Video Scene Indexing With Behavior IDs and Person Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to search for scenes showing specific behavior of a specific person in videos using simple keyword searches, as they cannot associate specific behavior with a search key.
Innovation Solution
A video processing system that generates skeleton information from body regions, converts it into behavior IDs, specifies person IDs from facial regions, and registers these IDs and scene-related information in a database for easy retrieval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If facial image search system stores only facial features and basic attributes in database, then database storage is simple, but cannot search for specific behavior of specific person
Solution Approach 1:
The patent segments video information into multiple components: facial features (for person identification), body skeleton information (for behavior recognition), and attributes. This segmentation allows the system to store and search for specific behavior of specific persons by combining facial feature data with skeleton-based behavior data in the database.
2Measurement precision
If store detailed behavior information for every scene, then search accuracy improves, but database complexity and storage requirements increase
Solution Approach 1:
The patent extracts only the essential behavior information from video data by generating skeleton information that represents key body parts and their movements. This extracted skeleton data is then converted into behavior IDs and stored in the database, providing accurate behavior search capability while maintaining database simplicity by storing only essential features rather than complete video data.
3Measurement precision
If use complex behavior recognition algorithms, then behavior identification accuracy improves, but computational load increases
Solution Approach 1:
The patent replaces complex behavior recognition algorithms with a skeleton-based approach that models human body movements using simplified geometric representations (lines and joints). This substitution reduces computational complexity while maintaining behavior identification accuracy by focusing on key skeletal landmarks and their temporal variations rather than analyzing complete image data.
Data Source
AI summary
A video processing system (10) includes: an image acquisition unit (11) that acquires at least one frame image included in video data; a skeleton information generation unit (13) that generates skeleton information based on a body region of a person included in the at least one frame image; a behavior conversion unit (14) that converts the skeleton information into a behavior ID; a person specifying unit (18) that specifies, based on a facial region of the person included in the at least one frame image, a person ID for identifying features of a person estimated to be an identical person; and a registration unit (20) that registers the behavior ID, the person ID, and scene-related information related to the at least one frame image in a database in association with each other.