Tutorial: How to Complete an

Unsupervised Image Classification with QGIS and the SCP Extension

Unsupervised image classification is a remote sensing method used to classify raster images without predefined (human input) training data. It uses clustering algorithms to group raster pixels with similar spectral properties. The user can assign land-cover classes to each cluster once the process has finished. Image classification is used in land-use and land-cover mapping, environmental monitoring, disaster management, urban planning, natural resource management, agriculture, and forestry.

QGIS, with the help of the Semi-Automatic Classification Plugin (SCP) can complete the unsupervised image classification process in 3 easy steps, as outlined below. Detailed descriptions of each step are provided as you scroll down the page. There is also a video tutorial available.

Video tutorial on how to complete an unsupervised image classification in QGIS using the SCP extension

Step 1: Source Orthoimagery

If you already have an orthoimage, jump down to step 2. If you don’t have an orthoimage of your project location, you can quickly download one from Equator. Follow the steps below to download an orthoimage.

  1. Log in or create a new account with Equator
  2. Locate your project site by using the Search tab or simply zooming into your site.
  3. Create your site boundaries by clicking on the +NEW SITE button on the bottom left of the screen. Use one of the predefined boundary boxes or create a custom boundary. Don’t forget to name your project site.
  4. Under the Data tab, click on the Orthoimagery (GeoTIFF) option.
  5. The Orthoimagery dialog box will appear. Select your project site and any customizations. Then click Generate.
  6. The orthoimage for your project site will appear within the site boundary.
  7. To export the orthoimagery out of Equator, go to the Layers tab and click the download button beside the orthoimage layer.
  8. The Customize Download dialog box will appear. Select GeoTIFF as your format, set your coordinate system, and click Process.
  9. Your orthoimage (GeoTIFF file) will download into your Downloads file.
Export orthoimage (GeoTIFF) in Equator

Step 1: Source orthoimage (GeoTIFF) from Equator

Step 2: Add Orthoimage to QGIS

Once you’ve sourced your orthoimage, you need to add it as a raster file to QGIS.

  1. Open QGIS
  2. In the top menu, select Layer > Data Source Manager > Raster. Under the source heading, locate your orthoimage. Then click Add.
Import orthoimage into QGIS

Step 2: Import orthoimage into QGIS

Step 3: Complete Unsupervised Image Classification

Unsupervised image classification can be completed using the semi-automatic classification plugin (SCP). To install the SCP plugin, click on Plugins in the top menu. Select Plugins > Manage and Install Plugins. The dialogue box for Plugins will appear. Search for SCP, select it, and click install. Now follow the steps below to complete your analysis:

  1. Navigate to SCP on the top menu. Go to SCP > Show Plugin > Band Set. Under the Multiband Image List, add an image. Navigate to your orthoimage.
  2. Still in the SCP Plugin dialogue box, select Band Processing > Clustering from the left menu.
  3. Under the Input heading, define your parameters. You can select the method to make an unsupervised classification (either K-means or ISODATA), define the threshold, number of classes, number of iterations, and no data values.
  4. Once your parameters are set, click RUN.
Result of Unsupervised Image Classification from SCP

Step 3: Unsupervised Image Classification Results in QGIS with the SCP Extension

FAQ’s

QGIS is a powerful, free, open-source software application designed for working with geospatial data. It is used for a wide range of tasks related to geographic information systems (GIS) and geospatial analysis.

QGIS can be downloaded from: https://qgis.org/es/site/

The Semi-Automatic Classification Plugin (SCP) is an extension for QGIS, used for image processing and land cover classification. It allows users to perform semi-automatic and supervised image classification tasks on remote sensing data (satellite and aerial imagery).

  1. Open QGIS
  2. Navigate to Plugins on the top menu.
  3. Select Plugins > Manage and Install Plugins
  4. The dialogue box for Plugins will appear
  5. Search for SCP, select it, and click install

Supervised and unsupervised classification are the two main approaches to land cover classification in remote sensing. Supervised classification uses training data (human identified) to classify pixels in an image. The data consists of samples of each land cover type and the user manually selects these samples from the image. Supervised classification is generally more accurate because the user has control over the classification process; however, it requires more effort and expertise.

Unsupervised classification uses a clustering algorithm to automatically group pixels of similar spectral properties. The user does not need to provide training data; however, they will need to assign land-cover classes to each cluster after the analysis has been run. This method is a faster and easier approach; however it may not be as accurate as supervised because the algorithm does not contain any information about land cover types.

Looking for an orthoimage for your next project?
Check out the thousands of databases on Equator.
We might just have what you’re looking for.