|
The MR2 program from Rollei has been designed to perform photogrammetric analyses using a standard microcomputer running DOS. The system carries a "list" price of more than $20,000, not including a Rollei semi-metric camera and lens(es). Both 35 mm. and 2 1/4 square semi-metric cameras are available from Rollei to use with the computer program, and other manufacturers also make 2 1/4 square semi-metric cameras. It is also possible to use non-metric cameras; there may be some loss of accuracy, but there is a huge saving of money. (The semi-metric 35 mm. camera from Rollei, without lens, lists for $5967!)
A semi-metric camera (Rollei's or others) has a glass plate to hold the film flatter, and there are fiducial marks printed on the glass to create reference points on the negatives. In addition, the lens is calibrated so that the true center of the lens and any distortion can be specified. A non-metric camera has no glass plate, but it can be calibrated with the Rollei program (for lens center and distortion), though probably not as precisely as a camera calibrated at the factory. Since previous experiments showed that a larger negative was more important than the semi-metric features, CSA's experimentation has been with a 6 cm. by 7 cm. non-metric camera - a Mamiya RB67 - rather than a very expensive Rollei or comparable semi-metric camera. The camera was calibrated with the aid of the Rollei program.
To use the system, one takes a series of photographs of the area in question, taking them from divergent angles and noting the spots from which the photos are taken, the direction of aim, and the lens used. (The information about location need not be precise, but it provides a starting point for the program.)
The photographs are enlarged (the larger the better) and placed on a digitizing tablet, a kind of electronic drafting board.
One then gives the computer the information about each photo (the location information from the field notes) and digitizes common points on the photos. Some field data must be provided - coordinates of points, distances between points, and the like. There must be enough information to permit scaling and orientation, but the requirements are minimal. (This is one of the advantages of the Rollei system; it requires very few field measurements.) The system will then determine the precise locations of the camera as the photos were taken.
Once the locations of the camera have been defined, individual points can be located if they can be selected on more than one photograph. (As the number of photos on which a point can be located increases, the reliability of the point location increases.) As many as eight photographs can be on the digitizer at one time, providing many different angles and/or closeup, detailed views.
When a point has been located, the system provides an on-screen set of coordinates (x, y, and z). Those coordinates are put into a data file in sequence, but the on-screen information also includes an indication of accuracy based on the math model used. It is not an indication of real-world accuracy but only a measure of the ability of the math model to resolve the point as it was specified on the various photos.
Targets like the points on the roughly shaped stones of the Mycenaean wall used in this experiment are difficult to pinpoint; the precise point digitized in one image can be hard to find in another. On the other hand, corners of squared masonry are quite easy to specify.
The number, direction, and quality of the photographs are crucial to the operation of the program. The photos must be taken from divergent angles, and data points must be visible in at least two shots. (Three is much safer. The program will not always be able to resolve the location of a point located on only two photos.) Shadows must not obscure data points, and the lighting should not change significantly from one photo to another, since shadow lines may be used to help locate individual points.
When the surveying has been completed, the data file created can be altered to make an AutoCAD ® compatible file format, and the information may then be transferred directly into an AutoCAD drawing. (This process will be automated eventually to prevent errors and save time.)
Although the results, as reported above, were excellent in terms of efficiency and accuracy, using the Rollei system has been inexcusably difficult, something one might tolerate with inexpensive software but not software carrying this price tag. Those of us who have been using personal computers since the days of CP/M (before IBM entered the personal computer business) will remember programs like this one. The software itself works and produces good results, but it is utterly unreliable. MR2 fails altogether if the wrong version of DOS is used (without identifying the problem either with an error message or in the manual) or if some RAM is used by a TSR program (only during certain procedures and, of course, without an identifying error message). In one case, the program declared that the calibration data were incompatible with the photographs when, in fact, the only problem was that too many data points had been provided, something discovered by accident. One "advantage" of the program's numerous problems is that one ceases to blame oneself when things go wrong, assuming instead that the program has failed again.
The program also supports only a few specific digitizers, limiting one's ability to use equipment already owned. As if that weren't bad enough, one module of the latest update of the program will not operate with one of those specified digitizers, and another of the specified digitizers seemed not to function at all, though the digitizer was checked and working properly.
Data entry processes are, at times, unbelievably difficult, requiring the operator to use a menu on the digitizer, for instance, when the keyboard would be far quicker. Indeed, it is so time-consuming to enter and modify data that a text editor is necessary so that one may deal with files directly oneself, working around the Rollei program. (In fact, the basic set up requires the operator manually to modify a file to change serial ports and digitizers; so a text editor is truly necessary.) The best description of the user interface is "user-vicious." (Thanks to Warren Williams for the phrase.)
When errors are made, correction is invariably difficult. For example, there is no way to remove a bad point from the data file automatically; so one must interrupt the survey process to note the point number and coordinates for later removal from the data file with a text editor.
It is apparent that the program was written by several different persons. Although the menus have been made to look similar, the information screens can be quite different, and the data entry procedures vary in style if not awkwardness. Individual modules must load at odd times and the data files are different from module to module, making the manipulation of files even harder.
The manual is seriously incomplete, providing limited information in too many cases. All too predictably, it has no index. It also does not make clear what processes must be completed and what processes are optional, though that is an important matter. Nor is there any general information about the photographs - recommendations about divergence of angles, lighting, preparation, etc.
The serious user will, of course, learn about the quirks of the program; ultimately he will be able to use it well. The problems can be overcome, and most can be avoided with experience. The novice, on the other hand, will find the frustration to be significant, insurmountable without help, until some experience has been gained. Indeed, were it not for my prior experience with the program at the office of the American distributor, Terra Metric, Inc., Foster City, CA, and frequent telephone conversations with their personnel, operation of the system would not have been possible.
The negative aspects of this program would not be so regrettable were it not for the fact that the program does a job of great potential benefit for archaeologists and architectural historians. Using MR2, one can survey in detail with both speed and accuracy - and preserve the "raw material" of the survey for further study at a later date.
There are other photogrammetry programs nearing readiness, and one hopes that they will be more usable without being less accurate. Until they appear, however, the Rollei system seems to be the only choice, and it can only be recommended to the more computer sophisticated in the scholarly community. What a pity.
- H. Eiteljorg, II
For other Newsletter articles concerning the applications of photogrammetry and CAD modeling in archaeology and architectural history, consult the Subject index.
Table of Contents for the August, 1990 issue of the CSA Newsletter (Vol. III, no. 2)
Table of Contents for all CSA Newsletter issues on the Web