Multi-Sensor Subject Tracking With Live and Dormant Anchors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-camera tracking technologies face challenges in real-time multi-subject tracking due to scalability issues and complexity in synchronizing and calibrating multiple cameras, especially in densely populated areas, leading to overburdened compute resources and inaccurate trajectory tracking.
Innovation Solution
A real-time location system (RTLS) processes synchronized optical image streams in micro-batches to initialize and maintain anchors for each subject, using behavior embeddings and hierarchical clustering to manage live and dormant anchors, enabling efficient and continuous tracking with sub-second response times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple cameras are used for multi-subject tracking, then tracking coverage and subject detection capability are improved, but system complexity and compute resource requirements increase significantly
Solution Approach 1:
The system divides the monitored area into multiple zones, each handled by a dedicated camera. Subjects are tracked within their respective zones rather than across all cameras simultaneously, reducing the computational complexity of multi-camera coordination while maintaining comprehensive coverage.
Solution Approach 2:
The patent introduces an intermediary coordinate system that standardizes position data from multiple cameras into a common reference frame. This mediator layer simplifies the integration of multi-camera data by providing a unified spatial representation, reducing the complexity of synchronizing and calibrating multiple camera views.
2Measurement precision
If synchronized optical image streams from multiple sensors are processed, then real-time multi-subject tracking accuracy is improved, but processing time and computational load increase
Solution Approach 1:
The system pre-establishes anchor points in the monitored area before tracking begins. These pre-defined anchors provide reference positions that simplify real-time subject location calculations, allowing the system to achieve high tracking accuracy without performing complex computations on every incoming image frame.
Solution Approach 2:
The patent processes only the necessary portions of image data required for tracking decisions. Rather than analyzing entire high-resolution frames, the system extracts and processes only the relevant features and regions containing subjects, reducing computational load while maintaining tracking precision.
3Measurement precision
If complex clustering and matching algorithms are applied to all anchors, then subject identification accuracy is improved, but processing speed decreases
Solution Approach 1:
The system applies different processing strategies to different types of anchors based on their characteristics. Active anchors (those currently tracking subjects) receive full clustering and matching processing, while dormant anchors use simplified verification. This localized quality approach ensures high identification accuracy for active tracks while maintaining overall processing speed.
Solution Approach 2:
The patent dynamically adjusts processing parameters based on anchor state and subject activity. When subjects are stationary or in low-motion states, the system reduces the frequency and complexity of clustering operations. When motion detection indicates active tracking needs, full processing algorithms are applied, optimizing the balance between accuracy and speed.
Data Source
AI summary
In various examples, multi-sensor subject tracking for monitored environments for real-time and near-real-time systems and applications are provided. A location system performs multi-subject tracking using streaming data from multiple sensors. Subject tracking may be based on individual anchors and behavior states that are initialized for individual subjects using representations (e.g., behavior embeddings) derived from the streaming data. Clustering may be used to generate behavior clusters that individually represents a trackable subject. Behavior states for live anchors may identified based on continuity of trajectory and tracked by iteratively propagating their behavior states forward over time. Clusters lacking continuity of trajectory may be used to initialize new anchors, or matched to dormant anchors that may be reclassified as live anchors and propagated. Propagated behavior states may be updated using behavior data represented by the behavior embeddings.