Multi-Camera Object Tracking via Segmented Association Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing moving object tracking systems face challenges in obtaining an optimum association result due to dependency on camera combinations and sequences, leading to errors and reduced accuracy, especially with an increasing number of cameras.
Innovation Solution
A moving object tracking apparatus that collectively processes information from multiple cameras to associate the same moving object, using a method that calculates feature values and similarity matrices to classify and cluster tracking information, thereby achieving accurate and consistent association across multiple cameras.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If association is performed repeatedly between a pair of cameras, then a movement locus can be obtained, but the association result varies depending on camera combination and sequence, causing errors to propagate
Solution Approach 1:
The patent segments the association process into two distinct stages: (1) pairwise association between adjacent cameras to obtain preliminary movement loci, and (2) clustering association across multiple cameras to correct and optimize results. This segmentation allows each stage to focus on specific tasks, reducing error propagation and improving overall consistency.
Solution Approach 2:
The patent implements a feedback mechanism where the clustering association stage uses the preliminary results from pairwise association as input, evaluates them across multiple cameras, and generates corrected movement loci that feed back into the system. This feedback loop enables continuous optimization of association results, ensuring consistency regardless of camera combination or processing sequence.
2Area of stationary object
If more cameras are added to increase coverage, then monitoring range expands, but association complexity and error propagation increase
Solution Approach 1:
The patent divides the multi-camera association problem into manageable segments by first performing pairwise association between adjacent cameras, then applying clustering association across the expanded camera network. This segmented approach prevents combinatorial explosion and maintains computational tractability even as the number of cameras increases.
Solution Approach 2:
The patent introduces a temporal dimension to the association process by processing cameras in a predetermined sequence and using clustering across time-expanded data. This dimensional transformation converts a complex spatial association problem into a more manageable spatiotemporal problem, reducing overall system complexity.
3Measurement precision
If pairwise association is performed between all camera combinations, then comprehensive coverage is achieved, but processing time and computational load increase significantly
Solution Approach 1:
The patent extracts and utilizes the spatial-temporal overlap relationships between adjacent cameras, performing association only where necessary. By focusing computation on overlapping regions and using clustering to fill gaps, the system achieves comprehensive coverage without requiring exhaustive pairwise comparison of all camera combinations.
Solution Approach 2:
The patent performs preliminary pairwise association between adjacent cameras before executing the clustering association across multiple cameras. This preliminary action establishes baseline movement loci that guide the subsequent clustering process, reducing the computational search space and accelerating overall processing.
Data Source
AI summary
According to an embodiment, a moving object tracking apparatus includes an acquiring unit, an associating unit, and an output control unit. The acquiring unit is configured to acquire a plurality of pieces of moving object information representing a moving object included in a photographed image. The associating unit is configured to execute an associating process for associating a plurality of pieces of the moving object information similar to each other as the moving object information of the same moving object for three or more pieces of the moving object information. The output control unit is configured to output the associated moving object information.