Principles and evaluation of autostereoscopic photogrammetric measurement

Stereoscopic perception is a basic requirement for photogrammetric 3D measurement and accurate geospatial data collection. Ordinary stereoscopic techniques require operators wearing glasses or using eyepieces for interpretation and measurement. However, the recent emerging autostereoscopic technology makes it possible to eliminate this requirement. This paper studies the principles and implementation of autostereoscopic photogrammetric measurement and evaluates its performance. We first describe the principles and properties of the parallax barrier-based autostereoscopic display used in this study. As an important metric property, we quantitatively present the autostereoscopic geometry, including viewing zones and the boundary of a viewer's movement for autostereoscopic measurement. A toolkit AUTO3D is developed that has common photogrammetric functions. The implementation principles are described by addressing the differences compared to the ordinary stereoscopic technology. To evaluate the performance of the autostereoscopic measurement, images at a resolution of 25 mu m and 50 mu m are measured by a group of seven (7) operators, who are asked to digitize 18 well-defined roof points and 18 ground points. These results are evaluated by comparing the same measurements obtained from a popular stereoscopic photogrammetric workstation. It is shown that the precision of autostereoscopic measurement is about 16 percent to 25 percent lower than the conventional stereo workstation.