Academic journal article Cartography and Geographic Information Science

A New Geospatial Overlay Method for the Analysis and Visualization of Spatial Change Patterns Using Object-Oriented Data Modeling Concepts

Academic journal article Cartography and Geographic Information Science

A New Geospatial Overlay Method for the Analysis and Visualization of Spatial Change Patterns Using Object-Oriented Data Modeling Concepts

Article excerpt

Introduction and problem definition

Many change-analysis approaches are documented in the literature--see for example, the overview articles from Coppin et al. (2004), Lu et al. (2004), Singh (1989), and Chen et al. (2012). Some of this research deals with post-classification methods, often termed as post-classification comparison (PCC). This paper focuses on PCCs based on classification results from remotely sensed imagery of different time stamps, but, in principle, it addresses change comparisons of any geospatial classification data sets. For such comparisons, pixel-by-pixel or object-by-object comparisons are usually conducted. Traditional geographic information system (GIS) overlay or intersect routines induce complete (and complex) matrices of change, which results in specific challenges such as generation of binary change images with no recognition of major and minor landscape changes, or a complex class-by-class-based evaluation (Ahlqvist 2008). Many processing steps have to be performed to reach aggregated and - what is often needed more high-level results of changes. Even if these steps are established as repeatable models or protocols, the analysis is limited by the restriction of relational data models in most GIS (cf. De Smith et al. 2007; Longley et al. 2005).

In this study, I make use of another approach which is theoretically formulated but not available in conventional GIS (cf. Chen, Wang, and Chen 2013): Egenhofer and Frank (1992) presented in a survey article concepts of object-oriented modeling applied to geographic data. Based on the original work by Brodie (1984), they described how an object-oriented data model is built on the four basic concepts of abstraction, namely classification, generalization, association, and aggregation. Based on these concepts, a comparison of different data sets (i.e., different classifications such as land use/land cover data sets) would theoretically overcome the aforementioned problems. The following selected concepts of object-oriented data modeling are taken from Egenhofer and Frank (1992) and are examined for their potential to improve the overlay processing in PCCs:

* Concept of generalization: Generalization in object-oriented modeling should not be confused with the same term used in cartography. In the former, it expresses the idea of grouping several classes of objects with common properties or operations into a more general superclass. For a PCC, such a super-class could represent a "change class" that has subsumed certain changes of interest (e.g., if the land-cover classes forest or scrubland are changing into bare soil or grassland, the concept of generalization can help to define super-classes such as "degradation", "urban sprawl," or similar). Such a concept of generalization needs to be flexible, that is, the subsuming of classes to a superclass should be easily changed if needed.

* Concept of association: Relationships between objects, also called grouping or partitioning. According to Egenhofer and Frank (1992), an example of an association in the GIS domain is neighborhood (e.g., there may be a relationship between a land parcel and an adjacent house lot). In the presented case of PCC, it could also be a relationship among objects through time, for example, a homogeneous land cover object is divided on the second time stamp into several objects (different classes) but the objects are still related and can be accessed (e.g., through a query).

* Concept of aggregation: Similar to association, aggregation is the combination of objects to form a higher level object (in the semantic sense). Egenhofer and Frank (1992) termed this higher level object an aggregate or composite object, with the aggregated parts keeping their own functionality. An example would be the aggregation of changes to a higher level information object. Such an object could be an artificial unit (for instance, a regular gridded or hexagon layer), and also an administrative unit where the temporal changes of the aggregated objects are "reported". …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.