Feature-Vector Loitering Detection Through Tracking Interruptions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for loitering detection in captured scenes are prone to errors due to temporary obscuration or re-identification of moveable objects, leading to inaccurate determination of their presence over a given period.
Innovation Solution
A method using feature vectors extracted by a machine learning module to iteratively update an indicator based on similarity, ensuring robust detection of moveable objects remaining in a scene for a predetermined portion of time, even with temporary interruptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If tracking of moveable objects is used to measure presence time, then detection of objects in scene can be achieved, but tracking may be temporarily interrupted due to obscuration or object leaving/returning to scene
Solution Approach 1:
The patent creates a copy of the object's identity information through feature vectors that can be stored and retrieved even when tracking is interrupted. Instead of relying on continuous tracking data, the system extracts and stores characteristic features of the object that serve as a persistent identifier, allowing reconstruction of presence information after interruptions.
Solution Approach 2:
The patent introduces feature vectors as an intermediary between object tracking and presence determination. These feature vectors act as a mediator that bridges gaps in tracking continuity, enabling the system to associate pre-interruption and post-interruption tracking data through similarity comparison rather than direct continuous tracking.
2Reliability
If re-identification algorithms are used to maintain object identity after tracking interruption, then object association can be restored, but risk of erroneous re-identification increases in scenarios with many objects
Solution Approach 1:
The patent transforms the re-identification problem from a discrete classification task to a continuous similarity measurement task. Instead of assigning fixed object IDs that must be exactly matched, the system uses feature vectors and calculates similarity metrics, allowing for gradual degradation of match quality and more robust handling of ambiguous cases with multiple objects.
Solution Approach 2:
The patent performs excessive feature extraction and similarity comparison by evaluating feature vectors against multiple potential matches and using iterative updating with threshold checks. This partial matching approach, where multiple candidate associations are evaluated rather than committing to a single re-identification immediately, reduces erroneous associations in crowded scenes.
3Measurement precision
If continuous tracking is maintained to ensure accurate presence time measurement, then object location data can be accurately recorded, but system complexity increases to handle interruptions and re-identification
Solution Approach 1:
The patent segments the complex tracking problem into independent feature extraction and similarity comparison tasks. Instead of maintaining a single complex continuous tracking system, the method divides the problem into extracting feature vectors from individual frames and then comparing these vectors to determine presence, simplifying each sub-task while maintaining overall accuracy.
Data Source
AI summary
A plurality of feature vectors for moveable objects are received in a sequence of image frames captured during a first period. The feature vectors are received from a machine learning module trained to extract similar feature vectors in different image frames. An initial value is assigned to a first feature vector , or to a cluster of feature vectors identified in a second sequence of image frames preceding the first image frame. The indicator indicates whether vector is alive. For each subsequent image frame , the indicator is updated by: updating the value based on a similarity between the feature vector and the first feature vector or the cluster of feature vectors. If the value of the indicator indicates that the vector is alive, determining that there is a moveable object that is located in the captured scene at least the predetermined portion of the given period of time.