OpenDroneMap / ODM

A command line toolkit to generate maps, point clouds, 3D models and DEMs from drone, balloon or kite images. 📷
https://opendronemap.org
GNU Affero General Public License v3.0
4.71k stars 1.08k forks source link

Does OpenDroneMap support semi global matching #184

Closed JzHuai0108 closed 7 years ago

JzHuai0108 commented 8 years ago

Hello,

It seems to me OpenDroneMap only supports patched based methods like CMVS and PMVS to densify point clouds. Following generating point cloud with PMVS, semi global matching (SGM) was reported to improve the density of point cloud. It seems both Agisoft PhotoScan and Pix4D support SGM.

Is OpenDroneMap capable of performing SGM after PMVS? If not, is there a plan to incorporate SGM into OpenDroneMap?

Thanks,

dakotabenjamin commented 8 years ago

Hi,

I have been researching SGM and looking at ways to incorporate it into our code. Right now, however, there are no concrete plans.

smathermather commented 8 years ago

Good tip, @JzHuai0108. This is definitely something to look into given how dense Photoscan and Pix4D clouds are relative to ODM.

smathermather commented 8 years ago

SGM is definitely on the wish list. It's lack of implementation here is probably the greatest gap in the current toolchain.

edgarriba commented 8 years ago

maybe you can take a look here https://github.com/Itseez/opencv/blob/master/modules/calib3d/src/stereosgbm.cpp

smathermather commented 8 years ago

Yes, OpenCV has it. Here's some notes from @paulinus I got via e-mail:

"The original paper

or it journal version

are still the reference for this as far as i know.

There is an implementation of the method in opencv, but only for the case of 2 images.

For drone images, we would need to handle more images and to split each image into tiles since SGM eats a lot of memory for high resolution images."

Also from recommended reading from Charles Toth at Ohio State:

Hirschmüller H. Stereo processing by semiglobal matching and mutual information. IEEE Transactions on Pattern Analysis and Machine Intelligence 2008, 30(2), 328-41.

dakotabenjamin commented 7 years ago

389 supports SGM