Relevance-based template matching for tracking targets in FLIR imagery

Area of Science:

• Computer Vision
• Image Processing
• Embedded Systems

Background:

• Automatic target tracking requires low computational footprint, especially for real-time embedded applications.
• Forward-looking infrared (FLIR) sensors offer distinctive features suitable for small-footprint tracking techniques.
• Existing template-based tracking algorithms can be computationally intensive, limiting their use in resource-constrained environments.

Purpose of the Study:

• To enhance the computational efficiency of template-based target tracking algorithms.
• To reduce execution time and resource usage in automatic target tracking.
• To maintain tracking robustness while improving algorithmic speed.

Main Methods:

• Proposed a novel method to increase computational efficiency in template-based target tracking.
• Utilized dynamic thresholds to dynamically narrow down computations.
• Leveraged target intensity profiles from forward-looking infrared (FLIR) images.

Main Results:

• The proposed dynamic threshold approach significantly reduced the number of computations required.
• Demonstrated a reduction in both execution time and resource utilization compared to reference algorithms.
• Achieved comparable robustness to existing target tracking techniques.

Conclusions:

• The dynamic threshold method offers a computationally efficient solution for template-based target tracking.
• FLIR image intensity profiles are effective features for low-footprint tracking.
• The approach is suitable for real-time applications with limited computational resources.