Feature Extraction “The identification of geographic features and their outlines in remote-sensing imagery through post-processing technology that enhances feature definition, often by increasing feature-to-background contrast or using pattern recognition software. (ESRI GIS definitions] “ Typical photo interpretation to digitise features

data are not well understood (was) expensive classification accuracies The extraction of features from aerial photography is the most tedious part of mapping, why hasn't Remote sensing been used more to update GIS data? data are not well understood (was) expensive classification accuracies complexities of reality insufficient resolution

BC: updating of Vegetation Resource Inventory where does it work? Feature simplicity Feature homogeneity Feature certainty

Process for creating feature vectors from image data : Select bands and channels to maximise feature contrast Classify (multispectral) or threshold (single channel) Create single DN channel or bitmap Clean results -> sieve or filter (generalise) RTV -> Raster to Vector conversion smooth lines and export to GIS

Project example to automatically map fire polygon

ISODATA classification showing 50 classes Clusters 5, 7, 12, 15, 20, 25, 37 -> burn extent EASI modeling ->feature extraction Sieve

Smoothing the pixel ‘jaggies’ …

Feature Points: Experiment to extract trees as points to avoid digitising – M.Sc forestry project Green / red ratio Threshold and sieve Final points

Extracting non-forested areas near Kitimat

Supervised classification Using 5,4,3 and 4/3 as input

Extracting glaciers and cutblocks in Kakwa TM 3/5 ratio -> <-TM 4/3 ratio

Mt. Edziza Provincial Park Extracting current glacier extents - accumulation / ablation, compared to earlier vectors Mt. Edziza Provincial Park

Land cover boundaries

Image fusion to increase resolution Landsat ETM+ 15m PAN 30m MS fused 15m MS We can try this with your ALI data (10 v 30 m data) ….

http://www.ems-i.com/wmshelp/Introduction/Modeling_Guidelines/DEM_Guidelines.htm

http://www. gis. unbc. ca/courses/geog432/projects/2006/jaminf/index http://www.gis.unbc.ca/courses/geog432/projects/2006/jaminf/index.htm

Automatic Road Extraction from Remotely Sensed Imagery http://www.icrest.missouri.edu/Projects/NASA/FeatureExtraction-Roads/index.htm

Feature extraction – other applications http://www.ualberta.ca/~szepesva/CMPUT412/ip2.pdf

Process for creating feature vectors from image data : Select bands and channels to maximise feature contrast – lab 4 Classify (multispectral) or threshold (single channel) - lab 3/4 Create single DN channel or bitmap – lab 3/4 Clean results -> sieve or filter (generalise) – lab 3 RTV -> Raster to Vector conversion – lab 6 smooth lines to avoid the jaggies / export to GIS – lab 6 This is also a template for possible projects – though not required

Kelowna, 2010

Urban change: Kelowna, 1999-2000