Research On 3D Target Reconstruction Technology Based On Ghost Imaging

Ghost imaging,also known as correlation imaging,is a new imaging method.Compared with traditional optical imaging,ghost imaging can achieve images with sampling less than Nyquist limit and diffraction limit,etc.These advantages and characteristics make it more and more concerned and become a research hotspot in the field of optical imaging.Ghost imaging can also be applied to 3d imaging.However,it is easy to be affected by noise in the process of 3d imaging.More importantly,the choice of initial growing position of 3d reconstruction will also significantly affect the quality of reconstructed images.These shortcomings greatly reduce the possibility of the application of 3d ghost imaging.This paper focuses on the research of noise immunity in current 3d ghost imaging technology.The main content and innovation points are as follows:1)Combine photometric stereo with computational ghost imaging,the projected light pattern is modulated by Hadamard matrix,and three 2d shading images can be achieved by three spatial separated bucket detectors at the same time.The normal vectors and gradient map are obtained from shading images,so as to get the 3d image of object;2)In this paper,sub-pixel displacement technique is applied to 3d computational ghost imaging.low-resolution patterns are generated by modulating light source with Hadamard matrix to get low-resolution images.Then four low-resolution images are registration to get a high-resolution 2d shading image for 3d image reconstruction.Experiment results show that this method can improve the SNR of shading image by 10%.In the 3d image reconstructed from 2d images,the folds are obviously reduced,which has better smoothness and noise immunity;3)The optimal initial growing position selection algorithm is proposed.Calculated the gradient partial derivatives of the target surface,Set the threshold value to select out the candidate initial growing position,the standard deviation of the normal vector z component of the candidate initial growing positions are calculated,and the position of the minimum value is selected as the optimal initial growing position to reconstruct the 3d shape of the object.Experiment results show that this method can more accurately describe the 3d features of the object.Multiple representative points on the object are taken as error probes.When only device noise exists in the system,the error between the reconstruction result and the actual height is about 4mm.Even if the noise increases to 8%,the error is only 5mm.Simulation and experiment results show that the sub-pixel displacement technique applied to 3d computational ghost imaging method can obtain smooth 3d images in noise environment,and the optimal initial growing position method can maintain the 3d features of the object.Compared with the traditional method,the proposed scheme can make the reconstructed 3d image more smooth and more accurate,and improve the noise immunity of the 3d imaging system.This research will promote the application and development of 3d computational ghost imaging in noise environment.