Clustering

Algorithms for detecting objects as clustered points in the foreground scene.

Table 1. Available fields in blickfeld.percept_pipeline.config.Clustering
Field	Type	Default	Unit
db_scan	DBScan	-	-
db_scan	Use db-scan for clustering the foreground

D BScan

Density-based spatial clustering algorithm using the euclidean distance between points

Table 2. Available fields in blickfeld.percept_pipeline.config.Clustering.DBScan
Field	Type	Default	Unit
min_points	uint32	-	-
min_points	Minimum amount of points required to define an object-cluster Reasonable default: 10
radius	float	-	m
radius	Controls how far away points belonging to the same object-cluster can be. Reasonable default: 0.2
distance_dependent_radius	float	-	m
distance_dependent_radius	Multiply the given radius with the range of a point and add it to the radius. This significantly improves the clustering for close-range scenes. Reasonable default: 0.025