Using 2D video analysis and model based image matching to measure joint angles for forensic biomechanical analysis

Forensic injury biomechanics involves the use of all relevant data in order to conduct analysis and draw conclusions about an incident under investigation. Video recordings of the scene can be especially helpful due to their objectivity and the wealth of information which can be garnered from them. We compared the accuracy of different analysis techniques and multiple camera views for measuring human joint angles. The analysis techniques were reverse projection and model-based image matching (MBIM). For this purpose, 8 static postures performed by a human subject were recorded with 4 cameras. One camera was placed so that its focal plane was aligned with the plane of the subject's movement and 3 cameras were placed at various out-of-plane locations. The knee and elbow joint angles were first measured using a goniometer and then compared to angle measurements made through reverse projection and MBIM employing both single and combined camera views. Overall, the results indicated that multi-camera solutions and single, in-plane camera views produced joint angle reconstructions with the highest accuracy when comparted to the single out of plane camera views. Moreover, there was no significant difference between in the MBIM and reverse projection techniques in regard to joint angle accuracy.