Video Retrieval via Object Contextualization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video search systems based on physical contents and semantic objects of video frames do not perform adequately in configuring frame indexes or extracting core features, often due to high-dimensional feature extraction or complex algorithm implementation.
Innovation Solution
A video search system using object contextualization, which includes an object contextualization server that detects objects in video frames, generates contextualization data including object types, attributes, and locations, and stores this data as text files for big data storage. The video search server then performs text-based searches on this data to find videos meeting specific search conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video search systems use high-dimensional feature extraction or complex algorithms to improve search performance, then search accuracy is improved, but device complexity and computational resource requirements increase
Solution Approach 1:
The patent extracts only the essential semantic information (object types, attributes, and relationships) from video frames rather than processing complete high-dimensional features. This extraction approach maintains search accuracy by capturing key semantic content while significantly reducing computational complexity and resource requirements.
Solution Approach 2:
The system creates simplified textual representations (copies) of video content that capture the essential semantic meaning. Instead of storing and processing complex visual features, the patent generates text-based descriptions that can be efficiently searched while preserving the core searchability of video content.
2Reliability
If video search systems store and analyze complete video data to improve search capability, then search completeness is improved, but storage capacity requirements increase
Solution Approach 1:
The patent extracts only the necessary semantic information from complete video data for storage and analysis. By storing only object types, attributes, and contextual relationships rather than full video frames, the system maintains search completeness while dramatically reducing storage requirements.
Solution Approach 2:
The system uses lightweight textual representations instead of heavy video data for storage. These simplified data structures serve the purpose of video search without requiring the continuous retention of large video files, enabling efficient storage and retrieval.
Data Source
AI summary
A search system using object contextualization according to an aspect of the present disclosure stores object contextualization data obtained by contextualizing a detected object and attribute information of the object with respect to one or more video files for each frame, in the form of a text file in a big data storage, and searches for a video that meets a search condition with respect to the stored object contextualization data.