Camera Network Topology Inference via Object Tracking

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for tracking objects across multiple cameras, especially non-overlapping ones, face challenges in accurately determining object links and topology due to the need for pre-informed motion patterns and external identifiers, which increases installation and maintenance costs and reduces tracking accuracy.

Innovation Solution

An apparatus and method that includes an object extractor, a haunting data generator, and a topology inferrer to track object movements by extracting moving objects, determining appearing and disappearing cameras and times, and inferring network topology using these data points, allowing for accurate measurement of camera distances and object tracking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If pre-informed motion patterns and external identifiers are used to track objects in non-overlapping multiple cameras, then object tracking accuracy is improved, but installation and maintenance costs increase

Engineering Contradiction:

Improveobject tracking accuracyVSAvoidinstallation and maintenance costs

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs self-calibration by automatically determining camera topology and relative positions through analyzing object movement patterns and transition times between cameras. This eliminates the need for manual calibration procedures, external identifiers, or pre-informed motion patterns, thereby reducing installation and maintenance costs while maintaining tracking accuracy.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces mechanical/calibration-based approaches with an automated computational approach. Instead of using physical calibration objects or manual configuration, the system uses image processing and temporal analysis of object transitions to automatically infer camera topology, substituting physical calibration mechanisms with algorithmic solutions.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If manual calibration and overlapping Fields of View are used to track objects, then object tracking is achieved, but device complexity and installation costs increase

Engineering Contradiction:

Improveobject tracking capabilityVSAvoidcamera configuration complexity

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The system automatically determines camera overlap relationships and topology by analyzing when objects appear and disappear from different camera views. This self-calibration process eliminates the need for manual configuration of camera parameters, overlapping FOV setup, and calibration procedures, making the system easier to deploy and operate.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system performs preliminary automated calibration by analyzing historical image data to establish camera topology and overlap relationships before actual object tracking begins. This preliminary analysis phase automatically configures the system parameters that would otherwise require manual setup, simplifying subsequent tracking operations.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If transition times and pre-informed motion methods are used for tracking, then object links are determined, but the system requires external information increasing complexity

Engineering Contradiction:

Improveobject link determinationVSAvoidsystem configuration requirements

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system self-determines object links and camera relationships by automatically analyzing transition times and object movement patterns between cameras. It generates its own calibration data from observed object trajectories, eliminating the need for external motion pattern information or pre-configured link data, thereby maintaining reliability while reducing complexity.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS8300102B2Apparatus and method for inferencing topology of multiple cameras network by tracking movement

Publication Date: 2012.10.30 AJOU UNIV IND ACADEMIC COOP FOUND

AI summary

Provided are an apparatus and a method for tracking movements of objects to infer a topology of a network of multiple cameras. The apparatus infers the topology of the network formed of the multiple cameras that sequentially obtain images and includes an object extractor, a haunting data generator, and a haunting database (DB), and a topology inferrer. The object extractor extracts at least one from each of the obtained images, for the multiple cameras. The haunting data generator generates appearing cameras and appearing times at which the moving objects appear, and disappearing cameras and disappearing times at which the moving objects disappear, for the multiple cameras. The haunting DB stores the appearing cameras and appearing times and the disappearing cameras and disappearing times of the moving object, for the multiple cameras. The topology inferrer infers the topology of the network using the appearing cameras and appearing times and the disappearing cameras and disappearing times of moving objects. Therefore, the apparatus accurately infers topologies and distances among the multiple cameras in the network of the multiple cameras using the cameras and appearing and disappearing times at which the moving objects appear and disappear. As a result, the apparatus accurately track the moving objects in the network.