0

I want to know if there is any implementation of GMM (Gaussian Mixture Model) for PCL library? Actually I have searched but I couldn't find any library or source code! The main idea is using this GMM technique to detect moving object's point cloud acquired by Kinect sensor in the scene.

Finfa811
  • 618
  • 1
  • 8
  • 28
hadi saadat
  • 13
  • 4

1 Answers1

0

Try using segmentation module, concretely GrabCut algorithm, which can be useful for background/foreground estimation and uses GMM for that purpose.

Finfa811
  • 618
  • 1
  • 8
  • 28
  • Thank you for your answer. As I understood GrabCut needs kind of supervisor to grab the foreground object! am I right? if so it's not that automatic algorithm to be applied. but I couldn't understand how to imply the GMM on GrabCut could you explain a little more? – hadi saadat May 20 '16 at 16:57
  • You don't need to use the whole `GrabCut` algorithm. You just can go through the code in `grabcut_segmentation.cpp` and take the implementations that you need. Perhaps you can use the GMM related functions to create your MoG. If you are trying to detect foreground I guess your scenario is a stream of organized point clouds from a fixed camera pose. So you need to have a loop to read frame by frame and compare every point with the mean of the gaussian components and update the parameters (weight, sd, mean) depending if the point is considered as background or foreground. – Finfa811 May 21 '16 at 18:16