Research On Depth Illumination Estimation For Augmented Reality

Augmented Reality is a technology that superimposes virtual reality,realizi-ng the enhancement of real-world information,realizing the authenticity of virtual objects from the visual effect,and enhancing human perception of the real world.In the augmented reality system,the estimation and simulation of li-ghting have a great influence on the authenticity of the system,so how to a c-hieve the same lighting in the virtual object is a problem in the augmented r e-ality system.Aiming at the problem of illumination consistency in the augme nt-ed reality system,the main work and innovations of this article are as foll-ows:1.This article first addresses the problem that the traditional low-illuminati-on image enhancement algorithm has insufficient brightness and unclear details.Based on the feasibility of deep learning-based illumination estimation,a low illumination image enhancement algorithm based on improved generative adversarial network is proposed.This method can perform illumination estimation on a single low-illumination image to enhance the illumination of the low-illumin-ation image,to obtain a more realistic high-illuminance image,and lay the foundation for the realization of the consistency of illumination in the subseque-nt augmented reality.2.Afterwards,this paper aims at the problem of the need for auxiliarymarkers or auxiliary equipment and the need to obtain scene geometry informat-ion in the traditional augmented reality lighting estimation method.In order to verify that deep learning-based lighting estimation can be applied to augmente-d reality,a method based on illumination estimation method of the residual de-nse network only estimates the illumination of pairs of pictures,and renders virtual objects so that they can be applied to augmented reality to solve the pr-oblem of illumination consistency in augmented reality.In this paper,by applying deep learning technology to solve the problem of lighting consistency in augmented reality,an image lighting estimation met-hod based on deep learning is proposed,which provides methods for the enh-ancement of low-illuminance images and the lighting of virtual objects in augmented reality.Verification,the illumination estimation method based on deep learning can better solve the problem of illumination consistency in augmented reality.Compared with the traditional algorithm,the illumination estimation t-echnology based on deep learning does not require any auxiliary markers and auxiliary equipment,nor need to obtain The geometric information of the scene has better accuracy and development potential.Therefore,the research on t he depth illumination estimation of augmented reality proposed in this paper has certain research and practicality.