4 Progressive Techniques
69
The set of spatial objects that need to be represented in a map depends on its level
of detail. In this context, the term ‘detail’ refers both to the amount of entities contained in the data set and to the information stored about such entities, for example
their geometry [24]. Therefore, changes of detail can be caused by the addition or
elimination of some entities, as well as by the refinement or coarsening of the representation of existing entities (e.g. change of dimension or shape).
Limited bandwidth makes the (one-step) transmission time of large, detailed vector maps unacceptable for real-time applications. However, as downloaded maps may
contain more detail than users require, progressive transmission has been identified
as a viable solution: a less detailed version can be sent first and then refined upon
user request.
A coarser map version can be generated by applying map generalization techniques to a fully detailed map [9, 31, 32, 34, 38, 42, 43, 48]. This may lead to a
remarkable reduction of data. If the extracted representation satisfies users’ requirements, an improvement is obtained both from a transmission point of view (a smaller
file is sent) and from an information overload point of view (only relevant data is
received and displayed).
However, there are still several impediments to the online application of map
generalization: first, the cartographic principles on which it is based have not been
completely formalized yet. Secondly, it is a very complex and time-consuming process that currently still involves interaction between semi-automatic solutions and
expert cartographers (see [48] for a survey of map generalization techniques and
operators).
Several tools for map generalization have recently been developed, based on the
application of object-oriented principles, expert systems, neural networks, and casebased reasoning [31, 42, 49].
A critical issue in map generalization is preservation of consistency (e.g. topological constraints) while decreasing the level of detail. For example, random selection
of objects (to be eliminated or simplified) does not usually generate a topologically
consistent representation at coarser detail. Consistency is an essential property for
data usability (see also Chap. 8): the answer to a query on the generalized map must
be consistent with the answer obtained when the same query is applied to the original
(fully detailed) map. Most of the proposed generalization algorithms do not guarantee the preservation of this property. Therefore they often require the application of
a posteriori checks to rectify introduced inconsistencies.
With these issues in mind, researchers have studied the realization of Web mapping systems based on different types of solutions.
Cecconi and Galanda [11] have summarized the different scenarios in which
users may issue a request to a Web mapping system. These include the following:
• On-the-fly Mapping Request: The user wants to download a first map for an
overview of the requested information. The waiting time should be short and
generalization only partially applied.
69
The set of spatial objects that need to be represented in a map depends on its level
of detail. In this context, the term ‘detail’ refers both to the amount of entities contained in the data set and to the information stored about such entities, for example
their geometry [24]. Therefore, changes of detail can be caused by the addition or
elimination of some entities, as well as by the refinement or coarsening of the representation of existing entities (e.g. change of dimension or shape).
Limited bandwidth makes the (one-step) transmission time of large, detailed vector maps unacceptable for real-time applications. However, as downloaded maps may
contain more detail than users require, progressive transmission has been identified
as a viable solution: a less detailed version can be sent first and then refined upon
user request.
A coarser map version can be generated by applying map generalization techniques to a fully detailed map [9, 31, 32, 34, 38, 42, 43, 48]. This may lead to a
remarkable reduction of data. If the extracted representation satisfies users’ requirements, an improvement is obtained both from a transmission point of view (a smaller
file is sent) and from an information overload point of view (only relevant data is
received and displayed).
However, there are still several impediments to the online application of map
generalization: first, the cartographic principles on which it is based have not been
completely formalized yet. Secondly, it is a very complex and time-consuming process that currently still involves interaction between semi-automatic solutions and
expert cartographers (see [48] for a survey of map generalization techniques and
operators).
Several tools for map generalization have recently been developed, based on the
application of object-oriented principles, expert systems, neural networks, and casebased reasoning [31, 42, 49].
A critical issue in map generalization is preservation of consistency (e.g. topological constraints) while decreasing the level of detail. For example, random selection
of objects (to be eliminated or simplified) does not usually generate a topologically
consistent representation at coarser detail. Consistency is an essential property for
data usability (see also Chap. 8): the answer to a query on the generalized map must
be consistent with the answer obtained when the same query is applied to the original
(fully detailed) map. Most of the proposed generalization algorithms do not guarantee the preservation of this property. Therefore they often require the application of
a posteriori checks to rectify introduced inconsistencies.
With these issues in mind, researchers have studied the realization of Web mapping systems based on different types of solutions.
Cecconi and Galanda [11] have summarized the different scenarios in which
users may issue a request to a Web mapping system. These include the following:
• On-the-fly Mapping Request: The user wants to download a first map for an
overview of the requested information. The waiting time should be short and
generalization only partially applied.