Methods and apparatus for clustering templates in non-metric similarity spaces
View Patent ↗Clustering biometric templates is performed by determining fiduciary templates and cluster seed templates, both from a gallery of biometric templates. Similarity vectors are formed by comparing members of the cluster seed templates to the fiduciary templates. The gallery is then partitioned into clusters based upon the similarity vectors, and the clusters are populated from the remainder of the gallery. Partitioning may be performed by a classifier that implements a supervised machine learning algorithm that is trained with the similarity vectors, such as a multi-decision tree classification system. Matching may be accommodated by accessing a probe template, determining a cluster neighborhood for the probe template, and searching the cluster neighborhood to determine whether the gallery includes a match corresponding to the probe template. The same similarity metric is used both to partition a gallery into clusters, and in matching a probe template to the so-clustered gallery.
1. A non-transitory computer readable medium having stored thereon instructions which cause a processor to implement a method for clustering biometric templates, the method comprising:
determining a set of fiduciary templates from a gallery of biometric templates, the biometric templates being stored in a database;
determining a set of cluster seed templates from the gallery of biometric templates, the membership of the set of cluster seed templates being relatively large in number as compared to the set of fiduciary templates;
forming a plurality of similarity vectors by comparing members of the set of cluster seed templates to the set of fiduciary templates, wherein individual ones of the similarity vectors indicate the similarity of a given member of the set of cluster seed templates to respective ones of the set of fiduciary templates; and
partitioning the gallery of biometric templates into clusters based upon the plurality of similarity vectors.
2. The non-transitory computer readable medium of claim 1 , further comprising instructions for:
populating the clusters using a remainder of the templates in the gallery other than the fiduciary templates and the cluster seed templates.
3. The non-transitory computer readable medium of claim 1 , wherein partitioning is performed by applying a classifier to the plurality of similarity vectors, and the classifier implements a supervised machine learning algorithm that is trained with the plurality of similarity vectors.
4. The non-transitory computer readable medium of claim 1 , wherein partitioning is performed by applying a classifier to the plurality of similarity vectors, and the classifier implements a multi-decision tree classification system.
5. The non-transitory computer readable medium of claim 4 , further comprising instructions for:
accessing a probe template and determining a set of leaf nodes that the probe template is assigned to according to the multi-decision tree classification system;
determining a cluster neighborhood for the probe template to be those members of the gallery of biometric templates that are assigned to the set of leaf nodes; and
searching the cluster neighborhood to determine whether the gallery includes a match corresponding to the probe template.
6. The non-transitory computer readable medium of claim 1 , further comprising instructions for:
accessing a probe template;
determining a cluster neighborhood for the probe template, the cluster neighborhood corresponding to the partitioning of the gallery; and
searching the cluster neighborhood to determine whether the gallery includes a match corresponding to the probe template.