You are on page 1of 2

Susan T.

Williams
Geog 883

LAB 7: CHANGE DETECTION


Methodology
I set up a container for LAS Datasets in ArcCatalog, calculated Statistics, and verified coordinate systems.
I checked the Properties to see that (2) is the class used for Ground. Then I loaded the LAS datasets into
ArcGIS, creating one for the 2008 las and one for the 2010 las.
I loaded the orthophoto and created a DEM for the 2008 data. This was done by first creating a LAS
Dataset Layer File with appropriate class codes selected (2, in this case). Then a raster surface model
was created (LAS Dataset to Raster) using Natural Neighbor interpolation. An appropriate color ramp
was selected and a Hillshade generated. This process was repeated to create a DEM for the 2010 data.
Then I created a DSM from the 2008 layer to represent above-ground features. This was done by first
creating a LAS Dataset Layer File consisting of only the first returns. Then a raster surface model was
created (LAS Dataset to Raster) using Natural Neighbor interpolation. An appropriate color ramp was
selected and a Hillshade was generated. This process was repeated to create a DSM for the 2010 data.
Then I created a DTM from the 2008 layer. This was done by first creating a LAS Dataset Layer File
consisting of only the last returns. Then a raster surface model was created (LAS Dataset to Raster)
using Natural Neighbor interpolation. An appropriate color ramp was selected and a Hillshade was
generated. This process was repeated to create a DTM for the 2010 data.
I spent some time with the Swipe tool to examine the various layers.
I opened eCognition in RuleSet Mode, opened a workspace and began a new project. In Load and
Manage Data view, I loaded the orthophoto imagery and the six files created from the LiDAR data (a
DEM, DSM, and DTM each for 2008 and 2010), and renamed all bands and layers accordingly.
Switching to View 4, I created an nDSM (DSM minus DEM) for 2008 and 2010, and an nDTM (DTM minus
DEM) for 2008 and 2010. Using layer arithmetics, I subtracted the 2008 nDTM from the 2008 nDSM to
help separate buildings from trees and then performed a multiresolution segmentation.

The next step was Classification. I assigned all unclassified objects with an NDVI greater than 0.5 as
Vegetation, and all remaining unclassified objects were classed Impervious. Impervious objects
with a 2008 nDSM value greater than 12 (feet in height) were reclassified as Buildings. Buildings
with a 2010 nDSM value less than 12 (feet in height) were then reclassified as Buildings Removed.
I attempted to refine the data by merging among Buildings, Impervious, and Buildings Removed classes.
The final step was to Export the Buildings Removed as a point shapefile.

Strengths and Weaknesses


One potential weakness of the dataset is that any larger structure that was not successfully merged into
one singular object appears as multiple points in the shapefile, which may be confusing to field
inspectors if not properly explained.
One potential strength of the dataset is the ease and speed with which it can be loaded into ArcMap and
overlaid on a city map to help easily identify parcels in need of field inspection.

You might also like