Video Conferencing Tracking Filters for Reflection False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video conferencing systems incorrectly identify reflections of people off glass walls or reflective surfaces as real individuals, leading to miscounting, misidentification, and incorrect zoom adjustments due to tracking and detection software confusion.
Innovation Solution
Implementing three filters: a confidence threshold filter, a similarity filter, and a delay filter to distinguish between real and reflected objects by assigning confidence scores, comparing sub-images for similarity, and delaying the processing of newly detected objects to differentiate between physical and reflective objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If tracking and detection software is used to identify objects in video streams, then object detection capability is improved, but false positive identifications of reflections as real persons occur
Solution Approach 1:
The patent segments the object detection process into multiple independent filtering stages: initial detection, confidence scoring, similarity comparison, temporal consistency checking, and final validation. Each stage processes detection results independently and applies specific criteria, breaking down the complex problem of distinguishing real persons from reflections into manageable sequential steps that collectively improve reliability without sacrificing detection capability.
Solution Approach 2:
The patent introduces intermediary validation mechanisms between detection and final identification: confidence score thresholds act as intermediaries to filter low-quality detections, similarity comparison serves as an intermediary to verify detection plausibility against known patterns, and temporal consistency checking acts as an intermediary to validate sustained detection across frames. These intermediaries prevent false positives from propagating to final results.
2Measurement precision
If confidence scores are assigned to detected sub-images, then detection quality assessment is improved, but processing complexity increases
Solution Approach 1:
The patent transforms the qualitative assessment of detection quality into a quantitative parameter (confidence score) that ranges from 0 to 1. This parameter change enables automated threshold-based filtering, where detections below a specified confidence threshold are automatically rejected. The confidence score parameter consolidates multiple quality factors (image clarity, object completeness, detection certainty) into a single measurable metric that simplifies decision-making while maintaining comprehensive quality assessment.
3Reliability
If similarity comparison between sub-images is performed, then false positive filtering is improved, but computational load increases
Solution Approach 1:
The patent performs preliminary filtering using confidence scores and temporal consistency checks before executing computationally intensive similarity comparisons. Only detections that pass these preliminary filters are subjected to similarity analysis. This preliminary action reduces the number of pairs requiring comparison, significantly lowering computational load while maintaining effective false positive filtering for the subset of detections that reach the similarity comparison stage.
Solution Approach 2:
The patent applies similarity comparison selectively rather than universally - only to detections that meet specific confidence and temporal criteria. This partial action approach focuses computational resources on the most suspicious or uncertain detections where false positives are most likely, rather than wasting resources comparing all detected objects. The similarity threshold is set to require high similarity values, meaning only near-perfect matches are accepted, providing strong false positive filtering with minimal comparisons.
4Reliability
If delay filtering is applied to newly detected objects, then transient detection errors are reduced, but response time to detect real objects increases
Solution Approach 1:
The patent implements dynamic detection thresholds and validation criteria that adapt based on detection history and context. For objects that appear suddenly without temporal continuity, stricter validation is applied. For objects that maintain consistent detection across multiple frames, validation requirements are relaxed. This dynamic approach allows rapid confirmation of persistent real objects while imposing delay filtering primarily on transient detections that are likely false positives, thereby minimizing overall response time impact.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method including detecting, in a digital image, a set of sub-images matching a selected object type. The method also includes generating a first confidence score that a first sub-image in set of sub-images matches a selected object type. The method also includes generating a second confidence score that a second sub-image in set of sub-images matches the selected object type. The method also includes generating a similarity measure by comparing the first sub-image to the second sub-image. The method also includes removing, responsive to the similarity measure exceeding a similarity threshold value and the first confidence score exceeding the second confidence score, the second sub-image from the set of sub-images. The method also includes processing, after removing, the digital image using the set of sub-images.