|
|
|
|
Copyright © 1988 CSA
Important notice: The CSA web site was re-designed in August of 2010. Some documents then available were out of date; so they were not included in the re-design and were not updated. This is one of those documents. Information about dates of posting and revision remains here, but there will be no revision of any kind after August, 2010.
This WWW document is based on the booklet, Surveying for Computer-Assisted Drafting and Design: Experiments in Three-Dimensional Techniques, which was written in 1988. This version of the text was revised in early 1996 before being posted here. Corrections, suggestions for additional information, and other recommendations for the next revision are welcome. This booklet has been written as an aid for those archaeologists who want to use computer-assisted drafting and design programs (CAD) to record archaeological material, and implicit throughout the booklet is the assumption that CAD is the preferred method for recording that material. The CSA booklet, Computer-Assisted Drafting and Design: New Technologies for Old Problems , sets forth the case for CAD and should be read before this booklet by those unfamiliar with CAD.
Although computer-assisted drafting and design programs have been devised for use by draftsmen and designers who arbitrarily determine dimensions as part of the design process, scholars who intend to draw existing three-dimensional objects with the aid of those programs can determine the dimensions of their material and enter those dimensions accurately into the CAD system with a minimum of time and labor.
Archaeologists are accustomed to taking careful dimensions in the field, but those dimensions have generally been taken with the aid of two-dimensional sketches and intended for two-dimensional media. Furthermore, the making of fully three-dimensional measurements in the field is difficult, bordering on the impossible in many cases, without the aid of sophisticated (i.e., expensive) equipment. As a natural result, it has often been necessary to be content with two dimensions only or with two on one view and two on another view such that all three are ultimately available but difficult to reconcile. In addition, assumptions rather than measurements often have sufficed for a missing third dimension - e.g., assumptions that surfaces are plane or that they are truly vertical or horizontal.
Although gathering three-dimensional data in the field is not an easy task - one which becomes more and more complex as the remains become more irregular - it is crucial to the successful use of computer-assisted drafting and design systems. Convenient, efficient, and accurate techniques for collecting this information must be available to the scholars using CAD if it is to become widely used.
Two techniques for solving this problem will be described and evaluated in this booklet. A third was originally planned but proved to be inadequate and will be described only briefly.
II. Surveying with Photography
III. Low-Tech Three-Dimensional Surveying
IV. Combining Total Stations and Photogrammetry
V. Comparisons and Conclusions
Table one - Total Station and Directly Measured Dimensions Compared
Table three - Comparison of Dimensions from Rollei System (with Standard 35 mm camera) and Directly Measured Dimensions
Table four - Comparison of Dimensions from Rollei System (with Medium Format Camera) and Directly Measured Dimensions
Table five - Comparison of Accuracy of Total Station and Rollei System
Table six - Speed Comparisons - Total Station versus Rollei System
In order to examine the survey methods under discussion here, a series of tests has been undertaken. There are some underlying assumptions which should be made clear, and one should also understand the manner in which the comparison tests were carried out.
The most important underlying assumption is that good measurements at an archaeological site should be related directly to a grid system, not to individual objects under study. That is, measurements should not be of objects in terms of length, width, or height; instead, all important points on the objects should be located in terms of three coordinates (normally one for the east-west dimension, one for the north-south dimension, and one for the elevation). All points in such a system are measured relative to the east-west and north-south axes and to a benchmark for elevation (corresponding to the x, y, and z axes in a Cartesian system). This is necessary to provide both accuracy and convenience.
Implicit are the assumptions that the realizable level of accuracy must be kept as high as possible and that accuracy is the primary criterion by which any method should be assessed, although there may be levels of accuracy beyond which one accomplishes nothing useful.
It has also been assumed that cost - both cost of purchasing equipment and the true cost of operating in the field, including labor time and use of specialized personnel - is a significant matter which should be addressed here.
* * * * *
In order to carry out the tests, an appropriate area was first chosen for measurement. The area was an unused doorway into Thomas Library at Bryn Mawr College. The building is built of both roughly cut stone (the basic walls) and finely cut stone (moldings, thresholds, door jambs, etc.) The area chosen includes both varieties of stone, as well as carefully carved moldings, arches, and windows. It presents many problems such as one might encounter in field work.
The area was drawn (not at scale) to provide the base sketch which would be used by any student of the area, regardless of the survey method used to follow up. That process was timed. The area was later surveyed with a highly automated surveying instrument; all parts of the survey process were also timed. The data obtained in the survey process was then fed into a CAD system, and a data file was created with the points as established by the surveying. A drawing was made from the data points. The data entry process was not timed, because it would be automated in a proper systematic approach in the field.
The second process began with the drawing also, but the drawing was only a safety device; photographs provide the real data, and measurements are taken directly from them with the aid of a computer program. The photographs were taken, developed, and printed. That was timed. They were then used to provide the coordinates for the same data points which had been surveyed, and the process was timed. The coordinates obtained were entered into a CAD data file; the process of entering the coordinates into the computer was not timed. A drawing was created from the data file. A number of actual measurements was taken (twice for purposes of comparison), and the measurements were compared with those obtained with the surveying methods. This was an attempt to determine accuracy, but it is difficult to know what should serve as a standard, and one must bear in mind the difference between accuracy and repeatability. Time was also compared, as was cost.
The total station is a combination theodolite(1) and electronic distance-measuring device; the two are normally attached to a data collector. Although the data collector is not technically part of the total station, it is crucial to the fully automatic functioning of the system.
The total station itself consists of three interconnected items. First, there is a sighting device through which one may view a target and aim precisely at that target; magnification, focusing capacity, and cross-hairs permit very accurate aiming of the instrument. Second, there are sensors which determine the vertical angle away from the horizontal and the horizontal angle from true North for the sighting device when it has been aimed at a target. These two parts perform the function of the theodolite (but with electronic assists). The third part of the total station is an electronic distance-measuring device which determines the distance between the station and the target at which the instrument is aimed (normally a prismatic reflector). This device uses timing calculations to determine the distance to the prism(2). (Used separately an electronic distance-measuring device is sometimes called an EDM.) The distance and angle measurements may then be stored in the data collector.
The data collector is a small calculator/computer which accepts electronic data directly from the total station; it may store the data about each point as received (distance and angles) or apply the appropriate trigonometric functions, directly calculate the three coordinates of each point, and then store the coordinates (an east-west dimension, a north-south dimension, and an elevation). In either case, it also holds various notes and identifiers so that the user will be able to relate the coordinates to the points which were measured, remember how they are related, and record other matters of importance to the individual surveyor. The information stored in the data collector may be either dumped directly to a computer via cable or telephone line, or it may be printed out (or both).
In actual practice, the user begins with a sketch of the area to be surveyed. During the surveying, each surveyed point is given a number on the sketch corresponding to that in the data collector, and careful notes are taken to make certain that the data can be properly interpreted. One person aims and activates the total station and its data collector. A helper must position the reflector. A third person may be needed take the notes; maximum efficiency may be obtained with two or three persons (in some long-distance work, four) depending upon a variety of factors, such as the number of points that are close together, the distance from the instrument to the points, and so on. In some cases, a walkie-talkie is desirable for communications between the person(s) at the instrument and the person(s) with the prism.
Positioning of the reflector presents problems under certain circumstances. The reflector (that is, the prism's optical point of reflection) is the only point which is directly measured by the total station; everything else is calculated from that dimension and the angles recorded at the time the dimension is taken. Therefore, the prism must be exactly in the position for which a measurement is sought. That is often impossible, but, fortunately, it is possible to apply the trigonometric functions and still obtain accurate results if the prism is offset from the point to be located, so long as the offset is specified and is truly vertical and/or directly in the line of sight from instrument to point(3). Therefore, data collectors have been designed to permit the operator to enter as a known quantity the dimension corresponding to the distance from the point for which the measurement is sought to the point of reflection and an indication of whether the offset is in the vertical direction or along the line of sight. Then the necessary adjustments may be made automatically, and the location of the object rather than the prism will be found.
When used by surveyors, the reflector is normally placed on a simple rod with a pointed tip; the rod is sometimes called a prism pole. Then an elevation offset is entered into the data collector corresponding to the height of the reflector above the bottom of the pole; sightings are made with the tip of the prism pole placed on the point to be surveyed. The pole is held in a strictly vertical orientation (with the aid of a built-in bubble level) so that the distance will be measured to a point directly above the surveyed point; of course the reflector is aligned so as to be roughly perpendicular to the prism pole (it can be swiveled for better alignment) in order to produce a good reflection back to the instrument.
When a surveyor is locating a point on a vertical surface, however, the prism pole may not work so easily, since the prism pole cannot be placed in a vertical plane with its tip on the point to be surveyed. There are ways to deal with this problem. One is to remove the prism from the pole and place the prism itself on the point, facing the instrument. However, not all prisms are designed for that, and a true reading can only be made if the prism housing has a tip designed to be placed on the survey point(4). There are small prisms, called mini prisms or peanut prisms, designed to work well in such circumstances, but even the small prisms are difficult to use, especially on irregular surfaces, since they cannot be positioned at acute angles relative to the walls.
A second approach is to use reflective material - tape or a bicycle reflector. Reflective tape works well at short distances; bicycle reflectors are problematic, since the actual reflective plane cannot be determined. So their accuracy is limited to a centimeter or so.
One total station manufacturer, Sokkia, has another way to overcome this problem. They have produced a two-legged prism pole with two prisms in freely rotating housings. One prism is at the end of one leg of the pole where the two legs meet; the other is half way down that same leg. If the tip of the leg with the two prisms is placed on the survey point, and the tip of the other leg positioned anywhere that will permit the operator to hold it steadily, the surveyor can take a reading of both prisms. Then the total station can calculate the position of the survey point. Of course, special software is required as well as the special prism pole.
The last of the possibilities for surveying vertical surfaces requires the prism to be mounted in line with the prism pole rather than at right angles to it. With the tip of the pole then held on the point to be surveyed, the reflector can be moved until it is directly on a line between that point and the total station. This requires that the instrument operator lock the viewing cross-hairs on the point to be surveyed before the reflector is put in place. Then the assistant must hold the tip of the prism pole on that spot and manipulate the pole so that the reflector is precisely in line with the cross-hairs while the tip stays on the proper spot. (This process could be carried out with a mini prism to make it simpler.)
It should be mentioned that there are survey instruments that can make readings with no prism at all. However, the readings, when accurate enough for archaeological or architectural history use, require good lighting conditions, relatively reflective material, and a very expensive instrument. These conditions are unreasonable; so such instruments are not yet to be recommended, though they are likely to get better and less expensive in the not-too-distant future.
Once the measurements have been taken and stored in the data collector, the data may be directly transferred to a CAD system, transferred to an intermediate computer file for calculation, or simply printed out. Various calculations may be performed on the data if that is appropriate. Programs exist for transferring the data directly to surveying packages and for making a variety of calculations.
The Experiment
The doorway at Thomas Library, Bryn Mawr College (described above), was surveyed with the total station. A sketch was prepared, and the total station was brought to the area, positioned, and its position obtained by sightings from known points (5). The points to be located were then surveyed, with the appropriate notes made in the process. Two surveyors from the Philadelphia Electric Co. volunteered their free time to perform the experiment. The author is very grateful to Walt Payne and Ed Davis for that assistance, without which the experiment would have been far more difficult to accomplish.
For this experiment, the prism was generally placed directly against the points which were to be surveyed without the use of any prism pole; in some cases, the prism was placed on the pole and a new offset used. The total station was aimed carefully at points to be surveyed; then the prism was placed on the point and the reflector brought into line so that the reading could be made. The depth of the prism housing was entered into the data collector so that the coordinates of the point, not the prism, would be calculated. The data collector directly calculated the three coordinates of each point after the sightings were made. One of the surveyors operated the instrument; one held and positioned the prism pole; the author noted the points on the sketch and indicated the points to be sighted.
For the experiment, the data for the points were printed out and manually transferred to a CAD system. A computer link would be used in practice if this were the survey method employed in the field; so the time required to enter the data was not included in the comparisons with the other experiment.
Only about 50 dimensions were made, because the equipment which was available did not include the miniature prism pole and prism which would have been needed to take more sightings on the rough stone wall.
The time required for this work was as follows(6):
| sketching | 20 min. |
| setting up and locating total station
(two persons, 15 min. each) total (2 x 15 min.) | 30 min. |
| taking 46 measurements
(three persons, 58 min. each) total (3 x 58 min.) | 174 min. |
| TOTAL | 224 min. (3:44 hrs.) |
The number of points measured without moving the total station could easily have been in the hundreds. That would have added some additional time for sketching, and the time required per point (in total worker hours) would remain at more than four minutes per sighting; working with two persons (the note taker also holding the prism) could reduce the total worker time per sighting, but working with the necessary care with two persons would probably not make a significant difference in the number of worker hours required.
Accuracy of total stations is very high, and they have been in use over a period of time long enough to justify confidence in their accuracy and reliability. In the experiments, the following dimensions were obtained from the total station (after the data had been put into the CAD system) (7) and from direct measurement. The figure in parentheses is the measurement in feet provided by the total station; the same measurement, converted to meters, follows. The largest discrepancy between the measured distance and that obtained from the total station was .005 m., and the average (for 30 measurements which could be compared) was .0015 m.; the standard deviation from the mean of the discrepancies was .0014 m.
| Object | Total Station Measurement | Direct Measurement | Discrepancy | |
|---|---|---|---|---|
| BLOCK ONE | ||||
| left side | (3.883 ft) | 1.184 m | 1.185 m | .001 m |
| right side | (3.880 ft) | 1.183 m | 1.184 m | .001 m |
| bottom | ( .300 ft) | .096 m | .097 m | .001 m |
| top | ( .331 ft) | .101 m | .102 m | .001 m |
| BLOCK TWO | ||||
| left side | (3.879 ft) | 1.182 m | 1.187 m | .005 m |
| right side | (3.879 ft) | 1.182 m | 1.187 m | .005 m |
| bottom | ( .128 ft) | .039 m | .040 m | .001 m |
| top | ( .142 ft) | .043 m | .041 m | .002 m |
| BlOCK THREE | ||||
| left side | ( .844 ft) | .257 m | .258 m | .001 m |
| right side | ( .844 ft) | .257 m | .256 m | .001 m |
| bottom | ( .162 ft) | .049 m | .051 m | .002 m |
| top | ( .159 ft) | .048 m | .050 m | .002 m |
| MORTAR BETWEEN BLOCKS THREE & FOUR | ||||
| left | ( .039 ft) | .012 m | .008 m | .004 m |
| right | ( .037 ft) | .011 m | .010 m | .001 m |
| BLOCK FOUR | ||||
| left side | (1.482 ft) | .452 m | .455 m | .003 m |
| right side | (1.484 ft) | .452 m | .453 m | .001 m |
| bottom | ( .161 ft) | .049 m | .050 m | .001 m |
| top | ( .161 ft) | .049 m | .050 m | .001 m |
| MORTAR BETWEEN BLOCKS FOUR & FIVE | ||||
| left | ( .021 ft) | .006 m | .005 m | .001 m |
| right | ( .018 ft) | .005 m | .005 m | --- |
| BLOCK FIVE | ||||
| left side | (1.485 ft) | .453 m | .453 m | --- |
| right side | (1.486 ft) | .453 m | .452 m | .001 m |
| bottom | ( .154 ft) | .047 m | .048 m | .001 m |
| top | ( .165 ft) | .050 m | .050 m | --- |
| EDGE OF MOLDING | ||||
| left | ( .196 ft) | .060 m | .061 m | .001 m |
| center | (2.681 ft) | .817 m | .818 m | .001 m |
| right | (2.632 ft) | .802 m | .802 m | --- |
| BLOCK SIX | ||||
| right side | (1.126 ft) | .343 m | .348 m | .005 m |
| bottom | ( .376 ft) | .115 m | .116 m | .001 m |
| top | ( .413 ft) | .126 m | .126 m | --- |
Measurements were taken twice by tape for comparison purposes. The average discrepancy between measurements for the two sets was less than .001 m., and the largest discrepancy was .006 m., the next largest .003. The standard deviation from the mean of the discrepancies was .0012 m.
It is impossible to put an exact price on the total station equipment. Manufacturers' prices vary substantially; so do their discount policies for academics. However, $7-20,000 would be required for a complete system with the best combination of price and discount; diferences arise from differing levels of accuracy of the theodolite and the EDM, as well as maximum range. (8) (Costs of maintenance cannot be predicted, but repairs should not be frequent, though regular maintenance may be desirable if the instrument is used heavily.)
Photogrammetry has long been used to aid in the measuring of archaeological and architectural material. Stereo pairs have also been used - especially underwater - to make possible good, accurate drawings. Determining accurate three-dimensional coordinates from stereo photographs, however, is a much more difficult task, requiring very expensive computer equipment and sophisticated viewing equipment. As a result, for general use, photogrammetry is either too limited or too expensive for field use when 3-D CAD drawings are the ultimate aim.
Rolleiflex, the German camera company, has attempted to overcome the cost limits of photogrammetry and has begun marketing a group of computer programs to make possible accurate determination of coordinates from photographs. The camera/lens used must be calibrated to provide accuracy, but the photographs do not need to be stereo pairs taken from known positions and with carefully controlled orientation. Indeed, the system works better when the photographs used are from quite different viewpoints so that the calculations of converging angles may be made from divergent points.(9)
This system requires a group of photographs of any area to be surveyed. (There must be some points which have been measured in the photographs; no more than ten such measurements are needed.) The photographs must be taken with a calibrated camera/lens combination. The best results are obtained, not surprisingly, with Rolleiflex cameras which have been manufactured for this work and have therefore been calibrated in advance (10). These cameras also have glass plates with precisely etched cross-marks in front of the film plane to assure film flatness and to mark the film at the time of exposure for increased accuracy.
Once exposed, the film must be normally processed and the appropriate negatives enlarged. Enlargements are arranged on a digitizer, and set-up information about each is fed into a linked computer (lens and calibration, known points, edges of the enlargement). As many as eight enlargements - of the same area but from different viewpoints - may be on the digitizer at the same time, and each point to be dimensioned must be visible on at least two (preferably three) of the enlargements. Then the computer can calculate the location of any individual point which can be located on the enlargements (11).
The ExperimentThe area - again the doorway at Thomas Library, Bryn Mawr College - was sketched for record-keeping purposes, and a number of photographs was taken. The camera used was a 35 mm. Rolleiflex model (lent to CSA for the purposes of this experiment); a 15 mm. lens was used. Photographs were taken from eight locations - some directly in front of the area under study and some from angles off to either side. Photos were taken from both close to the building and farther away so as to provide three different levels of available detail. From each location at least one photograph was taken from a low vantage point and one from a relatively high one. A total of 36 exposures was made. The photographs were taken in the morning, with the building in full sunlight in order to have strong shadows to assist in the picking of common points on more than one photograph. The positions and exposures were pre-planned so that the time required for the actual photography could be very short and the shadows, consequently, as nearly as possible in the same place from first to last photo (12). Careful records were kept to make certain that each photograph could be identified by camera location and angle.
Two wooden poles were carefully positioned and marked to provide known points for calibrating the system. A few dimensions were also taken for the same purpose.
The film was processed and the negatives printed by the author, not a photo lab. All enlargements were made on Ilford Multi-Speed paper; the film was Kodak's T-MAX 100 film. (Additional enlargements - on Kodak paper - were printed by a laboratory in California during the digitizing work.)
The enlargements were taken to the offices of the company that was then the distributor of the software system, Terra Metric, Inc. The author worked there with William Archibald, one of the principals of Terra Metric, to learn how to input the data. When he had become familiar with the system, the author repositioned the enlargements, set up the computer again, and located the key points for the experiment. The process was timed.
The resulting data file was printed out and taken back to CSA's offices for entry into the CAD system. The time required for the data entry process, which would be automated in a functioning system, was not recorded.
The time required for this work was as follows:
| sketching | 20 min. |
| making control measurements | 30 min. |
| setting up and taking photographs | 45 min. |
| developing film | 45 min. |
| printing enlargements | 30 min. |
| setting up system with photos on digitizer | 45 min. |
| digitizing points (203 points) | 62 min. |
| TOTAL | 277 min. (4:37 hrs.) |
These times may be somewhat deceptive. The photographic processes gain substantial efficiency as more photos are taken and developed at one time. (And additional experiments showed that only 12 photos as a maximum were needed.) Therefore, the times for developing and photographing would be reduced in the field. As it stands, however, this timing experience indicates that efficient use of the system would reduce the time per point to about one minute. Of course, the more points one records from a given set of photographs, the more efficient the process is. A selection of 500 points from these photographs would require only about three-quarters of a minute per point.
In the experiments, the following dimensions were obtained from the Rollei system (after the data had been put into the CAD system) and from direct measurement. The largest discrepancy between the measured distance and that obtained from the Rollei system. was .013 m. (not including the measurement of the left edge of the molding, about which, see note below), and the average (for 21 measurements) was .0044 m.; the standard deviation from the mean of the discrepancies was .003 m.
| Object | Rollei Measurement | Direct Measurement | Discrepancy |
|---|---|---|---|
| BLOCK ONE | |||
| left side | 1.179 m | 1.185 m | .006 m |
| right side | 1.176 m | 1.184 m | .008 m |
| bottom | .093 m | .097 m | .004 m |
| top | .096 m | .102 m | .006 m |
| BLOCK TWO | |||
| left side | 1.180 m | 1.187 m | .007 m |
| right side | 1.185 m | 1.187 m | .002 m |
| bottom | .038 m | .040 m | .002 m |
| top | .040 m | .041 m | .001 m |
| BLOCK THREE | |||
| left side | .258 m | .258 m | --- |
| right side(13) | .256 m | ||
| bottom(13) | .051 m | ||
| top | .047 m | .050 m | .003 m |
| MORTAR BETWEEN BLOCKS THREE & FOUR | |||
| left | .014 m | .008 m | .006 m |
| right(13) | .010 m | ||
| BLOCK FOUR | |||
| left side | .447 m | .455 m | .008 m |
| right side | .446 m | .453 m | .007 m |
| bottom | .050 m | .050 m | --- |
| top | .049 | .050 m | .001 m |
| MORTAR BETWEEN BLOCKS FOUR & FIVE | |||
| left | .009 m | .005 m | .004 m |
| right | .009 m | .005 m | .004 m |
| BLOCK FIVE | |||
| left side | .451 m | .453 m | .002 m |
| right side | .449 m | .452 m | .003 m |
| bottom | .050 m | .048 m | .002 m |
| top | .053 m | .050 m | .003 m |
| EDGE OF MOLDING | |||
| left(14) | .039 m | .061 m | .029 m |
| center | .811 m | .818 m | .007 m |
| right | .797 m | .802 m | .005 m |
| LOW UPRIGHT FACE | |||
| right side | .340 m | .348 m | .008 m |
| bottom | .103 m | .116 m | .013 m |
| top | .124 m | .126 m | .002 m |
The experiment was not conducted with the benefit of markers positioned on the wall in advance to serve as targets. Such markers are recommended to provide clear, easily identifiable measuring points; they would enhance accuracy significantly, making it easier to pick the control points which must be common to all the photographs. The use of markers would be advisable, when possible, for those points which are crucial, serve as control points, or may be difficult to select in photographs.
The Rollei programs include a module which permits the system to deal with photographs from an uncalibrated camera/lens. Another experiment was therefore conducted with a standard 35 mm. single-lens-reflex camera (a Minolta Maxxum with 50 mm. lens). The photographs were taken again, from various positions again; the negatives were developed and enlarged. (The camera lens was taped to keep the lens focusing mount from moving while the photographs were made; this makes it certain that the true focal length will remain constant.)
The results were again obtained and compared. The average discrepancy between the measured dimensions and those obtained with the standard 35 mm. camera was .005 m., and the standard deviation from the mean of the discrepancies was .004 m. The largest discrepancy was .014 m. (In this instance, the author did not personally enter the data; the photographs were sent to the distributor. Fewer data points were recorded; so there were only 24 dimensions to be compared.)
| Object | Rollei Measurement | Direct Measurement | Discrepancy |
|---|---|---|---|
| BLOCK ONE | |||
| left side | 1.172 m | 1.185 m | .013 m |
| right side | 1.170 m | 1.184 m | .014 m |
| bottom | .090 m | .097 m | .007 m |
| top | .101 m | .102 m | .001 m |
| BLOCK TWO | |||
| left side | 1.176 m | 1.187 m | .011 m |
| right side | 1.174 m | 1.187 m | .013 m |
| bottom | .035 m | .040 m | .005 m |
| top | .041 m | .041 m | |
| BLOCK THREE | |||
| left side | .253 m | .258 m | .005 m |
| right side | .251 m | .256 m | .005 m |
| bottom | .044 m | .051 m | .007 m |
| top | .044 m | .050 m | .006 m |
| MORTAR BETWEEN BLOCKS THREE & FOUR | |||
| left | .011 m | .008 m | .003 m |
| right | .013 m | .010 m | .003 m |
| BLOCK FOUR | |||
| left side | .449 m | .455 m | .008 m |
| right side | .446 m | .453 m | .007 m |
| bottom | .048 m | .050 m | .002 m |
| top | .046 m | .050 m | .001 m |
| MORTAR BETWEEN BLOCKS FOUR & FIVE | |||
| left | .009 m | .005 m | .004 m |
| right | .007 m | .005 m | .002 m |
| BLOCK FIVE | |||
| left side | .449 m | .453 m | .004 m |
| right side | .450 m | .452 m | .002 m |
| bottom | .045 m | .048 m | .003 m |
| top | .048 m | .050 m | .002 m |
Yet a third experiment was conducted - this time with an uncalibrated 2 1/4" square camera (a Bronica SQ with 80 mm. lens borrowed from Webb:Taylor, Inc., in Philadelphia for the experiment). The previous experiments had made the advantages of larger negatives clear; it seemed that bigger enlargements would speed the process and make greater precision possible at the same time. (The photographs were again sent to the distributor for data entry so that fewer data points were recorded.) In this particular instance, the larger enlargements meant that only two photographs could be placed on the digitizer at once; so the points could only be picked from two photos. Furthermore, the negatives were found not to be properly rectangular; so the initial orientation was not as precise as would be desired. Nonetheless, the results were excellent.
The results were again obtained and compared with the measurements taken by tape. The largest discrepancy between the dimensions derived from the computer system and those obtained by tape on the site was .005 m., and the average discrepancy was .0021 m. The standard deviation from the mean was .0014 m.
| Object | Rollei Measurement | Direct Measurement | Discrepancy |
|---|---|---|---|
| BLOCK ONE | |||
| left side | 1.186 m | 1.185 m | .001 m |
| right side | 1.185 m | 1.184 m | .001 m |
| bottom | .096 m | .097 m | .001 m |
| top | .097 m | .102 m | .005 m |
| BLOCK TWO | |||
| left side | 1.188 m | 1.187 m | .001 m |
| right side | 1.186 m | 1.187 m | .001 m |
| bottom | .040 m | .040 m | --- |
| top | .038 m | .041 m | .003 m |
| BLOCK THREE | |||
| left side | .254 m | .258 m | .004 m |
| right side | .254 m | .256 m | .002 m |
| bottom | .050 m | .051 m | .001 m |
| top | .049 m | .050 | .001 m |
| MORTAR BETWEEN BLOCKS THREE & FOUR | |||
| left | .012 m | .008 m | .004 m |
| right | .012 m | .010 | .002 m |
| BLOCK FOUR | |||
| left side | .452 m | .455 m | .003 m |
| right side | .452 m | .453 m | .001 m |
| bottom | .049 m | .050 m | .001 m |
| top | .048 m | .050 m | .002 m |
| MORTAR BETWEEN BLOCKS FOUR & FIVE | |||
| left | .009 m | .005 m | .004 m |
| right | .007 m | .005 m | .002 m |
| BLOCK FIVE | |||
| left side | .448 m | .453 m | .005 m |
| right side | .450 m | .452 m | .002 m |
| bottom | .050 m | .048 m | .002 m |
| top | .049 m | .050 m | .001 m |
The price of the Rolleiflex system varies, depending on the camera used. The basic computer system costs $15,000 (in 1989), but it requires the Rolleiflex camera, and the price of the 35 mm. Rolleiflex camera is about $7,000. In order to use a standard camera, additional computer software, which costs $7,000, is required (in 1989). The superiority of the larger negatives for this work make the extra software a better choice than the calibrated camera, and discounts for the software will reduce the price somewhat. Of course, the software can be used on data from many different cameras; so the cost can be spread over several users.
A recent Newsletter of the Institute of Nautical Archaeology contained information about a three-dimensional surveying system which could be used under water. Acoustical devices were used to determine the distances from three fixed and known points to an unknown fourth point. Using trigonometry and geometry, the precise position of the unknown point could then be calculated. Although there is no analogous system for use on land (total stations are simpler and less expensive), the idea that three dimensions from known points could be used to determine the coordinates of a fourth point prompted some study of the possibilities of using a simple tape-measure-based surveying method for archaeological work.
A simple system was attempted. Three fixed points were established as the starting points for measurements, and the points were carefully related to one another to simplify the mathematics. For each point to be surveyed three dimensions were taken, one from each fixed point. The dimensions were noted and put into the a spreadsheet for calculating and the crucial dimensions figured. The results were so inaccurate that a fuller experiment was not attempted.(15)
Although this system should succeed in theory, the need for measuring accuracy is very high, and the demands on the system are such that the necessary level of accuracy is not easily obtained.
A from of photogrammetry, more correctly called photo rectification, involves surveying with only one photograph. In this case, however, the material to be surveyed must be on a single, flat plane.
A photograph of such a plane - a wall or floor, for instance - can be used to determined the positions of points on the plane. One needs to survey four points on the plane so that they may be used to determine the angles from which the photograph was taken. A computer, with the four points specified and located, can calculate the position of any other point on the plane.
Combining total station data with this photo rectification is an ideal way to obtain large quantities of survey information at reasonable cost. A wall can be surveyed, assuming it is flat, by taking a photograph and making total station readings of four points on the wall. Then the photograph can be rectified in a CAD program (AutoCAD, for instance) or a single-photo photogrammetry program. Any points on the wall can then be added.
The points located along the wall, to continue the example, will not be as accurate as if they had been survey in three dimensions, but they will be more accurate than if they had been simple measured with crossing tapes, as is commonly done. In either case, there is an assumption of flatness that is invalid, but such an assumption is reasonable in the sense that it allows good data to be obtained at reasonable cost.
This method was used in Pompeii in 1994, and it worked very well. Sections of walls were surveyed, and the walls were photographed. We had a few problems caused by strong sun and deep shadow in the same photo, but, by and large, good information could be obtained in a short time. If a total station is available for a project, this combination of the total station and photogrammetry is an ideal method when large, relatively flat surfaces need to be carefully studied. Even without a total station, photo rectification can be used, but the base survey data must then be supplied by standard survey or by some other means.
There are several scales upon which to compare these approaches to three-dimensional surveying - accuracy, ease of operation, speed, expense, personnel required, potential for error - and there are some intangibles as well.
Accuracy
Accuracy is the first and most important criterion for making a judgment. It is the sine qua non of the system. For total stations and three-dimensional photogrammetry there has been a direct check back to the actual stones. Therefore, the systems may be compared to a standard and to one another. One must be certain as well to compare the results to one's true needs. Millimeter accuracy may be a worthwhile goal, but many archaeologists would consider such accuracy to be unnecessary, if not illusory.
A comparison of the levels of accuracy of the total station and the Rollei photogrammetric system shows that the total station is more accurate. The difference, however, may well be a difference of no consequence for field work, since even the Rollei photogrammetric system with a standard 35 mm. camera can yield dimensions well within a centimeter of the measured dimensions.
| Device | Discrepancies between measured and obtained dimensions | ||
|---|---|---|---|
| Maximum | Mean | Standard deviation | |
| Total Station | .005 m | .0015 m | .001 m |
| Rollei System 35 mm. semi-metric camera |
.013 m | .0044 m | .003 m |
| 35 mm. standard-camera | .014 m | .0053 m | .004 m |
| medium-Format standard camera | .005 m | .0021 m | .001 m |
Ease of Operation
Ease of operation is an important consideration. It affects speed, likelihood of error, and personnel requirements. The simplest of these systems to use in the field is the photographic one. However, the laboratory portion of the photo system is considerably more demanding and would require training. Since the operator would probably be the operator of the CAD system, there would not be an inordinate amount of such added training required. The total station is more complex to operate in the field, but, of course, there is no laboratory phase for the total station system. It is very likely that an excavator would want the architect to operate the total station, which would reduce training time.
The differences here are not critical. The total station is the more demanding in the field, but its demands are not beyond reason. There is a difference, however, in the kind of personnel required by the two systems. The total station requires a level of training and/or background significantly greater than that required by the photogrammetric system, and errors would be difficult to correct if not found in the field. The photogrammetric system, on the other hand, requires that the photographer understand the needs of the system, and omitted photographs could cause data to be unobtainable.
There is another important advantage to the use of photogrammetry. The photographs, if properly preserved, provide a record to which one may return for additional survey information at any time in the future. Using a total station, on the other hand, means that, once one has left the site, the survey is finished.
Speed
A direct comparison may be made for speed of operation.
| total time and time per point (in hrs.:min.:sec.) required
to obtain coordinates of points from a single set-up or set of photographs - based on extrapolation from timing experiences noted above | ||||
|---|---|---|---|---|
| Device | 100 points | 500 points | ||
| total time | time per point | total time | time per point | |
| Total Station | 7:08:00 | 4:17 | 32:21:00 | 3:53 |
| Rollei System | 4:06:00 | 2:27 | 6:08:00 | :44 |
As is clear from the table, the Rollei system requires less time when a large number of measurements is sought, but there is an assumption that large numbers of data points may be wanted from one set of photographs or instrument vantage point. There are occasions when this may be a valid assumption, but in many excavations there may not be that many points in one area which require precise location.
This may be an important factor in another fashion. Speed may encourage the gathering of more data. Once the photographs have been oriented, one may tend to locate points which, were it more difficult, would have been less carefully located and which would not have been located with a total station because of the extra time required in the field.
Although no time recording was used, it is safe to say the using photo rectification for survey of flat surfaces would be the fastest method of obtaining data.
Expense
A fairly direct comparison may be made for expense, but there are some variables which cannot be controlled. The cost of the total station system is substantial - $7-12,000 (ignoring the higher-priced possibilities) for the total station itself (including the data collector) and some costs for links to the computer. Some of the high cost of this system is not in the price of the equipment, but in the operating time, since at least two persons are needed for all operations. No precise dollar value can be placed on that time factor, but one should examine the speed section with that in mind.
Adding photo rectification has essentially no additional cost, since it is built into some CAD systems.
The cost of the photo system is also substantial. The computer program costs $15,000 plus $7,000 (in 1989; I have no current prices) if one is to use a standard camera. In addition, one must have a digitizer, but that would be part of the CAD system anyway. At this time it is not clear what discounts would be available for academic users of the system.
An excavator or architectural historian might consider the possibility of taking photographs at a site and returning the photos to a service center for analysis. This might make it economical to use the photogrammetric system on a small site. In fact, the Rollei computer system may be used on any number of sites at the same time. Only the actual photographic equipment needs to be duplicated (cameras at each site, of course), while the computer system is used to analyze all the photographs at one central laboratory. Similarly, total stations may be rented as well as purchased for use on small sites, but, of course, one machine may only be in use at one place at a time.
Many excavators and architectural historians will react to the costs of these systems with shock and look no further. One must, however, bear in mind the true, long-term costs of these systems. They should bring substantial economies to the entire process by automating a great deal of the early work. When properly coupled with CAD programs, it should be possible for an scholar to have all the drawings and the basic catalogs finished within weeks of returning from an project if he has taken a computer along. Without the computer in the field, drawings should be available within a matter of a few months at most.
Personnel Required
As discussed above, for the photo system there are no needs for specialized extra personnel in the field, but the laboratory work will require some training for the digitizing of the photos. The total station, of course, requires trained operators, although there is no reason why other personnel cannot be trained to use the total station.
Potential for Error
The total station is so fully automated that the potential for error should be considered very low (16). The operator must carefully enter a few numbers, but there are relatively few opportunities for error. The most significant problems arise when determining the true position of the instrument. This requires training in surveying techniques and considerable care, as does setting up the instrument at the outset.
Errors should be extremely rare with the photo system. Careful recording of camera/lens combinations used is required, and a mistake could be damaging, but the error would be obvious in the operation of the program. Otherwise there is little possibility for error. More important, the original sources of the data - the photographs - are always available; so the data may always be reconstructed.
Intangibles
The single most attractive feature of the photo system is the permanence of its data source. So long as the necessary number of photographs is available, the potential for making a measurement remains. This is of great value.
It is the immediacy of the total station which makes it so desirable. One may obtain locations for points on the spot and without delay if the instrument is set up appropriately. Those points may also be input directly into the computer without requiring any copying of numbers.
The Nature of the Site
The nature of the building under study of the site being excavated has a substantial impact on the value and ease of use of each of these systems. For example, a site with a single high vantage point from which a total station may survey the entire site is an ideal one for the total station, and a complex site with many features inaccessible to the prism is ideal for a photogrammetric system.
On a more mundane level, the complete absence of electricity makes the total station less desirable because of the difficulties in keeping it charged. Furthermore, without a computer of some sort at the site, a total station is risky, since one must regularly move the data to a more permanent storage device. Of course, there are many economic and environmental concerns when one considers taking a computer into the field, and with a computer in the field the photogrammetric system can very nearly approach the immediacy of the total station.
It should be clear that the author has not made a choice between the total station and the photogrammetric system. This is not simply an attempt to appear unbiased. In fact, there is no obvious choice because there are conditions which may favor one or the other, and the individual excavator must chose on the basis of his own needs and conditions.
In general, the combination of a total station and photo rectification seems the most desirable, flexible, and economic system. However, when significant portions of the material do not lend themselves to photo rectification, three-dimensional photogrammetry is required.
In order to assist excavators in dealing with these choices, personnel from CSA are prepared to assist in the examination, evaluation, and selection of surveying systems/equipment, CAD systems, and related equipment or programs for use by other scholars. Training in the use of surveying systems, CAD, and related techniques will also be available from CSA. This booklet is part of CSA's assistance, but more individualized help is available on request. Inquiries should be directed to: Harrison Eiteljorg, II, Director, Center for the Study of Architecture, P.O. Box 60, Bryn Mawr, PA 19010 (email user-name nicke at (@) domain-name csanet.org; tel.: 484.612.5862).
It has been suggested that CSA offer as a service to excavators the photogrammetric analysis of photographs taken in the course of excavation. Interested excavators should contact CSA.
NOTE 1: A theodolite is a standard surveying implement which consists of a sighting device mounted on two scaled bases, one horizontal and one vertical. The operator sights a target and reads from the scales to determine the angles from the horizontal and from true North for his aimed direction. The target is normally a ranging rod which is held on the point to be surveyed and kept strictly vertical. The rod is marked with height measurements, permitting the operator to read the markings through his sighting device and to use them to determine relative elevations. Return to body of text.
NOTE 2: Since the distance measurement is dependent upon very precise measurements of time, there is a base limit on accuracy which is unrelated to the distance from the electronic distance-measuring device to the prism. The resulting margin of error, which amounts to one to five millimeters, depending on the is the minimum margin within which measurements will fall, regardless of the distance measured. Other factors will also affect accuracy when longer distances are involved. Return to body of text.
NOTE 3: If the prism lies directly on the line of sight from the instrument to the point, the operator need only enter the amount of the offset into the data collector (or the computer program which operates on the raw data), and the offset will be added to the measurement obtained with the EDM. Then the calculations begin with that secondary dimension. If the offset is in the vertical direction, that dimension may also be entered into the data collector. In that case, however, the calculations are performed and the x, y, and z coordinates of the prism are found; then the offset is subtracted from the z value to locate the point in question. (In theory, of course, any offset could be factored into the equations. In practice, however, only the line-of-sight and the vertical offsets will work in the field.) Return to body of text.
NOTE 4: The prism does not simply reflect the incoming beam back to the EDM. The path taken by the beam involves more than one reflective surface; so the total distance from EDM to prism is slightly less than the distance traveled by the infrared beam. It is possible to determine the relationships involved, but the housing for the prism is not designed to be placed on the survey point. It is designed to be screwed into the prism pole. A physical adjustment to the housing must be made to use the prism directly, or an offset must be entered into the data collector. Return to body of text.
NOTE 5: There was no attempt to find a true known point. A hypothetical location of a point was entered for a point of reference. As a result, the process closely resembled the process which would be used in an excavation after fixed points of reference had been surveyed at the very beginning of work. Return to body of text.
NOTE 6: Dr. David Romano of the University Museum, University of Pennsylvania, took a total station to Greece during the summer of 1988. Although he did not attempt to keep time records, he reported that his experience is not appreciably different. He did find more people to be desirable (a team of four), but he was working on more complex surveying tasks and large areas for at least part of the time.
During the summer of 1994 I used a total station in Pompeii, both with a prism pole and with reflective tape. Our work there was in a relatively confined area; so no communication problems arose, but we were working on standing architecture that was quite irregular. Timing was not recorded, and further efficiency can be expected as the work crew gains experience. However, I would estimate that the need to move ladders, take notes, discuss archaeological issues involved, and so on made the time required per measurement more like six or seven worker-minutes. Return to body of text.
NOTE 7: The points were entered into the CAD system (by typing coordinates) as the end points of lines. The CAD system was then queried for dimensions. As a check, calculations of dimensions were made directly from the base data, they agreed with the dimensions provided by the CAD system. Return to body of text.
NOTE 8: The higher end of the price range - above $12,000 or so - is unlikely to be required. Among the manufacturers of total stations are the following: Sokkia, Wild (Leica), Nikon, Kern, Pentax, and Topcon. Return to body of text.
NOTE 9: Other companies also make such photogrammetry programs now, and the category has come to be called close-range photogrammetry. Return to body of text.
NOTE 10: The calibration is only accurate for the specific focusing position for which the calibration is performed; so one must use a lens calibrated at a reliably repeatable focusing point (infinity is the simplest and best choice). Some Rollei lenses are calibrated at four focusing positions, each marked with a detent on the focusing mechanism so the user can be sure to be using one of the predetermined focusing marks without being limited to focusing at infinity. Return to body of text.
NOTE 11: The most effective pointing device on the digitizer is a so-called puck, which has a magnifier and cross-hairs to assist in precisely locating points and a button to indicate that the current location of the puck is to be marked. Return to body of text.
NOTE 12: Those shadows are crucial to seeing the same point on different photographs; therefore, one must take the photos rather quickly, lest the shadows move during the time between the first and last exposures. Return to body of text.
NOTE 13: The author neglected to obtain the location of a point which made it impossible to determine the three missing dimensions. Since the program was used in California, at the offices of the distributor, it was impractical to add the location of the missing point when its absence was noticed. Return to body of text.
NOTE 14: This measurement was made to a rounded corner and may be in error because of the difficulty of picking the same spot in the photos and in real life, when the stone may be examined. It has not been included in the calculations of the average or standard deviation discrepancy. Return to body of text.
NOTE 15: This initial experiment was carried out indoors in the CSA office to minimize extraneous problems at the outset. Return to body of text.
NOTE 16: If a total station is used without the data recorder, the possibilities for error rise significantly. In that case, survey points must be recorded by hand, introducing a virtual certainty that there will be occasional errors. Return to body of text.