Using uncertain conceptual spaces to translate between land cover categories

Abstract

To support diversified uses of geographical information there is a need for enhanced spatial data infrastructures to create interoperability between users and prod.ucers of geographic data. One important interoperability problem is caused by dlfTerences in data semantics, for example heterogeneous land use/land cover ~lassification systems. A critical review of an existing method for semantic mteroperability between land cover classifications is used to motivate and introduce a modified framework based on conceptual spaces and rough fuzzy sets. Land cover categories are defined by a set of defining attributes formally represented as a collection of rough fuzzy membership functions and importance weights. This parameterized representation is used to translate between the US Natural Vegetation Classification Standard and the European CORINE Land Cover system based on evaluations of different aspects of semantic similarity between categories. The results demonstrate that the use of different similarity metrics in a conceptual space, together with the explicit rough fuzzy uncertainty representation, increases the semantic separation between land cover categories. Diagrams and maps illustrate the information that can be gained from the semantic similarity assessment. These developments open new possibilities to explore semantic relationships between concepts, both within a classification and between classifications used in different contexts.

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation – Part 2: Validation

“…To reveal the inherent ill-posedness of “conceptual matching” between two categorical variables A and B investigated by Ahlqvist (2005) (refer to the Part 1, Chapter 4), one co-author of this paper, different from the independent human expert (refer to Acknowledgments), conducted a second inherently equivocal selection of “correct” entry-pairs in a binary relationship, R: A ⇒ B ⊆ A × B, to guide the interpretation process of the OAMTRX = FrequencyCount(A × B) instance shown in Table 4. This second experiment provided an mDMI pair of O-Q 2 Is equal to OA(OAMTRX = FrequencyCount(A × B)) = 97.28% ± 0% and CVPAI2(R: A ⇒ B) = 0.6731.…”

“…Chapter 2 describes materials including the SIAM computer program, the time-series of annual WELD image composites, the reference USGS NLCD 2006 map and the EPA Level III ecoregion map of North America. Methods, specifically, an original protocol to compare without sampling the test SIAM-WELD map and the reference USGS NLCD 2006 map of the CONUS, whose map legends do not coincide and must be harmonized (reconciled, associated, translated) (Ahlqvist, 2005), is proposed in Chapter 3. Experimental results are presented in Chapter 4 and discussed in Chapter 5.…”

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 1: Theory

“…Ahlqvist writes that “to negotiate and compare information stemming from different classification systems (Bishr, 1998; Mizen, Dolbear, & Hart, 2005)… a translation can be achieved by matching the concepts in one system with concepts in another , either directly or through an intermediate classification (Feng & Flewelling, 2004; Kavouras & Kokla, 2002)” (Ahlqvist, 2005). Stehman describes four common types of thematic map-pair comparisons (Stehman, 1999).…”

“…Whereas a large portion of the RS community appears concerned with the aforementioned first type of map comparisons exclusively, the protocol proposed in (Baraldi et al, 2014) focuses on the second type, which includes the first type as a special case. In Couclelis (2010), the author observed that inter-dictionary concept matching (“conceptual matching”) (Ahlqvist, 2005) is an inherently equivocal information-as-data-interpretation process (Capurro & Hjørland, 2003), see Table 3. In common practice, two independent human domain-experts (cognitive agents, knowledge engineers) are likely to identify different binary associations between two codebooks of codewords (Laurini & Thompson, 1992).…”

“…Chapter 3 reviews the long history of prior knowledge-based decision trees for MS color naming presented in the RS literature. To cope with thematic map legends that do not coincide and must be harmonized (reconciled, associated, translated) (Ahlqvist, 2005), such as dictionaries of MS color names in the image-domain and LC class names in the scene-domain, Chapter 3 proposes an original hybrid inference guideline to identify a categorical variable-pair relationship, where prior beliefs are combined with additional evidence inferred from new data. An original measure of categorical variable-pair association (harmonization) in a binary relationship is proposed in Chapter 4.…”