Research on improved evidence theory based on multi-sensor information fusion

Area of Science:

• Sensor Fusion
• Artificial Intelligence
• Information Theory

Background:

• Heterogeneous multi-sensor systems generate complex data.
• Existing information fusion models struggle with diverse data types and conflicting evidence.
• The Dempster-Shafer (DS) evidence theory provides a framework but has limitations with high conflict.

Purpose of the Study:

• To develop an improved Dempster-Shafer (DS) evidence theory algorithm for effective heterogeneous multi-sensor information fusion.
• To address limitations of standard DS theory in handling highly conflicting evidence.
• To enhance the accuracy and efficiency of multi-sensor data fusion.

Main Methods:

• Introduced a compatibility coefficient to quantify evidence compatibility.
• Developed a weight matrix for propositions based on compatibility.
• Redistributed basic probability assignments for improved evidence sources.
• Incorporated the concept of credibility and average support into the synthesis rule.

Main Results:

• The proposed algorithm effectively fuses heterogeneous multi-sensor information.
• Demonstrated improved handling of highly conflicting evidence compared to standard DS theory.
• Achieved more reasonable fusion outcomes.
• Showcased a faster convergence rate in the fusion process.

Conclusions:

• The improved DS evidence theory algorithm offers a robust solution for heterogeneous multi-sensor information fusion.
• The enhancements allow for more reliable data integration, particularly in challenging scenarios with conflicting data.
• The algorithm presents a significant advancement in multi-sensor data processing efficiency and accuracy.