Pre-set estimation-based in-silico silhouette-based methodology for improving the robustness to viewing direction...

Area of Science:

• Forensic Science
• Computer Vision
• Biomechanical Analysis

Background:

• Forensic gait analysis aids legal evidence presentation by examining walking patterns.
• Increasing CCTV footage necessitates accurate analysis of pedestrian movement.
• Current silhouette-based methods struggle with accuracy due to varying viewing directions.

Purpose of the Study:

• To develop a novel in-silico silhouette-based analysis method for forensic gait analysis.
• To overcome the limitations of existing methods concerning viewing direction discrepancies.
• To enhance the accuracy and applicability of gait analysis in real-world forensic cases.

Main Methods:

• Proposed a new in-silico silhouette-based analysis technique.
• Expanded the number of viewing direction settings to 900, significantly more than previous methods.
• Developed software tools for computer-based execution of the analysis procedures.

Main Results:

• The proposed in-silico method demonstrated accuracy comparable to the established calibration-based method.
• Experimental results confirmed the method's effectiveness under existing viewing direction differences.
• Practical comparisons in actual consultations validated the technique's real-world applicability.

Conclusions:

• The novel in-silico method offers improved accuracy for forensic gait analysis.
• This approach effectively addresses challenges posed by varying camera viewpoints in CCTV footage.
• The method is anticipated to enhance analysis accuracy in real cases and potentially replace previous techniques.