Sayısal ortamda kartografik genelleştirme

Abstract

Sayısal genelleştirme, çevresel ve özniteliksel dönüşümler vasıtasıyla bir veri kaynağından, işaretsel ya da sayısal olarak kodlanmış bir kartografik veri grubu türetme işlemi olarak tanımlanabilir. Bu türetme işleminin amaçlan; haritadaki ya da kodlanmış durumdaki verilerin miktarını, türünü ve kartografik gösterimlerini önceden belirlenen amaca ve amaç edindiği kitleye uygun olarak azaltmak ve hedef ölçekte gösterimin açıklığım korumaktır. Bu tanım hem genelleştirme işleminin felsefesini hem de ulaşılmaya çalışılan amaç ve hedefleri kapsamaktadır [1]. Uzun yıllar manuel olarak gerçekleştirilen genelleştirme işlemi bilgisayar teknolojisindeki gelişmelerle birlikte sayısal ortamda ele alınmaya başlanmıştır. Bu alanda önemli çalışmalar yapılmış, genelleştirmenin belli işlemlerinin bilgisayar destekli olarak gerçekleştirilebildiği noktaya ulaşılmıştır. Bununla birlikte, halen yapılmakta olan araştırmalar işlemin tamamen otomatik olarak gerçekleştirilmesi konusunda umut vericidir. Bu çalışmada, öncelikle, problem hem varsayıma dayanan hem de gerçek yaşamdan örneklerle ortaya konulmuştur. Manuel genelleştirme ile sayısal genelleştirme arasında kısaca bir karşılaştırmanın da yapıldığı bu kısımdan soma sayısal genelleştirmenin felsefesine dayanan, McMaster ve Shea tarafından geliştirilen kavramsal bir genelleştirme modeli, objelerin vektör formatındaki çevresel bilgileri esas alınarak ayrıntılı şekilde incelenmiştir. Daha sonra uzman sistem yaklaşımının genelleştirme için uygunluğu ve olası bir Kartografik Genelleştirme Uzman Sistemi (KGUS) ya da tam anlamıyla bir Kartografik Uzman Sistemi'nin (KUS) yapısının ve karakteristiklerinin ne olabileceği konusunda uzmanların görüşlerine yer verilmiştir. Sonuçlar ve öneriler bölümünde ise yine bu konuda araştırma yapan kartografların görüş ve önerileriyle olaya biraz daha ışık tutulmaya çalışılmıştır. Ekler kısmında verilen, başarıyla gerçekleştirilmiş manuel genelleştirme örnekleriyle de konunun somut olarak sergilenmesi amaçlanmıştır.

In a digital environment, the generalization process supports a variety of tasks, including: digital data storage reduction; scale manipulation; and statistical classification and symbolization. Digital generalization can be defined as the process of deriving, from a data source, a symbolically or digitally-encoded cartographic data set through the application of spatial and attribute transformations. Objectives of this derivation process are: to reduce in scope the amount, type, and cartographic portrayal of the mapped or encoded data consistent with the chosen map purpose and intended audience; and to maintain clarity of presentation at the target scale. This definition embraces a philosophy which considers the manner in which the generalization process is achieved, as well as the goals and objectives sought through its application. Cartographers have struggled for centuries with the difficulties of map generalization and the representation of Earth features. It could be argued that the first published work that addressed the problem of cartographic generalization was produced in the early twentieth century by the German cartographer Max Eckert (1921). In his writings, Eckert took the position that cartographic generalization bridged between the artistic and scientific side of the field. Beginning with the theoretical work on map generalization by Perkal (1966) and Tobler (1966), the foundation for future efforts in digital generalization was established. In this study, firstly, the problem was introduced by means of a hypothetical and a practical example. After this part which also includes a brief comparison between manual and digital generalization, a comprehensive and conceptual generalization model based upon a philosophy of digital generalization and developed by McMaster and Shea was examined in detail for the objects' spatial information in vector format on principle. After that various authorities' opinions were mentioned about the appropriateness of the approach of expert system for generalization and what the framework and the characteristics of a possible Cartographic Generalization Expert System (CGES) or a full Cartographic Expert System (CES) can be. In the part of conclusions and suggestions there is a look based upon the opinions of the cartographers to the future. By means of the successfully realized manual generalization-examples, introduced in the appendix, it is aimed to make a clear illustration in a concrete way. VI Manual Versus Digital Generalization First, whereas the manual process is extremely labor-intensive, the digital process strives to free the cartographer from mundane, manually-intensive activities by substituting computer-driven manipulations. Also, the manual process is highly subjective, and as a result, is idiosyncratic in its selection of generalization methods and the degree to which those methods are applied. In contrast, digital generalization attempts to provide consistent application of generalization manipulations by nominally following a predetermined set of computer instructions. And, finally, the most significant difference between manual and digital generalization is that the manual process is holistic in its perception and execution. The Requirements of Generalization Generalization, as recognized by Muller, is initiated by four main requirements: - economic requirements, - data robustness requirements, - multipurpose requirements, and - display and communication requirements. A Comprehensive Conceptual Model This model developed by McMaster and Shea (1988) baseds upon a philosophy of digital generalization. The generalization process is decomposed to three operational areas: (1) a consideration of the philosophical objectives of why to generalize; (2) a cartometric evaluation of the conditions which indicated when to generalize; and (3) the selection of the appropriate spatial and attribute transformations which provided the techniques on how to generalize. Philosophical Objectives The first component of the conceptual model examines the intrinsic objectives of why cartographic generalization is conducted within a digital environment These objectives include (a) an adherence to general, intuitive cartographic principles (theoretical elements), (b) attendance to the specific requirements of the generalization problem being considered (application-specific elements), and (c) consideration of existing computing technology demands and capabilities (computational elements). From a theoretical perspective, generalization techniques help counteract the undesirable consequences of scale reduction. To guide the generalization process in the digital domain, six theoretical elements may be distinguished: - reducing complexity, vıı - maintaining spatial accuracy, - maintaining attribute accuracy, - maintaining aesthetic quality, - maintaining a logical hierarchy, and - consistently applying generalization rules. The level of generalization must ultimately meet the requirements of a final published map or graphic display. Three application-specific elements may be identified for the final application: - map purpose and intended audience, - appropriateness of scale, and -retention of clarity. The computational perspective of generalization is of significant importance in the digital domain. Here, a cartographer generalizes to balance the relationship between sampling interval of data, data complexity, storage availability and requirements, and CPU-needs. Three computational elements should be considered: - cost effective algorithms, - maximum data reduction, and - minimum memory/disk requirements. Cartometric Evaluation The situations in which digital generalization are required arise ideally due to the success or failure of the map product to meet its stated goals; that is, during the cartographic abstraction process, the map fails "...to maintain clarity, with appropriate content, at a given scale, for a chosen map purpose and intended audience". The when of generalization can be examined from three distinct viewpoints by identifying: (1) the geometric conditions under which generalization procedures would be invoked; (2) the spatial and holistic measures by which that determination was made; and (3) transformation controls of the generalization techniques employed to accomplish the change. Six geometric conditions that will occur under scale reduction may be used to determine a need for generalization: - congestion, - coalescence, - conflict, - complication, - inconsistency, and - imperceptibüity. Conditional measures are assessed by examining basic geometric properties of inter-and intra-feature relationships. Many of these measures are summarized vm below. Although this list is by no means complete, it does provide a beginning from which to evaluate conditions within the map which do require, or might require, generalization. - density measures, - distribution measures, - length and sinuosity measures, - shape measures, - distance measures. - Geştalt measures, and - abstract measures. The generalization process is accomplished through the application of a variety of generalization operators -each attacking specific problems- each of which can employ a variety of algorithms. To obtain unbiased generalizations successfully, the order in which the generalization operators are applied becomes as critical as the selection of the algorithms employed by those operators. In addition, the input parameters required to obtain a given result at a given scale plays a significant role in affecting generalization transformations. Concomitantly, there may be permutations, combinations and iterations of operators, each employing the same convoluted structure of both algorithms and parameters. The three transformation controls critical to generalization are: - generalization operator selection, - algorithm selection, and - parameter selection. Spatial and Attribute Transformations The final area of discussion considers the component of the generalization process that actually performs the actions of generalization in support of scale and data reduction. This how component of generalization is most commonly thought of as the operators which perform the generalization process. Spatial and attribute transformations are those modifications made to the digital data. The two types of transformations -spatial and attribute- are not necessarily independent and in many cases are intricately related. In considering the process of digital cartographic generalization, nearly all applications of the process have as their first step the selection of objects and attributes from the initial database for representation. Although the selection process conceptually is not part of generalization, it must be considered a necessary preprocessing step to the spatial and attribute transformations. Before geographical objects or their statistical attributes can be manipulated by the generalization operators, a decision must be made to either include or exclude the object and/or attribute in the generalized map. Once a object or attribute is initially selected, the generalization process continues by the application of spatial or attribute transformations, respectively. IX Geographical generalization involves the geometric manipulation of the object's spatial information, either in vector or raster format Statistical generalization involves the process of either classification and/or symbolization. Spatial transformations are those operators that alter the data representation from a geographical or topological perspective. Here, the focus is primarily on the locational aspects of the data and, for the most part, ignores the associated statistical component The generalization of cartographic features to support scale reduction must obviously change the way features look in order to fit them within the constraints of the map. Data sources for map production and geographic information system applications are typically of variable scale, resolution, projection, and accuracy. Ten spatial transformations have been identified which control this graphic modification: - simplification, - smoothing, - - aggregation, - amalgamation, - merging, - collapse, - refinement, - exaggeration, - enhancement, and - displacement Attribute transformations manipulate the underlying statistical characteristics of a feature, with the subsequent spatial changes necessary only to depict the changes in attribute information. Two attribute transformations have been identified: - classification, and - symbolization. An Approach of Expert System Cartographic Generalization Expert System (CGES) can imitate the thinking process and reasoning of a cartographic expert and use the special knowledge (including the models and algorithms) of the expert to solve the problems of the cartographic generalization. This way will sum up the cartographic method and experience of excellence in cartographic process and let the experience to become the wealth used by all cartographers. This way can also unified the knowledge of all cartographers and guide the practice of cartographic generalization by these experience and theories. A full Cartographic Expert System (CES) would be capable of producing, without human intervention, maps of all types. The system would produce maps for any data set and situation. Maps made by the system should be as effective as the maps produced by professional cartographers, and base-map information would be derived from a single database. CES would allow the user to specify scales, projections, colors, symbols, and other map elements, but would make good decisions about defaults for any of these if the user chose not to specify them. Ideally, such a system should even make an appropriate choice of the type of cartographic representation if the user did not wish to make such a decision. Conclusions Hitherto, efforts associated with computer-assisted generalization are focused on the development of algorithms by means of classic procedural programming languages like FORTRAN 77. In informatics, several more flexible programming techniques are now available. They have evoked great interests in the development of generalization. Capability of these techniques lies in the explicit formulation of cartographic expert knowledge on purpose of establishing rulebases. Several systems for semi-automated generalizations are already available in the market Some Geographical Information Systems (GIS) provide limited functions for generalization, including line simplification, area dissolve and aggregation. More comprehensive systems offer the possibility of batch generalization of the streets, housing and city blocks at larger scales, such as the system CHANGE which is commercialized by Zeiss; other systems are entering the market which provide the tools for interactive generalization, such as MGE Map Generalizer developed by Intergraph. The increasing application of GIS-technology in many areas requires a more efficient computer-assisted generalization. According to Grünreich one of the solutions is to integrate generalization functions into the cartographic editor of a GIS-software package. A generalization editor emerges in this way and it is able to support the cartographer efficiently by doing interactive revision on the results of automatic generalization. According to Kilpelâinen and Sarjakoski, it has been widely discussed which kind of additional information, e.g. attribute data, should be stored in the GIS databases, as well as what kind of structure the dala should have. But there is a lack of studies how we can make real use of this attribute data. Most algorithms in the literature to date have focused on the generalization of line features in map projections; they are designed to operate in a two-dimensional Cartesian coordinate system. Cartographic products production systems that "capture" map features in non projected coordinates (latitude, longitude) or that produce cartographic databases as products in addition to projected map graphics require simplification and smoothing methods that are able to process non projected coordinates. According to Christian, one approach for generalizing non projected data is the adaptation of existing projected data line simplification and smoothing algorithms by the substitution of certain geometric computations. XI Furthermore Kilpelainen and Sarjakoski also mention that generalization tasks are often expressed in two dimensional domains, but in modelling reality, our attention focuses on the third dimension, which can be gathered from databases. Perhaps the next epoch for generalization research will include 3-dimensional generalization. In conclusion, as it can also be seen by the above explanations based upon authorities' opinions, there are more respects of generalization which must be considered in digital environment Each one can be said to be a stage to reach the goal of expert systems. The aspect of expert system will be embraced in accordance with feasibility of these. All of these, of course, is directly related to developments in computing technology. It is promising that there are enough researchers studying on this subject