Clipping raster image using coordinates from CSV file with Python

Question

I am trying to clip a raster image into a small image using coordinates in a CSV file. In the CSV file, I have minx,maxx,miny,maxy as the coordinates which makes the bounding box and clips the images.

I have a total of 100,000 coordinates and it will create 50,000 images. I am trying to number the images in ascending numbers. Example 0,1,2,3.... till 50,000. The problem is that when clipping it's escaping some images and only clipping 26,000 images. I just cannot find the bug in my code.

I thought it was the coordinate but when checking the coordinates it seems fine, the bug somehow is in my code.

import pandas as pd 
from osgeo import gdal

df = pd.read_csv("G:\\satellite\\align\\satellite.txt")
fs = gdal.Open('G:\\satellite\\align\\satellite.tif')

i, j = 0, 0

while True:
    minx = df.iloc[i][0]
    miny = df.iloc[i][1]
    maxx = df.iloc[i+1][0]
    maxy = df.iloc[i+1][1]
    
    out_tif = r"G:\\satellite\\training_images\\satellite_rename\\{0}.tif".format(j)
    ds = gdal.Translate(out_tif, fs, projWin = [minx, miny, maxx, maxy])
    
    j = j+1
    i += 2
    if i > 100000: # The iteration depends on how many co-ordinates you have
        break

Answer 1

According to gdal python docs:

projWin --- subwindow in projected coordinates to extract: [ulx, uly, lrx, lry]

uly would be 'upper-left' which corresponds to a maximum for y. Try the following for your projWin:

projWin = [minx, maxy, maxx, miny]

It could also explain why your script stops making crops. If the raster is not square (they rarely are) then at some point in your iteration the coordinates may be out of bounds, so to speak, and thus no crop.