Research On Monocular Polarization 3D Imaging Technology

In the Industry 3.0 era,three-dimensional data is in urgent needs since the traditional two –dimensional data can't provide sufficient information due to the lack of depth information.Therefore,three-dimensional data measurement should be fully developed to benefit applications including medical diagnosis,face recognition,and driverless driving,etc.However,limited by imaging distance,environment and cost,it's difficult for traditional 3D imaging technologies to meet the requirements of low-cost and high-precision 3D imaging.For this sake,in this study we focus on exploiting polarization 3D imaging technology which employs both intensity and polarization information of light to recover 3D target surface.Besides,polarization 3D imaging system is simple in structure and high in cost performance,making it a promising technique for 3D imaging.Based on the law of reflection and refraction,this thesis analyzes the relationship between surface normal gradient field and polarization characteristics of diffuse reflection light from target surface,then studies the 3D reconstruction method based on normal gradient field.The target surface can be recovered according to polarization information.However,the direction of surface normal gradient field obtained from polarization information may be uncertainty,thus will lead to the surface distorted.In addition,3D imaging based on diffuse polarization acquires the surface normal direction rather than authentic modulus,so it cannot recover the true 3D information of target surface.Aiming at solving the ambiguity problem of normal gradient field direction in 3D imaging based on diffuse polarization,this thesis proposes a normal gradient field correction method based on shape from shading.The method utilizes dichromatic reflection model to separate the specular component for diffuse reflected component.On this basis,the diffuse reflected component is used acquire the prior knowledge of surface normal gradient field by shape from shading method.And normal gradient field is then corrected according to the prior information.Experimental results show that the method can accurately determine normal gradient field direction,which can used to restore the accurate target surface and refrain from target surface distorting.After normal gradient field correction,the reconstruction error due to normal azimuth error angle decreased about 35%.Besides,the reconstruction results for different types of reflected light show that the reflected light component from target surface would not lead to ambiguous flips,and the method has strong applicability.In order to obtain the true 3D information of target in 3D imaging based on diffuse polarization,this thesis proposes a real 3D information recovery method based on monocular vision.The method derives the target spatial distance acquisition method by combining camera calibration and camera measurement technology.The true 3D information can be recovered using the spatial distance and target surface.The experimental results show the measurement error in depth distance is less than 10% and that in plane distance is less than 2%.The method can recover the true 3D information and the relative error of real 3D information is less than 10%.Tests to targets in three different shapes show that the method can maintain high precision and is robust.Above all,this thesis mainly focuses on the normal gradient field correction and real 3D information acquisition method based on shape from shading.The developed method can accutately recover target shape and regain real 3D information.It provides theoretical basis and technical support for the high performance and low-cost 3D imaging.