moving points matching 2d laser scan
Hello everyone!
I start saying that I am a neophyte in using PCL but I have a problem I cannot figure out how to solve.
I am writing a pedestrian detector and tracker which uses combined informations from Lidar and Camera. The point clouds I am handling are 2D laser scans.
The detections are made in the Image plane and then I can retrieve the position of each detection in the point cloud without any clustering.
My problem is that in this way I cannot have a good motion estimation to feed the Kalman Filter (which is tracking points onto the ground plane) so I am looking to some good technique to estimate the motion of a point given the current position and some old point clouds saved in a buffer.
Basically, the problem is to associate one single point (i.e. the detection) with another point belonging to the previous point cloud to be able to compute its estimated speed and orientation.
How could it be done by using PCL or other libraries?
I have seen online that the Nearest Neighbour could be the simplest way but is it possible to use the PCL NN (working with Kd-tree I guess) to find the points in a neighbourhood of ONE single point? Moreover, I am working with only 2 coordinates (2D laser scan), could it be an issue?
Sorry for that but I am a bit confused about the topic.
Thank you
Best regards
