A Multi-Sensor Data-Fusion Method Based on Cloud Model and Improved Evidence Theory.

Area of Science:

• Computer Science
• Engineering
• Information Science

Background:

• Information-aware systems rely on heterogeneous multi-sensory devices.
• Multi-sensor data often presents ambiguity and contradictions, posing challenges for accurate analysis.
• Existing data fusion methods struggle with conflicting evidence.

Purpose of the Study:

• To propose a novel data-fusion method using cloud model and improved evidence theory.
• To address the limitations of traditional evidence theory in handling conflicting sensor data.
• To enhance the accuracy and reliability of multi-sensor information fusion.

Main Methods:

• Utilized the cloud model for converting quantitative sensor data to qualitative basic probability assignments (BPA).
• Combined Jousselme distance, cosine similarity, and Jaccard coefficient to measure evidence similarity.
• Employed Hellinger distance for evidence credibility assessment and integrated similarity with credibility to improve evidence before fusion.
• Applied Dempster's rule for the final data fusion process.

Main Results:

• The proposed improved evidence theory demonstrated superior convergence and focus compared to traditional methods.
• Achieved up to 100% confidence in the correct proposition in numerical examples.
• In early indoor fire detection, the method improved accuracy by 0.9-6.4% and reduced false alarm rates by 0.7-10.2%.

Conclusions:

• The developed multi-sensor data-fusion method is valid and feasible for real-world applications.
• The approach effectively handles ambiguous and conflicting data, leading to more reliable fusion results.
• Provides a valuable reference for advancing multi-sensor information fusion techniques.