Change Detection
Nick Clinton, Earth Engine Developer Relations
Ladies and gentlemen, start your Earth Engines!
Direct classification of change
Time 0 bands
Time t bands
…
classify()
Change categories
Post-classification comparison
Time 0
Time t
sample()
  | c1  | c2  | c3  | 
c1  |   |   |   | 
c2  |   |   |   | 
c3  |   |   |   | 
From
To
errorMatrix()
Index change
Time t
Time 0
Change
subtract()
Spectral distance
= pt
= p0
Distance = ||pt - p0||
Spectral angle
= pt
= p0
Angle =
acos( (pt · p0) / (||pt||*||p0||) )
Landscape metrics
Dominance = 1 - [ ∑k(-PklogPk) / logK ]
Contagion = 1 - [ ∑i ∑k(-PiklogPik) / 2logK ]
What changes will you detect?