Statistical Analysis and Visualisation of 3D Anthropological and Archaeological Materials

The expanding scale of 3D hominid fossils and archaeological artefacts have encouraged a growing use of rigorous scientific study to reveal mechanisms of human evolution and understand human activities and behaviour. Often statistical analysis and visualisation methods, e.g. geometric morphometrics, are applied to obtain what may appear to be reliable findings. However, owing to the diversity and complexity of research materials and questions, off-the-shelf methods may not work well. Thus, there are some important considerations that need to be borne in mind regarding the interdisciplinary study, and innovative approaches are required to enhance performance. There are multiple challenges that I have addressed. First, three different semilandmarking approaches on morphometric analyses and visualisation of mean and allometrically scaled surfaces are assessed. These approaches produce different semilandmark locations, which in turn lead to different results, although non-rigid approaches are broadly consistent. My second concern is about virtual restoration by means of reflection. A landmark-free method is presented to quantify gross and regional surface asymmetry, and then a landmark-based deformation method is developed to compute geometric models of missing data with the predicted boundary curve as a constraint. Finally, I explore craniofacial relationships between hard and soft tissues among modern humans, and develop computerised methods to recreate probable faces of archaic humans and Homo sapiens based on the learnt relationships. Results suggest average dense facial soft tissue thickness depths contribute to enhancing the approximation accuracy. Although nasal (and oral) hard tissues have an effect on the corresponding soft tissues, some caution is needed when approximating soft tissue structures. Overall, this thesis makes contributions within the field of virtual anthropology and archaeology. I examine the effect of semilandmarking approaches on geometric morphometrics, and develop new methods for asymmetry detection, virtual restoration and facial approximation. These proposed methods have been applied to different case studies.