Radar-Aided Video Tracking for Redundant Track Elimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video tracking systems often create separate tracks for objects that are related, such as a person and a truck, leading to redundant tracks that can trigger unnecessary alarms or notifications.
Innovation Solution
A method that utilizes radar tracking to identify redundant video tracks by comparing video tracks with radar tracks based on similarity and classification, determining if the radar track corresponds to both the first and second video tracks for the same classification and not the second track's classification, thereby identifying the second track as redundant.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video tracking is performed on all detected objects, then detection completeness is improved, but the number of redundant tracks increases
Solution Approach 1:
Radar data is introduced as an intermediary to verify and filter video tracking results. The radar track serves as a mediator to determine whether a video track represents a main object or a related object, thereby reducing redundant tracks while maintaining detection completeness for relevant objects.
Solution Approach 2:
The patent replaces pure video-based tracking with a hybrid system that uses radar data substitution. Radar tracks are used to validate video tracks, substituting the need for purely visual detection and reducing false positives from related objects like passengers or cargo.
2Measurement precision
If separate tracks are created for related objects, then detection accuracy is improved, but false alarms increase
Solution Approach 1:
The system uses radar tracking data as feedback to verify video tracking results. By comparing video tracks with radar tracks and analyzing classification consistency, the system provides feedback to identify and eliminate redundant tracks, thereby reducing false alarms while maintaining accurate detection of relevant objects.
3Reliability
If radar and video data are integrated, then track reliability is improved, but system complexity increases
Solution Approach 1:
The integration process is segmented into distinct steps: obtaining video tracks, obtaining radar tracks, determining correspondence between tracks, and identifying redundant video tracks. This segmentation of the integration process manages system complexity by breaking down the integration task into manageable, sequential operations.
Data Source
AI summary
Video tracks include positions and a classification relative to respective objects detected in a sequence of video image frames captured of a scene during a time period and the radar track includes positions and classification in relation to of an object detected in radar data captured for at least a part of the scene during the time period. An indication is obtained for each video track of the video tracks whether the radar track and the video track of the video tracks correspond based on a similarity according to a similarity measure. A second video track is ignored or deleted on condition that the radar track and a first video track correspond, the classifications associated with the radar track and the first video track correspond, the radar track and the second video track correspond, and the classifications associated with the radar track and the second video track do not correspond.