Euclidean consistency-driven dual-layer information fusion framework for UAV-based traffic accident scene...

Area of Science:

• Photogrammetry and Remote Sensing
• Computer Vision
• Geospatial Analysis

Background:

• Accurate 3D reconstruction of traffic accident scenes is crucial for forensic analysis.
• Unmanned Aerial Vehicle (UAV) photogrammetry offers a viable solution but requires precise flight parameter optimization.
• Existing methods often lack interpretability and robustness in complex scenarios.

Purpose of the Study:

• To develop an interpretable dual-layer information fusion framework for UAV flight parameter optimization.
• To enhance 3D reconstruction accuracy in traffic accident scenarios using UAV oblique photogrammetry.
• To identify optimal flight parameters (altitude, overlap) for centimeter-level reconstruction accuracy.

Main Methods:

• Development of a dual-layer fusion framework integrating a linear module and a nonlinear support vector regression layer.
• Utilizing the Euclidean Consistency Index (ECI) to drive information fusion and assess inter-layer agreement.
• Conducting 27 experimental configurations of UAV oblique photogrammetry to evaluate reconstruction accuracy using elevation error, horizontal error, and distortion metrics.

Main Results:

• Flight altitude and image overlap were found to have coupled effects on reconstruction performance.
• An optimal parameter range was identified: 20-25 m altitude, 80-85% forward overlap, and 70-75% side overlap.
• The proposed framework achieved centimeter-level accuracy and stable geometric consistency, with high inter-layer agreement (ECI = 0.9133).

Conclusions:

• The developed interpretable fusion framework provides a robust foundation for UAV flight parameter selection in accident reconstruction.
• The identified optimal parameters significantly enhance 3D reconstruction accuracy and geometric stability.
• Preliminary validation in a real-world accident scene supports the practical applicability of the framework.