3D Reconstruction of High-reflective and Textureless Targets Based on Multispectral Polarization and Machine Vision

Abstract

With the rapid development of photogrammetry and machine vision technology, a higher universality of three-dimensional reconstruction is required. For high-reflective and non-metal targets with smooth surface or extreme simple texture, a large area of data void may appear on the reconstruction surface since the traditional 3D reconstruction methods depend on texture and reflective characteristics. To solve this problem, a 3D reconstruction method based on multispectral polarization imaging is proposed in this paper, which integrates photogrammetry and machine vision and achieves accurate reconstruction by obtaining accurate multispectral polarization characteristics of targets. The proposed method does not rely on the texture information on the surface, and it can solve the problem of joint estimating refractive index and zenith angle simultaneously, which cannot be achieved only by Fresnel theory. Due to straylights and diffuse reflection light have different polarization and spectral characteristics at different wavelengths, we can remove the highlight to improve the reconstruction accuracy. The 3D reconstruction method based on multiband polarization imaging is the guiding progress of 3D reconstruction after the fusion of photogrammetry and machine vision, which has a wider application range.