Better detect files with geotransform data that aren't geospatial

gdalwarp and gdal_translate can be used to transform images in pixel
space.  If such an image has no projection specified and cannot be
interpreted as latitude and longitude, assume it is not geospatial.

CHANGELOG.md

@@ -7,6 +7,7 @@
 - Format dates in item lists ([#1707](../../pull/1707))
 - Guard dtype types ([#1711](../../pull/1711), [#1714](../../pull/1714), [#1716](../../pull/1716))
 - Better handle IndicaLabs tiff files ([#1717](../../pull/1717))
+- Better detect files with geotransform data that aren't geospatial ([#1718](../../pull/1718))
 
 ### Changes
 

sources/gdal/large_image_source_gdal/__init__.py

@@ -979,7 +979,13 @@ def isGeospatial(path):
             if ds.GetProjection():
                 return True
             if ds.GetGeoTransform(can_return_null=True):
-                return True
+                w = ds.RasterXSize
+                h = ds.RasterYSize
+                trans = ds.GetGeoTransform(can_return_null=True)
+                cornersy = [trans[3] + x * trans[4] + y * trans[5]
+                            for x, y in {(0, 0), (0, h), (w, 0), (w, h)}]
+                if min(cornersy) >= -90 and max(cornersy) <= 90:
+                    return True
             if ds.GetDriver().ShortName in {'NITF', 'netCDF'}:
                 return True
         return False