Exploring the new features of Semi-Automatic Classification Plugin v.5.0 "Kourou"

I am glad to announce that the new Semi-Automatic Classification Plugin (SCP) version 5.0, code name "Kourou" (dedicated to the Europe's Spaceport in French Guiana where Sentinel satellites are launched, see http://www.esa.int/Our_Activities/Launchers/Europe_s_Spaceport/Overview_of_Europe_s_Spaceport) will be released on 1st of September.




Following the changelog:
-new interface
-new input file (extension .scp) which is a zip file containing the shapefile and signature list file
-possibility to create multipart ROIs (pressing CTRL + mouse click) and CTRL + Z for removing last part
-C ID is automatically incremented after saving a ROI
-function to merge ROI polygons from table
-new Land Cover Signature classification based on the range values of spectral signatures
-new tab for the definition of range values for Land Cover Signature classification
-in Landsat and Sentinel pre processing added option for calculations based on the band set
-new tab for PCA (Principal Components Analysis) of Band set
-in Band set added options for automatic calculation of virtual raster, stack of layers, build overviews, and Band calc expressions
-enhanced signature plot that allows for interactive definition of range values and new navigation (pan with left click and zoom with scroll or right click)
-enhanced Band calc allowing for the use of variables "#BLUE#" , "#RED#", and "#NIR#" in expressions
-Band calc option to use custom file names for multiple expression adding @fileName at the end of expression
-in Band calc added Decision rules for calculating raster based on conditions (e.g. "raster1 >0" or multiple rules separated by semicolon)
-added button for importing SCP file (polygons with automatic reprojection and spectral signatures), CSV, and external shapefile
-direct search of Landsat images from NASA CMR Search
-new tab for search and download of ASTER images L1T  from NASA Land Processes Distributed Active Archive Center
-new tab for conversion of ASTER images L1T
-new tab for manual raster editing
-new tab for classification sieve
-new tab for classification erosion
-new tab for classification dilation
-new tab for conversion from vector to raster (using a reference raster for pixel alignment)
-new tab for batch processing (Landsat conversion, Sentinel conversion,  ASTER conversion, Create band set, Classification, Split raster bands, Vector to raster, Clip multiple rasters, Accuracy, Land Cover Change, Classification report, Classification to vector, Reclassification, Classification sieve, Classification erosion, Classification dilation, Edit raster using shapefile, Band calc) and option for using a working directory (!working_dir!)
-new tab for editing RGB list
-improved settings of UL and LR points for area definition in several tabs (left click for UL point and right click for LR point)
-added service option for Sentinel-2 download mirrors (e.g. working with https://finhub.nsdc.fmi.fi , https://data.sentinel.zamg.ac.at)
-if available, Sentinel-2 granule preview are downloaded from the Amazon Web Services (http://sentinel-pds.s3-website.eu-central-1.amazonaws.com)
-results of image searching are added to the previous results in the table
-improved scatter plot
-ROI size (pixels) is calculated in the Signature details
-added button for calculation of spectral distances
-clip multiple rasters using shapefile and accuracy assessment now work also if shapefile and raster projections are different

In this new version I have tried to improve the usability and functionality of SCP, especially considering the availability of free images such as ASTER, Landsat, and Sentinel-2 that will foster land cover monitoring.
I have placed the main functions in one dock and one toolbar to make more space available for displaying the image in the map.



This new dock is composed of many tabs; the first is for setting the classification input.
One of the major changes is the new input file (extension .scp), which stores both ROIs and spectral signatures. 
In addition, the panel SCP news will display updates about the plugin and other information such as the state of image download services (the download of the news can be disabled in the settings).
single table will display ROIs and spectral signatures allowing for an easier management of the training input.


Commands that require frequent interaction with the image (color composite, ROI creation and classification previews) are placed in a toolbar



A new tab has been added for managing color composites in the RGB list.



It is possible to create multipart polygons during ROI creation by pressing CTRL and then adding the new part with mouse (also pressing CTRL + Z removes the last part). This allow to create a single spectral signature even if pixels are not contiguous in the image (the following images are © Copernicus Sentinel data 2016).



The Scatter plot and Spectral signature plot have been improved.
In particular, the scatter plot can be displayed for ROIs, temporary ROIs, and the entire image; color describes the frequency of pixel values, and decimal precision of calculation can be defined by the user.
It is also possible to draw a polygon inside the plot, and create a Scatter raster that is a temporary raster classified according to the intersection of scatter plots and drawn polygons. This can also be used for the definition of spectral signatures.


SCP now uses the QGIS proxy settings for connecting to internet.
The tool Landsat download has been updated to allow direct search (without the prerequisite of downloading a Landsat database). Moreover, the ASTER images L1T can be downloaded directly using the required USGS EROS credentials (https://ers.cr.usgs.gov).


Also, the tool Sentinel-2 download has been updated, to allow the search and download from mirrors (such as https://finhub.nsdc.fmi.fi , https://data.sentinel.zamg.ac.at). These services require to login with a free account.
Now color previews of granules are downloaded from the Amazon Web Services (http://sentinel-pds.s3-website.eu-central-1.amazonaws.com) if available.In addition, image cloud mask (a vector .gml file) is downloaded.


The tab Band set has been updated with automatic functions that can also be triggered after the download or image conversion to reflectance: calculation of virtual raster, stack of layers, build overviews, and calculate a custom expression in Band calc.


The tool Band calc has been improved with a new tab for the calculation of an output raster based on decision rules. Rules are verified from the first to the last row in the table; if the first Rule is false, the next Rule is verified for that pixel, until the last rule.


The Preprocessing of ASTER images has been added; also, a new tab allows for the PCA (Principal Components Analysis) of the Band set. A tool for the conversion from vector to raster has also been added.



For the Postprocessing, a new tool allows for the manual editing of rasters, setting directly pixel values using temporary ROIs or a vector. Moreover, tools for the refinement of classifications have been added, such as classification sieve (for removing small patches from a classification), classification erosion (for removing the border of a class patch), and classification dilation (for dilating the border of a class patch).


A tab for Batch processing has been added, allowing for the execution of SCP functions using a scripting interface. The following functions are available: Landsat conversion, Sentinel conversion, ASTER conversion, Create band set, Classification, Split raster bands, Vector to raster, Clip multiple rasters, Accuracy, Land Cover Change, Classification report, Classification to vector, Reclassification, Classification sieve, Classification erosion, Classification dilation, Edit raster using shapefile, Band calc. There is also an option for the definition of a working directory (!working_dir!) for input and output files.


I have developed a new classification algorithm called Land Cover Signature Classification which is based on the definition of spectral ranges (a pixel is labelled as class X if the spectral signature thereof is completely contained in the spectral region defined by class X). I am going to describe this in a specific tutorial.
Further information will be provided in the user manual.



Following a brief video about the new features of SCP v. 5.





There are also other minor changes that will be described in the tutorials that I am going to publish in the next few weeks. You are kindly invited to test this new version and  please contribute to SCP.
For any comment or question, join the Facebook group and the Google+ Community about the Semi-Automatic Classification Plugin.

Newer posts Older posts