Video Event Tracking Using Scene Cuts and Image Hash Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is difficult for users to simultaneously play video games and selectively record exciting or surprising content while reacting to the game, and summarizing game progress is challenging due to the vast amount of user-directed content generated during play.
Innovation Solution
An apparatus and method for video tracking that identifies predetermined events within a sequence of images by detecting scene cuts using perceptual hashes and comparing image hashes to a database of notable events, employing computational techniques to reduce resource overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If users attempt to simultaneously play video games and selectively record exciting content, then recording capability is improved, but user attention and gameplay experience deteriorate
Solution Approach 1:
The patent segments the video content analysis into distinct phases: scene cut detection identifies temporal boundaries, candidate event images are selected at these boundaries, and database matching identifies specific events. This segmentation allows the system to process only relevant frames rather than analyzing every frame, reducing computational complexity while maintaining automation.
Solution Approach 2:
The system performs preliminary action by pre-detecting scene cuts and identifying candidate event images before full event recognition. By pre-processing the video to identify potential event locations through scene change detection, the system reduces the computational burden of subsequent event matching operations.
2Measurement precision
If comprehensive video analysis is performed to identify all events, then event detection accuracy is improved, but computational resource consumption increases
Solution Approach 1:
The patent applies local quality by focusing computational resources only on candidate event images identified at scene cut boundaries. Rather than analyzing all video frames uniformly, the system concentrates processing power on specific localized regions (temporal boundaries) where events are most likely to occur, improving efficiency while maintaining detection accuracy.
Solution Approach 2:
The system performs partial action by analyzing only a subset of frames (candidate event images at scene boundaries) rather than all frames. This partial analysis approach achieves sufficient event detection accuracy for practical purposes while dramatically reducing computational resource requirements compared to comprehensive frame-by-frame analysis.
3Difficulty of detecting and measuring
If scene cut detection is performed to identify candidate events, then event detection capability is improved, but false positive matches increase
Solution Approach 1:
The patent implements feedback by comparing detected candidate events against a database of known events and using the matching results to refine event identification. The system evaluates whether detected scene cuts correspond to actual predetermined events through database comparison, providing feedback that filters false positives while maintaining detection capability.
Data Source
Figure 1
Figure 2A~2B
Figure 2C~2D
AI summary
A method of identifying a predetermined event within a sequence of images comprises the steps of obtaining a database of data items each representing one of a plurality of predetermined events, identifying candidate event images within the sequence of images, comparing data representing at least a first a candidate event image with one or more data items in the database, and identifying that a predetermined event has occurred within the sequence of images if a candidate event image matches a data item in the database to a predetermined matching threshold degree.