PCL : computing FPFH at given keypoints
Suppose i have a bunch of keypoints detected by 2D SIFT or SURF and i want to calculate FPFH descriptors at those keypoints. How can i do this?
ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
Suppose i have a bunch of keypoints detected by 2D SIFT or SURF and i want to calculate FPFH descriptors at those keypoints. How can i do this?
There is a PCL tutorial to show how to do it
You are right, I did not read the tutotial and only assumed it would work like the SHOT descriptor. Try to ask this question on pcl-users.org
There is a method on the fpfh estimation object to specify the indices of the source cloud points that will be used as keypoints:
PointIndicesPtr keypoints (new PointIndices);
// Compute NARF keypoints here ...
pcl::FPFHEstimation<pcl::PointXYZ, pcl::Normal, pcl::FPFHSignature33> fpfh;
fpfh.setInputCloud(sourceCloud);
fpfh.setInputNormals(normals);
fpfh.setSearchMethod(kdtree);
fpfh.setRadiusSearch(0.05);
fpfh.setIndices(keypoints);
fpfh.compute(*descriptors);
//...
This question is quite old, but I figured this might still be useful
Asked: 2013-03-04 09:47:16 -0500
Seen: 2,048 times
Last updated: Mar 04 '13
IDE for ROS (kinetic) code debugging
Tried Normal Distributions Transform with my own files (in correct PCD format) and it throws errors
Is raspberry pi 3 suitable for testing ros with pcl for computer vision purpose?
Nodelet crashes when using pcl::VoxelGrid<pcl::PointXYZ>, pcl::SACSegmentation<pcl::PointXYZ>
parallel plane from an obtained plane with PCL
QVTKWidget is sluggish in rqt in Kinetic 16.04, but fine in Jade 14.04