People

Project summary

Constructing such environments has traditionally been extremely labour-intensive and time-consuming. We are investigating automatic and semi-automatic computer-vision techniques to aid in the extraction of geometric and illumination information from image sequences.

The results of this research are being applied to scene-of-crime reconstruction for Police forensic analysis and training purposes. Building on a previous collaboration with the Computer Examinations Unit of the Greater Manchester Police, we are studying the potential use of these techniques for the reconstruction of virtual scenes-of-crime. For this application, fast and accurate model construction is essential so that the resulting virtual environment can be used during investigations. Specific uses include analysing data (such as visualising bullet trajectories), checking witness statements, helping witnesses to recall information (virtual reconstructions), visualising routes through a building (for example, determining what can be seen from different vantage points), briefing officers, and use as a tool for evaluating different scenarios. The facility to experiment with different lighting conditions also provides a powerful investigative tool which normal photography and video recording cannot offer

Research areas

• Interactive reconstruction of virtual environments from photographs (Simon Gibson).
• Recovery of surface material and illumination properties from images (Simon Gibson)
• Automatic Recovery of Geometry from Video Sequences (Dan Oram)
• Interactive Model Matching in Virtual Environments (James Sinnott). He's investigating model matching using superquadrics.

Project funding and partners