Out of memory tiling some maps

[GoogleCodeExporter]

What steps will reproduce the problem?
1. convert2gdal.py --get-cutline 18502_3.KAP
2. gdal_tiler.py -c -z "14 15 16" 18502_3.vrt

What is the expected output? What do you see instead?

Got python out of memory error after over 6 hours.  (Python 2.7_64 ... 8GB ram)

What version of the product are you using?

hg clone Sept, 30 2011

What was the operating system?

Windows

Please provide any additional information below.

I am attempting to find a solution myself for this and post if i am able to 
solve.  

Original issue reported on code.google.com by manim...@gmail.com on 1 Oct 2011 at 2:20

Attachments:

• 18502_3.KAP

[GoogleCodeExporter]

The direct tiling of the .kap file produces an expected output.
So, it seems to be there is some issue with generating .vrt file by 
map2gdal.py. I'll investigate it further.

BTW in the hg version of the tools some scripts were renamed and some 
parameters syntax was changed, so the proper command sequence should be: 

1. map2gdal.py 18502_3.KAP
2. gdal_tiler.py -c -z 14:16 18502_3.vrt
or 
2. gdal_tiler.py -c -z 14,15,16 18502_3.vrt

Original comment by vadp.d...@gmail.com on 2 Oct 2011 at 8:54

• Changed state: Accepted

[GoogleCodeExporter]

Meanwhile, there is a workaround for tiling the file attached with a borders 
removed:

map2gdal.py --cut-file 18502_3.KAP
gdal_tiler.py --cut --cutline=18502_3.gmt 18502_3.KAP

Original comment by vadp.d...@gmail.com on 2 Oct 2011 at 8:58

[GoogleCodeExporter]

The workaround above should work fine for files with the Mercator projection 
and WGS-84 (or equivalent) datum

Original comment by vadp.d...@gmail.com on 2 Oct 2011 at 9:04

[GoogleCodeExporter]

Perfect, work-around worked.

Original comment by manim...@gmail.com on 3 Oct 2011 at 6:06

[GoogleCodeExporter]

The problem is actually with the chart attached as it's actually poorly 
calibrated. gdal_tiler.py was using thin plate spline transformer for warping 
based on GCPs, but these wrongly aligned GCPs were causing the warping going 
nuts.

A new versions of map2gdal.py and gdal_tiler.py were pushed into the rep. These 
changes use for warping a geotransform by default. 

The result in this case looks now quite convincing. 

The thin plate spline transformer can be forced by gdal_tiler's "--tps" switch. 
Which can be useful for some other poorly calibrated charts.

Original comment by vadp.d...@gmail.com on 4 Oct 2011 at 12:03

• Changed state: Fixed

[GoogleCodeExporter]

Original comment by vadp.d...@gmail.com on 4 Oct 2011 at 12:05