Lidar data processing

1. Downsampling
voxel grid
2. Segmentierung (RANSAC Algorithm) (Random Sample Consensus)
select 3 points to build a plane, see how many points ist within the toleranz distance of the plane.
repeat the process, select the plane with 3 points with minimal Outliers.
3. Clustering (Euclidean Clustering)
kd-tree (x split, y split, x split ...) double pointer

Proximity(point,cluster):
    mark point as processed
    add point to cluster
    nearby points = tree(point)
    Iterate through each nearby point
        If point has not been processed
            Proximity(cluster)

 

EuclideanCluster():
    list of clusters 
    Iterate through each point
        If point has not been processed
            Create cluster
            Proximity(point, cluster)
            cluster add clusters
    return clusters

4. BoundingBox

looks at the min and max point values of an input cloud and stores those parameters in a box struct container
5. Visualisierung

Problem: Rechenzeit zu lange, ungefähr 100 ms, normalerweise 20 ms
Lidar: 5-15 Hz