Learning rules and detecting errors in OpenStreetMap data
summary
Downloads full OSM into a local PostGIS database, extracts ~11,000 Victoria bridges, computes topological relations with surrounding features, and uses PostGIS + Orange + Python and frequent itemset mining to learn common patterns. Result: ~75–80% of bridges both cross an obstacle and meet linear features on both sides, leaving ~2,500 Victoria bridges missing those relations as potential errors. A separate New Zealand run tests ~750,000 buildings with a ray intersection model, flagging ~17,000 (≈2%) as inaccessible to closest roads.
AI-written summary · about
chapters (8)
- 0:00Introduction to research
- 2:12Why detect OSM errors automatically
- 3:34Case study on bridges in Victoria
- 8:56Detected bridge errors in Victoria
- 9:59Case study on building accessibility
- 12:32Detected inaccessible buildings in New Zealand
- 13:57Interactive error detection tool
- 14:42Q&A session
related talks
25mUsing Python to Unlock Insights from OpenStreetMap Data at Scale
43mLocate all the things
37mProcessing Open Street Map Data with Python and PostgreSQL
12mRiveting insights! Bridge.Watch and America’s infrastructure
32mRunning Every Street in Paris with Python and PostGIS
17mLaying the Foundations for a National Building Outlines Dataset
23mDataEngBytes 2023 - BRI-T1-11 - Lizzie MacNeill
40mGIS capability of PostgreSql using POSTGIS & OSM DATA