Populating/Updating GIS: Automatically? ti Toni Schenk Department of Civil and Environmental Engineering and Geodetic Science The Ohio State University schenk.2@osu.edu T. Schenk IGERT Colloqium 2/22/08 1
Motivation Usefulness of GIS depends greatly on its information content. Information should be rich (satisfy current and future needs) accurate up-to to-date Keeping information up-to to-date may be biggest challenge how to detect changes how to update changes T. Schenk IGERT Colloqium 2/22/08 2
Motivation Populating and updating GIS manually is slow expensive Ever increasing amount of data poses challenge to extract t GIS information Can the processes of creating and updating GIS be automated? Your new GIS update is ready to be installed. Do you want to install it now? T. Schenk IGERT Colloqium 2/22/08 3
GIS Players surveying GIS photogrammetry t -populate remote sensing -update user data information knowledge Geospatial Information providers Geospatial Information managers T. Schenk IGERT Colloqium 2/22/08 4
Data sensor output, result of data acquisition raw signals or preprocessed data t is not an end product, nor does it directly provide answers to applications examples: pixels, laser points data needs to be (further) processed T. Schenk IGERT Colloqium 2/22/08 5
Information more specific than data to answer questions information can be retrieved from data information implicitly in data extracting information from data makes it explicit it feature extraction ti features are information primitives T. Schenk IGERT Colloqium 2/22/08 6
Knowledge elusive concept, ill-defined means different things to different people to deal withknowledgeitmustbe represented knowledge and its representation are closely related knowledge: facts, procedures, heuristics that t can be used to make inferences T. Schenk IGERT Colloqium 2/22/08 7
Data, Information, Knowledge semantic content data information knowledge signal representation symbolic T. Schenk IGERT Colloqium 2/22/08 8
Data, Information and GIS GIS contains explicit geospatial information about objects, e.g. a polygon is labeled with building, the vertices may be considered data The object is extracted from raw data, e.g. images Volume l of raw data >> volume of GIS information T. Schenk IGERT Colloqium 2/22/08 9
Example GIS of a rural county: 10,000 centerline segments 80,000 parcels 13,000 sewer line segments 130,000 spot elevation points 12,000 water line segments Input data: aerial images GSD 0.1 m County 25 x 25 miles 0.16 10e+12 pixels 3 x 2.5 x 0.16 10e+12 MB 180 MB 1.2 TB T. Schenk IGERT Colloqium 2/22/08 10
Data Deluge data information? T. Schenk IGERT Colloqium 2/22/08 11
From Data to Information and Knowledge in Human Vision i knowledge perception scene information data acquisition data T. Schenk IGERT Colloqium 2/22/08 12
Image Formation on Retina ~1 Mio ganglion cells raw information ~120 Mio photoreceptors raw data T. Schenk IGERT Colloqium 2/22/08 13
Paradigm data acquisition preprocessing raw data data feature extraction perceptual organization applications infor rmation kn nowledge T. Schenk IGERT Colloqium 2/22/08 14
Representational Issue image seen by computer image seen by human T. Schenk IGERT Colloqium 2/22/08 15
Spatial Extent: The Peephole 3x3 3 T. Schenk IGERT Colloqium 2/22/08 16
Connect-the the-dots Metaphor connect-the-dots algorithm Meaning of structure? C I I L N O T. Schenk IGERT Colloqium 2/22/08 17
Intelligent Data Acquisition add one new point: where should it be added to yield maximum information information gain 0 4-6 1 1? T. Schenk IGERT Colloqium 2/22/08 18
Example Image Ocean City, 6146 1368 x 1320 pixels pixelsize ~ 15 micron T. Schenk IGERT Colloqium 2/22/08 19
Examples of Edge Images Canny edge detector T. Schenk IGERT Colloqium 2/22/08 20
Examples of Edge Images T. Schenk IGERT Colloqium 2/22/08 21
Segmented Edges straight line segments length > 25 pixels T. Schenk IGERT Colloqium 2/22/08 22
Connecting Edge Segments connects edge segments with same line properties and small gap T. Schenk IGERT Colloqium 2/22/08 23
Connecting Edge Segments connects edges with larger gaps T. Schenk IGERT Colloqium 2/22/08 24
Parallel Edges distance between parallel l lines < 10 pixels T. Schenk IGERT Colloqium 2/22/08 25
Parallel Edges distance between parallel edges < 200 pixels T. Schenk IGERT Colloqium 2/22/08 26
Perpendicular Edges edges are grouped to perpendicular pairs, considering proximity T. Schenk IGERT Colloqium 2/22/08 27
Computed Corner Points corners are computed from perpendicular p edges T. Schenk IGERT Colloqium 2/22/08 28
Clustering Perpendicular Edges T. Schenk all perpendicular edges IGERT Colloqium 2/22/08 example of 4 clusters 29
Clustering Perpendicular Edges T. Schenk all perpendicular edges IGERT Colloqium 2/22/08 example of 3 clusters 30
From Data to Information and Knowledge in a GIS Environment data information query knowledg ge T. Schenk IGERT Colloqium 2/22/08 31
Information and Knowledge Generating Processes T. Schenk IGERT Colloqium 2/22/08 32
A World of Models graphics world visualiz zation GIS add. information reconstructed t world our world model world hypothesized world humans sensors physical world T. Schenk IGERT Colloqium 2/22/08 33
Concluding Remarks data < information < knowledge bottleneck in generating and updating GIS lies in information extraction from data and NOT in data collection end d products consist of information + knowledge from data to information by extracting features and their perceptual organization T. Schenk IGERT Colloqium 2/22/08 34
Concluding Remarks Over lifetime of GIS: cost of updating >> original creation Automate update Aspects of updating Detect changes Implement changes T. Schenk IGERT Colloqium 2/22/08 35
Concluding Remarks Additional updating considerations Update attributes of objects Precision Upgrading from 2-D to 3-D Missing information thank you T. Schenk IGERT Colloqium 2/22/08 36