Object Tracking Using Minimax Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking methods using particle filters are not robust enough to handle changing environments with illumination variations and occlusions, leading to inaccurate tracking performance.
Innovation Solution
An object tracking method employing multiple histograms and a Minimax estimation technique to select the feature with the maximum cost, using Bhattacharyya Distance for weight calculation, which enhances tracking robustness and accuracy in dynamic environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single histogram or traditional particle filter is used for object tracking, then the algorithm complexity is low, but the tracking robustness deteriorates under illumination changes and occlusions
Solution Approach 1:
The patent combines multiple histograms (color histogram, texture histogram, edge histogram) into a unified tracking framework. Each histogram captures different feature aspects, and their integration creates a more robust representation that maintains tracking reliability under varying conditions while managing complexity through systematic feature fusion
Solution Approach 2:
The patent creates a composite feature representation by integrating multiple histogram types (color, texture, edge) analogous to composite materials. This composite approach leverages the strengths of each individual histogram type to achieve superior tracking robustness compared to using any single histogram alone
2Measurement precision
If multiple histograms are used to improve tracking robustness, then the tracking accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent applies Minimax estimation to select the most reliable histogram feature dynamically, rather than always processing all features equally. This selective approach achieves high tracking accuracy by focusing computational resources on the most informative features while avoiding unnecessary computations from less reliable features in certain conditions
Solution Approach 2:
The patent dynamically adjusts feature selection based on Minimax estimation of reliability parameters. By changing which histogram features are actively used based on current tracking conditions and reliability assessments, the system maintains high accuracy while adapting computational complexity to actual needs
3Reliability
If traditional single-feature tracking is used, then the processing speed is high, but the tracking reliability under environmental changes deteriorates
Solution Approach 1:
The patent pre-computes and stores multiple histogram representations during an initialization phase. This preliminary action allows the system to quickly access and compare pre-prepared feature histograms during tracking, maintaining high processing speed while benefiting from multiple feature representations for improved reliability
Solution Approach 2:
The patent dynamically selects which histogram features to use based on real-time reliability assessment through Minimax estimation. This dynamic adaptation allows the system to maintain high processing speed by activating only the necessary features while ensuring tracking reliability through selective use of the most appropriate feature representations
Data Source
AI summary
Provided are an object tracking method and an object tracking system. The object tracking method uses a particle filter configured by combining a Minimax estimation technique with a Monte Carlo sampling method. In the object tracking method, for each of the particles, a feature having a maximum cost among measurement values of at least two or more features is selected, a weight for the particle is calculated by using the measurement value of the selected feature, and a position of the target, object in the frame is estimated by using the weight of each of the candidate particles.