There are two types of plugins supported by CloudCOmpare right now:
standard plugin (generally this is just a bridge to an existing library/algorithm developed by third-parties)
OpenGL filter (this corresponds to advanced shaders that require some processing and privileged access to the rendering context through frame buffer object, etc.)
Standard plugins
qPCV - Global illumination (ShadeVis algorithm - Borgo et al.) [STABLE]
qHPR - Hidden Point Removal (Katz et al.) [STABLE]
qPoissonRecon - 3D meshing based on Poisson reconstruction (Kazhdan et al.) [STABLE]
qRansacSD - RANSAC Automatic Shape Detection from point clouds (Schnabel et al.) (only works on Visual C++ for the moment)
qKinect - capture of 3D colored point clouds with the Microsoft Kinect camera (IN DEV.)
OpenGL filters
qEDL - Eye Dome Lighting - real time approximation of global illumination (Boucheny et al.) [STABLE]
In fact, this plugin is still a "work in progress" (to tell the truth, I almost forget about it ;). So if it has ever been released in the official version, it was an error. Or did you simply get it from the sources?
I never had the time to fully test it, and I was also wondering if any users could use it "as is". From what you say, I understand that it's "sufficient" to install the libfreenect drivers?
Anyway, it's good to see that it does work, and even on a totally different platform than mine :). i'll try to package this properly and ship this plugin in the official version "asap".
It's in the 2.2 version thats is still available as old stable one. You maybe forgot to remove it but well, that's great!
I just install drivers and press on this "strange" button qKinect and Talaaa: point cloud. I just reverse the axis for the sensor to get a depth map.
in fact with have a student working on Kinect here, in EDF so it was interesting to show it.
The main issue concerns the texture with an important offset between points and texture, but well, that's not so bad... not for reconstruction purposes...
In fact, the offset between points and texture comes from the fact that libfreenect couldn't give me the internal Kinect calibration information at that time. So I used the 'standard' information that can be found on the web. I don't know if there is a better way to do this now.
I was in a conference two weeks ago where photogrammetrists made a paper for the calibration of the kinect. As soon I have the proceedings I will send you the article if you want to have a look.
I know that Microsoft will release official SDK and drivers, results will certainly be better.
It's about the schnabel plugin. Will it be released soon, since it's a really important point for me to extract primitive shapes from point clouds in my Phd work (which is concerning registration).
I really hope to see this plugin in cloud compare!!!
Well it will also be useful not for my personal work too, of course :)
It's already available in the Visual 2008 version. For this, you just have to download the sources, and launch the project with the 'start_visual.bat' file (XP) or 'start_visual_x64.bat' (Vista/Seven 64 bits) in the 'trunk\qCC\msvc\VC9' folder. You'll also need a Visual compatible version of Qt and the Qt Visual Addin for MSVC 2008.
It is almost fully operative : it segments the cloud as Schnabel's library outputs, and it is alerady able to associate planar and cylindrical parts with the corresponding mesh primitive. The other primitives should be easy to add to CloudCompare (cone, torus and sphere).
However, the big issue is the compilation with MinGW ...
Daniel
Attachments
qRansacSD plugiin output (planes only)
cc_ransac_sd_planes.jpg (95.62 KiB) Viewed 24016 times
