In certain sites and images, the veg edge extraction process falls over when contours are attempted to be extracted. Specifically at:
File COASTGUARD/VegetationLine.py, line 232
contours_ndvi, t_ndvi = FindShoreContours_WP(im_ndvi, im_labels, cloud_mask, im_ref_buffer)
File COASTGUARD/VegetationLine.py, line 1030
t_ndi, _ = Toolbox.FindWPThresh(int_veg, int_nonveg)
File COASTGUARD/Toolbox.py, line 2013
model.fit(sample)
The code fails because model cannot take a variable of shape (0,1) (i.e. sample is empty). Tracing this back, this happens on very hazy/dark/poor quality images, which have somehow passed the initial catches for image quality (cloud masking, empty rasters). The reason the array going into FindWPThresh is empty is because not enough classes have been found in order to define a contour along; most likely all pixels have been classes as nonveg.
This needs to be fixed by either catching these images early and discarding based on reflectance histograms, or the slightly faster route of a conditional after the images have been classified which skips this image if the labels are all one class.
In certain sites and images, the veg edge extraction process falls over when contours are attempted to be extracted. Specifically at:
The code fails because
modelcannot take a variable of shape (0,1) (i.e.sampleis empty). Tracing this back, this happens on very hazy/dark/poor quality images, which have somehow passed the initial catches for image quality (cloud masking, empty rasters). The reason the array going intoFindWPThreshis empty is because not enough classes have been found in order to define a contour along; most likely all pixels have been classes asnonveg.This needs to be fixed by either catching these images early and discarding based on reflectance histograms, or the slightly faster route of a conditional after the images have been classified which skips this image if the labels are all one class.