I Symbolization J,1 II e L~ "-"-:"u"'dll... Table I: The kinds of CGs and their classification, (where, t - a local neighbourhood topology)

Transcription

1 POSTER SESSIONS 484 REPRESENTATION OF THE GENERALIZED DATA STRUCTURES FOR MULTI-SCALE GIS M.O.Govorov Dept. of Cartography,' Siberian State Academy of Geodesy Plahotnogo 10, Novosibirsk, , Rssia Abstraj!t This paper considers the rgency of maintenance. of the generalized mlti-scale or mltiresoltion data strctres (MSDS or MDS) and.the principles of its independent implementation. Ths,. the different kinds of cartographic generalization (CG) with the point of view of their representations in the mlti-scale database (MSDB) are examined and classified. The spatial objects and properties of MSDS are presented. The generalized data strctres, which reslt from different kinds of CG, are implemented in the conceptal model. 1 Introdction Now the efforts of many researchers. are directing to the design of inlti-scale GIS (MSGIS) [2, 5-10]. These. researchers have chosen different directions.to the resolving this task. Bt the main. problem is remained as the problem of fonnalization and contextal comprehending of atomated CG. The fol:malizatio.n of CG implies the. searching for axioms and bilding p the fonnal langage of CG, which is based on those axioms. Last problem is closely connected with fonner one and sally is discssed dring the design MSGIS. MSGIS can be defined as. a spatial infonnation system, which spports all fnction of conventional GIS and additionally maintains MSDS or mlti-detailed database strctres (MDS). MSGIS mst be capable to represent entities of the world at the different levels of details and to spport modification across resoltion levels (6]. The existing commercial GIS's software, which maintains the spatial models at a fixed scale level, sally contains the following main blocks: I) analysis management systeni; 2) graphical management system; 3) data base management system. There.are at least three ways of creation MSG[S. And above mentioned blocks of conventional GIS can plliy different roles in the constrction MSG[S. Also. there are different opinions of implementation of these problems. Most researchers mainly have paid attention on the exection of atomated CO in the first two blocks (3, 8]. [n crrent years some athors pointed ot on the needs of maintenance of a mlti-scale or mlti-detailed or mltiresoltion database (MDB) (5-7, 10]. The options for design MSOIS and their advantages and disadvantages are following: 1. The first way is creation of MSGlS, which contains the spatial atomatic CO processor and DBMS, which maintains a spatial data model in the single base scale. The smaller. scales can be retrieval fll atomatically at the time of the qery. This way is efficient for the captre and the storage of data. Bt crrently however, complete sccessfl design of sch systems is dobtfl on accont of the locational geometry of object in the detailed base scale do not define the geometrical topology of the object in smaller scale. Also sch systems can cased the data redndancy dring the perfonning of CO, that is not acceptable for the interactive infonnation system [6]. 2. The second option of MSGIS contains CGblock, which can maintain the intelligent compteraided CO design with ser interaction (2, 8]. The DBMS maintains, as in the first case; a spatial data model in the single base scale. The CO operations in this case don't create new data in database. The design of intelligent compter-aided generalization systems is recommended by the application of the expert system and knowledge based techniqes in order to bild software capable of intelligent CO decisions (8]. The disadvantages of these systems are: these systems can be sccessflly sed for compter assisted cartography design, that is for map compilation, bt these systems. don't gide an' application ser, who may be introdced not with the techniqes of CO; the second option is time delay for real time applications; the data spporting strctres are necessary for effective.interactive sing of the system [2]; the expert systems cannot find precision soltion to the ambigosly and nclearly fonnlated problems of CO [4]; sch systems can cased the data redndancy dring the perfonning of CG de to inadeqate treatment of scale-related variability within a file (5, 6, etc.]. 2491

2 3. The third way is sed the intelligent compter-aided generalization design system with ser interaction and DBMS, which maintains the generalized MSDS [5, 6J. In this case, the generalization operation block can be sed either for creation, maintenance and pdate of MSDS's strctre or for spatial mltiresoltion analysis, design ser's strctres of geographical phenomenon and maps. The third option gives the possibility to avoid same disadvantages of second option. The advantages of maintenance of mltiple data base are: redction of time and labor consming ser interaction for retrieval pre-generalized versions of object and obtaining closely version of appropriate representation accordance to ser needs; redction of processing time for retrieval data at the appropriate resoltion by comparison with a SDDB; redction of data dplication by comparison with the storage of several pre-generalized versions of the object; redction of the significance of prpose-orientation factor on the process of compter CG; increase of the field of ser applications of the system; doing the process of CG more phenomenon-based and impartial de to sing whole inner strctre of the particlar entity, etc. The disadvantages of maintenance ofmdb are: the process of realization, maintenance and pdate ofmdb is sophisticated; the assimilation for a ser is sfficiently complicated. In this article attention is paid on the implementation of spatial generalized MDS. For this goal, the CG is examined first of all, becase, as it was above mentioned, the nderstanding of compter CG plays a vital role for sccessflly atomation of this process and also mlti-detailed digital model mst spport the spatial and attribte transformations, which are sed in the map generalized design. 2 Generalization in digital environment A digital CG is the application of abstraction mechanism to the world's entities for the formlation of image of reality by means of distraction and spplement. There are several conceptal models for a CG in digital environment [8J. Bt these models are far from complete formalized description of atomation generalization process. The difficlties for the complete rigorosly soltion of these problems are cased firstly by the mltiformity of the CG. The main factors of CG can be classified as follows [8]: a) phenomenon-based factors, which are cased de to the conceptal natre of modeled entity (the essential of the analyzable phenomenon; the featres of territory; the relationship among entities, etc.); b) prpose-orientation factors (ser needs and prpose; contents; scale; technology of the compilation map, etc.); c) graphic media and format factors (the visalization prpose, the technology of compilation map; the kinds of cartographic object; the methods of cartographic representation; the rles of map design, etc.); d) comptational factors ( the effective of information system realization). CG in GIS environment can be divided on the two stages: 1. The modeling and CG of the real-world objects for the creation of spatial strctres in the digital forms. 2. The CG of digital spatial strctres for the creation of their cartographic representations. Above mentioned factors of CG are playing different roles and are having different significant for the CG process on the different stages. The factors a) and d) are more important for the stag'e I. The factors b) and c) have inflence on the stage 2. The factors a) and b) are more the difficlty to be formalized, former one - de to mltivariable of natre of phenomenons, last one - de to coexistence of several ser views for the same real-world object and it's cartographic representation. The factors c) and d) are sally programmable. Therefore, for the simplification of the problems of formalization of atomation of CG is sefl to maximm separate the stages 1 and 2. For this goal in MSGIS mst be spported the following demands'. 1. The inner mltiple representation of entity ( its geometry and inner and oter topology ) shold be embedded in database strctre. 2. The date strctre and operators shold be extensible. 3. The logical independence between multiple data architectre and the generalization processor shold be implemented. Accordance to the definition of CG, as kind of abstraction, and above mentioned spposition abot factors of CG the following operation and corresponding CG techniqes can be distingished: 1. Omitting of whole entity or parts of its geometry ( selection, simplification, collapse ). 2. Systematization ( aggregation, concentralization, association, absorption ): 2492

3 2.1. Aggregation of several homogeneos objects to one object on the same level of object's hierarchy Classification: Concentralization of several homogeneos objects into one object on the higher order object's hierarchical level [I] Semantic Association [1]: Typification is the process of groping objects in classes by their types Classification is the process of groping objects in classes according to their qantitative properties Absorption of objects into the other inhomogeneos objects. 3. Display operators ( exaggeration and nderestimation, displacement, smoothing, symbolization ).. The techniqes of the CG of entity and object, which make p the entity on maps or in a digital database are shown into the Table I: I! Applied to! Transforma-! Level of! Affeclon! Affect on I! p- points.! tion reslts:! inflence:! the data! thembr: I Kinds of! I - lines.! g - geometry.! 0 - object.! strctre:! e -exaggeration. I generalization! a - areas.! s - semantic..! e - entity! s ;strctr..! -nderestimation. I, at - attribte! t - t0j.10101ly,-' I d - display, d -displacement I Selection I Simplification I Concentralization I Aggregation I Absorption I Collapse I Exaggeration and! p. l.a! t! 1. at g o! p.l. a s. g. t e! p.l. a g. t p.l. a s. g. t l. a g. t I nderestimation I. ago d I Displacement p. I. a: god d I Association at s I Smoothing I god I Symbolization J,1 II e L~ "-"-:""'dll... Table I: The kinds of CGs and their classification, (where, t - a local neighborhood topology) 3 The ways of design spatial mltiresoltion database The sggestions abot creation of MSDB or MDB architectre are described in following articles by Jones, C.B., and Abraham, l..m. (1986), Jones, C.B. (1991), van Oosterom, P. (1991) [5, 6,10], et at Mainly these athors paid attention on the two kinds ofcg,.these are simplification and selection. There are at list three ways for the design of MDB for MSGIS. These ways of mlti-scale database organization and their advantages and disadvantages are following: I. The MDB can be organized as a storage of several pre-generalized versions of the entity. Advantages are [6]: the realization; pdate and maintenance of MSDB are simple; the efficient access is existing to the different scale representations. Disadvantages are: a storage of data is expense de to data dplication; the database strctre maintains not the whole model of geographic entity, bt its slices in the different scale levels. 2. The MDB can be organized as a storage of single MDS of the objects, which are contained the entity, with possibility to recover a cartographic 'representation at a reqired level of scale or resoltion. Advantages are: a data dplication is avoiding by comparison with the first option; a data redndancy is avoiding by.comparison with the single detailed database (SDDB); a data dplication for maintenance of different resoltion of entity is sfficient; a complete or mltiresoltion model of entity is stored, bt not its scale slices. Disadvantages are: the realization, pdate and maintenance of integrity are complicated. 3. The MDB can be organized as a storage of a SDDB and additional (axiliary) information, which spports.generalization operators. Advantages are same as for a SDDB. Disadvantages are: very 2493

4 detailed infonnation for resolved conflicts and for spporting a real-time CG of single detailed representation is necessary to store; the definition of infonnation, which is necessary to store for the rigorosly spporting of generalization process, is ncertainty. In this paper the attention is focsed in particlar on the maintenance of MOS. The following problems shold be defined on the stage of design of MDS: I. Type of a GIS architectre. 2. Method of implementation of the spatial object's index. 3. Methods of maintenance of integrity of database. 4. Methods of maintenance topological object relationships. 5. Techniqes of CG, which shold be maintained into a s\orage strctre and, which shold be exected on the stage of visalization. 6. The object-layer debate. 7. The tiling and edge matching problems. There are three different types of a GIS architectre, which has following characteristics [9]: I) separate sbsystems for storing and retrieving spatial data and thematic infonnation, which is stored in relational DBMS; 2) a single system for storing and retrieving the spatial and thematic data in the pre relational DBMS; 3) an extensible or object-oriented DBMS, which can spport more attribte types and access methods. This paper otlines only the stdy of geo-relational implementation for spatial MSDS. Sch model of data strctre, of corse, has same disadvantages by comparison with an object-oriented model, bt now even the implementation of conventional last one for GIS has been on the stage of investigation. The frther research of this topic is spposing to do for an object-oriented implementation. It can be considered what a geo-relational implementation of MSDB is the first step in the stdy of mltiresoltion data strctre. I I ARC II PN sa NBR I x F... OD~..."::TO NODE : 5 SD' B It PLiNE :3,... SDZ FSCL A.R.C, ARC2 ARC3ARC.4 A,~ so, PN' SD2 P... "t so. PN4 ;; ~D~ ~ C9 ARC10 ARC" ARC' 2ARC'~o1 ARC' I # P~L D.-- ':.:~:=:'=2'::::::~=:;::=::::=~.PUNE } {.. POL J Figre I: The conceptal schema of mlti-scale spatial featres: A. A representation of arc featre; B. A representation of polylinear featre; C. A representation of polygonal featre; D. A representation of object, (where, # - an 10; F - from; TO - to; { } - a sbset of elements; SCL - a scale; OBJ - an object; X, Y - rectanglar coordinates; SO - a scale denominator; PN - a point; NBR - a nmber; OR - an oriented; R - right; L - left; PUNE or PL - a polyline; POL - a polygon; LPN - a label point; SP PR - a spatial primitive; SQ - a seqence) The spatial primitives, object and their properties, which have been sed for design geo-relational model of MSDS are: I) a point featre is a one-dimensional spatial object with coordinates and a niqe identifier (10) within the database or a map. 2) an arc is an oriented seqence of ordered vertices, where the beginning of the line may have a special start node, and the end a special end node. Also an arc can be a part of zero or more polygons, and therefore has a left and right polygon identifiers. 3) a node is a jnction of one or more arcs. 4) a label point is sed for representation of point featres and for assign 10 to polygon. 5) a polyline consists of one or more arcs. A polyline is an oriented seqence of arcs. 6) a polygon consists of one oter and zero or more inner polylines. 7) an object is a seqence of spatial featres, which make p a representation of the real world entity on a map. 8) the cartographic entity or it's object can be existing on maps of intennediate adjacent scales. 9) dring of simplification is going to selection of vertices of the arcs, bt the nodes of these arcs are saving. 10) the following characteristic of line (arc) cartographic simplification is taken into consideration that points (vertices), selected for small scale representatlons, are always a sbset of those sed in larger scale representations. II) the arc, which is a part of nonzero polygon, can't be 2494

5 selected. 12) the displacement, exaggeration, nderestimation, smoothing and symbolization don't give rise of the topology modification, these operations are pre graphical. 13) the reslts of selections, con centralization, aggregation, absorption, collapse and' association mst be condcted in different relational tables for maintenance of integrity of database. The conceptal model of spatial featres is shown on the figre For this scheme has been bilt logical model, which has been implemented as shell software nder a commercial geo-relational database management system. The amont of storage in the mlti-scale geo-relational database is significantly redced de to integration mlti-scale storage of vertices in the line mlti-scale trees [6J, which minimized vertex dplication and also sing same records for arcs, polylines and polygons into the different scales of entity representation. Althogh this method introdces additional data in the fonn of seqence nmber of vertex, scale denominators, pointers on the files, the total amont of storage is redced significantly by comparison with the storage of several pre,generalized versions of entities. Conclsion The logical model of mlti-scale representations for the different kinds of generalized techniqes; the possibility of application of hierarchical methods for the cases ofa simplification and selection; the spatial access mechanism by means of the overlapping region:s scheme; the qestions of integr.ity of database; the spported demands and the contents of the MSDB wiil be present as poster on the conference and will be pblished elsewhere. Frther research efforts will be concerned with developing an object-oriented approach for mlti-detailed representation. Acknowledgment The athor shold like to express their appreciation to the China Postdpctoral Science Fondation and LIESMARS, WTVSM and professors W'Hehai, D Daosheng, who have provided financial, technical and consltative spport for the research. References [IJ Alaev,E.B., Social-economic geography (conceptyal,tenninologic dictionary).moscow,1983,350p. [2J Aasgaard, R., 1992, Real time Cartographic generaliiation, a reqirement for efficient presentation of geographic data. Proc. of the 5th Int Symp. on Spatial Data Handling, pp [3J Beard, M.K., 1986, How to srvive on a single detailed database. Proceedings, Ato-Carto 8, ACSM-ASPRS, pp. 2Ue220;.. [4] Bregger, B.P., andmller,j.-c., 1992, Mechanisms of geometric. abstraction. Proce. of the 5th Int. Symp. on Spatial Data Handling, Chareston, Soth Carolina, IGV Com. on GIS,pp.123c133.. [5]' Jones, C.B:, 1991, Database architectre for mlti-scale GIS. Proceedings, Ato-Carto 10; Baltimore, ACSM-ASPRS, pp [6] Jones,C:.B., and Abraham, l. M., 1986, Design consideration for a scale-independen~.cartographic database. Proceeding ofthesecondlnt. Symposim on Spatial Data Handling, Seattle, pp [7J Kilpelainen, T.., and Sarjakoski, T., 1993, Knowledge-based methods and mltiple representation as means of on-line generalization. Proc. of 16th Int. Cart. Conf, Cologne I Koln, rca, pp.21i-220. [8J Map generalization: making rles for knowledge representation. Edited by Bttenfield, B.P., and McMaster, R.B., 1991, I-st edition, Longman Grop UK Limited. [9J Vijlbrief, T., and van Ooster<lm,P., 1992, The GEO++System: an extensible GIS. Proc. of the 5th Int. Symp. on Spatial Data Haridling, Chareston, Soth Carolina, IGU Com. on GIS, pp [loj van Oosterom, P., 1991,' The,Reactive-tree - a storage strctre for a seamless,. scaleless geographic database. Proceedings, Ato~Carto 10, Baltimore, ACSM-ASPRS, pp