A closed form self calibration of one-dimensional light stripe feature width function for indoor navigation

Abstract

Light stripe projection (LSP) is one of useful methods to recognize 3D information in various vision applications. In general, light stripe feature (LSF) is detected by LOG (Laplacian Of Gaussian) operator. Because LOG variance parameter is corresponding to LSF width, improper parameter value causes severe performance degradation. Our previous work proposed a method predicting LSF width by modeling LSF irradiance as 2D Gaussian function [1]. Although the method could enhance the performance of LSF detection, calibration method included time consuming genetic algorithm-based optimization procedure. This paper proposes a closed form calibration method assuming LSF irradiance can be approximated as 1D Gaussian function if field of view (FOV) of light plane projector (LPP) is much wider than that of camera or the LPP generates uniform LSF. Experimental results show that derived function can correctly predict LSF width, and related parameters can be calibrated in a closed form.