Applications of AI in the geospatial domain
1 like1,562 views
The document discusses the increasing applications of artificial intelligence (AI) in the geospatial domain, highlighting the explosion of geospatial data due to various sources like IoT devices, satellites, and high-resolution imagery. It emphasizes using AI to manage this overwhelming data through applications such as real-time analytics in smart cities, automated mapping, and change detection in remote sensing. Finally, it offers various use cases demonstrating how AI can enhance decision-making processes and improve operational efficiency in geospatial contexts.
Applications of AI in the geospatial domain
- 1. Applications of AI in the Geospatial Domain 1 Erik van der Zee Geospatial Enterprise Architect (Geodan) AI in Practice | Startup Village Amsterdam Amsterdam | 6th November 2019
- 2. 2 Erik van der Zee Geospatial Enterprise Architect (Geodan) @erikvanderzee erik.van.der.zee@geodan.nl +31 6 10099691
- 3. Contents 3 1. The Geospatial Data Explosion 2. About Geospatial Algorithms 3. Applying of AI to Geospatial
- 4. The Geospatial Data Explosion 4
- 5. 5 The Geospatial Data Explosion Causes of geospatial data explosion • More geospatial data creators (humans, + smart things “IoT”, drones, satellites, etc.) • New geo data types (e.g. lidar point clouds, 360 panoramic pictures, CCTV, dashcams, bodycams, drones, etc.) • Higher spatial resolution (megapixel → gigapixel cameras, high resolution scanners) • Higher temporal (real-time) resolutions (image snapshots → continuous HR video) • Higher spectral resolutions (“bands”) (multi-spectral → hyper-spectral “data cubes”) • Derived data (information products, spatio-temporal model output e.g. weather and climate models)
- 6. 6 The Geospatial Data Explosion Real-time geo-enabled sensor data streams
- 7. 7 The Geospatial Data Explosion Remote sensing satellites (EU Sentinel, Worldview, etc. ) → Petabytes of data
- 8. 8 The Geospatial Data Explosion Hyperspectral Data (“Data Cubes”) captured by planes, drones, cars • From Multispectral • To Hyperspectral
- 9. 9 The Geospatial Data Explosion Near real-time remote sensing imagery of each place on earth (200+ Planet Satellites) • xxx
- 10. 10 The Geospatial Data Explosion GIGApixel Photos and Videos (360gigapixels.com)
- 11. 11 The Geospatial Data Explosion Thermographic Aerial Photos
- 12. 12 The Geospatial Data Explosion Panoramic (360°) Pictures and Videos
- 13. 13 UC 3 - Surveillance Cameras Cameras everywhere… (body cams, dash cams, surveillance cams, etc.)
- 14. 14 The Geospatial Data Explosion OpenStreetCam (open dash cam data)
- 15. 15 The Geospatial Data Explosion Generation of Lidar Pointclouds (from air by planes, drones, cars, smartphones, etc.)(1/2)
- 16. 16 The Geospatial Data Explosion Generation of Lidar Pointclouds (from air by planes, drones, cars, smartphones, etc.)(2/2)
- 17. 17 The Geospatial Data Explosion Location Enabled Social Media Content (georeferenced images, videos, tweets, etc.)
- 18. 18 The Geospatial Data Explosion Real-Time predictive model and simulation output
- 19. About Geospatial Algorithms A VERY brief introduction… 19
- 20. 20 Geospatial The power of location • All physical objects have a location in space and time (“XYZT”) • There are spatial relationships between objects • To know the location and status (by using sensors) of objects at any moment in time can be used to optimize business processes • Example 1 knowing the real-time positions of package pick-up and delivery vehicles and the positions of orders and cancellations can be used to do real-time planning and on the fly re-planning • Example 2 knowing the real-time fill % of recycle containers (with sensors) in the city can be used to optimize collection routes
- 21. 21 Spatial Relationships Relationships between geospatial objects (cross, disjoint, inside, overlap, …)
- 22. 22 Spatial Relationships Real-time spatial operators and geometry processors (e.g. geofencing) Real-Time Spatial Operators • Inside, Outside, Enter, Exit • Intersect, Disjoint, Touches, Contains, Crosses, Equals, Overlaps, Within Real-Time Geometry Processors • Buffer, Simplifier, Projector, Union, Difference, Cutter, Convex Hull
- 23. 23 Georeferencing Pictures/Video Georeferencing objects in panoramic pictures by combining image recognition with point clouds Source: velodynelidar.com
- 24. Geospatial AI Using AI to handle the geospatial data tsunami 24
- 25. 25 Geospatial Data Overload AI to the resque… • … Petabytes of BIG geospatial hyperspectral remote sensing data • … Billions of sensor measurements from smart objects (IoT) • … Needs to be analysed in real-time • It is simply impossible for humans to analyse it all… • AI to the rescue!
- 26. 26 Geospatial AI Handling the geospatial data flow using AI
- 27. 27 Machine Learning and Deep Learning Classification – Object Detection - Segmentation
- 28. 28 Some Use Cases for Geospatial AI Sheet voor alleen tekst, verdeeld over twee kolommen 1. UC 1 – Decision Making in Smart City Scenarios (real-time analytics of sensor event streams and subsequent taking of spatial actions (Geospatial IoT) 2. UC 2 – Automation of mapping (spatial object recognition and extraction, e.g. traffic signs, zebra crossings, etc.) 3. UC 3 – Surveillance cameras spatial object recognition (car license plate) or subject recognition (persons face) 4. UC 4 - Change detection in remote sensing and aerial photography data (PlanetLabs, Google Earth Engine)
- 29. 29 UC 1 Decision Making in Smart City Scenarios Real-time analytics of sensor event streams and subsequent initiation of spatio-temporal actions + predictions
- 30. 30 UC 1 – Decision Making in Smart City Scenarios Sensing – Analysis – Actuating (Geospatial IoT) Sensing Analysis and Prediction Act(uat)ing raw events meaningful events Sensing Actuating (tasking)
- 31. 31 UC 1 - Decision Making in Smart City Scenarios Geo-enabled sensor data streams • If This (and This and This and…) Then That (and That and That and…)
- 32. 32 UC 1 - Decision Making in Smart City Scenarios Recognition of events (e.g. accidents) based on activity clusters of sensor events or social media events → Location + content text/images
- 33. 33 UC 1 - Decision Making in Smart City Scenarios Location Intelligence Use Cases
- 34. 34 UC 2 Automation of Mapping Spatial object recognition and extraction, e.g. traffic signs, zebra crossings, etc.
- 35. 35 UC 2 – Automation of Mapping Automated Land Use Classification
- 36. 36 UC 2 – Automation of Mapping Detecting swimming pools
- 37. 37 UC 2 – Automation of Mapping Detecting roads
- 38. 38 UC 2 – Automation of Mapping Detecting zebra crossings
- 39. 39 UC 2 – Automation of Mapping Detecting traffic signs
- 40. 40 UC 2 – Automation of Mapping Road Feature Detection → Automated digitization
- 41. 41 UC 2 – Automation of Mapping Road Feature Detection and GeoTagging with Deep Learning
- 42. 42 UC 2 – Automation of Mapping Creating points of interest based on 360 panoramic pictures + text recognition • Xxx
- 43. 43 UC 3 Surveillance Cameras Spatial object (car license plate) or subject (persons face) recognition in surveillance cameras
- 44. 44 UC 3 - Surveillance Cameras Automated Number Plate Recognition (ANPR) cameras → Locations of cars in space and time
- 45. 45 UC 3 - Surveillance Cameras Spatial object (e.g. car license plate) recognition in surveillance cameras
- 46. 46 UC 3 - Surveillance Cameras Surveillance Cameras (static and moving) capturing high-resolution footage
- 47. 47 UC 3 - Surveillance Cameras OpenStreetCam (open dash cam data)
- 48. 48 UC 3 – Surveillance Cameras Scanautos (parkeerbeheer) → Can also be used for garbage detection → Cleaning tasks
- 49. 49 UC 3 - Surveillance Cameras Spatial subject (e.g. person clothes / walk) recognition in surveillance cameras
- 50. 50 UC 3 - Surveillance Cameras Spatial subject (e.g. person face) recognition in surveillance cameras
- 51. 51 UC 4 Change detection Change detection in remote sensing and aerial photography data (PlanetLabs, Google Earth Engine)
- 52. 52 Change Detection Object change detection (Planet Labs)
- 53. 53 Change Detection Object change detection (Planet Labs)
- 54. 54 Change Detection Land Cover change detection (Munich airport)
- 55. 55 September 17, 2016November 10, 2019November 11, 2019 Change Detection Area changes and moving objects
- 56. 56 Change Detection Area changes (construction) February 12, 2016July 19, 2016October 6, 2016
- 57. 57 Change Detection Change of business (opened, closed, moved) based on Streetview or social media images
- 58. End of Presentation … Q&A Time! 58
- 59. President Kennedylaan 1 1079 MB Amsterdam 020-5711 311 Buitenhaven 27-A 5211 TP ‘s Hertogenbosch 073 – 6925 151 Contact Vestiging Amsterdam Vestiging Den Bosch info@geodan.nl 59
- 60. Colophon Title Applications of AI in the Geospatial Domain Publication Geodan (www.geodan.nl) Author Erik van der Zee (Geospatial Enterprise Architect) Date 06-11-2019 Version 1.0 Status Final Information Erik van der Zee Email erik.van.der.zee@geodan.nl Phone +31 (0)6 10099691 Twitter @erikvanderzee Geodan President Kennedylaan 1 1079MB Amsterdam The Netherlands Phone 020 5711311 Twitter @GeodanNL Youtube GeodanNL channel
- 61. Applications of AI in the Geospatial Domain 61 Erik van der Zee Geospatial Enterprise Architect (Geodan) Amsterdam | 6th November 2019