Research On Image Processing Technology Of Aperture Coded 3D Imaging System

3D imaging technology has been widely used in various fields.The traditional active or passive 3D imaging technology has complex general structure,high precision requirements for devices,single imaging environment,which is difficult to meet the needs of small structure,high concealment performance and long working distance in military field.In this paper,the optical computing imaging technology is used to study the single-objective passive 3D imaging system.The complex field of wavefront is encoded by using the chiral phase mask template,and the 3D imaging of long-distance target is realized by digital decoding algorithm.The main contents of this paper are as followsThe theoretical research of aperture coding single order 3D reconstruction technology is completed.The coding theory and decoding theory of aperture coding are analyzed briefly,and the three-dimensional reconstruction principle of aperture coding system is analyzed.The imaging characteristics and performance analysis of the single-eye passive three-dimensional imaging optical system.The imaging performance of the designed optical system is analyzed.The imaging characteristics of the phase mask template composed of wedge prism are analyzed.The relationship between the point diffusion function and the target distance is established and analyzed.The influence of different focal length and wedge angle on the ranging accuracy of the system is analyzed.The decoding of the encoded image is realized based on the cross-correlation NCC algorithm.The principle of NCC algorithm is briefly explained,and the algorithm is improved.The edge feature of the coded image is improved by using image filtering technology,so as to improve the robustness of NCC algorithm.Then,the image correlation is improved by sacrificing the coding characteristics in Y direction.Then,the sub images of coded images are divided,their correlation is calculated and the coding information is extracted,The decoding of the encoded image is completed,and the target is reconstructed in three dimensions.The decoding of the coded image is realized based on SIFT feature point matching.The principle of deep learning is analyzed briefly.First,the relative depth information of the target is obtained by using the deep learning,but the absolute distance cannot be obtained.Then,SIFT algorithm is used to extract the feature point matching of two coded images,so as to obtain the information of the point diffusion function carried in the coded image,calculate the absolute distance information by using the detected feature point matching,assign the absolute distance information of the feature points to the relative depth image,and linear fitting other regions depend on their gray value,Thus,the distance information of all regions is obtained,and the sky area is identified and removed,so as to realize the 3D reconstruction of the target.