Unleashing the Power of Advanced Analytics in ArcGIS
Anonym
At NV5 Geospatial, we are always looking to provide more people with access to better analytics. Here we explain how GIS users can benefit from a wider range of analytic tools based on the latest ENVI technology, using a hyperspectral example for illustration.
A recent help article describes the improvements we made to ENVI and ArcGIS® interoperability with the release of ENVI 5.5 in March 2018. Once you install the ENVI Py® for ArcGIS® Python client library over an existing ENVI installation, 10 new image-processing tools will be added to ArcGIS that you can use right away. Plus, you can create your own set of tools with more options than before.
Each ENVI tool published to ArcGIS is based on an ENVITask that performs a specific data-processing operation. We tested and verified that at least 80 ENVITasks can be successfully published as standalone tools in ArcGIS. With an ENVI Crop Science license, 16 additional ENVITasks are available for use. As we continue to develop new ENVITasks, you will able to create even more tools in ArcGIS. See the Running ENVI Analytics in ArcGIS Pro tutorial to learn more.
Not all ENVITasks can be published as standalone tools in ArcGIS, primarily because they use data types that are not supported in ArcGIS. Examples include regions of interest (ROIs), point clouds, and ENVI raster series data. However, the good news is that you can create a model that includes these tasks, then package and deploy the model as a tool in ArcGIS. Models created with the ENVI Modeler can internally manage data formats that are not supported in ArcGIS, plus you can take advantage of the entire set of ENVITasks to create powerful image-processing workflows, including hyperspectral analysis.
Example
Suppose that you want to use Spectral Angle Mapper (SAM) classification to map different types of trees in an AVIRIS scene of a forested area:
AVIRIS image of a boreal forest in northern Minnesota, USA. The red polygon shows the area of interest that will be used for tree classification.
The SAMClassification task requires reference spectra (ground truth) in the form of spectral libraries or ROIs, neither of which are recognized data types in ArcGIS. So you would not be able to directly add the SAMClassification task to ArcGIS. However, you can create a model that includes:
- The SAMClassification task
- A reference to a known spectral library or ROI file on disk (or network location)
- Tasks to resample the library spectra to match the image spectra
- A task that extracts metadata and properties from the input image
Example SAM classification model
Plot showing reflectance curves of various tree types. We collected reference spectra from U.S. Geological Survey (USGS) and SPECCHIO spectral libraries, then used the ENVI Spectral Library Builder to create a new library with only the spectra we are interested in.
After building the model, select the Generate Metatask menu option in the ENVI Modeler, then choose to publish the task to ArcMap or ArcGIS Pro.
As long as the "Toolboxes" folder of your ArcGIS installation has write permissions, the new tool will be automatically added to ArcGIS. (A tutorial provides additional steps if you encounter permissions issues.)
When you run the tool in ArcGIS, you are only prompted for an input image and an optional output filename and location. All of the data-processing steps and the locations of reference spectra are managed internally. Here is the result of running the "Forest SAM Classification" tool on the AVIRIS subset. The classes represent different tree species.
Here are some other ENVITasks designed for use with hyperspectral data from which you can create tools for ArcGIS:
- LinearSpectralUnmixing
- MatchedFiltering (requires a model)
- MinimumNoiseTransform
- MixtureTunedMatchedFiltering (requires a model)
- PixelPurityIndex
- SpectralAdaptiveCoherenceEstimator (requires a model)
- SpectralAngleMapperClassification (requires a model)
With the availability of a wider range of ENVI analytics, you can incorporate even more powerful image-processing capabilities into your GIS environment.
Acknowledgements
AVIRIS data are available from the NASA/Jet Propulsion Laboratory AVIRIS Data Portal at https://aviris.jpl.nasa.gov/alt_locator/ .
Clark, R., G. Swayze, R. Wise, K. Livo, T. Hoefen, R. Kokaly, & S. Sutley. (2007). USGS Digital Spectral Library splib06a, U.S. Geological Survey, Data Series 231.
Hueni, A., J. Nieke, J. Schopfer, M. Kneubühler, & K. Itten. (2009). The spectral database SPECCHIO for improved long term usability and data sharing. Computers & Geosciences. DOI: 10.1016/j.cageo.2008.03.015.