Towards and adaptable spatial processing architecture

268 views

Published on

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
268
On SlideShare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
9
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Towards and adaptable spatial processing architecture

  1. 1. TOWARD AN ADAPTABLE AND INTEROPERABLE SPATIAL PROCESSING ARCHITECTURE By Dr. J. Armando Guevara (c) 1989aspa 1989
  2. 2. IntroductionA geographic information system (GIS) is composed of a set of building blocks termedgeographic information system procedures (GISP). A GISP is an abstract algorithmicfunction of a GIS that allows one to input, process and output describable elements froma spatial data structure (SDS) and/or spatial data base. Based on this, a GIS can bedefined as a model composed of a set of objects (the spatial data structures) and a set ofoperators (the GISP) that perform transformations and/or queries on the spatial objects.The elements modeled by a GIS are generally imbedded in two-dimensional space and, insome instances, in two and one-half, and three-dimensional space. In addition to theelements it manipulates being spatially located, the elements themselves possess a set ofattributes that can be qualitative or quantitatively defined. These attributes can not onlygive a description of the spatial elements, but can also become a time component(changes in time) of the spatial data base. Given this particular nature, GISP must be ableto both query/transform the spatial elements and the attributes associated with thoseelements.GISP are categorized to be the primitive operators of a GIS. In this sense, GISP can be: a) SELECTORS - Capture and select spatial data. b) RECOGNIZERS - Structure/search spatial data. c) PROCESSORS - Process spatial data. d) TRANSFORMERS - Output spatial data.Part of the complexity of designing a GIS and its associated GISP comes from themultiple target applications that can exist for such a system. To control this complexityand maintain its genericness, a GIS must suffice answers to the following interoperabilityconcepts: 1. Functional interoperability: hardware and OS functional independence. 2. Data base interoperability: data format independence. 3. Data structure interoperability: the coexistence of vector, lattice, and raster data structures under one data model. 4. Knowledge interoperability: the utilization of artificial intelligence techniques to create data base model usage schemes and create application procedures. 5. Human interface interoperability: GUI platform independence. 6. Data transfer interoperability: the ability of a GIS to exist and transfer data independent of the hardware platforms.The above notions, each one, are a complete treaty dealing with digital spatial knowledge,digital spatial representation, digital spatial data structures, and digital spatial algebra’s. Iemphasize the term digital because I feel that, although human cognition of space canaspa.doc 2(c) J. Armando Guevara - 1989
  3. 3. give insight into its digital representation, it may not necessarily yield the appropriateanswers needed, for example, to robotic perception.The manner in which a human finds a path can be drastically different from the way analgorithm/heuristic will navigate and discriminate a set of geometric objects to find itsway.As knowledge has been gained on the behavior of spatial algorithms, a functionalcategorization emerged (functional breakdown) that has established the basis forfunctional interoperability.Functional interoperability refers to the ability of a GIS to be able to access any data set(or portion thereof, in a seamless data base) and apply operators without the system losingtrack of the data environment and history of the operations performed. Functionalinteroperability would allow access to all GISP within one process environment.Although this has a tremendous power, it is not without its user interface complications.Further, functional interoperability has implications that at design level force to separatethe functional domain of input, process, and output, into three distinct and separatefunctional blocks that “talk to each: via some pre-established or “standard”communications protocol.This functional categorization gives way to the basic architecture of a GIS: data input,data base management, analysis and output. Within each category, different means havebeen created to handle the user interface. Menu and command driven functions havebecome the main ways of interaction. These functions have a proper protocol tointernally deal with processes. Some are action driven while others are environmentdriven.Action driven functions produce an immediate feedback to the user (e.g., draw a map).Environment driven functions have a cumulative effect that ends in an action drivenfunction. Action driven functions are easy to explain and use. Environment drivenfunctions pose a variety of user interface problems.A functional interoperable GIS would be mostly environment driven. Such a systemwould require knowledge of environment and function tracking procedures. Recognizingthe importance of user interaction with a GIS is the concern of the Human InterfaceInteroperability Principle and is key to the life appreciation or depreciation of the system(how easy or complicated it is to learn and use).2. GIS Architectural ModelsOne of the most important concepts introduced in GIS that allows the user to digitallymodel spatial relations is that of topology. The various data structures introduced toaspa.doc 3(c) J. Armando Guevara - 1989
  4. 4. handle geographic data (Morton sequences, Peano curves, quad trees, R-trees, B-trees,etc.) and their operational data definition (vector, raster, lattice) are mechanisms to bridgeconcepts of space and processes to its implementation.To achieve true interoperability at the functional and the data levels, the design of ageneric GIS system must take into account the integration and management of all datatypes (data structures). This would allow a GIS system to handle planimetric data,surface data and imagery data. Additionally, this notion of spatial interoperability wouldhide the physical representation of the model, allowing for multiflexible ways of handlingthe model.The ultimate task of a GIS is to model some aspect of a spatial reality. The model shouldinclude enough information that would allow its user to obtain answers to queries andinfer situations that otherwise would not be possible. Two types of architectures can beidentified:a) a generic functional modelb) a specific derived modelA generic functional model (GFM) is an architecture made of basic spatial operators thathandle basic spatial primitives like points, lines, areas and surfaces. The model holdsdescriptive data about the primitives but does not know about existing relationships. Themodel is functionally driven i.e., any further inference about the data aside from primitivelocation and basic description is obtained via the spatial operators. The GFM is an openmodel that requires only very basic knowledge about the spatial primitives being stored.Access to this model architecture is done via an internal access programming interface(API) with a well defined protocol of input and output.A specific derived model (SDM) is an architecture created by building new constructs(higher level data types and more complex operators) from established relationshipsamong the spatial primitives and a linkage is created among compounded spatialprimitives and their descriptive data. The SDM requires a well-understood knowledge ofhow the GIS is going to be used and what it is going to model. The SDM in principlereflects the domain of the target application. The mechanism to enable the SDM is a highlevel programming language that implements specific constructs that emulate a “spatiallanguage”. This spatial language also has its own API.The relational approach to spatial data handling falls under the GFM category, while theobject oriented approach falls under the SDM. It is important to understand that thesetwo models are not mutually exclusive, i.e. a GFM can be used to support a SDM.However, the absence of an underlying GFM in a SDM raises flexibility and performanceissues.aspa.doc 4(c) J. Armando Guevara - 1989
  5. 5. The GFM has the following characteristics:1. It comprises a complete taxonomy of spatial operators. This taxonomy is based upon the spatial data primitive domains and is aggregated toward more complex domains. For example, a line is made of points. Applying operators to these lines and points can yield constructs such as path networks or irregular triangular networks. The domain of the operators have a basic taxonomy that has at its main nodes in the software tree structure input, process and output. Input can then be further categorized as digitizing table, scanner, file, etc. Yet the details of the input stream are “hidden” from the Input API.2. It should allow for dynamic relationship construction via spatial operators. The case above of points/lines => TIN, or points/splines => COGO. Topology is another example of building explicit relationship via operators.3. Compounding of spatial primitives should be done efficiently without restrictions or constraints. The compounding can still yield a more complex GFM.Internally, the GFM follows a similar structure to that described in the Digital Line Graph- extended, with the exception that at the lowest level of the model, relationships are not explicitly stored.The SDM has the following characteristics:1. Relationships between spatial primitives are pre-established in the model based on behavioral, procedural, and transactional facts. These facts make the SDM schema, the process rules or the application requirements.2. Mutations on the spatial primitives must be done efficiently via API using high level standard languages. Mutations such as aggregation (compounding) and disaggregation (uncompounding) can still yield a more complex or simpler SDM.The SDM would be the basic model for object oriented transactions, user interfaces basedon graphical components and seamless environments where the application must access avariety of data types across multiple platforms. Both the GFM and the SDM must allowfor the following types of data base queries:a) Spatial Context: given an unambiguous geometric definition extract from the data base, all elements selected by the geometric definition.b) Spatial Conditional Context: given an unambiguous geometric definition and a condition expressed in terms of the stored descriptive data, extract from the data base all elements selected by the geometric definition and that suffice the descriptive condition given.c) Descriptive Context: given a descriptive data element, extract for the data base all elements that match the one given.d) Descriptive Conditional Context: given a descriptive data element and a condition expressed in terms of the given element, extract from the data base all elements selected that suffice the descriptive condition given.aspa.doc 5(c) J. Armando Guevara - 1989
  6. 6. The conjunction of a GFM and a SDM would give the user the ability to perform spatialoperations at various levels of complexity and integration. GFM and SDM bring theability for a GIS to be flexible and application independent.Finally, both the GFM and the SDM must maintain the data base interoperability concept(i.e., preserve the notion of a continuous physical space underlying the data model acrossmultiple platforms).3 Toward an Adaptable Spatial Processing ArchitectureA modern GIS is expected to be able to integrate a different variety of data sources, whichwill be used in many ways and also under a wide range of support decision making. Thenature of separate user views of the same data base accompany a series of (sometimesconflicting) demands to the GIS designer that must somehow be met to guarantee theusefulness and longevity of the system. In synthesis, a GIS is a multidisciplinary tool thatmust allow for interdisciplinary support. Specialized spatial information systems are notmultidisciplinary tools, thus are very restrictive in regards to what can be done with them.An Adaptable Spatial Processing Architecture (ASPA) is what is needed to meet thedemands of both multidisciplinary and specialized applications. ASPA fundamentals arebased on a GFM that has a set of functional (GISP) primitives clearly defined that allowsthe automatic construction of a SDM. ASPA would be an expert monitor based on a highlevel language consisting of spatial operators that have definable hierarchical constructs.These spatial operators can be organized following a programmable schema that wouldallow them to generate the SDM. ASPA would work in conjunction with a data basemanagement system (DBMS). The DBMS would respond to both spatial and nonspatialoperators. The heart of ASPA and the DBMS would be a GFM.The spatial operators and spatial data structures that ASPA are built upon are based onthe five basic software engineering principles of modularity, encapsulation, localization,uniformity, and confirmability applied through the concept of abstraction at the designlevel of the SDM. Key to ASPA is functional and data interoperability.The interactions across the GFM and the SDM will be accompanied by rules governingthe interrelations between them. There are two important rules borrowed from theconcept of levels of abstraction. The first concerns resources: each level has resourceswhich it owns exclusively and which other levels are not permitted to access. The secondinvolves the hierarchy: lower levels are not aware of the existence of higher levels and,therefore, may not refer to them in any way. Higher levels may appeal to the externalfunctions of lower levels to perform tasks and also appeal to them to obtain informationcontained in the resources of the lower levels. This abstraction is effectively implementedand managed via a well defined API.aspa.doc 6(c) J. Armando Guevara - 1989
  7. 7. In this respect, the lowest level of abstraction is composed of the GFM, a clearly definedset of spatial operators (selectors, processors, recognizers, transformers) and a DBMS.A GIS must be data and system (OS, Hardware) independent. Multiple applicationtargets should be allowed between the GFM, SDM, and any external to the GIS functionaland data source. Similarly, the levels of abstractions induced in the GISP should allowthe GIS to perform identically on different computers with no user intervention whendoing the application targets.By separating input, process and data, and establishing a clear communication protocol(API) the GFM and SDM architectures provide for a mechanism to target end-users andapplications that are not purely GIS in nature but require of some kind of spatialoperations.4. ConclusionGIS technology has finally taken off in the eyes of the world; but it really is only thebeginning of a great adventure!As GIS users become more knowledgeable and demanding, the strength and goodness ofa design is truly discovered. The notions of interoperability presented above are veryimportant issues that need to be covered for a successful design. In my experience alongwith the internal algorithmic robustness and data base consistence and integrity,flexibility and user friendliness (magic words) are today the most relevant points to beconsidered from the outcome of what is to be an adaptable spatial processing architecture.We should avoid making the mistake made during the 70’s when authors entrenchedthemselves about whether raster data structures were better than vector structures. Noneand both were the answers. As we move toward more sophisticated systems and users,we must not lose track of the flexibility geographic information systems must have. GISare multidisciplinary tools. Fixed schemes will hinder GIS usage.The GFM and SDM architectural approaches, implemented with spatial operators thatcommunicate to the input (interface of the application) and output (say cartographicproduction) via a well defined API, should provide the most portable, interoperable andflexible mechanism for spatial tool construction. It is most likely that as GIS evolves, thenature of spatial data handling will evolve into more everyday type of applications. Forthat to happen, new views must be taken into consideration on how we design and buildthese spatial tools.aspa.doc 7(c) J. Armando Guevara - 1989

×