Stereo matching of light-spot image points in light pen in binocular stereo visual

Abstract

One of the difficulties in light pen 3D vision measurement is the stereo matching of the light-spot image points. Therefore, the SoftPoist algorithm is adopted to determine the corresponding relations of the light-spot image point in the left and right images to the spatial light spots through double-layer circulation iterative optimization, and then solve the stereo matching problem. The SoftPoist algorithm improves the POSIT algorithm so that the improved algorithm can be applied to huge amounts of light-spot feature points where the corresponding relations of the image points to its object unknown. The iterative optimization of the matching and pose estimation parameters can be computed if only there are four noncoplanar feature points, so the applied scope of the improved algorithm is much wider. The experiment results show that the algorithm can keep the measurement accuracy and shorten the time of the program operation to enhance matching efficiency and the improved algorithm has the practical value.