Validation of Volunteered Geographic Information with Manual and AI-assisted Mapping Toshikazu SETO* and Yuichiro Nishimura** A02 KOMAZAWA UNIVERSITY Visual Identity Guidelines ΫʴจϩΰλΠϓ ,ϚʔΫʴจϩΰλΠϓͷΈ߹Θͤ ɺ ࠨͷछͰ͢ɻ ԣ جຊܗͱ͠ɺ ༏ઌతʹ༻͠·͢ɻ ԣ ϫϯϙΠϯτͳͲɺ ʮԣʯ ͕ஔ͠ʹ͍͘ ߹ʹ༻͠·͢ɻ ॎ ॎܕαΠϯͳͲɺ ࡉ͍ஔʹ༻͠·͢ɻ ඞͣϚελʔσʔλΛ༻͍ͯͩ͘͠͞ɻ ࠨهҎ֎ͷΈ߹ΘͤΛ࡞͠ͳ͍Ͱ ͍ͩ͘͞ɻ ܗʣ ϙΠϯτʣ ॎ *: Associate Professor: Komazawa University Center for Spatial Information Science, the University of Tokyo **: Professor: Nara Women’s University [email protected]
OSM/VGI quality assessment • Quattrone et al., 2016 • Analysis of country-level update frequency and tag (uid, changeset, timestamp, version, lat, lon, taglist) • Anderson et al., 2019 • Focus on corporate editors using OSM-QA tiles • Raifer et al., 2019 • Development analytical platform and ohsome-API for OSHDB: Data Aggregation, Data Extraction, User stat • Minghini and Frassinelli, 2019 • “Is OSM up-to-date” date of creation (first edit), date of last edit, number of versions (revisions), number of different contributors who edited that node or way, frequency of update → Participatory GIS, community geography…
in OSM in recent years https://www.openstreetmap.org/user/mvexel/diary/400035 • Quality improvement and monitoring as geospatial data • Provision of automatic mapping technology by AI and other means and improvement of accuracy • Data assessment, especially for local daily activities • Response to vandalism data destruction (Juhász et al., 2020) • Emergence of diverse actors and acceptance in the active community • Commercial: rise of corporate members and advancements in AI technology (Anderson et al., 2019) • Humanitarian: gender gap issues, local contributors and young development (Solís & Zeballos, 2023)
(OSMCha) https://github.com/OSMCha/osmcha • Python package to automatically detect suspicious OSM changesets. • It was designed to be used with osmcha-django, but also can be used standalone or website via License: GPL-3.0.
period/data: Changeset (195,416 records) for one year from January 1 to December 31, 2023 • Extraction range: Area that includes Japan in the scope of coverage • Geometry indicates the edit range (rectangle) of the changeset. • Although there are a few that include the area around Japan and the entire world in the scope, for analysis purposes, it is limited to logs targeting Japan. • Acquire changeset logs for the relevant period from the API in GeoJSON format and map them using QGIS, etc. • https://osmcha.org/api/v1/changesets/ • In addition to basic OSM attributes (number of data edits, OSMuser, edit date/time, satellite images used, etc.), and OSMCha's own detection reasons (fragment) is included. • In this study, we mainly focus on OSMCha's “suspicious OSM editing” tag and explore its characteristics in terms of trends of OSM editing in Japan.
OSMCha (in part) • Possible Import • Tag a changeset as a possible import if the number of created elements is greater than 70% of the sum of elements created, modified and deleted and if it creates more than 1000 elements or 200 elements case it used one of the powerfull editors. • Mass Modification • Consider a changeset as a mass modification if the number of modified elements is greater than 70% of the sum of elements created, modified and deleted and if it modifies more than 200 elements. • Mass Deletion • All changesets that delete more than 1000 elements are considered a mass deletion. If the changeset deletes between 200 and 1000 elements and the number of deleted elements is greater than 70% of the sum of elements created, modified and deleted it's also tagged as a mass deletion. • New mapper • Verify if the user has less than 5 edits or less than 5 mapping days. • User has multiple blocks • Changesets created by users that has received more than one block will be flagged. https://github.com/OSMCha
14 51 70 % typ % 3% 20 D% % D% C% 2% te 01 01 % 01 01 A% Fore example, “mapwithai” case • A case where a check user pointed out that a “large number” of buildings have been added using Rapid and that the base map is out of date, resulting in misalignment. • There are also overlapping buildings and road/building intersection errors.
example, “harmful” case It was pointed out that the road data was edited in the wide area and the tag indicating the direction of travel was incorrect.
Works • Japan is characterized by a large contribution from New Mapper, it is undeniable that it may contain a certain amount of errors due to non-training. However, not all edits by New Mapper are suspect. • AI mapping by Rapid has also been actively incorporated, but even with these, errors have occurred. However, the spatial editing extent tends to be smaller than other errors. • In addition, there were some differences in the detection tendency of “suspicious editing” online mapping and field surveys. • Detection rules by OSMCha are detailed, so further case studies are important. • After the period under study, Japan experienced a major disaster and crisis mapping was conducted. This trend needs to be analyzed using similar methods. • AI implementation and automation in OSM is still in its infancy, and accumulation of research in the field of geography is expected in terms of mapping with geographical knowledge.
tosseto
yatabe
vyadav
debug_mode
robintux
territorium
sciencetokyo
akiraasano
signer
datarian
sabderemane
speakerdeck
jcasabona
tenderlove
kneath
rkendrick25
akademiya2063
reverentgeek
davidcarrasco
thirion
sergeychernyshev
soysaucechin
