Three-dimensional reconstruction method and device of moving target under variable illumination condition

Abstract

The invention discloses a three-dimensional reconstruction method of a moving target under a variable illumination condition. The three-dimensional reconstruction method comprises the following steps of: collecting images of the moving target in multiple visual angles under the variable illumination condition provided by a light source; establishing an initial three-dimensional model of the moving target according to the images of the moving target in multiple visual angles, wherein the images are collected by multiple cameras at each moment; obtaining multiple illumination images of the moving target at each moment through utilizing movement estimation and movement compensation; recovering by uniting the initial three-dimensional model and multiple illumination images to obtain normal information of the initial three-dimensional model; and carrying out combined optimization on geometrical information of the initial three-dimensional model and the normal information of the initial three-dimensional model to obtain an optimized three-dimensional model of the moving target. The invention also discloses a three-dimensional reconstruction device of the moving target under the variable illumination condition. In the method and device provided by the invention, advantages of multiple-visual-angle three-dimensional and luminosity three-dimensional methods are combined and the movement estimation and the movement compensation under the periodical variable illumination condition are combined so that the reconstruction of a high-precision three-dimensional model of the moving target is realized.

Description

technical field [0001] The invention relates to the technical fields of computer graphics and computer vision, in particular to a three-dimensional reconstruction method and a three-dimensional reconstruction device of a moving target under variable light conditions. Background technique [0002] Obtaining accurate 3D models of dynamic scenes from multi-view video sequences is a hot research issue in computer vision, computer graphics and other disciplines. The traditional image-based multi-view stereo reconstruction method is difficult to establish an accurate 3D model in the area lacking texture, but it can build a full-view 3D model. The traditional photometric stereo method is to use the multi-illumination image of a single viewing angle to establish a 3D model of a single viewing angle by estimating the normal information of the model surface and then estimating the depth map. Although the photometric stereo method can build a more detailed 3D model, the The model is a...

AI Technical Summary

Problems solved by technology

The traditional image-based multi-view stereo reconstruction method is difficult to establish an accurate 3D model in areas lacking texture, but it can establish a full-view 3D model

The traditional photometric stereo method is to use the multi-illumination image of a single viewing angle to establish a 3D model of a single viewing angle by estimating the normal information of the model surface and then estimating the depth map. Although the photometric stereo method can build a more detailed 3D model, the The model is a 3D model with a non-full perspective

As a result, it is impossible to directly collect multi-illumination images at the same time for moving targets

Images