SCP Questions of This Month

This post is a collection of questions and answers about the Semi-Automatic Classification Plugin (SCP) and remote sensing which were discussed in the Facebook group and the Google+ Community this month.
These questions vary from supervised classification technique to software issues, and can be useful to the readers of this blog for solving issues about the use of SCP.

Question

I keep getting an error when I try create the training input file. It gives an error saying no ratser file is selected even when I have selected one. When I try to import the new training input file I get this error:
Traceback (most recent call last):
File ".qgis2/python/plugins\SemiAutomaticClassificationPlugin\core\utils.py", line 5134, in getEPSGVectorQGIS
id = int(id.replace("EPSG:", ""))
ValueError: invalid literal for int() with base 10: 'USER:100001'
Any ideas?

Answer

That the error is related to the reference system. Please try to project all the data to the same coordinate reference system before creating the training input.


Question

While working through the estimation of land-surface temperature with Landsat image, I wondered if it's possible to obtain the surface albedo of this image, considering that the processed image already has the reflectance values.
For instance I would like to figure out if a roof of a building is reflecting more radiant energy than another another building. Albedo would be important since I want to make sure that the energy is actually leaving Earth's surface and not being reflecting to another building. Please help me to clarify what is occurring. Thank you again for all your help!
Edit: I should probably add for anyone who might not need as high of resolution albedo values, I believe MODIS has a product that has a resolution of 500 meters every 16 days.
Also I have found on Yale's Center for Earth Observation the formula for converting DNs to Albedo, so I guess the question then becomes what is the best way of accomplishing this with SCP?
http://yceo.yale.edu/how-convert-landsat-dns-albedo

Answer

DOS1 is used as simple atmospheric correction, based on the image dark pixels (http://semiautomaticclassificationmanual-v5.readthedocs.io/en/latest/remote_sensing.html#dos1-correction). This can be considered surface reflectance (other atmospheric correction methods can provide more accurate results, such as the USGS Landsat Surface Reflectance High Level Data Products. See https://fromgistors.blogspot.com/2015/01/landsat-8-surface-reflectance.html).
About the calculation of planetary albedo, you can enter in the SCP Band calc (http://semiautomaticclassificationmanual-v5.readthedocs.io/en/latest/main_interface_window.html#band-calc) the expression
 ((0.356*B1) + (0.130*B2) + (0.373*B3) + (0.085*B4) + (0.072*B5) -0.018) / 1.016
as the described in the link you shared, replacing B1, B2, B3, B4, and B5, with the Landsat bands converted to TOA reflectance (without DOS1).




Question

The problem is that the training input is not saved in the scp panel.

Answer

The problem is solved uninstalling the previous version of QGIS and installing QGIS 2.18.


Question

In Basic Tutorial 2: Land Cover Classification of Sentinel-2 Images, why bands 1,9, and 10 were unchecked?

Answer

The main reason is that those bands have 60m spatial resolution which is low compared to the other bands. Also, the spectral range (the wavelength of bands) is very sensitive to atmospheric effects, and the information is not very suitable for land cover classification.


Question

Can we use the same training input for  the classification of multiple images?

Answer

It is not advisable to use the same spectral signatures for classifying multiple images, because in general with 10m or 30m pixels you have mixed pixels, and the signatures can vary substantially with the study area (e.g. different materials) and season (e.g. different vegetation growth).
In case the image are acquired by the same sensor, with the same resolution and during the same season, it is possible to load the same training input (spectral signatures) in SCP for multiple images.
However, the atmospheric disturbance of images acquired in different dates could cause differences in the reflectance values, so you should use very precise atmospheric corrections (the DOS1 available in SCP could be not accurate enough).


Question

Please help me by suggesting a tutorial that allows me to know the areas of each of the ROIs using the Semi-Automatic Classification Plugin.

Answer

In QGIS you can right click the training input in the layer list, and then export the training input to shapefile (command Save as). Then you can open this shapefile in QGIS and calculate the Area field (see http://gis.stackexchange.com/questions/23355/how-to-calculate-polygon-areas-in-qgis).
Alternatively, you can use the Identify tool of QGIS (http://docs.qgis.org/2.0/en/docs/user_manual/introduction/general_tools.html) over ROIs.


Question

In Sentinel-2 download, is there an option to specify a tile reference (e.g. 50JMM) as a search option as an alternative to the coordinate option.

Answer

SCP relies on SciHub APIs for searching and downloading Sentinel-2 images (ses https://scihub.copernicus.eu/userguide/5APIsAndBatchScripting)
At the moment the APIs don't allow for searching by tile ID. However, you can filter the results in the plugin entering the ID in the field Filter


For any comment or question, join the Facebook group and the Google+ Community about the Semi-Automatic Classification Plugin.

Newer posts Older posts

Follow

RG