Video Target Search Using Relative Ranking and Space-Time Profiling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for visual search and object re-identification in multi-source visual data face challenges due to the lack of stable and distinctive visual features, especially in scenarios with varying view angles, lighting, and occlusions, making it difficult to associate objects across different camera views effectively.
Innovation Solution
A method that involves receiving a target selection, identifying and ranking potential matches, confirming or rejecting matches through user interaction, and iteratively refining the search by updating the ranking model using relative distance comparison techniques and space-time profiling to improve target finding and tracking across distributed camera views.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If visual features are extracted from multi-source video data for target matching, then target identification capability is improved, but reliability of matching deteriorates due to appearance changes from view angle variations, lighting changes, background clutter and occlusion
Solution Approach 1:
The patent segments the target object into multiple parts or components, extracting features from different segments rather than treating the object as a whole. This segmentation approach makes the feature extraction more robust to appearance changes, as individual segments maintain their characteristics even when the overall object appearance varies due to viewing angle, lighting, or occlusion.
Solution Approach 2:
The patent transforms visual features into a different parameter space or representation format that captures intrinsic object properties rather than appearance-specific characteristics. By changing the parameter representation, the system achieves more reliable matching across different viewing conditions, lighting scenarios, and occlusion states.
2Extent of automation
If automatic object attribute extraction is performed across distributed camera views, then search automation is improved, but measurement precision deteriorates due to lack of stable and distinctive visual features
Solution Approach 1:
The patent implements feedback mechanisms where the system continuously refines its attribute extraction based on matching results and validation against known target characteristics. This iterative feedback process improves measurement precision by correcting errors and adjusting extraction parameters based on actual performance across distributed camera views.
Solution Approach 2:
The patent performs preliminary actions by pre-processing video data and pre-extracting potential features before the actual matching process. This includes pre-segmentation, pre-feature extraction, and preliminary filtering of candidate attributes, which prepares the data in advance to improve both automation efficiency and measurement precision during the actual search and matching operations.
3Loss of information
If multiple independent data sources are integrated for comprehensive target search, then information completeness is improved, but device complexity increases due to challenges in associating information from disparate sources
Solution Approach 1:
The patent merges multiple independent data sources by integrating their outputs into a unified feature representation and matching framework. This merging process combines information from disparate sources while maintaining a coordinated association mechanism that reduces complexity through unified processing rather than separate handling of each source.
Solution Approach 2:
The patent develops a universal data association framework that can handle multiple types of data sources and multiple target types through a single integrated system. This multi-functional approach reduces device complexity by using the same core association mechanisms across different data sources rather than requiring source-specific processing pipelines.
Data Source
AI summary
Method and processor for searching for a target within video data comprising the steps of receiving a target selected from within video data. Identifying a current selection of target matches for the selected target within further video data. Ranking the current selection of target matches. Receiving a signal confirming or rejecting one or more of the ranked target matches. Identifying a further selection of target matches for the confirmed target matches from the further video data. Indicating portions of the further video data containing the further selection of target matches.