abhijeet3922 / change-detection-in-satellite-imagery Goto Github PK

(base) D:\Downloads\Change-Detection-in-Satellite-Imagery-master\Change-Detection-in-Satellite-Imagery-master>python PCAKmeans.py
Operating
(1600, 1600) (1600, 1600)

Both images resized to [1600 1600]

vector_set shape (102400, 25)
Traceback (most recent call last):
File "PCAKmeans.py", line 119, in
find_PCAKmeans(a1,b1)
File "PCAKmeans.py", line 91, in find_PCAKmeans
FVS = find_FVS(EVS, diff_image, mean_vec, new_size)
File "PCAKmeans.py", line 49, in find_FVS
FVS = np.dot(feature_vector_set, EVS)
MemoryError

You have a critical bug in find_vector_set().
i = I + 1 needs to be moved into the while k < ... loop. Your code as is creates one feature vector in vector_set[0,:] and everything else is zero because j is never reset, so the j and k while loops are only entered one time. The code still works, but you're basically running PCA on a single sample, so results are majorly suboptimal. You really don't need the i while loop at all once you make the change.

Word of warning that the code is in python 2, and some functions are deprecated in recent releases of some of the dependencies.

Hi,

This code works well for smaller datasets . but, for bigger images, what shall be resize settings as it is throwing warning of decompression bomb dos attack and error of could not broadcast input array from shape(100) into shape (25) line 19.
Images are resized to [20025 11590 0] and vector set shape to (2320897, 100)

2. Second part Can we generate stats like how much area is impacted in terms of sq meters etc
3. Is it possible to measure changes in classified sections like urban area, land, vegetation etc?