A center-line extraction algorithm of laser stripes based on multi-Gaussian signals fitting

Abstract

The three-dimensional objects recognition has been the focus of research in computer vision, in which the 3D scanning system as the data acquisition means is the priority of research among priorities. In 3D scanning system, how to calculate the laser stripe center will directly affect the accuracy of three-dimensional reconstruction. By analyzing the distribution of the laser stripe images and drawing on the principle of wavelet analysis, this paper proposes the algorithm of the center of laser stripe based on multiple Gaussian signals fitting. More specifically, we use Gaussian signals with different expectation and variance as bases, according to certain rules of combination, to fit the cross section of the laser stripe image to the maximum extent. When the best fitting results have been found, we can achieve the optimal estimation of the laser stripe center in terms of the symmetry center of each Gaussian signals. The results of the scanning and reconstruction experiments prove that the proposed algorithm can maintain its effectiveness even though the symmetry of the pixels distribution cannot be met.