Binh Tan AI
Illegal Construction Detection App
TARGETS
ISSUES
TECHNICAL SOLUTIONS
Low accuracy build-up area classification by ArcMap
Long update interval of satellite imagery
Depend on Google Earth source
Potential illegal construction areas detection is still manual
Adjust satellite images coordinates using ArcMap
Build-up area classification
Current solution
Use Maximum Likelihood Classification tool + signature file --> raster
Use Raster to Polygon tool in ArcMap --> Polygon Vectors
Imagery resolution
Aerial imagery sources
Free
SPOT 5
Google Earth/Maxar
LIDAR Projects
Range: 2009-2022
Range: 2010
Best: 0.3 m
Fee
Range: 2012 & 2019
PlanetScope
SPOT 6
Acceptable: 3 m
SPOT 7
Resolution: 1.5 m
Resolution: 0.1 m
Resolution: 0.3 m
Resolution: 10 m
Good: 1 m
Depend on specialist to generate build-up area map
Find satellite imagery source with high resolution for the years before 2009
High accuracy build-up area classification algorithm
Develop an AI algorithm that allows government staff to generate build-up area maps
Develop an AI algorithm that detects potential illegal construction areas automatically
Future solution
Collects training samples --> signature file
Minimal data labelling cost
Can use a better algorithms that can detect more than 70% illegal constructions
More interactive web app
Can export data to CAD/GIS file formats
Reuse data label when changing to a new data source
Easy to change a new algorithm in the future
IKONOS
Quick Bird
Resolution: 1.5 m
Resolution: 0.8 m
Resolution: 3 m
Rapid Eye
Resolution: 5 m
Resolution: 0.6 - 0.72 m
2009 – 2019
$1.28/km2
2002 – 2013
$17.50/km2
Từ 2015
$4.75/km2
Từ 2012
$4.75/km2
$1.80/km2
Từ 2017
2000 – 2014
$10.00/km2