Multi-Camera Trajectory Generation with Pose-Based Tracklet Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-camera tracking systems struggle to accurately generate trajectory information due to issues with intra-camera and inter-camera tracking, leading to incomplete and inaccurate integration of tracklets, especially when information is lacking.
Innovation Solution
A trajectory generation program that includes preprocessing, global optimization processing, and error suppression techniques to enhance the accuracy of trajectory information generation by integrating tracklets through ID transfer error removal, optimized hierarchical clustering, pose-oriented grouping, and error suppression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-camera tracking systems use conventional intra-camera and inter-camera tracking methods, then trajectory information can be generated, but the accuracy and completeness of trajectory information is insufficient due to integration errors and information loss
Solution Approach 1:
The patent segments the trajectory generation process into distinct stages: intra-camera tracking generates tracklets within individual cameras, then inter-camera tracking integrates these tracklets across multiple cameras. By dividing the complex multi-camera tracking into manageable segments (intra-camera processing followed by inter-camera processing), the system can apply optimized clustering algorithms at each stage to preserve tracklet integrity and reduce integration errors, thereby improving both accuracy and information completeness
Solution Approach 2:
The patent introduces optimized hierarchical clustering as an intermediary mechanism between intra-camera tracking and inter-camera tracking. This clustering process serves as a mediator that properly matches tracklets from different cameras based on spatial and temporal consistency, preventing erroneous integrations while maintaining complete trajectory information across the multi-camera system
2Measurement precision
If anchor-guided clustering is used to improve tracking accuracy, then clustering reference is provided, but trajectory information accuracy remains insufficient due to threshold-based limitations
Solution Approach 1:
The patent changes the parameters used in clustering by introducing optimized hierarchical clustering that considers multiple parameters simultaneously (spatial distance, temporal overlap, appearance features) rather than relying on a single threshold parameter. This multi-parameter approach allows for more reliable trajectory accuracy assessment by evaluating tracklet integrations against multiple consistency criteria, thereby improving both tracking accuracy and overall reliability
3Productivity
If simple clustering methods are used for integrating tracklets, then processing is efficient, but trajectory information accuracy is compromised due to insufficient error correction
Solution Approach 1:
The patent implements continuous error suppression processing that operates throughout the trajectory generation pipeline. Rather than performing error correction as a separate post-processing step, the system continuously applies error suppression during tracklet integration and clustering, maintaining processing efficiency while progressively improving trajectory accuracy through multiple refinement passes
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
A trajectory generation program causes a computer (100,200) to execute a process including generating a fragmentary trajectory based on the acquired analysis result calculating a feature amount regarding an appearance of the person based on a plurality of image frames associated with the generated fragmentary trajectory generating a plurality of representative clusters by clustering a plurality of specified fragmentary trajectories using a similarity of representative values of the calculated feature amounts of the plurality of image frames generating a plurality of subclusters obtained by clustering each of the plurality of clustered trajectories in each of the generated representative clusters using representative values of a plurality of poses of the person and setting the plurality of trajectories corresponding to the subcluster as a fragmentary trajectory of the same person based on a result of the plurality of generated subclusters.