Research On Accuracy Of Photon-counting Laser Three-dimensional Imaging For Non-line-of-sight Vision

Along with the development of component technology, especially the emergence of detector with high sensitivity and high temporal resolution, imaging laser radar is not limited to target detection in convention and it has been applied to the target detection and identification in some special conditions or environment.Non-line-of-sight imaging gets more and more attention from researchers, which can be developed to battlefield surveillance, search and rescue, medical diagnosis, etc.By means of the high-performance imaging laser radar, echo signal can be got from hidden target, used for three-dimensional reconstruction of hidden target, which is the key of non-line-of-sight imaging. In this dissertation, 3D reconstruction method for hidden object in Lambertian scene and imaging accuracy are preliminary investigated in theory and simulation.Based on the ideal scene, the multiple diffuse transmission equation of the laser signal used for non-line-of-sight imaging is deduced. Due to non-line-of-sight imaging belonging to weak light imaging, the photon-counting laser three-dimensional non-line-of-sight imaging system is designed and the transmission theory of the laser signal used for non-line-of-sight imaging is established. In view of photon-counting is a probability event, a Monte Carlo Simulation model can be built to study transmission signal. By means of simulation, the return photons reflected by hidden target form a waveform in time domain.An algorithm for non-line-of-sight imaging can be used to reconstructed hidden object validly, including the process that time reversal to space by ellipsoidal path, at the same time, the algorithm bring noise to acquired images. A three-dimensional Laplace filter algorithm is put forward to processing the original image. The results show that when the dimension of Target in decimeter-scale, the accuracy of reconstructed image can reach centimeter-scale.We investigate the accuracy of photon-counting laser three-dimensional imaging for non-line-of-sight vision. For a single hidden point target, the accuracy of reconstructed location is high. Horizontal accuracy can reach sub-millimeter and the longitudinal accuracy can reach millimeter. The imaging resolution of two hidden point is not only restricted by Rayleigh Criterion, but affected by the noise broughtfrom imaging algorithm, about which we put forward a research method. Research has shown that the resolution of imaging system is 3cm. The parameters of imaging system such as laser pulse duration, sampling point density and sampling range can affect reconstructed image quality. The scheme to improve imaging quality and shorten the operation time has been propounded, which have a positive guidance for the future experiments. In order to show the actual situation of photon counting, we take random error and noise into account in the process of imaging. The results show that the impact of random error greater than dark count and background noise.