Entropy-functional-based online adaptive decision fusion framework with application to wildfire detection in video

Area of Science:

• Computer Vision
• Image Analysis
• Machine Learning

Background:

• Decision fusion is crucial for improving the accuracy and robustness of automated systems.
• Existing methods often lack adaptability to dynamic environments or require extensive training data.
• Online learning and adaptive fusion are needed for real-time applications.

Purpose of the Study:

• To develop a novel entropy-functional-based online adaptive decision fusion (EADF) framework.
• To enhance image analysis and computer vision applications through adaptive decision integration.
• To evaluate the EADF framework's performance in a practical scenario.

Main Methods:

• Developed an EADF framework combining multiple subalgorithms.
• Employed an active fusion method with online weight updates using entropic projections.
• Incorporated an oracle (human operator) for feedback and validation.
• Utilized a video-based wildfire detection system for performance evaluation.

Main Results:

• The EADF framework demonstrated effective online adaptation of decision weights.
• The system achieved reliable performance in sequential image data processing.
• Simulation results validated the framework's efficacy in a real-world inspired application.

Conclusions:

• The EADF framework offers a robust and adaptive solution for decision fusion in image analysis.
• Online adaptive fusion with human feedback can significantly improve computer vision system performance.
• The developed system shows promise for applications like wildfire detection.