Un-calibrated Photometric Stereo Algorithm Based on Local Grayscale Information

Abstract

Aiming at the difficulty of light source parameter calibration and the weak practical application of traditional photometric stereo vision in 3D reconstruction of target object, an uncalibrated photometric stereo vision algorithm model based on local gray information of image is studied. Firstly, an adaptive clustering segmented is designed to generate mask image from the original photometric image and extract the target region; Then, singular value decomposition (SVD) is used to decompose the original image into the product of the near initial normal vector matrix and the initial illumination matrix, initialize the general shallow relief (GBR) transformation matrix, and construct the initialization model of the target object surface vector; The file set is constructed with the local gray maxima of Lambert reflection, and the GBR parameters are optimized by particle swarm optimization; Determine the accurate normal vector matrix and illumination matrix, calculate the depth map and reconstruct the target object. Experimental results show that the proposed algorithm has advantages in accuracy, generalization and convenience of target object reconstruction.