Web Processing Service for assisted land cover classification

EGU General Assembly 2013 - Vienna, Austria - 2013.04.08

Jérôme Gasperi (1), Charles Peyrega (2), Sébastien Dinot (2), Quentin Boileau (3), Alexis Manin (3), and Vincent
Heurteaux (3)
(1) CNES, Toulouse, France (jerome.gasperi@cnes.fr), (2) CS Systèmes d’Information, Toulouse, France, (3) Geomatys, Montpellier, France

The Orfeo Toolbox (OTB - http://www.orfeo-toolbox.org/) is an Open Source Remote Sensing Image Processing software library developed by CNES. The aim of the toolbox is to gather a large number of state of the art algo- rithms for building processing chains for satellite images. Using the constellation server (http://www.constellation- sdi.org/), we exposed the main OTB processing chains as Web Processing Services (WPS). The WPS provides rules for standardizing inputs and outputs for invoking geospatial processing services. These services are managed from a web browser using the mapshup web client (http://mapshup.info). mapshup supports both synchronous and asynchronous processes and offers direct visualisation of results. The whole system provides user a complete and comprehensive image processing chain to produce land cover classification from satellite orthoimagery.

Transcript

1. by Jérôme Gasperi Web Processing Service for assisted land cover classification EGU - ESSI 2.12 Vienna, Austria - April 8th, 2013
2. Objective Architecture Orfeo Toolbox mapshup Land Cover classification What's next ?
3. Objective
4. From pixel... Remote sensing image
5. From pixel... Remote sensing image ...to information Land cover thematic map
6. Architecture
7. Keep it simple.
8. Processing Keep it simple.
9. Processing Web application Keep it simple.
10. WPS a Processing Web application Keep it simple.
11. mapshup Web application http://mapshup.info WPS Constellation http://www.constellation-sdi.org Orfeo Toolbox Processing http://orfeo-toolbox.org
12. Orfeo Toolbox Processing http://orfeo-toolbox.org
13. Orfeo Toolbox Remote sensing image library Open Source Developped by the French Space Agency
14. Orfeo Toolbox More than 70 high level processing chains orthorectification segmentation classification etc.
15. Orfeo Toolbox More than 70 high level processing chains orthorectification segmentation classification etc. Supervised learning (land cover is computed from a set of "well known areas" given by user) Based on SVM (http://en.wikipedia.org/wiki/Support_vector_machine)
16. Web application mapshup http://mapshup.info
17. mapshup Improves access to geospatial data HTML + javascript Based on OpenLayers and jQuery
18. mapshup Supported data sources WMS WFS CSW WPS OpenSearch etc. Flickr Youtube GeoRSS KML Wikipedia Google maps Bing maps OpenStreetMap MapBox WMTS
19. mapshup Supported data sources See WPS demo https://vimeo.com/57101606 WMS WFS CSW WPS OpenSearch etc. Flickr Youtube GeoRSS KML Wikipedia Google maps Bing maps OpenStreetMap MapBox WMTS
20. Land cover classification
21. WPS a WPS a Web application Orfeo Toolbox mapshup
22. WPS a WPS a Web application Orfeo Toolbox mapshup
23. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Orfeo Toolbox mapshup
24. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Orfeo Toolbox mapshup
25. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Classification process description 2 2 - Set up MMI from process description Orfeo Toolbox mapshup
26. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Classification process description 2 2 - Set up MMI from process description Orfeo Toolbox mapshup
27. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Classification process description 2 2 - Set up MMI from process description + Image Well known areas 3 3 - Execute an asynchronous "Classification" request Orfeo Toolbox mapshup
28. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Classification process description 2 2 - Set up MMI from process description + Image Well known areas 3 3 - Execute an asynchronous "Classification" request Orfeo Toolbox mapshup
29. WPS a WPS a Web application 1 1 - DescribeProcess request for "Classification" Classification process description 2 2 - Set up MMI from process description + Image Well known areas 3 3 - Execute an asynchronous "Classification" request Land Cover 4 4 - Display result retrieved as a WMS layer Orfeo Toolbox mapshup
30. WPS endpoint
31. WPS endpoint Not a production server...so send me an email to get the url - jerome.gasperi@cnes.fr
32. WPS endpoint Not a production server...so send me an email to get the url - jerome.gasperi@cnes.fr And don't crash the server !!! I mean don't even try to crash the server :)
33. https://vimeo.com/63320692
34. What's next ?
36. Quality map The land cover classification processing includes a quality information computation i.e. the relative confidence per pixel that this pixel is effectively in the right class. This quality information could be returned as a WMS layer
37. Quality map The land cover classification processing includes a quality information computation i.e. the relative confidence per pixel that this pixel is effectively in the right class. This quality information could be returned as a WMS layer Active learning Based on the quality information, the application should propose user to validate the classification of the worst classified areas (i.e. areas with the lowest quality confidence). Updates from user should be returned to the server in order to compute a better classification...and so on
38. jerome.gasperi@cnes.fr