Video Tracking via Overlapping Time-Window Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The sequential correspondence-determination method for tracking target objects in video data is prone to interruptions due to occlusions or non-detections, and errors are not corrected, while the overall optimization method consumes excessive memory.
Innovation Solution
A tracking apparatus that extracts overlapping time-window frames from video data, integrates tracking results across these frames based on overlapping frames, and determines correspondences by counting votes to correct errors and reduce memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the sequential correspondence-determination method is used for tracking, then the memory usage is reduced, but tracking is interrupted when occlusion or non-detection occurs and errors cannot be corrected
Solution Approach 1:
The patent segments the tracking process into multiple time-window frames, where each time-window frame processes a limited set of consecutive frames. This segmentation allows the system to maintain tracking results within each segment while using overlapping segments to correct errors and maintain continuity, thereby improving reliability without requiring excessive memory resources.
Solution Approach 2:
The patent uses overlapping time-window frames where previous tracking results are retained and used as preliminary information for subsequent tracking. This preliminary action allows the system to quickly resume tracking after occlusion or non-detection events, improving tracking continuity without storing all historical data.
2Measurement precision
If the overall optimization method is used for tracking, then tracking accuracy is improved by correcting errors, but a large amount of memory is used
Solution Approach 1:
The patent divides the video sequence into multiple time-window frames with overlapping segments. Each segment performs local optimization to correct tracking errors, achieving high tracking accuracy without needing to store and process the entire video sequence simultaneously, thus reducing memory requirements.
Solution Approach 2:
The patent applies optimization only to overlapping regions between adjacent time-window frames rather than the entire sequence. This partial action is sufficient to correct errors and maintain tracking accuracy while significantly reducing the computational and memory resources required compared to full-sequence optimization.
3Measurement precision
If non-overlapping time-window frames are used, then memory usage is reduced, but tracking accuracy decreases due to loss of correspondence information
Solution Approach 1:
The patent segments the tracking process into time-window frames with controlled overlaps. The overlap portion contains reference frames that are used to determine correspondences between adjacent segments, ensuring accurate tracking while limiting memory usage to only the necessary overlapping region.
Solution Approach 2:
The patent uses overlapping frames as intermediary elements between adjacent non-overlapping time-window frames. These intermediary frames contain correspondence information that bridges the gap between segments, enabling accurate tracking across the entire sequence without requiring all frames to be stored simultaneously.
Data Source
AI summary
An extraction unit (21) treats two or more consecutive frames of a plurality of frames constituting video data as a time-window frame, and extracts a plurality of time-window frames from the video data so as to cause at least one frame to be overlappingly included in adjacent time-window frames. A tracking unit (22) treats each time-window frame as a target time-window frame, and tracks target objects in the target time-window frame to generate a tracking result. An integration unit (23) integrates tracking results generated in individual time-window frames, based on a frame overlappingly included in more than one time-window frame, to generate an integration result.