Spectral Index Optimizer¶
This algorithm finds the optimal spectral two-band index to estimate a measured variable via linear regression. All band combinations are tested and the one with either the lowest RMSE, the lowest MAE, or the highest R^2 is selected.
In the field “Features”, specify the input hyperspectral image.
In the field “Labels”, specify an image the same size as the feature image that contains the training data.
In the field “Index Type” you can select how the index is calculated from two bands:
- Normalized Difference: [(a - b) / (a + b)]
- Ratio Index: (a / b)
- Difference Index: (a - b)
In the field “Performance Type” you can select the accuracy measure:
- RMSE (Root mean squared error)
- MAE (Mean Absolute Error)
- R^2 (R-squared, coefficient of determination)
In the field “Raster”, specify a hyperspectral image that is to be used for estimating the variable. Usually, this is the same image as the feature image, but you can also apply the model developed on the feature image on another data set with the same spectral bandset.
You can specify the output file name in the field “Output prediction”, otherwise the output will be stored in a temporary file.
An HTML report will be generated showing the selected settings, the resulting best bands, the equation to generate the output image from the index, and the map of performance where you can identify spectral regions of high or low correlation to the variable of interest.
Narrow-band indices have been used frequently in remote sensing. Testing all possible band combinations to find the optimal spectral index can be found in numerous papers, e.g. Schlerf et al. (2005), Buddenbaum et al. (2012), or Buchhorn et al. (2013).
Schlerf, M., Atzberger, C. & Hill, J. (2005). Remote sensing of forest biophysical variables using HyMap imaging spectrometer data. Remote Sensing of Environment, 95, 177-194. DOI: 10.1016/j.rse.2004.12.016
Buddenbaum, H., Stern, O., Stellmes, M., Stoffels, J., Pueschel, P., Hill, J. & Werner, W. (2012). Field Imaging Spectroscopy of Beech Seedlings under Dryness Stress. Remote Sensing, 4, 3721-3740. DOI: 10.3390/rs4123721
Buchhorn, M., Walker, D., Heim, B., Raynolds, M., Epstein, H., & Schwieder, M. (2013). Ground-Based Hyperspectral Characterization of Alaska Tundra Vegetation along Environmental Gradients. Remote Sensing, 5, 3971-4005. DOI: 10.3390/rs5083971