An improved gray weighted method for sub-pixel center extraction of structured light stripe

Abstract

Center extraction of structured light stripe is an essential problem for the development of line structured light sensors (LSLS). To obtain the sub-pixel center coordinates precisely, an improved gray weighted method (IGWM) is proposed with an adaptive sampling region. Firstly, the center of the stripe is computed using gray weighted method (GWM) for each pixel column. Then these center points are fitted using moving least squares algorithm to estimate the tangential vector, the normal vector and the radius of curvature. For each center point, a rectangular region is defined with two sides parallel with the normal vector. The other two sides that parallel with the tangential vector alter their length automatically according to the radius of curvature. After that, the center coordinate at this point is recalculated based on the GWM, but in the normal vector direction and only takes into account the pixels within the rectangular region. The experimental results show that this method is not only suited for the center extraction of smooth laser stripes, but also the ones with sharp corners. The noise can also be obviously suppressed than that of the traditional GWM.