2

I would like to check if my set of NDVI images with min/max values from -1 +1 is good for at least 80%. I took my cue from this other topic: here. Considering that the nodata is -32768 But the values I get on multiple images are all the same.

with rasterio.open(NDVI[0], 'r') as src:
    meta = src.meta.copy()
    meta.update({"nodata":-32768})
    data = src.read(1)
    arr=data[data>-1.0]
pct_valid = 100 * (arr != meta['nodata']).sum() / (meta['width'] * meta['height'])

print('%.4f'%pct_valid)

OUTPUT:
-6.87526

How should I interpret it for my image?

To select only those images that are 80% good instead?

if pct_valid > 80.0:
    write(raster)
1
  • 1
    Could you please clarify what you mean by "good for at least 80%"? Are you trying to make sure at least 80% of all pixels are not NoData values?
    – Aaron
    Aug 31 '20 at 22:27
2

Here is an option that uses numpy to check if NoData values account for >20% of the total pixels.

The code is designed to be included in a for loop while iterating through a directory of images. If the routine detects an image that meets the 80% threshold, it returns a True, else False. Otherwise it would be easy to convert the function to populate a list of valid images.

import os
import rasterio
import numpy as np

img = '/path/to/geotiff.tif'
nodata_value = -32768 # enter the NoData value here


def check_image(img, nodata_value):
    # Read in image as a numpy array
    src = rasterio.open(img).read(1)
    
    # Count the occurance of NoData values in np array
    nodata_count = np.count_nonzero(array == nodata_value)
    
    # Get a total pixel count
    total_count = array.shape[0] * array.shape[1]
    
    # Get a % of NoData pixels
    val = nodata_count/total_count * 100
    
    # Check if image meets threshold of 80% valid pixels
    if val < 20:
        print("Image {0} has >= 80% of pixels that are valid".format(os.path.basename(img)))
        return True
    else:
        print("Image {0} has <80% of pixels that are valid".format(os.path.basename(img)))
        return False
    
if __name__ == "__main__":
    check_image(img, nodata_value)
2
  • I tried your code and it seems to work, doing a print(val) I see that all my images have a value of 0.0, does this mean that there are no knotted pixels in the image?
    – vins_26
    Sep 1 '20 at 9:00
  • I would test this approach on an image you know has <80% valid pixels to make sure the results are what you expect.
    – Aaron
    Sep 1 '20 at 12:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.