Multi-Camera Target Tracking Using Feature Banks for Real-Time Global IDs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cross-camera multi-target tracking methods are not suitable for real-time or online tracking, lacking the necessary accuracy and speed to identify and associate targets across multiple cameras in a timely manner.
Innovation Solution
A method involving local trajectory determination, feature bank updating, and clustering operations to determine global target identifications using anchors, performed at a predetermined clustering period, to associate local trajectories across multiple cameras.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If anchor-based cross-camera multi-target tracking method is used, then tracking accuracy is improved, but real-time processing capability deteriorates
Solution Approach 1:
The patent segments the tracking process into two distinct phases: offline phase for comprehensive anchor generation using full video data, and online phase for real-time tracking using pre-generated anchors. This segmentation allows complex computations to be performed offline while maintaining real-time performance during actual tracking operations.
Solution Approach 2:
The patent performs preliminary action by generating anchors offline before real-time tracking. The offline anchor generation process pre-computes trajectory associations across cameras, creating a reference framework that enables fast real-time tracking without repeating complex computations during live operations.
2Measurement precision
If comprehensive feature analysis is performed across all frames, then tracking accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs comprehensive feature analysis in advance during the offline phase, extracting and analyzing trajectory features from all video frames before real-time tracking. This preliminary analysis creates pre-computed anchor data that can be quickly referenced during online tracking without repeating time-consuming feature extraction.
Solution Approach 2:
The patent extracts essential trajectory features and anchor information from the complete video data during offline processing, separating the computationally intensive feature analysis from the real-time tracking process. Only the extracted anchor data, not the entire feature set, is used during online tracking.
Data Source
AI summary
According to an embodiment, a method for target tracking comprises: determining a plurality of local target trajectories having local target identifications based on a plurality of current frames at a current timing provided by a plurality of cameras; updating a first feature bank including a sub-tracklet feature of a recent sub-tracklet of each local target trajectory and a second feature bank including a sub-tracklet feature of an early sub-tracklet of each local target trajectory based on the plurality of local target trajectories; and performing, in a case where the current timing satisfies a time requirement for a predetermined clustering period, operations of: determining a plurality of current anchors having corresponding current cluster appearance features by clustering features in the union of the updated first feature bank and the updated second feature bank; and determining a global target identification of a detected target based on the plurality of current anchors.