[vtkusers] Weird data and SEGFAULT when using vtkIterativeClosestPointTransform

[Ronald van Duren]

Hi,


I want to create a point cloud by moving and rotating my depth sensor(Structure Sensor). I figured out ICP is probably the way to go since I can get a tranformation matrix which I can use to shift the sensors frame left, right up or down. Otherwise the points would be drawn on top of each other in static frame of 640x480.


After some time I got the vtkIterativeClosestPointTransform to work, I used the example provided by VTK. But the transformation matrix has weird data in it, sometimes it has several NaN values in it and the overal data in the matrix just doesn't seem right. My linear algebra skills are intermediate so I could be wrong. Also, when I move the sensor with a speed that is anything more than extremely slow a segfault is thrown. Because the source code is compiled I cannot trace down the segfault.


Any chance somebody knows what's wrong?


Thanks in advance!


Ronald van Duren.


Technical information:

  *   ?I use Android and specifically the NDK so everything is in C++.
  *   I use the compiled dynamic library for Android, Armeabi-v7a.
  *   I use the Structure Sensor.
  *   VTK version is 2.4.9.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtkusers/attachments/20160307/082ee833/attachment.html>