Soil Cluster Mapping with UAVs

1) Data-stack

RGB Camera¹:

• Red
• Green
• Blue

NIR Camera²:

• Red
• Green
• Nir

multispectral Camera ³:

• Red
• Green
• Rededge
• Nir

For reasons of spatial resolution and that the cameras did not all capture exactly the same area, two stacks were created (see Code GetChannel.ipynb). These were brought to the same spatial resolution

1. Red¹, Green¹, Blue¹, NIR², NDVI (calculated)², Rededge ³
2. Red¹, Green¹, Blue¹, NIR³, NDVI (calculated)³, Rededge ³

2) Training data

2.1 Test and training data were recorded in QGIS (as points and polygons), which is important for model training and validation. (see Code Create_TestTrainData.ipynb)

2.2 The stacks have been cut to smaller areas where the clay is definitely visible on bare fields.

3) Pre-Processing and Filtering

==> increase Saturation/contrast; decrease lightning

4) Object based image classification

Create Classification based on OBIA, the stack and the collected training data (see Code OBIA.ipynb)

5) Validation

Validate the model with the collected test data using a confusion matrix and calculated indices. (see Code Validation.ipynb)

6) Model Transfer

Apply the model to other areas and check with the eye whether the classification is successful. (see Code Model_Transfer.ipynb)

7) Provide Results