UAV Imagery Processing with ESA SNAP

Introduction to Tutorial

Airborne imagery – particularly UAV acquired imagery – is increasingly available for remote sensing research, but how to utilize this resource effectively can be a challenge. In this session, we will look at basic operations for handling UAV imagery in the European Space Agency Sentinel Application Platform (ESA SNAP) program, a powerful tool for analysing remote sensing imagery. We will look at operations such as NDVI calculation and using basic unsupervised classification techniques. Finally, in the next session, we will incorporate the ground spectroscopy data processed in the earlier session to conduct unsupervised classification of the imagery handled in this session, as well as conduct more advanced operations such as spectral unmixing and resampling.

Overview of Field Site

The data within the workshop folder,2024_Tyninghame_Survey.dim, was collected in March 2024 at Tyninghame Links, a coastal habitat encompassing forest, salt marsh, and sand dunes located 40 km East of Edinburgh, UK. A multispectral imagery UAV survey, taken using a MAIA S2 Multispectral Camera mounted on a DJI M210 UAV, was taken over the area, covering a variety of different landscape features, including the coastal beach, the sand dune system, the salt marsh bog, and grassland. Reflectance spectra of each of these components were recorded using an SVC HR-1024i field spectrometer, and which you have processed in the last tutorial to produce a spectral library. We therefore have a multispectral UAV image, paired with ground reflectance data.

Tutorial Steps

Part One – Importing Data into ESA SNAP, reprojection of data, and viewing RGB imagery

The European Space Agency Sentinel Application Platform (SNAP) is a desktop geographic information system (GIS) application specifically developed in mind for handling Sentinel-1, Sentinel-2 and Sentinel-3 data products. SNAP is designed to work best with .dim files, which is the standard I/O product format for the Sentinel Toolbox and the Scientific Tools. The format was chosen because it has a very simple and easy-to-understand structure. However, by default and with plug-ins we can read all types of file formats that you will encounter in remote sensing.

ESA SNAP can be downloaded here. For this session, we will only be using ESA SNAP with the standard Sentinel-2 Toolbox.

  1. Open ESA SNAP
  1. In the Main Toolbar, click File, then Open Product, and then navigate to where you have saved the 2024_Tyninghame_Survey.dim and Tyninghame Data folder.
  2. In the Product Explorer window, we can look at the imported data set. Look at the Bands folder, you will see the 9 MAIA bands, each with their associated centre wavelengths.
  1. The image imported is currently in the WGS 1984 co-ordinate system. ESA SNAP allows us to reproject the image in a different co-ordinate system using Raster tools. We will convert this image to the OSGB 1936 British National Grid system. To do this, with the image selected in the products window, select Raster, then Geometric, and then Reprojection. In the window that appears, select Reprojection Parameters, then Predefined CRS. Press select, and search “British”. Select OSGB 1936, and select Run. A new product will appear in the product explorer, called 2024_Tyninghame_Survey_reprojected.dim.
  1. Right click the 2024_Tyninghame_Survey_reprojected.dim product name.
  2. The RGB image will appear in the main window. Note that ESA SNAP will, by default, display only pixels within a range based on the histogram distribution of pixels. This means that “extreme” pixels i.e those 2 standard deviations from the median reflectance will be displayed at the maximum reflectance value of those pixels within that range. What this means for our image is that the 99.9\% reflectance panel that we used for image calibration will not appear in the window. To change the display settings, select – in the Colour Manipulation tab – the “Auto adjust to 100% of all pixels” button. Repeat this for both green and blue. Notice the change in the output.
  3. Under the Optical tab in the Main Toolbar, select Spectrum view. This opens a window giving the spectral data for what ever pixel is centred on in your active window.

  4. Select Open RGB Image Window. In the Window that appears, select the bands to use for your image.

    • Question: – What Sentinel-2 band numbers are equivalent to red, blue and green, and what are their centre wavelength?
      Answer Red is Band number 4, and has its centre wavelength at 664.6 nm. Green is Band number 3, and has its centre wavelength at 559.8 nm. Blue is Band number 2, and has its centre wavelength at 492.4 nm.
  5. Float your cursor over pixels in the RGB image window. Zoom in and out using the mouse scroll window, and pan by left clicking and dragging. Centre your cursor over the reference targets which are beside the “Survey Base Home” marker to have a look at their reflectances.

Part Two – Generating an NDVI Mask

SNAP comes with inbuilt functions that allows you to conduct index generation with your images. There are many functions to choose, and more with plug-ins, but for now we will use the standard Normalized Difference Vegetation Index (NDVI) which is in-built to SNAP.

  1. In the Main Toolbar, select Optical, then Thematic Land Processing, then Vegetation Radiometric Indices, and then select NDVI Processor.
  1. In the I/O parameters tab, keep all settings as suggested by SNAP. Note that you can save the output as many different formates, including ENVI, GeoTiff, and csv, which can be useful if you want to conduct further processing outwith SNAP.

  2. In processing parameters, set the Red and NIR Source Bands.

    • Question: – What Sentinel-2 band number is equivalent to broad near infrared (NIR), and what is its centre wavelength?
      Answer The band number associated with broad NIR is band 8, and its centre wavelength is 832.8 nm.
  3. Select Run. Once complete, a new product will show in your Product Explorer window.
  4. Select from the drop down of the new product the Bands sub-folder, then double-click the ndvi band. A new image window will appear.
  1. Let’s first synchronize our windows (i.e. if we move our cursor or pan across one image, then it will pan and centre on the other images). Press the divided section symbol in toolbar to show your NDVI map alongside your RGB image.
  2. In the Navigation window, select the icons “Synchronizes view across multiple windows”, and “Synchronizes cursor positions across multiple windows”.
  3. Activate your NDVI window by clicking anywhere within the NDVI image.
  4. We would like to select pixels over 0.4 NDVI. We can do this using the mask manager. In the main toolbar, select View, then Tool Windows, then Mask manager.
  1. Click the f(x) symbol.
  2. In the Expression window, type ndvi > 0.4, and press OK.
  1. Your NVDI image window will now display a red mask, with each pixel in red highlighting that the NDVI value for that pixel exceeds 0.4

Wrapup

In this tutorial, you have learned how to import UAV data, in .dim format, into ESA SNAP, reproject the image to a different co-ordinate system, and perform basic imagery operations, including NDVI calculation and mask generation. In the next tutorial, which will continue to use the imagery used and products generated in this tutorial, we will discuss how to conduct an unsupervised classification technique to UAV imagery, as well as how to conduct spectral mixing.