Analysis of structured highlight stereo imaging for shape measurement of specular objects

Structured highlight stereo (SHS) is a technique for the measurement of specular objects, and is based on viewing the reflection of a structured highlight pattern from multiple directions. This paper reports the results of a simulation and experimental analysis of the approach in order to further understand the accuracy and limitations of SHS. In particular, this paper describes two important findings related to this technique. Firstly, it is shown that the search region containing a consistent set of candidate highlights may be minimized using epipolar constraints. Secondly, an analysis is presented to quantify the accuracy of the SHS approach for different surface curvatures. The results indicate that, although it is possible to accurately measure the surface height of planar specular objects, the accuracy is reduced for objects with curved surface profiles. The measurement error occurs because in general it is not possible to measure the same surface location with both cameras, which then leads to an inconsistency between the surface normal estimates for each camera. The measurement error increases in relation to both the distance between measurement locations on the surface and the surface curvature.