Using control data to determine the reliability of volunteered geographic information about land cover

Comber, Alexis; See, Linda; Velde, M. van der; Perger, Christoph; Foody, Giles M.

doi:10.1016/j.jag.2012.11.002

Cited by 118 publications

(100 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The remaining 30% of tiles were used to validate the obtained LULC map. Here, we compare the classification result at each tile of the validation location [13]. Table 2 shows the structure of a confusion matrix between the classification results from LULC classes from crowdsourced data and LULC classes of the Landsat TM5 image.…”

Section: Application Of Majority Class Filter Determination Of Crowdsmentioning

confidence: 99%

Exploring Land Use and Land Cover of Geotagged Social-Sensing Images Using Naive Bayes Classifier

et al. 2016

View full text Add to dashboard Cite

Abstract:Online social media crowdsourced photos contain a vast amount of visual information about the physical properties and characteristics of the earth's surface. Flickr is an important online social media platform for users seeking this information. Each day, users generate crowdsourced geotagged digital imagery containing an immense amount of information. In this paper, geotagged Flickr images are used for automatic extraction of low-level land use/land cover (LULC) features. The proposed method uses a naive Bayes classifier with color, shape, and color index descriptors. The classified images are mapped using a majority filtering approach. The classifier performance in overall accuracy, kappa coefficient, precision, recall, and f-measure was 87.94%, 82.89%, 88.20%, 87.90%, and 88%, respectively. Labeled-crowdsourced images were filtered into a spatial tile of a 30 m × 30 m resolution using the majority voting method to reduce geolocation uncertainty from the crowdsourced data. These tile datasets were used as training and validation samples to classify Landsat TM5 images. The supervised maximum likelihood method was used for the LULC classification. The results show that the geotagged Flickr images can classify LULC types with reasonable accuracy and that the proposed approach improves LULC classification efficiency if a sufficient spatial distribution of crowdsourced data exists.

show abstract

Section: Application Of Majority Class Filter Determination Of Crowdsmentioning

confidence: 99%

Exploring Land Use and Land Cover of Geotagged Social-Sensing Images Using Naive Bayes Classifier

et al. 2016

View full text Add to dashboard Cite

show abstract

“…Hecht et al analyzed the completeness of building footprints in OSM by comparing the OSM data with the official data in Germany [10]; Fan et al evaluated the quality of OSM building footprint data in terms of completeness, semantic, position and shape accuracy using ATKIS data as reference data [11]. Comber et al evaluated the reliability of volunteered land cover using GLC-2000, GlobCover and MODIS V5 as control data [12]. In the above analyses, almost all of the researchers concluded that although OSM can offer a large amount of useful data with high responsiveness and flexibility, its main limitation is the irregularity of the data's completeness.…”

Section: Related Workmentioning

confidence: 99%

Dynamically Integrating OSM Data into a Borderland Database

Zhou

Zeng

et al. 2015

IJGI

View full text Add to dashboard Cite

Spatial data are fundamental for borderland analyses of geography, natural resources, demography, politics, economy, and culture. As the spatial data used in borderland research usually cover the borderland regions of several neighboring countries, it is difficult for anyone research institution of government to collect them. Volunteered Geographic Information (VGI) is a highly successful method for acquiring timely and detailed global spatial data at a very low cost. Therefore, VGI is a reasonable source of borderland spatial data. OpenStreetMap (OSM) is known as the most successful VGI resource. However, OSM's data model is far different from the traditional geographic information model. Thus, the OSM data must be converted in the scientist's customized data model. Because the real world changes rapidly, the converted data must be updated incrementally. Therefore, this paper presents a method used to dynamically integrate OSM data into the borderland database. In this method, a basic transformation rule base is formed by comparing the OSM Map Feature description document and the destination model definitions. Using the basic rules, the main features can be automatically converted to the destination model. A human-computer interaction model transformation and a rule/automatic-remember mechanism are developed to interactively transfer the unusual features that cannot be transferred by the basic rules to the target model and to remember the reusable rules automatically. To keep the borderland database current, the global OsmChange daily diff file is used to extract the change-only information for the research region. To extract the changed objects in the region under study, the relationship between the changed object and the research region is analyzed considering the evolution of the OPEN ACCESS ISPRS Int. J. Geo-Inf. 2015, 4 1708 involved objects. In addition, five rules are determined to select the objects and integrate the changed objects with multi-versions over time. The objects' change-type evolution is analyzed, and seven rules are used to determine the change-type of the changed objects. Based on these rules and algorithms, we programmed an automatic (or semi-automatic) integrating and updating prototype system for the borderland database. The developed system was intensively tested using OSM data for Vietnam and Pakistan as the experimental data.

show abstract

“…This generated a value in the range [0, 1] at each of the 80,073 grid locations for each class, and the class with the greatest value was assigned. This approach is a smoothing approach similar to that used by Comber (2013a) to determine fuzzy memberships distributions.…”

Section: Figure 2 the Kde Surfaces For 4 Classesmentioning

confidence: 99%

“…Geo-Wiki collects land cover data from volunteers and a number of applications have been developed to, for example, assess the quality of existing land cover products (Fritz et al, 2009), determine their uncertainties (Fritz et al, 2011) and generate hybrid global land cover maps . However, one of the critical issues associated with the use of citizen data relates to its quality (Foody et al, 2013;Comber et al, 2013a). One key problem is that different contributors or volunteers may have different underlying conceptualisations of the features that observed and thus recorded in crowd-sourced data.…”

Section: Introductionmentioning

confidence: 99%

COMPARING NATIONAL DIFFERENCES IN WHAT PEOPLE PERCEIVE TO BE THERE: MAPPING VARIATIONS IN CROWD SOURCED LAND COVER

Comber

Mooney

Purves

et al. 2015

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:This paper describes a simple comparison of the distributions of land cover features identified from volunteered data contributed by different social groups -in this case comparing two groups of Geo-Wiki campaigns. Understanding the impacts on analyses of citizen science data contributed by different groups is critical to ensure robust scientific outputs and to fully realise the potential benefits to formal scientific research. It is well known that different people, with different backgrounds and subject to different cultural factors, hold varying landscape conceptualisations. This paper analyses volunteered geographical information on land cover to generate land cover maps. It uses a geographically weighted approach to generate land cover mappings. The mappings generated by different groups (in this case a from a specific unnamed country) are compared and the results show how the predicted land cover distributions vary, with large differences in some classes (e.g. Barren land, Shrubland, Wetland) and little difference in others (e.g. Tree cover). This suggests that for some landscape features cultural and national differences matter when it comes to using crowdsourced data in formal scientific analyses and highlights the potential problems of not considering contributor backgrounds in citizen science. This is important because such data re now routinely being used to develop global land cover data, to generate uncertainty estimates of existing global land cover products and to generate global forest inventories. These in turn are being suggested as suitable inputs to such things as global climate models. A number of critical research directions arising from these findings are discussed.

show abstract

Using control data to determine the reliability of volunteered geographic information about land cover

Cited by 118 publications

References 26 publications

Exploring Land Use and Land Cover of Geotagged Social-Sensing Images Using Naive Bayes Classifier

Exploring Land Use and Land Cover of Geotagged Social-Sensing Images Using Naive Bayes Classifier

Dynamically Integrating OSM Data into a Borderland Database

COMPARING NATIONAL DIFFERENCES IN WHAT PEOPLE PERCEIVE TO BE THERE: MAPPING VARIATIONS IN CROWD SOURCED LAND COVER

Contact Info

Product

Resources

About