Beyond Spectral and Spatial data: Exploring other domains of information: 5 GEOG3010 Remote Sensing and Image Processing Lewis RSU
Domains of Information Spectral angular multi-temporal distance-resolved spatial
Spatial Information ‘texture’ / spatial dependency / context typically use measures of texture –size of objects –orientation –spacing and arrangement
Spatial Information Directional texture
Spatial Information Use: calculate texture measures –use to discriminate / classify –relate to physical properties (tree spacing etc.)
Spatial Information
Baringo, Kenya ‘Textures’ from tree density - dense to sparse
Measure texture using statistical measure of spatial dependency semivariance
Spatial Dependency points at a small distance apart (A-B ; B-C; C-D) are more likely to lie on the same object (have the same properties) than points further apart (A-C; B-D; A-D). Geostatistics
Spatial Dependency geostatistics - measure/model spatial dependencies using semivariogram Geostatistics
Spatial Dependency at some ‘lag distance’ (h) (spacing between points) the semivariance is: –half of the average (mean) squared difference between the property values at the sample points. Geostatisticssemivariance
Spatial Dependency Geostatistics - h = 1
Spatial Dependency Geostatistics - h = 2
Spatial Dependency Geostatistics - h = 3
Spatial Dependency Geostatistics - h = 14 Range of spatial dependency
Sill nugget variance
Features of the semivariogram Range –range of spatial dependency in data Sill –semivariance at and beyond range (half the scene variance) Nugget variance –extrapolated semivariance at lag 0 –variation at sub-measurement unit
Baringo, Kenya ‘Textures’ from tree density - dense to sparse
Summary spatial –‘texture’ information –may be directional row spacing –spatial dependancy - geostatistics semivariogram –range »size / spacing of objects –sill »variance at range - cover –nugget »sub-pixel variation / noise