Post your theoretical questions / usage questions here.
Hello,
I'm trying to rigidly register a point cloud describing a portion of a bone with a model of this bone so the transform is purely rigid.
Point clouds are pre-registered so that ICP (which is my ground truth algorithm) converges and gives me a good transform.
To my uderstanding, TEASER requires that point clouds have the same number of points.
I tried to complete the smallest point cloud with points of coordinates=(0.,0.,0.) and the results is really bad
I also tried to complete the smallest point cloud with points of coordinates (+inf or -inf) (random choice) and the results is also bad
I wonder if someone has succeeded to register point clouds with different number of points with another strategy to complete the smaller point cloud ?
May be by duplicating some of the points ?
@VincentDaanen TEASER takes in correspondences, not point cloud, so the poor performance is due to incorrect inputs. To generate correspondences, I recommend using learning-based approaches.
Have you read the documentation?
Post your theoretical questions / usage questions here. Hello,
I'm trying to rigidly register a point cloud describing a portion of a bone with a model of this bone so the transform is purely rigid.
Point clouds are pre-registered so that ICP (which is my ground truth algorithm) converges and gives me a good transform.
To my uderstanding, TEASER requires that point clouds have the same number of points.
I tried to complete the smallest point cloud with points of coordinates=(0.,0.,0.) and the results is really bad I also tried to complete the smallest point cloud with points of coordinates (+inf or -inf) (random choice) and the results is also bad
I wonder if someone has succeeded to register point clouds with different number of points with another strategy to complete the smaller point cloud ? May be by duplicating some of the points ?
Thank you
Vincent