This model finds differences in two different dates of ‘info-stack’ raster files, and produces a raster file showing positive or negative changes. These changes usually correspond to buildings or other infrastructure being added or removed. In addition, areas of removed trees are mapped.
‘Info-stack’ is a term used to describe a raster data set made up of layers which are not necessarily all "imagery", and in this context refers to a raster file with blue, green, red and NIR optical bands that have been radiometrically corrected, and a fifth band that specifies elevation for each pixel. The elevation band was generated automatically from the stereo imagery using a technique called semi global matching. Creation of the info-stack must be performed previously and is not part of this model.
The model primarily uses the 5th band (elevation). The part of the model that maps removed trees also uses the optical bands.
Below: Images left to right – Info-stack for 2010, info-stack for 2012, and changes detected by the model overlaid on the 2012 info-stack. This example shows medium and large positive change (yellow and orange) representing where houses have been added, and removed trees (green). The info-stack files are displayed using a 4/3/2 to red/green/blue band combination.
Below: Neighborhood Area submodel
Below: Assign Attributes submodel
Earlier info stack: (filename with path) of the earlier date input raster info-stack file. Info-stack is described at the beginning of this document.
Later info stack: (filename with path) of the later date input raster info-stack file. Info-stack is described at the beginning of this document.
Output results file: (filename with path) of the output raster file. This file is a single band thematic raster file with various colors noting the type of change. The meaning of the colors is described earlier in this document.
Minimum elevation change for building detection (feet): (float)(default 8.0) Minimum elevation difference between dates for non-vegetation change. Only negative change (trees removed) is considered by the model.
Minimum area size for buildings (pixels, sq meters): (integer)(default 100) Non-vegetation change areas must be at least this size. The sample data (info-stacks) pixels equal 1 square meter, so, in this case pixels are equivalent to square meters.
Minimum elevation change for trees (feet): (float)(default 5.0) Minimum elevation difference between dates for vegetation change.
Minimum area size for trees (feet): (integer)(default 50) Vegetation change areas must be at least this size. The sample data (info-stacks) pixels equal 1 square meter, so, in this case, pixels are equivalent to square meters.
Noise filtering amount (0-none, 1-low, 2-med, 3-high): (integer)(default 2) In effort to reduce noise, various levels of noise filtering are available to the model.
The following example info-stack data is for El Paso, Texas. Info-stack is described at the beginning of this document. The sample data supplied with this model is courtesy of HxIP - Hexagon Imagery Program (http://www.leica-geosystems.us/en/HxIP-Hexagon-Imagery-Program_106454.htm).
The original images of El Paso, Texas were acquired on August 26, 2010 and August 30, 2012. Leica XPro software used stereo pairs of each date to generate a DSM (digital surface model). The software created a 5-band file where the first 4 bands are imagery (orthorectified using the DSM), and the 5th band is the DSM (elevation units of decimeters).
This 5-band data set was then preprocessed in ERDAS IMAGINE, converting the image pixel values to ground reflectance. A small subset of this resulting 5-band file is provided as example data. The band order is blue, green, red, NIR, and the DSM.
2010a_info_stack_subset.img – West El-Paso info-stack for the year 2010.
2012a_info_stack_subset.img – West El-Paso info-stack for the year 2012.