20220824FOSS4G2022

Analysis of the spatiotemporal accumulation process of Mapillary data and its relationship with OSM road data: A case study in Japan
https://talks.osgeo.org/foss4g-2022-academic-track/talk/9KGNRA/

Transcript

1. 2022/08/24 FOSS4G 2022 Academic Track 1
   Analysis of the spatiotemporal accumulation process of
   Mapillary data and its relationship with OSM road data:
   A case study in Japan
   Toshikazu Seto* Ph.D. (tosseto)
   and Yuichiro Nishimura
   *: Associate Professor Department of Geography, Komazawa University
   OSGeo.JP (OSGeo Charter member) /OSMFJapan
   DOI: 10.5194/isprs-archives-XLVIII-4-W1-2022-403-2022
   A02
   KOMAZAWA UNIVERSITY
   Visual Identity Guidelines
   ϩΰλΠϓ
   ,ϚʔΫʴ࿨จϩΰλΠϓͷ૊Έ߹Θͤ
   ͸ɺ
   ࠨͷछͰ͢ɻ
   ԣ
   جຊܗͱ͠ɺ
   ༏ઌతʹ࢖༻͠·͢ɻ
   ԣ
   ϫϯϙΠϯτͳͲɺ
   ʮԣʯ
   ͕഑ஔ͠ʹ͍͘
   ৔߹ʹ࢖༻͠·͢ɻ
   ॎ
   ॎܕαΠϯͳͲɺ
   ࡉ௕͍഑ஔʹ࢖༻͠·͢ɻ
   ඞͣϚελʔσʔλΛ࢖༻͍ͯͩ͘͠͞ɻ
   ࠨهҎ֎ͷ૊Έ߹ΘͤΛ࡞੒͠ͳ͍Ͱ
   ͍ͩ͘͞ɻ
   ॎ
   

   View Slide

2. 2022/08/24 FOSS4G 2022 Academic Track 2
   Overview of Mapillary Data in Japan
   Property Description Type Range of value
   key Keycode String 41,765,634
   ca Cardinal points Real 0 – 360 degrees
   camera_make Device maker String 91 types
   camera_model Device name String 488 types
   captured_at Timestamp Date
   from 2014-01-01
   to 2020-09-31
   pano
   Panoramic
   shooting
   Boolean
   Yes: 5,413,030
   No: 36,352,604
   sequence_key Sequence code String 41,765,634
   user_key Account code String 1,513
   username Account name String 1,513 users
   Mapillary API (v. 3, 2017)
   

   View Slide

3. 2022/08/24 FOSS4G 2022 Academic Track 3
   Database Construction
   1. Get data from Mapillary Search Image API (v. 3, 2017)
   ▶ GeoJSON（Zoom7 – total: 14h15m10s）
   2. Convert to GeoJSON ▶ PostGIS ▶FlatGeobuf (.fgb)
   • PostGIS: 1.79-GB ▶ .fgb: 11.73-GB with 41,765,634 records
   • Reading & processing .fgb was three times faster than local PostGIS
   3. Spatial Join with OSM road data attribution via QGIS
   • All OSM road data (.fgb): 3.2-GB with 8,313,281 records
   • the maximum search radius was set to 50 m of centerline
   4. Output 17.6-GB point database in .fgb (with 23 attribution)
   + Aggregate with polygon for each municipality
   (232.2 MB; 1903 records with 32 attribution)
   + Aggregate with 1-km Grids
   (126.5 MB; 388,762 records with 20 attribution)
   • https://github.com/tossetolab/mapillary-analysis-japan
   

   View Slide

4. 2022/08/24 FOSS4G 2022 Academic Track 4
   • 1500 unique Mapillary contributors
   during the six years, with the top 10
   users contributing approximately 90% of
   the images in Japan.
   • The proportion of Mapillary data
   compared to OSM road data has
   reached about 16% in some areas.
   • Approximately 40% of Mapillary related
   OSM roads have been updated since
   the Mapillary images were taken, which
   can help update OSM road data in Japan.
   Number of Mapillary
   postings per contributor
   OSM roads-Mapillary
   coverage rates
   Monthly Mapillary
   photos taken and
   number of contributed
   users.
   

   View Slide

5. 2022/08/24 FOSS4G 2022 Academic Track 5
   Last update date of OpenStreetMap road before
   and after Mapillary image taken by 1-km grid
   (unit: days)
   

   View Slide