Occluded Object Tracking Using Group-Proximity Re-Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking systems struggle with occlusions when objects move in groups, leading to the loss of object tracks and the inability to re-identify them accurately due to time-out periods, especially in multi-camera settings.
Innovation Solution
A method that reconnects inactive object tracks by applying a less strict similarity requirement when a detected object instance is spatially proximate to an active object track, allowing for the reactivation of temporarily occluded objects by associating them with connected active tracks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a strict similarity requirement is applied for re-identification, then the accuracy of object re-identification is improved, but the ability to track occluded objects in group movements deteriorates
Solution Approach 1:
The patent applies different similarity requirements to different object tracks based on their local context. Specifically, it identifies groups of objects moving together and applies a relaxed similarity requirement only to objects within the same group that are spatially proximate, while maintaining strict similarity requirements for other objects. This local differentiation resolves the contradiction by allowing flexible tracking within groups without compromising overall re-identification accuracy.
Solution Approach 2:
The patent dynamically adjusts the similarity requirement based on the current tracking situation. When an object is detected to be part of a moving group and is spatially proximate to other tracked objects, the similarity threshold is relaxed. This dynamic adjustment allows the system to adapt to occlusion scenarios in group movements while maintaining accuracy for isolated objects, thus resolving the contradiction between strict accuracy and reliable track continuity.
2Reliability
If the timeout period for inactive tracks is extended, then the ability to re-identify occluded objects is improved, but the system complexity and computational load increase
Solution Approach 1:
Instead of extending the timeout period for all inactive tracks uniformly, the patent applies a localized approach by identifying groups of objects moving together. Only objects within these identified groups that are spatially proximate receive extended timeout treatment, while other tracks maintain the standard timeout period. This reduces overall system complexity compared to a universal extension approach.
Solution Approach 2:
The patent performs preliminary identification of object groups and their spatial relationships before applying timeout extensions. By pre-establishing which objects belong to moving groups and their proximity relationships, the system can efficiently manage extended timeouts only for relevant tracks without increasing overall complexity. This preliminary classification simplifies subsequent track management decisions.
3Productivity
If object tracks are deleted after timeout, then the productivity of the tracking system is improved, but the loss of information about occluded objects increases
Solution Approach 1:
The patent selectively preserves track information for objects within identified moving groups that are spatially proximate, while allowing deletion of other tracks after timeout. This localized preservation maintains object information for groups where occlusion is likely, improving productivity by not unnecessarily retaining all tracks while preventing information loss for relevant occluded objects.
Solution Approach 2:
The system performs preliminary identification of moving groups and spatial proximity relationships before timeout deletion occurs. This allows the system to make informed decisions about which tracks to preserve and which to delete, maintaining productivity by efficiently managing track storage while preventing information loss for objects that are part of coordinated group movements.
Data Source
Figure 1~2A
Figure 2B~2C
Figure 3~5
AI summary
An apparatus and a method for tracking a temporarily occluded object in image frames of objects and associating detected instances into tracks. An active first track of a first object and an active second track of a second object are identified, which objects are moving together. Upon determination that the first track, but not the second track, has become inactive connecting it to the second track. Thereafter, a detected object instance is associated with an inactive track when a similarity score between the detected instance and the inactive track fulfills a similarity requirement. The requirement is less strict when the instance is spatially proximate to the second track and when the detected instance is compared to the inactive first track connected to the second track as compared to when the detected instance is compared to an unconnected inactive track.