Tracking Assistance Device for Multi-Camera Video Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring systems face challenges in efficiently checking and correcting tracking results for moving objects across multiple cameras, leading to increased burden on monitoring personnel and inefficiencies in error handling.
Innovation Solution
A tracking assistance device and method that displays videos from multiple cameras, using a tracking target setter, confirmation video presenter, candidate video presenter, and tracking information corrector to refine and correct tracking information by identifying the highest and next highest possible videos of the tracking target, allowing for simple error correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a monitoring person sequentially checks videos from each camera to track a moving object, then the tracking work can be performed, but the burden on the monitoring person increases and efficiency decreases
Solution Approach 1:
The system segments the tracking task by automatically dividing video streams from multiple cameras into discrete tracking candidates with confidence scores. The confirmation video presenter extracts and presents only the most relevant video segments for verification, breaking down the overwhelming full-video review into manageable, prioritized portions.
Solution Approach 2:
The system performs preliminary tracking and candidate extraction before presenting videos to the monitoring person. The tracking process executes in advance to identify potential tracking targets and generate confidence scores, so that when the monitoring person needs to verify tracking results, the work is already partially completed with high-probability candidates pre-selected.
2Reliability
If all videos from multiple cameras are displayed for tracking verification, then complete tracking information is available, but the complexity of the display system increases and error identification becomes difficult
Solution Approach 1:
The system applies local quality by treating different video candidates differently based on their tracking confidence scores. High-confidence candidates are presented as primary confirmation videos, while lower-confidence candidates are presented as alternative options. This differentiated presentation quality helps the monitoring person focus verification efforts where they are most needed.
Solution Approach 2:
The confirmation video presenter extracts only the most relevant video segments from the complete set of camera feeds. Instead of displaying all videos simultaneously, the system extracts and presents individual high-probability tracking candidates, making the verification process simpler while maintaining reliability through selective presentation.
3Productivity
If the system automatically selects tracking targets without confirmation, then processing speed increases, but the accuracy of tracking identification decreases
Solution Approach 1:
The system incorporates feedback by presenting tracking candidates to the monitoring person for confirmation. The monitoring person can verify whether the automatically identified tracking target is correct, providing human feedback that corrects automated tracking errors. This feedback loop maintains high processing speed while improving identification accuracy through human-in-the-loop verification.
Solution Approach 2:
The system performs partial automation by automatically extracting high-confidence tracking candidates but leaving final confirmation to human operators. Rather than fully automated tracking that might make errors, or fully manual tracking that would be slow, the system performs the easier portion automatically and reserves human judgment for cases requiring verification, achieving a balance between speed and accuracy.
Data Source
AI summary
Included are tracking target setter that sets a person as a tracking target in response to designation by a monitoring person, confirmation video presenter that extracts a video of a moving object with the highest possibility of being the moving object which is the target, based on tracking information, and displays only the videos as confirmation videos, candidate video presenter that displays as a candidate video, a video of a moving object that may be the moving object set as the tracking target is the next highest of the moving object of the confirmation video having an error, where there is an error in the confirmation video, and causes selection of a corresponding candidate video by the monitoring person, and tracking information corrector that corrects the tracking information so that the moving object corresponding to the selected candidate video is associated with the moving object set as the tracking target.