Research On Passive Non-Line-of-Sight Image Reconstruction Based On Microfacet Model

The hidden scene information outside the field of view has strong application value in many aspects such as national defense,unmanned driving,disaster rescue,etc.,and the nonline-of-sight is the technology that born to obtain information outside the field of view.Nonline-of-sight can be divided into active non-line-of-sight and passive non-line-of-sight according to whether active controllable light sources are used.Passive non-line-of-sight has attracted the attention of many scholars in recent years due to its relatively cheap experimental equipment and easier implementation of experimental scenes.The main research content of this paper is passive non-line-of-sight image reconstruction.Firstly,this article analyzes and summarizes the principle and the model solving algorithms of passive non-line-of-sight image reconstruction.The construction and solution of the forward transmission model of light in the passive non-line-of-sight experimental scene is discussed,the effect of the occluder between the relay surface and the screen is analyzed,and expounds that the forward transmission matrix of light is an important factor affecting the forward transmission model of light.Several solution algorithms of commonly used forward transmission model of light are summarized.Secondly,this article analyzes the actual reflection of the relay surface and in order to solve the problem that the forward transmission model of light in the previous passive nonline-of-sight image reconstruction method only considers the diffuse reflection but ignores the influence of the specular reflection.To solve this problem,this article proposes a passive nonline-of-sight imaging method based on the microfacet model,this method uses the microfacet model to model the specular reflection of the relay surface and brings it into the forward transmission matrix of light,so as to improve the forward transmission model of light to make it more in line with the actual reflection of the relay surface Condition.In addition,this article also carries out the selection experiment of model parameters,and selects the model parameters with the best effect and the most suitable for the actual situation of the relay surface.At the same time,in order to verify the superiority of the method in this paper compared with the original method,the method model and the original method model were respectively brought into several different model solving algorithms to compare and analyze the reconstructed images.The final experimental results show that the passive non-line-ofsight image reconstruction method based on the microfacet model proposed in this paper can effectively improve the image perception and image indicators of the reconstructed images compared with the original method.Finally,in order to remove the occluder between the relay surface and screen and improve the robustness to external light sources of passive non-line-of-sight imaging methods,this paper proposes the Nlos U-Net neural network for passive non-line-of-sight image reconstruction.And in order to obtain passive non-line-of-sight image data under the interference of external light sources,this paper builds a passive non-line-of-sight experimental platform to collect image data.Image reconstruction results show that the Nlos U-Net neural network has better robustness to external light source interference than the original U-Net neural network,and the reconstructed images also have a significant improvement in image perception and image indicators.