Human height estimation from highly distorted surveillance image.

Area of Science:

• Forensic Science
• Image Analysis
• Photogrammetry

Background:

• Video surveillance cameras (VSCs) are vital for forensic evidence extraction.
• Lens distortion in wide-field-of-view cameras can introduce significant errors in human height measurements.
• Accurate camera calibration is essential for reliable forensic measurements.

Purpose of the Study:

• To analyze human height extraction from VSC frames.
• To identify and mitigate error sources, particularly lens distortion.
• To evaluate camera calibration techniques for forensic applications.

Main Methods:

• Analysis of human height extraction from VSC video frames.
• Focus on camera calibration to correct lens distortion.
• Comparison of Amped FIVE software with single view self-calibration techniques.
• Application to fish-eye camera frames and real-world forensic cases.

Main Results:

• No significant difference between calibration approaches in controlled test fields with fish-eye cameras.
• Evidence of differences in height measurements between calibration methods in actual forensic cases.
• Validation using known heights of individuals in test and real-world scenarios.

Conclusions:

• Camera calibration is critical for accurate human height measurement from VSCs.
• While effective in test fields, calibration methods may yield different results in complex real-world forensic scenarios.
• Further research is needed to refine calibration techniques for diverse forensic imaging conditions.