Object Tracking Across Non-Overlapping Cameras
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object-tracking technologies in multiple non-overlapping cameras require a training phase and are ineffective when optical characteristics change, leading to differences in color histograms that prevent accurate tracking.
Innovation Solution
The method involves dividing color rendering values into sub-color regions, converting RGB pixels to hue values, generating and extending color histograms, and comparing them to determine object identity without a training phase, while considering changes in saturation and intensity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional color histogram methods are used for object tracking in multiple non-overlapping cameras, then object tracking can be performed, but a training phase is required and the system fails when optical characteristics change
Solution Approach 1:
The patent transforms the color representation from RGB color spaces to HSV color space, specifically utilizing the Hue parameter which is invariant to illumination changes. This parameter transformation allows the system to compare color histograms across different cameras without requiring training, as the Hue values remain consistent despite changes in lighting conditions or camera optical characteristics.
Solution Approach 2:
The patent introduces an intermediary calibration object with known, distinct color properties that is placed in the field of view of all cameras. This calibration object serves as a reference to establish corresponding relationships between color histograms from different cameras, enabling the system to learn the mapping between camera color spaces without requiring a full training phase for each tracking scenario.
2Area of stationary object
If multiple cameras with different optical characteristics are used to expand coverage, then more areas can be monitored, but color histograms for identical objects differ significantly
Solution Approach 1:
The patent converts color data from RGB to HSV color space and specifically uses the Hue parameter for comparison. The Hue parameter represents the dominant wavelength of color and is independent of illumination intensity and color temperature, allowing consistent color identification across cameras with different optical characteristics and lighting conditions.
Solution Approach 2:
The calibration object acts as a common reference point visible to all cameras. By capturing images of this known-colored object from each camera and establishing color histogram relationships, the system creates a mapping that enables accurate color comparison across all camera views, effectively bridging the gap caused by different optical characteristics.
3Reliability
If conventional training-based methods are applied, then initial object tracking can be established, but retraining is required when environmental conditions change
Solution Approach 1:
The patent uses HSV color space transformation with Hue parameter extraction, which inherently provides illumination invariance. This parameter transformation eliminates the need for retraining when lighting conditions change, as the Hue values remain stable across different illumination scenarios, allowing continuous accurate tracking without time loss.
4Ease of manufacture
If simple algorithms are used for embedded systems, then implementation is easier, but handling complex color variations across cameras becomes difficult
Solution Approach 1:
The patent applies HSV color space transformation and Hue parameter extraction, which is a computationally efficient operation compared to full RGB color space analysis. This simple parameter transformation provides robust illumination invariance, enabling embedded systems to handle color variations across multiple cameras with minimal computational overhead while maintaining high adaptability.
Data Source
AI summary
Disclosed herein is an object-tracking apparatus and method in the environment of multiple non-overlapping cameras. Color rendering values are divided into a plurality of sub-color regions. RGB pixels of objects in a first image and a second image are converted into first hue values and second hue values, respectively. The first hue values are assigned to corresponding sub-color regions, and then a first color histogram is generated. The second hue values are assigned to corresponding sub-color regions, and then a second color histogram is generated. An area of the first color histogram is extended. It is determined whether the second color histogram is included in the extended area of the first color histogram. It is determined that the object in the first image is identical to the object in the second image if the second color histogram is included in the extended area of the first color histogram.