Evaluation of registration error through optical modeling of the display eye system

This thesis examines the system formed by the combination of the human eye and an optical see-through head mounted display (HMD) for Augmented Reality (AR) applications. It is critical that developers of AR systems know what expectations can be placed on the hardware of the system. What level of registration accuracy can we expect between the virtual content as displayed in an HMD and real content in the world? What errors are introduced? In the past, the error has been measured by comparing points displayed in the HMD with the location a user perceives those points. This gives a measurement of the total error in the system but does include other variables such as noise from the tracking system and human error in the pointing process. This thesis develops a model for the correspondence of points in an optical see-through HMD with points in the physical world as seen through the display, taking into account the parameters of the physical human-computer system and motion of the eye relative to the display. Typically the graphics that are generated in an augmented reality system are created using a pin hole camera model. The eye is much more complicated, thus this camera model may not be sufficient. A model of the human eye will be presented and used to analyze what happens to light that enters the eye, be it from a real world object or one presented in the HMD. Using this information we be able to model the registration of virtual content and real content in the HMD/eye system.