A Novel Fuzzy Distance-Based Minimum Spanning Tree Clustering Algorithm for Face Detection

Solving a clustering algorithm can usually be simplified into an optimization problem. Using relevant knowledge in graph theory, many optimization problems can be transformed into solving minimum spanning tree problems. Minimal spanning trees are also widely used in areas closely related to cognitiv...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Cognitive computation Jg. 14; H. 4; S. 1350 - 1361
Hauptverfasser: Li, Yang, Zhou, Wenju
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.07.2022
Springer Nature B.V
Schlagworte:
Access control
Algorithms
Ambient intelligence
Artificial Intelligence
Clustering
Cognition
Computation
Computation by Abstract Devices
Computational Biology/Bioinformatics
Computer Science
Data analysis
Datasets
Face recognition
Graph theory
Methods
Neighborhoods
Optimization
Set theory
Fuzzy distance
Face recognition
Clustering algorithm
Minimum spanning tree
ISSN:1866-9956, 1866-9964
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Solving a clustering algorithm can usually be simplified into an optimization problem. Using relevant knowledge in graph theory, many optimization problems can be transformed into solving minimum spanning tree problems. Minimal spanning trees are also widely used in areas closely related to cognitive computing such as for face recognition by face cognition and gene data analysis by gene cognition. However, the minimum spanning tree has the shortcoming of the distance between neighbours because of which the minimum spanning tree algorithm cannot cluster unbalanced data. Thus, the face recognition rate is low, and facial expression cognition is difficult. In this paper, a minimum spanning tree algorithm based on fuzzy distance is proposed for the shortcomings of the minimum spanning tree (FCP). First, a relative neighbourhood distance measure is proposed by introducing neighbourhood rough set theory; the neighbourhood matrix is obtained based on the distance. Second, the minimum spanning tree is solved by the prim algorithm and the neighbourhood matrix. Finally, the minimum spanning tree is partitioned to realize clustering of the minimum spanning tree. In this paper, the UCI dataset and Olivetti face database are selected to verify the performance of the algorithm, and the algorithm is evaluated by three evaluation criteria. The experimental results show that the proposed algorithm can not only cluster data of any shape but also deal with unbalanced data containing noise points. Especially in face cognitive computing, the values of ACC, AMI, and ARI can reach 0.852, 0.843, and 0.782, respectively. In this study, the algorithm can obtain very good clustering results for data with good geometric structure, and the overall performance is better than other algorithms. In face recognition detection, the improved cognitive computing of faces makes it possible to accurately recognize different expressions from the same person.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1866-9956
1866-9964
DOI:10.1007/s12559-022-10002-w