Proceedings of the 21st International Cartographic Conference (ICC) Durban, South Africa, 10 � 16 August 2003�Cartographic Renaissance� Hosted by The International Cartographic Association (ICA)ISBN: 0-958-46093-0 Produced by: Document Transformation Technologies

CONSTRUCTION OF DATABASE PLATFORM FORINTERACTIVE GENERALIZATION ON LARGE

SCALE TOPOGRAPHIC MAP

Cai, Z., Du, Q., Wu, H. and Liao, C.

School of Resources and Environmental Science, Wuhan University,Wuhan, 430079, P. R. China. E-mail: caisoft@public.wh.hb.cn

ABSTRACT

Map generalization is a procedure involving much intellective reasoning action, with very wide domain. It is also adifficult problem in the field of cartography in the world. This paper makes a study on the interactive digitalgeneralization, where map generalization can be divided into intellective reasoning procedure and operationalprocedure, which are done by human and computer respectively, and an interactive map generalization environment forlarge scale topographic map is then designed and realized.

The research includes:! the significance of the researching an interactive map generalization environment! the feature of large scale topographic map and interactive map generalization! the construction of map generalization-oriented database platform

A number of tests have proved that map generalization can be successfully and interactively done with the cooperationof human and computer if the procedures of map generalization are wisely decomposed. Compared with the traditionalmanual method, interactive map generalization can shorten the working time to 1/4 or even. Besides, the work willbecome less intensive with higher precision. Based the model generalization of the map, this environment cannot onlyfinish map generalization but also derive digital data at the same time. Therefore, this technology can, to some degree,meet the requirement of GIS to multi-scale spatial data.

1. INTRODUCTION

Traditional manual cartographic generalization, which requires the performers not only to master enough professionalknowledge and work skills but also to possess certain work experience, proves to be very difficult to be fulfilled.Furthermore, this method brings about low efficiency, a long time consuming, as well as big contrived error. With thetransformation of user�s consumption concept, as far as map production is concerned, a swift reaction to market isdemanded. Obviously, being its poor efficiency, this kind of map production reigned by paper map is difficult to meetthe users� demands, in addition, it can not keep up with the quick development of information at all.

As we all know, there have been more than thirty years since the introduction of computer technology into cartography.Owing to the speedy development of computer assistant cartography technology, the manner of map production isgradually becoming perfect; especially owing to the building and application of map database and GIS, a historictransformation from manual method to automatic method is also coming into being step by step. It is most noticeablethat this transition takes place at all aspects of cartography, such as map data, map production, cartographicaltechnology, cartographical software and so on. Under the background of digital map, the key technology, to achieveautomation and integration of �map design--map compilation--film output�, is map design and map compilation, inparticular the automatic compilation of digital map.

However, presently to a great extent manual method still prevails in map generalization. That is to say, on one hand, wehave achieved automatic cartography, but on the other hand, we are adopting manual cartographic generalization, it willbe too difficult for the performers to simultaneously work under two different environments, which will also destroy thesystem of automation and integration. In addition, the quick update of large-scale topographic map (especially urbanlarge-scale topographic map) is another reason.

Bypassing the great difficulty of reconstructing database, an advisable method is to derive data of smaller scale fromdata of larger scale (relatively) through cartographic generalization. Accordingly, it is urgent to realize automatic orsemi-automatic cartographical generalization based on the cooperation of human and computer.

In addition, performing cartographical generalization under a new digital environment is also an approach to helpcartographer free from burdensome and low efficient work. To sum up, in order to meet new requirement presented bydigital environment, it should be a development direction of map making to adopt new technology to improve theefficiency of map compilation.

The paper starts with the necessity and feasibility of interactive cartographic generalization based on large-scaletopological map, the second section describes the detailes of building data model, the third presents results from tests ofthe system.

2. LARGE-SCALE TOPOGRAPHIC MAP AND INTERACTIVE CARTOGRAPHIC GENERALIZATION

This section firstly describes the characteristics of large-scale topographic map, and then is devoted to the method forsolving cartographic generalization based on large-scale topographic map.

2.1 The characteristics of large-scale topographic mapBy means of graphics decomposition, large-scale topographic map may be understood to consist of the following twoparts: map graphics and character annotation. The former includes three kinds of graphics elements: point, line and area.A digital topographic map must cover all spatial information and attribute information of six types of features: singlefeature, residential area, drainage basin, relief, pipe line and border, vegetation. In terms of large-scale topographic map,because almost all human and natural trivial elements on the earth surface require to be precisely and really reflected,added to the considerable amount of information and the quick changing and updating speed, its verisimilitude is veryworse, which embodies more obviously when man-made elements are represented.

In contrast to small scale map, in large-scale map, the entities represented by features are difficult to form objects, themajority of them consist of complex geometric structures, and this kind of extraordinarily elaborate express, to certainextent, affects simplification and abstraction of data model based on objective world. When added into database, thiskind of data representation is little similar to a full replication; scattered structure of entity world still remainsunchanged. For example, in 1/1k scale map, a house is not only denoted as a simple polygon, but attaches some otherlinear structures, such as balcony, porch and its pole, ladder, stairway and so on. That is to say, the majority of objectsare compound, which increase the difficulty of cartographic semantics identification, making operators designed forcartographic generalization have to excessively focus on the generalization of geometric information and the disposal oftopological relations between elements. Urban large-scale topographic map is characteristic of numerous man-mademap elements (such as house, street, pipe net, construction facility, etc). Different from natural features, these featurestake on obvious human traits. For instance, the borders of streets can�t embody fractal characteristics as natural features(rivers, coastlines, etc) do. On the whole, the features distributing with human traits develop a unique structure. Forexample, the angel of a house polygon is a right angel, and the houses lying beside the two sides of roads are regularand form arrays.

2.2 Interactive generalization based on large-scale topographic mapLooking through the characteristics of large-scale topographic map mentioned above, we can conclude that itsrepresenting of content elements is more detailed, and the relations among them are also more complex. Presently, thebasic theory of automatic generalization, as we all know, is still not mature. Therefore, realizing purely automaticgeneralization based on such complex information relations is hardly possible at all. Taking insight into thecharacteristic of large-scale topographic map, it is not difficult to find out the majority of characteristics and relations ofmap content elements, while represented, employ the relations of geometric graphics.

Interactive generalization, however, is a human-computer cooperative working style. During the course ofgeneralization, resorting to the map knowledge and map product experience what he master, the performer is required toparticipate in such works as selecting generalization operators, setting parameters, feeding back the executing results,etc. And computer, depending on certain software arithmetic, execute basic generalization operation. Analyzing thedifferent roles acted by human and computer, the generalization process as a whole may be summarized as the problemof 3W+H. That is, When---when does the user put forward the condition of generalization and simplification to executegeneralization transformation of shortening scale; Where---where do the spatial occupancies generate conflicts, andwhere are the features too dense; Who---which features are important; How---how to execute cartographicsimplification, replacement, and conflation etc.

In these problems, 3W problems belong to the deeds involving in powerful intellective generalization, reasoning, andjudgment, presently they can only be solved by hand. But H problems can be accomplished by perfect generalizationoperators. Indeed, in the whole task, the disposals of 3W problems account for a larger proportion, however, our brains�thinking, as is known to us, can swiftly arrive at an answer, which consumes a little of time. H problems, although onlyaccount for a minor proportion, but deal with the compilation and maintenance of geometric graphs, should becategorized into physical labor. Provided they are solved by hand, it might need a very long time consuming.

So adopting this kind of generalization method, human and computer achieve a kind of mutual complement throughcooperative work, compared to traditional manual operation, its efficiency proves to be an evident improvement, it willbe a better generalization method relatively to current research level.

Consequently, it can be seen that the design and construction of data structure and data model directly determinewhether the characteristics and demands of interactive generalization based on large-scale topographic map can berightly and flexibly reflected or not, in addition, the research and development of following generalization operatorsmust as well sufficiently pay attention to these characteristics.

3. THE CONSTRUCTION OF MAP GENERALIZATION-ORIENTED DATABASE PLATFORM

The aim of this section is to provide all the technical details designing this system.

3.1 The hierarchical organization of generalization objectsWhile designing basic platform of map database (includes designing the database model and organizing data structure),the development of the subsequent application functions should be sufficiently taken into account, the designedplatform should present data objects with all kinds of structure relations that map generalization operations need, amongwhich, hierarchical organization method is regarded as a most important method.

Firstly, hierarchies of generalization objects should be partitioned logically. In traditional cartographic generalization,we execute it on different hierarchies. This idea is still of significance in software designing, it is essential to constructversatile map layer manager, which is used to provide compilation functions orienting operations. After the hierarchiesof objects to be generalized are partitioned, the generalization operators respectively toward these hierarchies and thecorresponding control parameters will be determined eventually, which are the foundation of application moduledevelopment in the subsequent generalization process.

During the course of partitioning hierarchies, such respects as the basis of classifications of map content elements(natural elements include relief, drainage basin, vegetation, and social elements include road, residential area, land use,pipe network facility, etc), geometric features (including point, line, area, net, etc), spatial relativity and so on, shouldbe given enough consideration.

The generalization element hierarchy should approximately embody three characteristics: operational sequence,structural singularity and element hierarchies� overlapability. Operational sequence suggests that while solving spatialconflicts among elements of different hierarchies, the priority of generalization must be taken into account, keeping thefeatures with the higher priority immobile, and deleting, cutting, replacing the features with less priority. The purpose ofstructural singularity is to meet the operational requirements of generalization operators, because most of operators onlyserve the objects with single structure, such as the graph structure of road constructing, the polygon structure of arealake or reservoir constructing, the Voronoi diagram constructed by all elevation points, the Delaunay triangulationstructure rising from a group of buildings. Element hierarchies� overlapability is designed to realize all the elementhierarchies� overlapping after they are respectively generalized, so that we can adjust spatial relations among them andeliminate the conflict contradictions.

Secondly, in the software system, how to organize the hierarchies relations before generalization and aftergeneralization is also a problem that should be solved. Generalization operations, not the same as those commongraphics editing operations, which may either directly substitute new objects for old objects, or resume the original stateby using undoing operations, involve in considerable complicated computations. Furthermore, under most situations,the operational object is always towards multi-item. In the interactive undertakings, in fact, whether a generalizationoperation is in accordance with criterion is mainly justified by performer. Because this kind of justice is achievedthrough comparing the state before the generalization with the one after the generalization, it is entirely wrong to adoptthe editing operations similar to replacement. The generalization results ought to be derived from the base map data, foradapting to this requirement, it is necessary to simultaneously store two hierarchies of data before generalization andafter generalization.

3.2 The design of system based on Object-Oriented (OO) methodObject-Oriented method is deemed as a new method to analyze and solve problems. To a great extent, it isapproximately similar to human thought styles. OO method takes object as the most fundamental element, overcomingthe disadvantages that the relation between data structure and behavior is not very compact, meanwhile, OO methoddevelops such outstanding peculiarities as modularity, information encapsulation and hiding, abstraction inheritance,polymorphism and so on, which offered a most valid instrument and approach for managing large software, andadvancing software reliability, reusability, expansibility and maintainability.

Figure 1. System Design.

When building large-scale topographic map database, generally, we pay little attention to the need of cartography, butfocus on considering it as a geographical information database, thereby it is spatially vital to execute the abstraction ofdata type. This system database platform, based itself on the characteristics of cartographic generalization and GIS,CAD technology, can be abstracted as graphics, layer, object, geometrical class (including point class, line class, areaclass, annotation class, path class, region class, and group class). From up to down the hierarchical structure can bedescribed: graphic->layer->object-> geometrical class, combining with the operation class in relate to cartographicgeneralization, this system design frame may be interpreted in the figure 1.

Object class of this system mainly consists of four parts: system interface, map database management platform, basicoperators and arithmetic, element generalization process, besides it still includes graphic symbol design and datainterface, etc.

Map database management platform class includes graphics class, point class, line class, area class, annotation class,path class, region class, and group class (see Table 1).Element generalization process class includes building class, drainage basin class, vegetation class, and relief class (seeTable 2).

Generalization operator class may as well be divided into Delaunay Triangulation Network class, overlay analysis class,etc (see Table 3).

While carrying out module design, this system makes the best use of inheritance and polymorphism, for example,element generalization class, through multi-parent inheritance, can obtain inheritances of all kinds of operations definedby a set of generalization operators.

Table 1. Classes of Map Database Platform.

Table 2. Feature Generalization classes.

Object class Descriptions of main attributes Main operations

Building class

Coordinates of building polygon, thelayers of building, structure ofbuilding, adjoining buildings, shape,the smallest boundary rectangle

Partition of buildings groups, contiguousbuildings recognition, replacement ofbuilding, simplification of shape, deletion,conflation, evaluation

Drainage basin class

Coordinates of polygon, properties oftriangulated network, minimumbounding rectangle, the descriptionof shape, the relations of polygons

Recognition, filtration, deletion of smalllake, bi-line river is converted into single-line river, elimination, conflation,simplification, and evaluation of islands

Road class

Coordinates, length, properties ofroad, the descriptions of part convex,contiguity relations, thecharacteristics of bends

Deletion, conflation, join, replacement,extraction of axis, simplification, summaryof bends properties, elevation

Relief class

Coordinates, characteristics,elevation, contiguity relations ofcontour lines, valleys, ridges,elevation points

Filtration, join, deletion of contour lines,simplification of bends characteristics,constructing Voronoi, smoothness

Vegetable class

Coordinates, area, perimeter,attribute characteristics, contiguityrelations, the characteristics of bendson boundary, and shape of polygon

Deletion, simplification, combination,replacement, constructing DelaunayTriangular network, evaluation

Object class Descriptions of main attributes Main operations

Graphics class

scale denominator, graphics name, scope,underling layers and geometric elements ofspatial index, coordinate system, saved filename

create, data input, data output, indexconstructing, delete, read and write, show,save

Layer classState, layer name, operation characteristics,geometric properties, underling geometricelements

create, copy, move, delete, show, data input,construction maintenance

Geometric objectclass

Geometric coordinates, attributes, state,keyword, boundary rectangles, index grids,structure relations

add, delete, move, read and write, show,register, grid index building, topologyorganizing

Table 3. Generalization operator classes.

3.3 Generalization-oriented map database logical organizationWith regard to research and development of this software, it is the first to build a small map database to carry out datamanagement of cartographic generalization, meeting the demands of index on object information and structureinformation in generalization process, sequentially realizing data� s highly efficient storage and maintenance. Multi-sheet, large amount of data, multi-scale, and multiple coordinate systems, and multiple spatial structures are typical ofthis database.

Logistic hierarchical structure of database is organized in accordance with the system: graphics->layer->element class->object->geometric attribute description, which may be expressed as Figure 2.

Figure 2. Logical Construction of Database.

The idea of from up to down adopting tree structure to build database, ensures the consistence between physical storageof program realization and application-oriented logic structure, that is, as early as the time of storage and managingdatabase, hierarchical relations is embodied, so we can directly acquire the information we retrieve with no excesssearch calculations needed, improving the usage efficiency of database.

As shown in figure 2, all objects in this database, in addition to being built semantic hierarchical structure mentionedabove, meantime, are also built a spatial grid index, which is used to register, manage, and maintain all objects indatabase, sequentially quicken the feedback speed of object identification and retrieval.

Similar to the hierarchical organization of database, we may build the hierarchical structure of object classes based on

Object class Descriptions of main attributes Main operations

Triangulatedirregular networkclass

Coordinates of group points, conditionsof triangulated network, the vertexes oftriangle, neighboring triangles, the centerof triangular gravity

Contraction of Delaunay triangulated network,predisposal of data, extraction of axis,contraction Voronoi diagram, triangulatenetwork maintenance

Overlay analysisclass

Boundaries of polygons, the attributeconditions, islands, contiguity relations

Overlay analysis, computation of minimumbounding rectangle, difference combination,simplification, uniting, conversion betweenvector and raster

object-oriented design idea. In every class, encapsulation are exerted on the descriptions of class and the operations todata members, meanwhile, according to affiliations of data members and object characteristics, we build a serial ofinheritance relations among graphics, layer, element class, object.

In the seven kinds of element class defined above, point, line, area, and annotation belong to simple object types, whichare used to describe and save such simple entity objects as single facility, road, communication line, building,vegetation, lake, illumination text, enterprise name, and so on; but path, region, and group are used to expresscompound objects. Path is used for the storage of drainage basin network and road network; region is used for thestorage of such group polygons structure as the buildings group and lakes group, etc, and group is used to express thosecompound structures of objects, which may be any type. However, here it is emphasized that path, region and grouponly provide compound objects with the storage of structure frame, because the compound objects mentioned above areall derived from simple objects by means of relational operators and additive information, we need not save the simpleobjects a second time, but only save the basic structures of these compound objects in database, certainly, through thesebasic structures we can get concrete data. For example, in the buildings generalization, according to spatial contiguityrelations, some simple polygon structures of buildings are identified forming a group structure, at this time a compoundobject is derived from it, and we only need save it using region group structure.

The design of path and region structure in this paper shares the idea of using network to analyze path and using regionto analyze region in Arc/Info system, the kind of compounding relation between simple objects and compound object isalso approximately coincident with the idea in Arc/Info system.

3.4 Coordinate systemCoordinate systems involved into this database include: geodesic coordinate system, drawing coordinate system beforegeneralization, drawing coordinate system after generalization, database coordinate system, output device coordinatesystem, etc.

Map database generalization is the operation on graphics using virtual geodesic coordinate, but it is necessary to decidethe executions of operations through the user�s vision under drawing coordinate system, the index rules of generationare also described under drawing coordinate system, for example, we rule the minimum interval among houses is 1mm,twenty single facility objects per sq.dm, etc. In AutoMap software (AutoMap is a interactive generalization softwarebased on large-scale topographic map, it is developed by the Recourse and Environment Science School of WuhanUniversity, now this software have already been used to perform cartographic generalization by many surveyingenterprises and proved to be very good), the logic descriptions of data, in drawing coordinate system beforegeneralization, are designated to adopt mm as unit, and during the course of physical storage of data, the coordinatesdescribing data are transformed into database coordinates, the discrepancy between the two kinds of coordinates is onlya multiple which takes the resolution in system as coefficient, the coordinate origin lies in the center of drawing.

3.4.1 The transformation from geodesic coordinate to database coordinateThe transformation formula from geodesic coordinate to data coordinate is:

X = 2R [x � ( xmax + xmin)]

（1）Y = 2

R [ y � ( ymax + ymin )]

R = M U1000 S （2）

Here (X, Y) is database coordinate; (x, y) is virtual geodesic coordinate; (xmin, xmax, ymin, ymax) is the range of virtualgeodesic coordinate; R is the zoom coefficient of the transformation from geodesic coordinate to database coordinate,(formula 2); U is the unit of geodesic coordinate, such as meter. S is the system resolution; M is the denominator ofscale.

3.4.2 The transformation from database coordinate to drawing coordinate The transformation formula from database coordinate to drawing coordinate is:

X = S x

（3）

Y = Sy

Here (X, Y) is database coordinate; (x, y) is drawing coordinate, S is system resolution.

3.4.3 The transformation from base layer coordinate to generalization layer coordinateThe transformation formula from base layer coordinate to generalization layer coordinate is:Here (X, Y) is generalization layer coordinate; (x, y) is base layer coordinate, M2 is denominator of scale beforegeneralization; M1 is denominator of scale before generalization.

X = x2

1

MM

（4）

Y = y2

1

MM

Because Automap software is mainly used to carry out cartographic generalization based on large-scale topographicmap, it is not necessary to consider the projection transformation. While outputting data, this system will question theselection from the geodesic system, drawing system before generalization, and drawing system after generalization. Thedata type of this system database is integer type, owing to the characteristic of real 32 bit in Windows NT system, theexpression of int type data is 4 bytes, which ensure enough accuracy, meantime, in drawing coordinate system, weselect 0.01mm as resolution, both the two accuracies can meet the demand of cartographic generalization aboutaccuracy.

3.5 Spatial grid indexIn order to improve the speed querying the objects in this database, presently, spatial grid index technology is widelyadopted in spatial database development. The bottom query in relate to spatial localization includes two aspects: one isquerying which objects there exist in a grid, the other is querying which grids a object lies in, the former mainly appliesto objects identifying and retrieval by window, but the latter mainly applies to objects registering and grids indexmaintenance after some objects are deleted. To realize bi-directional query, a good method is to build a bit matrix thattakes the serial numbers of grids as the rows and takes the keywords of objects as the lines, however, in this software,due to the enormous amount of objects (if 25 1/1k map sheets are joined, the number of pints, lines, and areas will addup to morn than 80,000), the amount of storage needed by matrix is so large that bi-directional query is difficult toachieve.

This system selects single-directional storage, that is, for the n× n grids, we do not register which grids a object goesthrough, but only register the objects keywords contained in every grid, which meet the demands of objects identifyingand retrieval by window mentioned above, when the user delete a object, firstly ,this system executes real-time registercalculations, then logouts corresponding keywords contained in the grids that this object goes through, because registercalculations need not spend much time at all, this kind of real-time calculation has no effects on the running efficiencyof the system as a whole.

When the system registers objects, the point objects, according to the grid locations they lie in, are registered pointobjects keywords; line objects are registered line objects keywords in terms of the girds serial numbers they go through;but for area objects, this system registers their keywords in the grids that their minimum bounding rectangles cover;annotation objects are registered with respect to the little square determined by the localization point and the size ofevery word; and path, region, group objects are not built grid index.

3.6 Physical storage of map databaseThe logic structure of database is complicated, for its structure relations are hidden in database. What the users see areonly its files stored in external memory, this system uses three files stored in external memory to store and manage allinformation in a map (see Figure 3).

*.amg the file of including the total description information about database and layer and the header description ofobjects, is a multi-fields file(amg comes from Automatic Map Generalization);*.xy the integrated file of all coordinates strings, is bi-field structure; and*.key the file of embracing such information of uncertain length as topological structure, grid index, character stringand so on, is single-field structure;

The partition of the three files is in terms of the fields� definition structure of records, the same structural data areintegrated together and saved in the same file, after they are read into memory, this system decomposes them to attaintheir logic meanings, and carries out respective memory buffer management.

Considering of the security of data, when reading or writing data, system will make copies of the three files mentionedabove: tmp*.gma, tmp*.yx,, tmp*.yek.

3.7 Building and maintenance of map databaseThe data source building database is approximately data files containing topological structure and attribute information,and the operations building database are mainly completed by correlative functions defined in Map class, here we onlymake several rough rules for the process building database.

It includes:! Reading information content files of data source;! Regrouping topological relation, getting the information about the arcs forming polygon and the external ring of

islands;! Performing the register of point, line, area, and annotation objects, and registering them in grid index;! Saving the series of keywords, coordinate strings and header information of objects.

Figure 3. Physical Structure of Database

After data are read into Automap, it is not immediate for the system to create external memory files, but to save all ofthem in memory buffer to perform management operations, only when the storage operations are activated, are theexternal memory files: *.amg, *.xy, *.key created and saved.

It is a complicated process to maintain database, after the user deletes an object, corresponding operations include:! Removing the object keyword;! Removing the registers of this object in correlative grid indexes;! Maintaining the information of the relations between this object and other objects, taking arc as example, after an

arc is deleted, the polygon containing this arc will not exist, either. It must be emphasized that all objects� deletionoperations don�t include immediately calling back their storage spaces, but only add sign 0 to their keywordsituations, as the user performs undoing operations, the system needs only change 0 into 1.

3.8 The design of database queryQuery is the most primary function with which a database provides the users, and its efficiency is also a main standardjudging whether a database�s function is strong or not. Compared with common relational database query, spatialdatabase query includes a more complicated process.

The query of this system is designed as the three manners: query according to spatial positions (localizationidentification, windowing, arbitrary polygon), query in terms of logic conditions (element code, layer, geometriccharacter class, area, perimeter, etc), and query with respect of structure relations, the query results of differentprocesses can be carried out AND, OR, XOR compounding for any times.

The query result employs bit strings to express; a integer of int type denotes 32 bit strings, then 6,000 integers of inttype can denote 192,000 bit strings, the position of bit represents the value of keyword, that the value of bit is 1 denotesthe keyword has already been selected, or it has not been selected, the method of bit express perfectly supports logiccalculations of the results.

In query class, such a group of bit operation functions are defined: READ bits, WRITE bits, MODIFY 1 or 0, AND,OR, XOR calculations, TRANSFORM bit strings into selected objects keywords, etc.

4. EXPERIMENTS AND ANALYSIS

Taking the 1/1k and 1/10k database data of Shenzhen city as examples, the author did a number of experiments; theresults proved that this kind of cartographic generalization oriented database platform took on high stability, and provedthis kind of human-computer interactive generalization environment based on this database can well help performerfinish generalization work under digital environment (see Figures 4 and 5). Though the automatic degree is not veryhigh, the effect of generalization basically accords with the demand of generalization, after automatic disposal, for theperformers, it needs only carry out some simple edit operations to come to the final demand. As far as the magnitude ofthe task is concerned, computer instead of human finishes burdensome and repeated labors, considerably reducing thelabor force of the performer, who hence can free from those simple, repeated labors. In addition, in terms of the periodof the performance, as was mentioned before, the total time can reduce to one fourth of original time or less. At thesame time, the performance accuracy also wins a big advance; what�s more, it is of more importance that the results ofgeneralization are data, which can be put into database. The results of the experiments are showed in Figures 4 and 5.

Table 4. Database Query.

The typeof query

The method ofquery Condition The illustration of function

NEW The first query

AND Adding object currently queried to current result, and carryingout OR calculation

RE Querying again based on current selected objects, andcarrying out AND calculation

Logic operations

UN Eliminating the objects currently queried in the selectedresults

POINT Identifying objects in light of appointed pointsBOX Selecting objects by rectangular windowing

CIRCLE Selecting objects by circular windowingLocalizationframework

POLY Selecting objects by arbitrary polygonal windowing

THROUGH Being viewed as selected objects that go through localizationframeworkLocalization

conditions WITHIN Being viewed as selected objects contained into the internallocalization framework

ALL All objects

LAYER Appointing some layer or several layers

Localization query

The range ofquery

TYPE The type of localization element (POINT/LINE/POLY/��)

Logic conditionThe same aslocalizationquery

The same as localization query

CODE Element code

LENGTH Length

AREA AreaThe items of theusers

USER_ITERM Attribute items the users define

> Bigger than the value given

= Equal to the value givenRelations

< Smaller than the value given

Logicquery

The range ofquery

The same aslocalizationquery

The same as localization query

a) Original buildings.

b) Buildings generalized.

Figure 4. Generalization of Buildings.

a) Original contours.

b) Contours generalized.

Figure 5. Generalization of Terrains (Contours).

5. REFERENCES

[1] Wu H H(1991). Cartography Database. Beijing: Surveying and Mapping Publishing House (In Chinese)[2] Wang J Y(1993). Generalization Principle of Topographical Map. Beijing: Surveying and Mapping Publishing

House (In Chinese)[3] Zhu G R(1983). Edit of Topographical Map, Beijing: Surveying and Mapping Publishing House (In Chinese)[4] Guo Q S(1998). Research on New Theories and Methods of Automatic Map Generalization: [Doctoral Paper].

Wuhan: Wuhan Technical University of Surveying and Mapping (In Chinese)[5] Cai Z L(1999). Research on Environment about Interactive Generalization On Large-Scale Topographical Map:

[Graduated Paper]. Wuhan: Wuhan Technical University of Surveying and Mapping (In Chinese)[6] Jaing A N(1982). Cartography Generalization. Beijing: Surveying and Mapping Publishing House (In Chinese)[7] Wang Q, Wu H H(1998).Research on Fractal Describe of Map Information and Automatic Generalization.

Wuhan: Wuhan Technical University of Surveying and Mapping Publishing House (In Chinese)[8] Miched, Worboys F., HiLaig M, Hearn Shaw and David J. Magaine, Object-Oriented Modelling for Spatial

Database, IJGIS, 1990, 4(4): 369-383[9] Weibel R., �Amplified intelligence and rule-based systems� In Map Generalization: Make rules for knowledge

Representation edited by Barbara B. and MCMaifer R., Longman Scientific and Technical, 1991,172-186[10] Jones C.B, Buady G.L., and Ware J.M.(1995) Map Generalization with a Triangulated Data Structure.

Cartography and Geographical Information System, 22(4): 317-331

CONSTRUCTION OF DATABASE PLATFORM FORINTERACTIVE GENERALIZATION ON LARGE

SCALE TOPOGRAPHIC MAP

Cai, Z., Du, Q., Wu, H. and Liao, C.

School of Resources and Environmental Science, Wuhan University,Wuhan, 430079, P. R. China. E-mail: caisoft@public.wh.hb.cn

Biography

CAI Zhongliang, male, 32,doctoral candidate. His major research interests include cartography, electronic map, mapgeneralization, and GIS. His typical achievements are the system for productive design on digital map (CartoSoft);authoring tool for multimedia electronic atlas (Atlas2000); map generalization software( AutoMap), etc.