Using ELM in GRYFN Processing Tool
GRYFN Processing Tool employs empirical line method (ELM) to match radiance (or digital number) to reflectance.
ELM in GRYFN Processing Tool
To use ELM, provide target files and perform target selection when creating a New Job (the pipeline should include hyperspectral processing).
Target files: The field measurements for the targets. At least two targets are required for the ELM.
Target selection: The user must manually draw a window on each target. The average of the pixel values within the window will serve as a data point for the linear regression.
The figure below shows the target selection tool, including three blocks: a) a user interface that shows the image for the user to draw windows, b) statistics (average/max/min) of the pixel values within each window, and c) R^2 for the linear regression based on the target selection.
Target Selection Tips
Guidelines for target selection:
Zoom in for a clear view of the targets.
Avoid using a cube with noticeable illumination difference compared to other cubes.
Ensure that the window is large enough for a robust average evaluation.
Avoid enclosing pixels close to the edges of the targets.
Check the R^2 feedback; Ideally, the R^2 after ~450nm should be higher than 0.99 (shown in green).
Target Selection Examples
The examples demonstrate various windows drawn on the rightmost target and the corresponding statistics.
Example 1: A good target selection
The window is well within the target and contains adequate pixels for averaging.
Example 2: A good target selection
The average pixel values within the window remain consistent with those from Example 1, indicating a robust evaluation.
Example 3: A bad target selection
The window contains pixels close to or on the edges of the target.
Example 4: A bad target selection
The window is too small for reliable averaging. One might get varying results when drawing a small window in different locations.
Last updated