This post has NOT been accepted by the mailing list yet.
I am trying to modify the code of the virtual scanner in order to get some views of a CAD Model from custom point of views. To be clear the ideal output of my code are N views of the object from different viewpoints that rotate around the object on the X and Y axis, but not on the Z axis, since I am not interested in top and bottom view of the object. First I tried to solve my problem using the renderView function, but I was able to produce only flat point clouds, so I thinked about modifying the pcl virtual scanner code using the -single_view option enabled and by passing to it some points of view.
Still I am not able to solve the problem since there are some points that are unclear to me:
- What exactly object coordinates mean? I think I have to save my .pcd in camera coordinates since the generated views are going to be a dataset for pose estimation. Am I doing anything wrong here?
- If I move the camera on the XY axes I should be able to rotate around the object assigning polar coordinates to the viewpoint and a fixed Z distance. However if I record the data in Object Coordinates the output pcd is always the same, while if I record it in Camera Coordinates I can't reproduce the result I want and my output is always translated from the origin.
Here are some picture that I hope explains what I am trying to say:
What I get:
What I want:
Sorry I am relatively new to PCL so maybe I am saying something wrong.