ГІБРИДНИЙ ПІДХІД ДО КЛАСТЕРИЗАЦІЇ ВІДЕОРЯДІВ РІЗНОЇ ДОВЖИНИ
A significant increase in the amount of data to be analyzed and processed requires the introduction of new efficient tools and methods for their collection and storage. This task is especially important when analyzing multimedia, in particular, video data, due to their great redundancy. One of the w...
Saved in:
| Date: | 2025 |
|---|---|
| Main Authors: | , , |
| Format: | Article |
| Language: | English |
| Published: |
V.M. Glushkov Institute of Cybernetics of NAS of Ukraine
2025
|
| Subjects: | |
| Online Access: | https://jais.net.ua/index.php/files/article/view/641 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | Problems of Control and Informatics |
Institution
Problems of Control and Informatics| Summary: | A significant increase in the amount of data to be analyzed and processed requires the introduction of new efficient tools and methods for their collection and storage. This task is especially important when analyzing multimedia, in particular, video data, due to their great redundancy. One of the ways to reduce the amount of information processed is clustering / segmentation of video sequences to isolate parts that are homo-geneous in content. This raises the problem of choosing the required number of clusters as source information. The article is devoted to the development of a hybrid clustering method for analyzing video sequences of various lengths. The method saves the advantages and eliminates the disadvantages of agglomerative hierarchical and fuzzy clusterings. To determine the similarity between segments of video sequences, the Levenshtein metric is used, which allows to calculate the distances between multidimensional sequences of different lengths. The criterion for the clustering process completion as a whole, and, accordingly, the result quality is determined by the Dunn index. The proposed hybrid approach to clustering video sequences is computationally simple to implement and allows solving the multidimensional time series analysis problems of arbitrary nature in the case when it is difficult to determine in advance the necessary number of clusters for splitting and under conditions of uncertainty about their possible overlap, i.e. in the case where the clustering result is the cover construction, and not data partitioning (exact cover construction). |
|---|