Package boofcv.alg.tracker.tld
Class TldNonMaximalSuppression
java.lang.Object
boofcv.alg.tracker.tld.TldNonMaximalSuppression
Performs non-maximum suppression on high confidence detected regions. A graph of connected regions is constructed.
Two regions are considered connected if their overlap is above a threshold. A region is considered a local maximum
if it has a score higher than all its neighbors. A weighted average is computed using all regions connected to
the local maximum.
NOTE: This is a completely different non-maximum algorithm from what was described in the paper. The algorithm
described in the paper only approximates non-maximum suppression.
-
Nested Class Summary
Modifier and TypeClassDescriptionstatic class
Contains connection information for a specific rectangular region -
Constructor Summary
ConstructorDescriptionTldNonMaximalSuppression
(double connectionThreshold) Configures non-maximum suppression -
Method Summary
Modifier and TypeMethodDescriptionvoid
Finds local maximums from the set of provided regions
-
Constructor Details
-
TldNonMaximalSuppression
public TldNonMaximalSuppression(double connectionThreshold) Configures non-maximum suppression- Parameters:
connectionThreshold
- Two regions are considered connected of their overlap is ≥ to this value. 0 to 1.0. A value of 0.5 is recommended
-
-
Method Details
-
process
Finds local maximums from the set of provided regions- Parameters:
regions
- Set of high confidence regions for targetoutput
- Output after non-maximum suppression
-
getConnections
-