EGIS (1994), copyright EGIS Foundation.
Ralf Bill Stuttgart University Insitute for Photogrammetry Keplerstraße 11 D-70174 Stuttgart Email ralf.bill@ifp.uni-stuttgart.de
The understanding of complex algorithms implemented in today's GIS packages is absolutely necessary for the interpretation of the results of spatial data processing. With classical teaching techniques such as blackboard or overheads complex algorithms are not easy to explain. A better learning success may be achieved when students can experiment with algorithms by themselves. Interactive GIS-teachware is one way of supporting these types of experiments. The paper describes criteria for the implementation of GIS-teachware and shows results.
Computer-aided learning (CAL), computer-aided instruction (CAI), computer-based training (CBT), intelligent tutorial systems (ITS) etc. -- new trends in education are more and more based on computers getting integrated into education and training. Many problems being taught may be much easier to explain by using computer-based methods. Especially if interactivity and graphic visualisation is needed, computer-based methods do have advantages compared to classical education techniques.
The GIS market is tremendously growing. The inevitable accompaniment of such an extremely rapid growth of GIS has been a considerable shortage of skills at all levels of GIS use (J. Raper and N. Green 1992). Parallel to this a shortage of educators in the GIS-sector could be recognized. In the last years the education sector has noticed this and consequently offers a variety of courses at all different levels. Beside the classical university education continuos education and training during a professional career is needed. Thus, integrating computers into educational work may unburden the educator in certain areas. Of course, computer-based teaching will not replace the human being in GIS education.
GIS-technology is getting more and more complex. The understanding of the underlying algorithms is absolutely necessary and essential for the interpretation of the results. Teaching complex algorithms from computer geometry or graph theory, for instance, is not very easy. The teacher needs a strong background and experience and has to prepare a precise didactical concept. Thus, education in this area depends highly upon the qualification of the instructor. This dependency may be overcome by means of computers, so that a more objective way of teaching may be achieved. This paper discusses computer-based education techniques. Two examples illustrate a computer simulation approach for teaching complex algorithms.
[End Page 1051]
Still, classical teaching techniques and media such as blackboard, overheads and exercises are nowadays offered. They may succeed in many cases, but in some cases they can also fail. In higher education and professional training the concepts of mass-production - one teacher, many students will no longer be acceptable, because they are not able to react on the different background and intelligence of the students.
From a media-didactic point of view the visual presentation of information yields a faster perception and a more permanent memory compared to verbal information impartment. Thus, computer-based teaching techniques usually prefer visual aids to support teaching.
Teachware in general identifies the methods and means for education and training in the spectrum of data processing. There are different methods in computer-based teaching (W. Sacher, 1990):
* Tutorial programs are imitating a standard lesson including success control.
* Exercise programs are used to strengthen existing knowledge
* Computer simulation is used to gain experience in precustomized and modelled situations.
* Computer learning games facilitate the access to the lesson.
* Programming problems manually allow students to gain experience by trial and error approaches. * Toolboxes are employed for solving specific tasks.
These various methods may be implemented with today's available techniques such as authoring systems and learning programs, both being a platform for the teacher to develop a lesson without any knowledge of low-level programming.
Computer-based teaching and all other forms of teaching have its advantages and disadvantages, which should be discussed in the following (M. Sacher, 1990, K. Gotz and P. Hafner, 1991, G. Zimmer, 1991). The advantages are to be seen in the objectivity of the computer and the standardisation of learning contents and units. A further advantage lies in the individualisation: the student determines the speed and the sequences to study; it is his own responsibility; the system gives an immediate feedback and allows interactivity. Computer-based teaching may be economically efficient when being decentralized and demand driven. The integration of new media and animation effects may push the motivation and activate different senses.
On the other hand, there may be also disadvantages which can be seen in a fictitious dialogue between human and machine and in the hidden and secret syllabus. The education may be reduced to pure information procurement in a virtual world. The computer does not understand; its reactions are predetermined. Developing computer-based teachware is not always efficient, the effort for updating is often underestimated. Hardware and software equipment should be available in the standard environment, i.e. the computer laboratory or at the student's home. The didactical concept, and not the number of media, the colourfull human interfaces and the gags implemented decide on the success of a teachware.
Research investigations (W. Sacher, 1990) have shown that computer-based education may result in a more open-minded and more positive thinking about computer technology. This is especially
[End Page 1052]
important for GIS which heavily depends on a positive attitude of the user, who often has to work with poor and stupid GIS packages. With computer-based teaching techniques one may gain a third of education time; bad students profit more by using computers than good students; the motivation increases. On the other hand, things taught with computers are faster forgotten and are less transferable; students learn images but may not understand the context. This mainly depends on the software and tutorials itself. Good teachware is still missing and is not able to follow the fast hardware development process. Establishing teachware is a hard job; to prepare a one-hour-lesson may take between 100 and 400 hours preparation time. It may be profitable especially for topics which are closely related to computers and which will stay stable for a certain period of time. This is valid for GIS in general and for algorithms in a GIS in particular, which leads us to the idea of developping GIS-teachware for algorithms.
GIS-demo-software are available on the market (e.g. SPANS, ERDAS, ARCDEMO). Some GIS vendors are also delivering their GIS software together with tutorials (e.g. IDRISI or MGE). These tutorials include documentation, instructions and prepared data sets. The user is guided through the system solving a practical exercise. Such tools are supporting the access to proprietary GIS-software. They usually do not teach the general GIS principles, but users understand the operation and behaviour of a special system faster. Vendors tutors and demonstrators may be helpfull following after a course on GIS principles, but they are not able to replace it.
At certain universities tutorials were developped to be embedded into classical education. GISTutor is one of these (J. Raper and N. Green, 1992). Here, the idea is to extend the range of teaching tools available and to improve their quality. Topics which can be demonstrated with a computer are explained with this medium. These tutorials make use of recent developments in computer science such as hypermedia systems, video integration and animation. GISTutor includes stepped animation sequences to illustrate, for example, line thinning with the Douglas-Peucker algorithm. In our teachware approach, which is described in the following chapter, we concentrate especially on this type of animation and visualisation. The embedded algorithms are to be programmed manually.
The chosen programming tools are Gnu C compiler (or others) and SUIT as user interface, both being public domain software. Thus, the GIS teachware packages are portable. SUIT runs on PC under DOS or Windows as well as on workstations under UNIX and OSF-Motif. The software may be shared with others, who want to expand it for further algorithms.
SUIT (M. J. Conway, 1992) establishes a graphical human interface on top of standard window management systems such as Windows, X-Windows or Macintosh. The software is easy to learn and easy to use especially if one compares it to the alternative, namely to program with the standard window system itself. SUIT was developped at the University of Virginia from 1990 to 1992. SUIT applications create objects (widgets), which can then be displayed, The properties of the objects could be manipulated via source code or interactive during the application. Typical objects are buttons, menus, scrollable lists, text editors etc. SUIT supports hierarchies between objects, complex objects and the inheritance between objects, i.e. it is partially object-oriented. SUIT includes a graphics package to generate graphical display for both, vector and raster data. The SUIT event handling is linking the users interaction with objects and their underlying actions.
We follow a computer simulation approach for the described teachware. Algorithms are studied,
[End Page 1053]
restructured in small units and programmed. Strong emphasis is put on the stepwise decomposition taking into consideration the parallel visualization step by step. The algorithm is documented in text form on screen parallel to the execution and visualisation. The teachware packages themselves are self-explanatory and very easy to use. The programs do not include learning control, i.e. the educator is not replaced with this tool. The teacher may demonstrate the software in the course, the software may also be passed to a student's PC at home for playing with it. The learning control takes place during the standard lessons and examinations. In our courses we implement the software on a notebook PC, which we connect with an overhead panel and demonstrate the software during the lesson. The steps and the visualized results are then discussed. One can compare and discuss the differences, again visually prepared. One may also create special examples to demonstrate extreme cases.
Different algorithms for triangulation and raster-vector-conversion are implemented in a user-friendly manner. The teachware developped supports a stepwise illustration of the background and behaviour of the implemented algorithms without any additional guidance. Differences in the results because of different assumptions of various algorithms are visualized and discussed. Thus, the student will fairly easy understand complex algorithms and will get a feeling for the mathematical background and its impact on spatial phenomena. Examples will be demonstrated in the presentation.
Triangulation is used in GIS for digital terrain models, for connectivity analysis and, in the case of Voronoi diagrams as a solution for the dual problem of a Delaunay triangulation, a neighbourhood analysis. Several algorithmic proposals and implementations exist. The mathematical and geometrical background of these algorithms is not easy to explain with standard education techniques. The results of the algorithms differ because of their different assumptions. In our teachware approach we want to explain the algorithms and we want to discuss the results and differences between the procedures. Three algorithms were selected for the implementation: Brute-force-method, radial sweep algorithm and Brassel-Reif-algorithm. The first two methods both try to set up a triangulation with a minimum sum of edge length between the points distributed in the plane. Still, the target functions differ: Brute force algorithm establishes a real minimum weight triangulation, whereas radial sweep algorithm creates a minimum weight triangulation for all combinations of four points in the point distribution. The Brassel-Reif-algorithm realizes a so-called Delaunay triangulation and a Voronoi tesselation in one step, i.e. the result could be used for different purposes such as terrain models and neighbourhood analysis.
The user of the package has a limited interactivity implemented in the software. He or she may run the program in steps, which are explaining and visualizing the algorithm step by step. On the other hand one may run the program following only the major steps and looking at the result. Different data sets from small to medium size are given, leading to an understanding of performance and to more realistic expectations when using commercial GIS packages.
Figure 1 is illustrating the human interface. Here the radial sweep algorithms is separated in three steps: the initial sweep, the creation of a convex hull and the final result after edge switching.
[End Page 1054]
[Insert Fig. 1]
[End Page 1055]
Raster-vector conversion is still an unsolved problem. Conversion software may achieve acceptable results for specific maps (linear phenomena in large scale), but for other maps (heterogeneous phenomena in medium and small scale) one may find examples where the conversion will fail all the time. Again there are a lot of different algorithms published with a variety of different assumptions behind them, which makes it fairly difficult for students to understand the principles and the problems. Stepwise animation combined with visual presentation makes it easier to follow the ideas and interprete the results. In this case a sequential approach was implemented illustrating all the individual steps to be carried out for the raster-vector conversion: read an image, clip and scale it, binarising the image, distance transformation, topologic skeletonizing, line extraction and polygon creation. All steps are visualized side by side (figure 2). One may repeat a step with a different parameter setting, a different type of neighbourhood or another distance function selected. This allows to discuss the results immediately. The student can start playing with the system and can conquer the behaviour of the algorithms without any mathematical background. The storage space needed for the raster matrices is very high, so that this teachware package works only for smaller examples. The package itself is designed as a toolbox which may be expanded for further algorithms. The following example is illustrating the human interface and the results of a raster-vector conversion.
[Insert Fig. 2]
Conway, M.J. (1992), Simple User Interface Toolkit (SUIT). Reference Manual, Version 2.3, University of Virginia.
[End Page 1056]
Cramer, M. (1993), Implementation von Raster-Vektor-Konvertierungsbausteinen als Basis fur eine GIS-Teachware. Diploma work at Stuttgart University (unpublished).
Gotz, K., Hafner, P. (1991), Computerunterstutztes Lernen in der Aus- und Weiterbildung. Deutscher Studien Verlag: Weinheim, 270 pages.
Raper, J., Green, N. (1992), Teaching the principles of GIS: Lessons from the GISTutor project. International Journal of Geographical Information Systems, Volume 6, No. 4, pp. 279-290.
Sacher, W. (1990), Computer und die Krise des Lernens. Klinkhardt Verlag: Bad Heilbrunn, 166 pages.
Trump, P. (1993), Untersuchung zur Erstellung von GIS-Teachware- Dargestellt am Beispiel der Dreiecksvermaschung. Seminar work at Stuttgart University (unpublished).
Zimmer, G. (Ed.,1991), Interaktive Medien fur die Aus- und Weiterbildung. Band 1 der Reihe 'Multimediales Lernen in der Berufsbildung'. BW Bildung und Wissen Verlag und Software GmbH: Nurnberg, 235 pages.
[End Page 1057]