Blob Tracker Classification for False Positive Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video analytics systems face challenges in accurately distinguishing between real objects and false positive blobs, leading to increased false alarms and reduced system performance due to factors like moving foliage, external elements, and lighting changes.
Innovation Solution
The system employs a method to maintain and differentiate between two types of blob trackers: frozen and jumping trackers, based on overlapping ratios, irregular size changes, and motion patterns, allowing for the removal of false positive blob trackers before they are converted to normal status.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If blob trackers are maintained and converted to normal status based on continuous association, then object tracking coverage is improved, but false positive rate increases
Solution Approach 1:
The patent segments blob trackers into two distinct types: frozen trackers (for stationary objects) and jumping trackers (for moving objects). This segmentation allows different evaluation criteria to be applied to each type, improving the ability to distinguish true objects from false positives while maintaining comprehensive tracking coverage.
Solution Approach 2:
The patent applies different quality criteria locally to different tracker types. Frozen trackers are evaluated based on alignment stability (overlapping ratio threshold), while jumping trackers are evaluated based on motion irregularity (size change and motion change thresholds). This local quality approach enables precise filtering of false positives without sacrificing tracking sensitivity.
2Productivity
If all blob trackers are converted to normal status after continuous association, then tracking completeness is improved, but false alarm rate increases
Solution Approach 1:
The patent performs preliminary classification of blob trackers into frozen and jumping types before conversion to normal status. This preliminary action allows the system to apply appropriate filtering criteria specific to each tracker type, ensuring that false positives are removed before they can generate false alarms, while maintaining tracking completeness for true objects.
Solution Approach 2:
The patent changes the evaluation parameters based on tracker type: frozen trackers use alignment threshold parameters (overlapping ratio), while jumping trackers use motion parameters (size change threshold, motion change threshold). This parameter change strategy enables the system to maintain high tracking completeness while selectively filtering false positives based on their specific characteristics.
3Object-generated harmful factors
If blob trackers are removed based on strict criteria, then false positive rate is reduced, but tracking sensitivity decreases
Solution Approach 1:
The patent implements dynamic tracker management by continuously monitoring tracker characteristics and transitioning trackers between states (new, frozen, jumping, normal, removed) based on their behavior. This dynamic approach allows the system to maintain high tracking sensitivity by keeping potential true objects in monitoring states while removing only those that clearly exhibit false positive characteristics.
Solution Approach 2:
The patent uses feedback mechanisms where tracker performance metrics (alignment overlap, size change, motion change) are continuously evaluated against thresholds. This feedback loop enables the system to adjust tracker status dynamically, removing false positives based on strict criteria while preserving tracking sensitivity for objects that meet the appropriate criteria for their tracker type.
Data Source
AI summary
Techniques and systems are provided for processing video data. For example, techniques and systems are provided for maintaining blob trackers for one or more video frames. A blob tracker can be associated with a blob generated for a video frame. The blob includes pixels of at least a portion of one or more foreground objects in the video frame. The blob tracker can be determined to be a first type of tracker or a second type of tracker. A first type of tracker has a first bounding box and a second bounding box with an overlapping ratio greater than an alignment threshold for the first type of tracker. A second type of tracker has an irregular size change or an irregular motion change over a threshold duration. The blob tracker can be removed from the plurality of blob trackers maintained for the one or more video frames when the blob tracker is the first type of tracker or the second type of tracker.