Robust and Accurate Surface Measurement Using Structured Light

Abstract

This paper proposes a robust and accurate method for measuring 3-D surfaces using a binocular system. To eliminate the effect caused by the distortion of projector lens, each structured light sheet is fitted to a conicoid. A curvilinear detector, which combines the zero-crossing detection algorithm with Steger's detector, is employed to detect the subpixel locations of the light stripes, thus preventing the Steger's curvilinear detector from failing to detect them in the endpoints of the stripes. The proposed coding method combines the information of the linked line with the gray code to avoid producing outliers caused by erroneous decoding and make the coding procedure more robust. Experiments showed that each structured light sheet fitted to a conicoid can effectively improve the measurement accuracy. The subpixel detection method can detect the exact subpixel locations of the stripes. Likewise, the encoding strategy results in the production of fewer outliers, while the reconstruction result becomes more perfect.