Quantcast

pairs of point clouds registration

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

pairs of point clouds registration

herman
This post has NOT been accepted by the mailing list yet.
I met a problem in points clouds registration,

the question is : I have used a 3d structured lighting scanner to get 12 point clouds(every 30° around the object), now I need to register these points cloud to get a complete object,my processes are  as below:

1、input the pair 1->pass through z ->statistic filter ->and we will get the scene points named scene
2、input the pair 2->pass through z ->statistic filter ->and we will get the object points now named obj
3、then for both scene and obj:
  downsampling->estimating surface normals->find the keypoints(using SIFT)->Feature         Estimation(using SHOTFeature)->Correspondence Esitmation-> Rough Align->ICP->we will get the result named complete
4、updata the complete to scene   and  repeat step2 to step 4 till all these 12 points cloud are processed
5、we will get a complete object at last
but as you see, this registration(one pair after one pair) is easy to cause  error accumulation, and the result is not so good
so I am looking for a better solution now.
thanks a lot!
herman
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: pairs of point clouds registration

vrshy
You can look into LUM and ELCH both of which are implemented in PCL, if you include them in your chain at the end after ICP your results will improve.
Loading...