Command line tool produces a black map

[LadyZayin]

Hi. Sorry for creating another issue so promptly, but I experienced another problem. I'm attempting to stitch 2 maps using the command line stitcher: the first one has been generated by slam (http://s9.postimg.org/6tmbr3dm7/map.jpg -not yet complete) and the other is the world map (http://s16.postimg.org/wf6oh1t51/real.jpg). Both maps have the same size and resolution. When I attempt to stitch them, the output is completely blank (black) map. Would you know how to fix this?

I believe the output is black because both maps are not similar enough. Indeed, I sometimes get better result after running my slam algorithm longer such that the map is more complete (so it does work sometimes). However, I really need the tool to function with less accurate maps such as the one above because I need to compare good versus bad mapping algorithms.

[pscholl]

I am not sure if the implementation can handle jpg file, you might want to try
with the maps converted to png/pgm.

You could also try playing around with the max_distance parameter, which filters
the corresponding points and is dependent on your map resolution. The basic idea
is to only use point pairs that have a distance less than max_distance, you can
see the selected points for transformation estimation when you supply the
verbose paramter.

On Wed, Jul 17, 2013 at 02:17:13PM -0700, LadyZayin wrote:

Hi. Sorry for creating another issue so promptly, but I experienced another problem. I'm attempting to stitch 2 maps using the command line stitcher: the first one has been generated by slam (http://s9.postimg.org/6tmbr3dm7/map.jpg -not yet complete) and the other is the world map (http://s16.postimg.org/wf6oh1t51/real.jpg). Both maps have the same size and resolution. When I attempt to stitch them, the output is completely blank (black) map. Would you know how to fix this?

I believe the output is black because both maps are not similar enough. Indeed, I sometimes get better result after running my slam algorithm longer such that the map is more complete. However, I really need the tool to function with less accurate maps such as the one above because I need to compare good versus bad mapping algorithms.

---

Reply to this email directly or view it on GitHub:
#2

http://www.ess.tu-darmstadt.de

[LadyZayin]

My bad, I should have mentioned that I had to convert the pgm files to jpg to upload them on the website. I do use pgm.

I will modify the max_distance parameter and tell you what happens! Thanks!
[Edit] I've played around with max_distance - I tried values from 1.0 to 100.0, but the result is always very similar to this: http://s23.postimg.org/wuwywy7q3/debug.png. I indicated with a red arrow where the room from the first map should be found on the second.

[pscholl]

Sorry for the late response, haven't seen your edit.

Besides that the corresponding point look a bit odd, it might definitely be a problem that the map you're aligning too is not generated by a laser scan. So you either align on a laser scanned map or find a way to remove the stuff you won't see on the laser scan (like the stairs, seats and outer walls) .

[LadyZayin]

No problem.

I understand. Then, I suppose mapstitch might not be the program best suited for my problem. I recently found a GIMP plugin named Image Registration which seems to work relatively well. Hopefully, using mapstitch and this tool should work well enough in most cases. Thanks again!

[pallavbakshi]

@LadyZayin I saw your post and I am having similar problem to solve. May I ask how did you solve it? I want to solve it automatically not manually. In case you uploaded the code to fix the problem then kindly direct me to it.

Thank you so much!!