Research On Single Photon Reflectivity And Depth Imaging Technology Based On Compressive Sampling

The Time-Correlated Single-Photon Counting(TCSPC)lidar is an application of TCSPC technology in pulsed lidar,which has the single-photon detection sensitivity and the picosecond time resolution.Single-pixel imaging uses Compressed Sensing(CS)theory and can obtain two-dimensional images by a point detector,which is not only low cost,but also has the advantages of ultra-high sensitivity and fast imaging.Combining the time-correlated single-photon counting lidar technology with singlepixel imaging technology can further improve sensitivity and shorten imaging time,so as to achieve long distance detection of extremely weak signals.This paper designs and builds a lidar based on single-pixel imaging technology to achieve reflectivity and depth imaging.The main content and research results are as follows:(1)A single-photon reflectance and depth imaging system based on compressed sampling is built,the system is mainly composed of a picosecond pulsed laser,a Photomultiplier Tube(PMT),a Digital Micro-mirror Device(DMD),a TCSPC module and a beam expander telescope.A precise synchronization control circuit based on Field Programmable Gate Array(FPGA)is developed.Realize the continuous measurement of photon arrival time composed of coarse time and fine time.The FPGA-based control circuit counts the laser pulses as the coarse time of arrival of the photon.The TCSPC module measures the detected photon and the next adjacent laser pulse.The pulse interval is used as the fine time of arrival of the photon.(2)A a single-photon reflectance and depth imaging system model based on compressive sampling is established,the Monte Carlo simulation method is used to simulate the process of reconstructing the reflectivity through photon counting and interpolating the time sequence of photon arrival time sequence to reconstruct the depth image.The effects of sampling times,imaging time,and noise level on imaging quality are studied.Furthermore,the Monte Carlo method is used to simulate the time-gated None-Line-of-Sight(NLOS)imaging mechanism and related parameters that affect imaging performance,laying a theoretical foundation for subsequent work.(3)The high-sensitivity reflectivity and depth imaging is established.The reflectivity and depth imaging results of different imaging scenes are analyzed.Experiment results show that the sampling rate has a significant impact on the reflectance imaging quality,and the depth imaging resolution is 4.696 cm.Aiming at the problem of long time-consuming and large amount of data for high resolution image reconstruction,a sampling and reconstruction joint optimization deep learning compression reconstruction network is designed.Compared with the existing network and traditional algorithms,the reconstruction quality and reconstruction time of the designed network are significantly improved.Furthermore,a single-photon compression imaging system based on multi-reflecting surfaces is built.The surface scattering of the object and the photon transmission process are analyzed,and the influence of time gating,different reflectivity objects and intermediary surfaces on the results is verified through experiments,which lays the foundation for NLOS imaging.