|
This post was updated on .
I'm extracting point clouds in PCD file format using Kinect sensor. My problem is that while the point cloud image I obtain is good, when I view it in pcd_viewer I get an image that is vertically flipped by 180 degrees. Like below[with a bottle as the object]:
![]() So I flipped the Kinect instead and now I get a normal straight pcd file, Like below [with a box and a bottle]: ![]() Is there something I'm doing wrong here? Thanks. |
|
Hi,
just try the key combination ALT + r. This should place the vtk camera to the pose given by the sensor origin/orientation in the point cloud. -Suat On 03/19/2012 02:44 PM, PriyankaP wrote: > I'm extracting point clouds in PCD file format using Kinect sensor. My > problem is that while the point cloud image I obtain is good, when I view it > in pcd_viewer I get an image that is vertically flipped by 180 degrees. Like > below[with a bottle as the object]: > > http://www.pcl-users.org/file/n3840467/flipped.png > > So I flipped the Kinect instead and now I get a normal straight pcd file, > Like below [with a box and a bottle]: > > http://www.pcl-users.org/file/n3840467/normal.png > > Is there something I'm doing wrong here? > > Thanks. > > -- > View this message in context: http://www.pcl-users.org/Orientation-of-point-clouds-obtained-by-Kinect-tp3840467p3840467.html > Sent from the Point Cloud Library (PCL) Users mailing list archive at Nabble.com. > _______________________________________________ > [hidden email] / http://pointclouds.org > http://pointclouds.org/mailman/listinfo/pcl-users _______________________________________________ [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
|
Administrator
|
Have the point clouds been saved using our OpenNI Grabber? If so, should the VIEWPOINT information be correctly stored
in the file? Cheers, Radu. -- http://pointclouds.org On 03/19/2012 06:05 PM, Suat Gedikli wrote: > Hi, > > just try the key combination ALT + r. This should place the vtk camera to the pose given by the sensor > origin/orientation in the point cloud. > > -Suat > > On 03/19/2012 02:44 PM, PriyankaP wrote: >> I'm extracting point clouds in PCD file format using Kinect sensor. My >> problem is that while the point cloud image I obtain is good, when I view it >> in pcd_viewer I get an image that is vertically flipped by 180 degrees. Like >> below[with a bottle as the object]: >> >> http://www.pcl-users.org/file/n3840467/flipped.png >> >> So I flipped the Kinect instead and now I get a normal straight pcd file, >> Like below [with a box and a bottle]: >> >> http://www.pcl-users.org/file/n3840467/normal.png >> >> Is there something I'm doing wrong here? >> >> Thanks. >> >> -- >> View this message in context: >> http://www.pcl-users.org/Orientation-of-point-clouds-obtained-by-Kinect-tp3840467p3840467.html >> Sent from the Point Cloud Library (PCL) Users mailing list archive at Nabble.com. >> _______________________________________________ >> [hidden email] / http://pointclouds.org >> http://pointclouds.org/mailman/listinfo/pcl-users > > _______________________________________________ > [hidden email] / http://pointclouds.org > http://pointclouds.org/mailman/listinfo/pcl-users [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
|
i am having the same issue: one of my pointclouds, after running the VoxelGrid and then calling savePCDFileASCII, is flipped vertically.
Any suggestion? |
|
I also get the problem after updating to ubuntu 12.04.. ALT+R don't fixe the problem!
|
|
I also have this problem. Images are flipped top down, as someone used flipud.m of matlab. I see that in the last ouptut visualization image of the "cluster recognition and 6DOF pose estimation using VFH descriptors" tutorial.
I am using PCL-1.5.1 version and my platform is Windows7,64Bit
On Wed, May 23, 2012 at 12:21 PM, Ivan Silva <[hidden email]> wrote: I also get the problem after updating to ubuntu 12.04.. ALT+R don't fixe the _______________________________________________ [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
|
Administrator
|
We are now using the viewpoint information stored in the PCD file/the cloud itself (sensor_origin and
sensor_orientation), instead of magically trying to infer one. Cheers, Radu. On 05/24/2012 01:22 AM, Inna Stainvas wrote: > I also have this problem. Images are flipped top down, as someone used flipud.m of matlab. I see that in the last > ouptut visualization image of the "cluster recognition and 6DOF pose estimation using VFH descriptors" tutorial. > I am using PCL-1.5.1 version and my platform is Windows7,64Bit > > > On Wed, May 23, 2012 at 12:21 PM, Ivan Silva <[hidden email] <mailto:[hidden email]>> wrote: > > I also get the problem after updating to ubuntu 12.04.. ALT+R don't fixe the > problem! > > -- > View this message in context: > http://www.pcl-users.org/Orientation-of-point-clouds-obtained-by-Kinect-tp3840467p4008278.html > Sent from the Point Cloud Library (PCL) Users mailing list archive at Nabble.com. > _______________________________________________ > [hidden email] <mailto:[hidden email]> / http://pointclouds.org <http://pointclouds.org/> > http://pointclouds.org/mailman/listinfo/pcl-users > > > > > _______________________________________________ > [hidden email] / http://pointclouds.org > http://pointclouds.org/mailman/listinfo/pcl-users [hidden email] / http://pointclouds.org http://pointclouds.org/mailman/listinfo/pcl-users |
|
Dear Radu, Thank you a lot!
Please can you help a little bit more! It could be nice to have somewhere description of these parameters. I am looking at cloud_properties.h and sensor_orientation.h. As I get sensor_origin is the camera origin in some "world coordinate system (CS)"; while sensor_orientation has 4 parameters, though 3 angles is sufficient to describe one CS w.r.t another one; so I don't really know what these 4 parameters stand for.... But as I understand if sensor_origin=(0,0,0). i.e. coincide with the camera origin, then all objects that are close to the camera will have smaller z and thus will appear below far away objects; so that objects lying on the table will appear as being under it.... Is it correct? And what really sensor_orientation is? I thought to understand by looking for some geometrical transorm, but did not do this yet. |
| Powered by Nabble | Edit this page |
