SlideShare a Scribd company logo

Crowd sourcing gis for global urban area mapping

•
•
Hiroyuki Miyazaki
Hiroyuki Miyazaki

1) Crowd-sourcing is proposed as a method to globally map urban areas by having an undefined large group of people interpret satellite imagery over the internet. 2) Developing such a system presents challenges including defining simple tasks, ensuring data quality, managing varied contributions, maintaining motivation, and providing reference information. 3) An experimental system was developed with web map and feature services to assign tiling tasks and collect ground information. Preliminary operation showed task completion times decreased with smaller tile sizes.

1 of 23
Crowd-Sourcing GIS for Global Urban Area Mapping Hiroyuki Miyazakia*, Satomi Kimijimab Masahiko Nagaib, Koki Iwaoc Ryosuke Shibasakia a The University of Tokyo, Japan b Asian Institute of Technology, Thailand c National Institute of Advanced Industrial Science and Technology, Japan
Needs on Global Urban Area Maps • Satellite-based urban area map enables – Monitoring without dependence on administrative district – comparing urban forming internationally – Disaster prevention & hazard assessment of broad areas Monitoring urban expansion Angel et al. (2005) Grid-based population estimation Bhaduri et al. (2002) Consistent definition and representation of urban area over countries and regions. Consistent geographical unit across countries and over time.
Sufficient ground truth data? • IGBP Land Cover Validation Confidence Sites (Muchoney et al., 1999) – # of urban sites: 44 / 966 • Global Land Cover Ground Truth database (Tateishi, 2002) – # of urban sites: 3 / 333 (Asia) • Degree Confluence Project (Iwao et al., 2006) – # of urban sites: 11 / 749 (Eurasia) • Too scarce for mapping urban area globally
Crowd Sourcing • A method to create & collect massive data by an undefined large group of people or community, the “crowd”, over the Internet. i.e. Wikipedia OpenStreetMap: tracing roads by GPS and visual interpretation of photos Geo-wiki: validating disagreements of global land cover maps using Google Earth Degree Confluence Project: posting ground information at the integer lat-lon grids.
Issues on GIS Crowd Sourcing • How to build GIS infrastructure with the Internet – Not only for showing maps – Also for editing maps • How to build a sustainable system – Not only IT infrastructure – Also work force management • How to manage data quality
Objectives • Identifying significant factors for effectiveness of crowd-sourcing system • System development of the crowd- sourcing GIS • System operation of the crowd-sourcing GIS
Challenges with Crowd Sourcing • Challenge 1. Defining tasks to be simple • Challenge 2. Quality assurance • Challenge 3. Managing various types of efforts • Challenge 4. Keeping motivation up • Challenge 5. Reference information for visual interpretation
Challenge 1. Defining tasks to be simple Mapping everything? Requiring knowledge/skills on visual interpretation of forest, agricultural field, barren, sandy land, water body urban … Operator Mapping urban areas with minimum requirement Visual interpretation Learning cost is heavy for beginners Requiring knowledge/skills on visual interpretation of urban areas OperatorMuch less learning cost Visual interpretation Easy project management
Challenge 2. Quality assurance • Technical background of operators varies with large diversity. Quality assurance is required for reasonable application of the ground truth data. • Assign experts as reviewer of the crowd’s works. The most trustful way to assure quality management. • Fixed map scale on the crowd- sourcing GIS Reviewer with expertise on remote sensing Operators Project manager Assign Crowd-sourced outputs Review / Revise
Challenge 3. Managing various types of efforts • Variety in Amount of contribution and motivation – By daily occupation, By technical interest, etc. • Partial efforts would complicate management. – Project manager have to consolidate partial efforts and reassign it Defining unit of task assignment being small for the efforts not to be partial.
Challenge 4. Keeping motivation up • Financial reward would be the most effective to stimulate motivation. But, we are not rich. • Intellectual stimulation is suggested to be a motivation in crowd-sourcing project. • Recognizing achievements is a good opportunity to identify what have been done by an operator’s hands. • Every assignments should be completed in a short time.
Challenge 5. Reference information for visual interpretation • Assuring data security: Just providing satellite image data would cause incidents with leakage of the data, especially serious in case of using paid image data. • Large amount of ref. images: Distributing composite images according to task assignments would be complicated. Effective control with Web Map Service
System development
System overview Satellite image archive (Thousands of GeoTiffs) Web Mapping Service (MapServer) User interface Reference images of requested extent Request for reference images with authentication Web Feature Service (GeoServer) Interpreted ground information Ground truth database (PostgreSQL & PostGIS) Catalog index Record with geometry WWW Display Interpret Crowd of the world Securely protected from the interpreters and the Internet Other map service Operator
Server-side System: Web-based Geographical Information Systems • Web Map Service (WMS) – The web service for generating and transferring map images by HTTP. – Request of map image is like: – Standardized by Open Geospatial Consortium* *an international organization of standardization of geographical data • Web Feature Service (WFS) – The web service for generating and transferring map vector data through WWW. – Standardized by Open Geospatial Consortium http://*&HEIGHT=512&WIDTH=512&XMIN=134.53 &XMAX=135.34&YMIN=32.05&YMAX=33.67
Client-side System: Web-GIS interface for visual interpretation Simplified Web-GIS interface with OpenLayers • Less learning cost • Effective work process • Reference information from public map service (Google Maps, Bing Map, Panoramio)Demo
Task management: Assign tasks by a tile scheme of Tile Map Service Demo Globally predefined tile unit with multi scale level Assign tasks by TMS tile scheme
System Implementation & Operation
Implementation Overview • Dedicated servers – Web Server: a dedicate server located at the University of Tokyo with Debian GNU/Linux 6.0.2 – Database Server: Amazon EC2 with Singapore node with Amazon Linux AMI • Server Software – Web server: Apache 2.2.16 – Database server: PostgreSQL 9.1.6 & PostGIS 1.5.4 – Data interoperability library: GDAL 1.6.3 – WMS software: MapServer 5.6.5 – WFS software: GeoServer 2.1.1 • Client software – Openlayers 2.12 – Dojo 1.7.1 – Google Maps API v3
Experimental Operation Overview • Period: February 2012 – August 2012 (7 months) • # of participants: 23 • # of tiles: – 80 km x 80 km: 12 – 20 km x 20 km: 38 – 10 km x 10 km: 92 (318 as of 27 November) • # of Features drawn: • Total work time: – 80 km x 80 km: 1413 hours – 20 km x 20 km: 260 hours – 10 km x 10 km: 161 hours
Size-time relationship 80 km × 80 km N = 12 20 km × 20 km N = 38 10 km × 10 km N = 92 ≈ 100 hour/tile ≈ 10 hour/tile ≈ 1 hour/tile
Discussion & Conclusion • Stability of the infrastructure: Implementation with cloud platform, such as IaaS (Infrastructure as a Service) and PaaS (Platform as a Service) • Size-time relationships: size has to be smaller than 10 km x 10 km for the working hours to be as long as one hour with casual participation. – conditions of assigned region, such as complexity of urban area and quality of satellite images. – operator’s background experience of remote sensing • Further issues: – Investigation of learning process on visual interpretation with operators and – Applying ‘Gamification’ approach, with which crowds are motivated for completing tasks of projects. – Development of methods for quality assessment on crowd-sourced data
Thank you! Please come & touch the demonstration at the WEBCON2 Hiroyuki Miyazaki, The University of Tokyo 東京大学 宮崎浩之 http://heromiya.net heromiya@heromiya.net heromiya@csis.u-tokyo.ac.jp

More Related Content

Crowd sourcing gis for global urban area mapping

  • 1. Crowd-Sourcing GIS for Global Urban Area Mapping Hiroyuki Miyazakia*, Satomi Kimijimab Masahiko Nagaib, Koki Iwaoc Ryosuke Shibasakia a The University of Tokyo, Japan b Asian Institute of Technology, Thailand c National Institute of Advanced Industrial Science and Technology, Japan
  • 2. Needs on Global Urban Area Maps • Satellite-based urban area map enables – Monitoring without dependence on administrative district – comparing urban forming internationally – Disaster prevention & hazard assessment of broad areas Monitoring urban expansion Angel et al. (2005) Grid-based population estimation Bhaduri et al. (2002) Consistent definition and representation of urban area over countries and regions. Consistent geographical unit across countries and over time.
  • 3. Sufficient ground truth data? • IGBP Land Cover Validation Confidence Sites (Muchoney et al., 1999) – # of urban sites: 44 / 966 • Global Land Cover Ground Truth database (Tateishi, 2002) – # of urban sites: 3 / 333 (Asia) • Degree Confluence Project (Iwao et al., 2006) – # of urban sites: 11 / 749 (Eurasia) • Too scarce for mapping urban area globally
  • 4. Crowd Sourcing • A method to create & collect massive data by an undefined large group of people or community, the “crowd”, over the Internet. i.e. Wikipedia OpenStreetMap: tracing roads by GPS and visual interpretation of photos Geo-wiki: validating disagreements of global land cover maps using Google Earth Degree Confluence Project: posting ground information at the integer lat-lon grids.
  • 5. Issues on GIS Crowd Sourcing • How to build GIS infrastructure with the Internet – Not only for showing maps – Also for editing maps • How to build a sustainable system – Not only IT infrastructure – Also work force management • How to manage data quality
  • 6. Objectives • Identifying significant factors for effectiveness of crowd-sourcing system • System development of the crowd- sourcing GIS • System operation of the crowd-sourcing GIS
  • 7. Challenges with Crowd Sourcing • Challenge 1. Defining tasks to be simple • Challenge 2. Quality assurance • Challenge 3. Managing various types of efforts • Challenge 4. Keeping motivation up • Challenge 5. Reference information for visual interpretation
  • 8. Challenge 1. Defining tasks to be simple Mapping everything? Requiring knowledge/skills on visual interpretation of forest, agricultural field, barren, sandy land, water body urban … Operator Mapping urban areas with minimum requirement Visual interpretation Learning cost is heavy for beginners Requiring knowledge/skills on visual interpretation of urban areas OperatorMuch less learning cost Visual interpretation Easy project management
  • 9. Challenge 2. Quality assurance • Technical background of operators varies with large diversity. Quality assurance is required for reasonable application of the ground truth data. • Assign experts as reviewer of the crowd’s works. The most trustful way to assure quality management. • Fixed map scale on the crowd- sourcing GIS Reviewer with expertise on remote sensing Operators Project manager Assign Crowd-sourced outputs Review / Revise
  • 10. Challenge 3. Managing various types of efforts • Variety in Amount of contribution and motivation – By daily occupation, By technical interest, etc. • Partial efforts would complicate management. – Project manager have to consolidate partial efforts and reassign it Defining unit of task assignment being small for the efforts not to be partial.
  • 11. Challenge 4. Keeping motivation up • Financial reward would be the most effective to stimulate motivation. But, we are not rich. • Intellectual stimulation is suggested to be a motivation in crowd-sourcing project. • Recognizing achievements is a good opportunity to identify what have been done by an operator’s hands. • Every assignments should be completed in a short time.
  • 12. Challenge 5. Reference information for visual interpretation • Assuring data security: Just providing satellite image data would cause incidents with leakage of the data, especially serious in case of using paid image data. • Large amount of ref. images: Distributing composite images according to task assignments would be complicated. Effective control with Web Map Service
  • 14. System overview Satellite image archive (Thousands of GeoTiffs) Web Mapping Service (MapServer) User interface Reference images of requested extent Request for reference images with authentication Web Feature Service (GeoServer) Interpreted ground information Ground truth database (PostgreSQL & PostGIS) Catalog index Record with geometry WWW Display Interpret Crowd of the world Securely protected from the interpreters and the Internet Other map service Operator
  • 15. Server-side System: Web-based Geographical Information Systems • Web Map Service (WMS) – The web service for generating and transferring map images by HTTP. – Request of map image is like: – Standardized by Open Geospatial Consortium* *an international organization of standardization of geographical data • Web Feature Service (WFS) – The web service for generating and transferring map vector data through WWW. – Standardized by Open Geospatial Consortium http://*&HEIGHT=512&WIDTH=512&XMIN=134.53 &XMAX=135.34&YMIN=32.05&YMAX=33.67
  • 16. Client-side System: Web-GIS interface for visual interpretation Simplified Web-GIS interface with OpenLayers • Less learning cost • Effective work process • Reference information from public map service (Google Maps, Bing Map, Panoramio)Demo
  • 17. Task management: Assign tasks by a tile scheme of Tile Map Service Demo Globally predefined tile unit with multi scale level Assign tasks by TMS tile scheme
  • 19. Implementation Overview • Dedicated servers – Web Server: a dedicate server located at the University of Tokyo with Debian GNU/Linux 6.0.2 – Database Server: Amazon EC2 with Singapore node with Amazon Linux AMI • Server Software – Web server: Apache 2.2.16 – Database server: PostgreSQL 9.1.6 & PostGIS 1.5.4 – Data interoperability library: GDAL 1.6.3 – WMS software: MapServer 5.6.5 – WFS software: GeoServer 2.1.1 • Client software – Openlayers 2.12 – Dojo 1.7.1 – Google Maps API v3
  • 20. Experimental Operation Overview • Period: February 2012 – August 2012 (7 months) • # of participants: 23 • # of tiles: – 80 km x 80 km: 12 – 20 km x 20 km: 38 – 10 km x 10 km: 92 (318 as of 27 November) • # of Features drawn: • Total work time: – 80 km x 80 km: 1413 hours – 20 km x 20 km: 260 hours – 10 km x 10 km: 161 hours
  • 21. Size-time relationship 80 km × 80 km N = 12 20 km × 20 km N = 38 10 km × 10 km N = 92 ≈ 100 hour/tile ≈ 10 hour/tile ≈ 1 hour/tile
  • 22. Discussion & Conclusion • Stability of the infrastructure: Implementation with cloud platform, such as IaaS (Infrastructure as a Service) and PaaS (Platform as a Service) • Size-time relationships: size has to be smaller than 10 km x 10 km for the working hours to be as long as one hour with casual participation. – conditions of assigned region, such as complexity of urban area and quality of satellite images. – operator’s background experience of remote sensing • Further issues: – Investigation of learning process on visual interpretation with operators and – Applying ‘Gamification’ approach, with which crowds are motivated for completing tasks of projects. – Development of methods for quality assessment on crowd-sourced data
  • 23. Thank you! Please come & touch the demonstration at the WEBCON2 Hiroyuki Miyazaki, The University of Tokyo 東京大学 宮崎浩之 http://heromiya.net heromiya@heromiya.net heromiya@csis.u-tokyo.ac.jp

Editor's Notes

  1. 土地被覆図を二種類並べる 都市と非都市だけのもの いくつかのクラスを持つもの さてどっちが簡単でしょうか シンプルなルーチンは習得コストの削減につながる。 プロジェクトを小規模にまとめることができる。 さらにもう一歩。都市と非都市は互いに排反なので、都市だけを描けば良い。