【作者】 王宇航；

【导师】 张子静；

【作者基本信息】 哈尔滨工业大学 ， 光学， 2023， 硕士

【摘要】 单光子探测器具有光子级的响应灵敏度和皮秒级别的时间分辨率,采用单光子探测器作为接收系统的单光子成像技术实现了低功耗、远距离探测。目前单光子成像系统最常用的探测器是盖格模式雪崩光电二极管(Geiger-mode Avalanche Photodiode,Gm-APD),它不输出强度信息,仅输出响应光子有无的0/1数字信号,无法判断探测到的光子信号属于探测信号还是噪声信号。目前单光子探测技术遇到的主要问题如下:背景光、后向散射光以及探测器的暗计数等因素对探测结果产生严重干扰,使得信号光子淹没于强噪声中,无法提取出有效信息;探测实时性要求探测过程的累计时间短,探测器能响应到的光子数稀疏。为了获得更好的成像质量,必须对噪声抑制和光子利用率提高开展进一步研究。首先对单光子成像系统的工作机理和性能影响因素进行分析,针对两种回波数据类型给出基于信息熵的单光子成像性能提升算法流程。针对强背景噪声环境,利用信号光子和噪声光子不同的到达时间概率模型,通过定义对光子计数直方图定义信息熵量化回波光子的随机性,使用滑窗操作遍历整个探测周期,寻找熵最小的窗口,结合卡尔曼滤波器进行精细深度估计;针对稀疏回波环境,通过单光子成像的泊松概率模型推导基于目标反射率和深度的似然函数,并结合目标图像的空间先验信息,引入信息熵作为正则项构建成本函数,通过稀疏泊松强度重建算法对其进行求解。然后通过仿真生成两种极端环境下单光子探测器响应结果,并研究噪声水平,累计时间等因素对所提算法性能的影响。仿真结果表明在强背景噪声环境下,本文提出的算法相较于匹配滤波法RMSE提升了22%,RSNR提升了55%;在稀疏回波环境下,本文提出的图像恢复算法相较于匹配滤波法所获得的反射率图像RMSE提升了61%,RSNR提升了8.7倍,所获得的深度图像RMSE提升了49%,RSNR提升了82%,可以实现平均2个信号光子的高效成像。更多还原

【Abstract】 The single photon detector has photon level response sensitivity and picosecond level time resolution,and the single photon imaging technology using the single photon detector as the receiving system can achieve low-power and longdistance detection.At present,the dector most commonly used in single photon imaging systems is the Geiger mode Avalanche Photodiode(Gm-APD),which does not output intensity information and only outputs a 0/1 digital signal that responds to the presence or absence of photons.It is impossible to determine whether the detected photon signal belongs to a detection signal or a noise signal.The main challenges of single photon detection technology are as follows: background light,backscattered light and the dark count of the detector seriously interfere with the detection results,causing signal photons to be submerged in strong noise and unable to extract effective information;The real-time detection requires a short cumulative time during the detection process,which cause the number of photons that the detector can respond to is sparse.In order to achieve better imaging quality,further research must be conducted on noise suppression and photon utilization improvement.This dissertation first analyzes the working mechanism and performance influencing factors of single photon imaging system,and gives the algorithm flow of single photon imaging performance improvement based on information entropy for two different echo data types.For strong background noise environments,we use the different arrival time probability models for signal photons and noise photons to quantify the randomness of echo photons by defining photon counting entropy,traversing the entire detection cycle using sliding window operation to find the window with the smallest entropy,and combining with Kalman filter to improve depth estimation accuracy;For the sparse echo environment,the likelihood function based on the target reflectivity information and depth information is derived through the Poisson probability model of single photon imaging,and combined with the spatial prior information of the target image,the information entropy is introduced as the regularization term of the likelihood function to construct the cost function,which is solved by the Sparse Poisson Intensity Reconstruction Algorithm.Then,this dissertation generated response results of single photon detectors in two extreme environments through simulation,and studied the effects of noise level,cumulative time,and other factors on the performance of the proposed algorithm.The simulation results show that in strong background noise environment,the proposed algorithm improves 22% RMSE and 55% RSNR compared with the matching filtering method.In the sparse echo environment,compared with the matched filtering method,the RMSE and RSNR of the reflectance image obtained by the proposed image restoration algorithm increased by 61% and 8.7 times,and the RMSE and RSNR of the depth image obtained increased by 49% and 82%,which can realize the efficient imaging of two signal photons on average.更多还原