Discovering Homogeneous Groups from Geo-Tagged Videos

Area of Science:

• Computer Science
• Data Science
• Geospatial Analysis

Background:

• Intelligent devices generate vast amounts of geo-tagged data (videos, images).
• This data captures spatiotemporal movements, offering insights into traveler behavior and enabling movement prediction.
• Existing methods lack efficient ways to analyze complex spatio-temporal data with directional information.

Purpose of the Study:

• To develop algorithms for discovering homogeneous groups from geo-tagged videos with view directions.
• To extend density clustering to incorporate fields-of-view (FoVs) for richer analysis.
• To propose an optimized model for efficient processing of this data.

Main Methods:

• Proposed algorithms for homogeneous group discovery from geo-tagged videos.
• Extended density clustering to support fields-of-view (FoVs).
• Developed a two-level grid-based index for optimization.

Main Results:

• Demonstrated the efficiency of the proposed algorithms.
• Showcased the effectiveness of the homogeneous pattern discovery approach.
• Validated the approach on both real and synthetic datasets.

Conclusions:

• The developed methods provide an effective way to analyze traveler behavior from geo-tagged video data.
• The approach enhances understanding of group dynamics and aids in movement prediction.
• The optimized model ensures efficient processing of large-scale spatio-temporal datasets.