Attached
Introduction: At times Forensic Investigators are presented with human remains that are highly
decomposed or mutilated andare beyond simple identification. At these instances the only
remaining option is the process of facial reconstruction. However since this process was not
carried out in Sri Lanka, we introduced a 3D computer graphics and facial anatomy based
forensic facial reconstruction approach to the local officials. This study related to the accuracy
assessment of 4 outputs which were obtained from the process introduced in our previous
studies.
Method: The skulls for the process were obtained from CT scan data of live subjects from two
medical institutes in Colombo. The computer modelling of the facial muscles were then
performed on the digital sculpting environment to arrive at the outputs. The 3D model of the
actual face of the subject was then obtained from surface reconstruction of the same CT scan
data which was utilized to render the skull. Once the fontal 2D images were captured of both the
final output model and actual face of the subject, 2D landmarks comparison was performed via
the interface developed for this project. This interface performed automated detection of the said
landmarks from the 2D images and variance calculations.
Results: This method resulted in close linear measurement values for most 2D facial landmarks
in consideration for the comparisons performed.The mean percentage variance was below 12.00
for most landmarks. However, comparatively high variances were present for landmarks
pertaining to the nose andmouth. j
Conclusion : Since forensic facial reconstruction is a complex and subjective process, thorough
evaluation is required to validate the process followed. The evaluations results illustrate that the
computerized digital sculpting method followed in this study produces acceptable outputs. Yet,
the results also showed that, facial features such as nose mouth can have high facial
reconstruction error.
Keywords: Forensic Facial Reconstruction, Digital Sculpting, Accuracy Assessment, 2D; Facial
Landmarks
