priori. The necessary information has to be obtained either from measurements in stereoplotters, or by com- putation, for instance, from an existing digital ter- rain model. The Wild AVIOPLAN OR1 performs the re- quired transformation. It is an instrument with op- tical projection. The optical devices such as a double zoom system for continuous change of I- U in; scale, a dove-prism for image rotation, and the ORIGINAL IMAGE RECTIFIED IMAGE (u,v) cross-slide of the photo-carrier for position- FIG.2. Grid transfer from aerial photograph to ortho- ing are servo-driven by a process computer. photo. MAP PROJECTION TRANSFORMATION

The principle is very simple. The "new" (trans- formed) projection shall be given in cartesian co- ordinates (x,y) as functions of h and latitude cp: i.e., x = fl (hcp), Y = f, (hcp). (1) The same shall be the case for the "old" (to be transformed) projection with the caltesian coordi- nates (u,v): i.e., u = h, (h,cp), o = h, (h,cp). (2) Next a cartesian grid (x,y) is defined in the "new" projection (Figure 4). With the inverse for- mulas of Equation 1, corresponding geographic coordinates (h,cp) are obtained: i.e.,

A = g, (x,Y), cp = g, (x,Y). (3) FIG.3. Wild AVIOPLAN OR1 with projection unit, process computer, magnetic tape station, and printer. By substitution of Equation 3 into Equation 2, it follows that

early interpolated to 0.01 mm by the process com- puter. In mathematical terms the OR1 performs a The result of this numerical transformation bilinear transformation for each grid mesh. (Equation 4) is illustrated in Figure 5. As was mentioned above these image coordi- The regular rectangular grid of the "new" (trans- nates have to be established first either by mea- formed) projection appears deformed and is super- surement or by computation in an external imposed on the "old" (to be transformed) map. computer. For the latter case a FORTRAN computer Consequently, a "rectification" of the grid of program SORA was developed at the Technical Figure 5 according to the orthophoto method will University at ViennaIAustria some years ago. lead to Figure 4 and will at the same time trans- (SORA stands for Software for Offline Rectification, form the entire map content into the new projec- Avioplan). (Otepka and Loitsch, 1976). tion. In analogy to the orthophotographic process, the GENERALMETHOD OF MAPTRANSFORMATION WITH "old" projection corresponds to the aerial photo- THE OR1 graph and the "new" one to the orthophoto. From the explanation above, the following pre- There are some additional points to be ob- sentation should very easily be understood. What served, however. A reproduction of the "old" map can be rectified by means of differential transfor- must be made on film (either negative or positive) mation can obviously also be distorted. The not larger than 9 by 9 inches. Even if this is done "distortion" may be a desired change of an original in a good reproduction camera, small projective if the appropriate mathematical nlodel can be for- displacements cannot be excluded. It is therefore mulated and is then imposed. The transformation advisable to measure the rectangular coordinates of a map with its whole content from one projec- (ur,u') of 10 to 15 well distributed points in the tion to another type of map projection is such a case. The mathematical model can be formulated, usually in rigorous terms. Reverting to Figure 2, the desired new map pro- jection is directly comparable with the "rectified" image (x,y). The map projection to be transformed stands in place of the "original" (u,u). Usually it cannot be the original map directly but a reduced transparent reproduction on film. This is in most cases necessary because the picture carrier of the OR1 is designed for aerial photography, while the format of the transformed map may be up to 750 mm bv 900 mm maximum. It is now furthermore assumed that the mathe- matical conversion formulas between both map projections are known. In particular, the coordi- nates of the "old" projection (u,u) must be ex- pressible in explicit form by the coordinates of the FIG.4. Cartesian grid in "new" (transformed) "new" (transformed) projection (x,y). projection. PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING, 1983

The following formulas were applied: Inverse Mercator, corresponding to Equation 3:

A = A,, + 180 m x T R cos cpc, 90 cp = 2 arc tan [tan (- + 45) . exp ---I my - 90 2 R cos cp(, Azimuthal (equal area), corresponding to Equa- tion 2:

fie07 Q sin (A - A,,) R I1 = .- v1 + sln Q sin Q,, + cos Q cos Q,, (A - A,,) "'

Iu fi[sin Q cos Q,, - cos Q sin Q,, (h - ,)I R v = . - FIG.5. Transformed grid in "old" (to be transformeuj Vl + sin Q sin Q,, + cos Q cos Q,, cos (h - A,,) "' projection. with the Eai-th taken as a sphere (angular values in degrees). repro film with a comparator and assign them a A Longitude principal distance. It can be an arbitrary value or, cp Latitude better, the principal distance (not length) of A(, Longitude of central point of azimuthal projec- the repro-camera if the results are to be inter- tion preted. cp,, Latitude of central point of aziinuthal projec- Following this, the length of the OR1 slit inask tion must be chosen. It determines the projection scan R Earth radius width and therefore also the number of points to m Scale factor (30 million) be computed for a specified format of the "new" map. A projection scan width (length of slit mask in A special version of the SORA program first com- the OR1) of 3 mm was chosen. In the case of the putes a spatial resection using the measured (ul,u') , this resulted in 73 projection and their corresponding (u,v) and transforn~sthe profiles each with a length of 408 mm. Conse- (x,y) grid into the (ul,v') plane. As a result the in- quently, nearly 30 000) (x,y) transformations were put data for the OR1 process-computer are ob- computed with the special SORA program in an ex- tained. ternal computer. The number has to be multipled The actual projection procedure in the OR1 is by a factor of 100 when the interpolation calcula- then identical with that for orthophoto production, tions of the OR1 process computer are also taken and the "new" (transformed) map is generated on into account. In other words, close to 3 million the film drum either in black and white or in color, pairs of (x,y) coordinates were computed to per- positive or negative, direct or mirror-reversed. form this transformation. For the spatial resection 20 (u1,v') coordinates were measured in the re- production. The Swiss atlas map, a ten-color print, is an The net time per projection amounted to ap- equal-area azimuthal projection with its central proximately 49 minutes as the drum rotation speed point at A,, = 20°E and A,, = 20°E and cp,, = 50"N at the periphery (film location) was ~urposelylim- (near Cracow, Poland). The scale is 1 : 15 million. ited to 10 mrnlsec for correct color exposure. It was decided to transform this map into a Mer- All projections were made twice, first on Kodak cator projection (conformal) with its length-true Vericolor IIL Color Negative Film, Process C41, parallel at cp, = 50°N. The curved meridians and and second on Kodak Ektachrome Duplicating parallels of the azimuthal projection were to be- Film 6121 (color postive), Process E6. come straight lines intersecting at right angles. For accuracy investigation 20 well distributed Large displacements in the north were to be ex- points were measured on the original 1 : 15 mil- pected. lion scale azimuthal map (paper sheet), on the l : The reproduction scales for both projections 30 million reduced scale azimuthal projection were chosen as 1 : 30 million. (film), and on the transformed 1 : 30 million Mer- For reasons of optimum equal-color reproduc- cator projection (film).Three modes of transforma- tion, both projections were made in the OR1 al- tions were then carried out: affine (six param- though the azimuthal map required only a eters), similarity (four parameters), and congruent reduction to half of the original scale. (three parameters). The measurements were in all

MAP PROJECTION TRANSFORMATION cases transformed against" the theoretical values at nature, they have proved that the method can be the nominal scale. applied very successfully and has a potential. The results showed in all three cases that there The main applications may perhaps not be the was practically no difference between the six-pa- transformation of complete maps but of certain rameter and four-parameter transformations. This features which cartographers may wish to transfer. indicates the absence of any affinity. The largest When entire maps are transformed, obviously errors were found for the original 1 : 15 million also the nomenclature will be changed in scale scale paper map. The positive scale factor of 0.3 and orientation. The principal applications will percent (original too small) indicates paper therefore most likely be in the revision of unanno- shrinkage or perhaps the assumption of a slightly tated maps. Geometrically, the possibilities seem different Earth radius. After reduction by this to be nearly unlimited. scale factor, the mean square error of a coordinate amounted to 0.19 mm at map scale. Surprisingly good were the results of the 1 : 30 million reduced azimuthal projection (Plate 1) as Kraus, K., 1976. Applications of a digitally-controlled well as of the transformed Mercator projection orthophoto instrument. Presented Paper, Commis- (Plate 2). With a fixed scale factor of 1, the mean sion IV, XI1 International Congress for Photogram- square error of a coordinate at map scale was metry, Helsinki. found to be 0.09 mm and 0.11 mm, respectively. Otepka, G. and J. Loitsch, 1976. A computer program for These excellent results speak not only for the digitally-controlled production of orthophotos. Pre- OR1 transformation process but also for the inter- sented Paper, Commission IV, XI1 International nal geometric quality of the original map, particu- Congress for Photogrammetry, Helsinki. larly if it is taken into account that the map was Stewardson, P. B., 1976. The Wild Avioplan OR1 designed and drafted many years ago when auto- orthophoto system. Presented Paper, Commission matic drawing tables were not available. 11, XI1 International Congress for Photogrammetry, Helsinki.

Although the map transformations carried out (Received 3 September 1982; accepted 15 December with the OR1 SORA system so far were of a test 1982; revised 18 March 1983)

Short Course Optical Science & Engineering

Doubletree Inn, Tucson, Arizona 9-20 January 1984

The purpose of the course is to acquaint both the specialist and the non-specialist engineer or scientist with the latest techniques in the design and engineering of optical systems. The course comprises 18 three-hour lectures; detailed notes will be supplied. The registration fee is $1200. The wide range of topics that will be covered includes geometrical and physical optics, optical system layout and design, Fourier methods, polarized light, radiometry, image quality, interferometry and optical testing, lasers, thin films, photodetectors, and visible and infrared systems. Address inquiries to Philip N. Slater Optical Systems & Engineering Short Courses Inc. P. 0. Box 18667 Tucson, AZ 85731 Tele. (602) 885-3798