Research On Methods And Application In Cave For Non-line-of-sight Imaging Based On Single Photon

Non-line-of-sight imaging is a new research in the field of computational imaging.It has important research and application value in many fields,e.g.,deep space exploration,automatic driving,medical diagnosis,search and disaster relief,counterterrorism operations and historical archaeology.Compared with the traditional imaging,non-line-of-sight imaging could reconstruct the targets beyond the field of vision,not only providing the information of hidden target,but also keeping the imaging equipment and operators far away from hidden scene.Non-line-of-sight imaging could improve the disadvantage of traditional imaging technique and expand the detection range effectively.These factors above make non-line-of-sight imaging have irreplaceable value in computational imaging and make it gradually become a hot research topic in the world,however in China this technology is still at the initial stage.Non-line-of-sight imaging treats ground or window as the diffuse wall,uses a laser to emit towards wall and a camera to receive the photons after three bounces,then extracts the information of hidden targets and use algorithm to reconstruct the hidden targets.Since the collected photons has been reflected at least three times,the amplitude and phase information has a large attenuation,therefore,the hardware for non-line-ofsight imaging needs to be extremely powerful and sensitive.When the environment is complex and the image range is large,it is difficult to use the weak signal to reconstruct the hidden target by solving the inverse problem,and the reconstruction precision and resolution cannot be effectively guaranteed.So currently all the research of non-line-ofsight imaging is at the laboratory level,with ideal diffuse wall,short distance and small targets,which is far from the actual application.Due to the defects of hardware and reconstruction algorithm,non-line-of-sight imaging cannot be applied to 3D reconstruction of distant targets in complex environments.Therefore,in order to promote the development of non-line-of-sight imaging and the application of non-line-of-sight imaging in real life,expanding the imaging range,increasing the complexity of the imaging environment,and solving the reconstruction problem of non-line-of-sight imaging in real life is imminent.Based on current defects of non-line-of-sight imaging,this dissertation is going to conduct different imaging models and establish 3D models which is consistent with real scene to analyze and do research on non-line-of-sight imaging.The main research contents and innovation work are as follows:1.To analyze the detector of hardware used for non-line-of-sight imaging,and propose a non-line-of-sight transient imaging method based on single photon detector array.This dissertation studys the single photon detector array,establishs the corresponding non-line-of-sight transient imaging model and describes the detail of this framework,meanwhile,this dissertation creates a reconstruction algorithm to solve the imaging model and use single photon detector array with different sizes and distances to make multiple experiments for single and multiple targets with different complexities to demonstrate the effectiveness and applicability of the imaging model and reconstruction algorithm.2.This dissertation proposes a model of non-line-of-sight imaging based on phasor field and virtual wave by cooperating with the University of Wisconsin-Madison.This dissertation analyzes the transmission and diffraction of light wave based on HuygensFresnel principle and Kirchoff's diffraction formula,applys the concept of light field to non-line-of-sight imaging and create the phasor field which is similar to light field to describe the transmission and character of wave in the scene of non-line-of-sight imaging.Combining with the transmission framework of phasor field,this dissertation proposes to construct the virtual wave to simulate the light transmission in the scene of non-line-of-sight imaging,treat the whole process of non-line-of-sight imaging as a linear system and treat the diffuse wall as the sampling wall to sample the virtual wave,then the hidden target could be reconstructed from the exacted amplitude and phase information by corresponding algorithm.Based on proposed algorithm,this dissertation uses single photon detector array with different sizes to reconstruct single and multiple targets with different complexities to demonstrate the ability for improving image quality.3.In order to make an effective combination of non-line-of-sight imaging and real scene,and provide theoretical supports and technological means for exploring other planets,this dissertation scans the cave in New Mexico by the Li DAR to obtain the 3D point cloud data of the cave and use implicit surface reconstruction algorithm of radial basis function to create the obj model of cave,this obj model has the similar shape and size compared with the real cave.Then using the virtual wave reconstruction algorithm of phasor field to reconstruct 2D and 3D images of the hidden scenes inside cave from different views to demonstrate the algorithm.Not only combining non-lineof-sight imaging with real scene,but also providing theories and ways for the application of non-line-of-sight imaging in universe.4.Based on the research of non-line-of-sight imaging,this dissertation proposes another application of non-line-of-sight imaging: to locate and track the hidden object.In this dissertation,a locating and tracking algorithm for the hidden object with single photon detector array is created.This dissertation uses different places as the experimental scene and the different moving hidden target with different trajectories is placed inside scene,then the proposed algorithm is used to mark each position of moving hidden target during the moving process and restore the motion trail of hidden target according to the time sequence,so as to verify the feasibility of the algorithm and expand the application of non-line-of-sight imaging.