Research On The Denoising Technology Of Photon Counting LiDAR Signal

Photon counting LiDAR is the frontier and hotspot development field of LiDAR.As it put single-photon detection technology to use,it can detect laser signals with a single photon-level energy,which can make the detection capability of LiDAR approaching the physical limit.This also makes this technology widely used in many fields such as: underwater target three-dimensional imaging,non-horizontal imaging,quantum communication,and space-based target ranging.However,because the single-photon detector only responds to the presence or absence of an optical signal,and does not reflect the energy intensity of the incident optical signal,and there is a dead time after triggering,a large number of detection cycles are required to accumulate the actual optical signal waveform.This leads to the accumulation of a large number of noise signals such as system hardware noise,environmental background noise,and Poisson noise caused by the detection mechanism of photon counting technology during the operation of photon counting LiDAR,which makes it difficult to accurately and efficiently determine the target position.To this end,this dissertation will focus on the photon counting LiDAR denoising technology,decompose the noise problem from multiple levels,and study different solutions.The main research contents and innovations of this dissertation are as follows:(1)To solve the jitter noise problem of the photon counting LiDAR system hardware itself,a photon counting signal denoising method based on down-frequency volumetric Kalman filtering is proposed.Firstly,based on the noise analysis of each hardware component of the photon counting LiDAR system,it is clarified that the biggest source of system hardware noise is the jitter of the synchronization signal.To this end,this method takes the jitter distribution characteristics of the synchronization signal of the photon counting LiDAR system as a priori knowledge,observes the synchronization signal in real time,and establishes the discrete nonlinear system observation equation based on volumetric Kalman filtering;and then defines an adjustable frequency reduction ratio.,change the proportional relationship between the synchronization signal and the laser pulse trigger signal,and construct a system state equation based on volumetric Kalman filtering to supplement the missing synchronization signal.Finally,the spherical radial approximation criterion in the volumetric Kalman filter is improved,so that it can adaptively select the fitting order according to the degree of jitter.In this way,the optimal estimation of the system state is realized,and the interference of the noise signal to the system is removed.The experimental results show that the ranging error caused by the time jitter of the synchronization signal generator can be significantly reduced by using this method.(2)To solve the environmental noise interference problem faced by the photon counting LiDAR system,a photon counting signal denoising method based on groupoptimal asymmetric generalized Gaussian is proposed.This method first evaluates the existing photon counting LiDAR echo models,and analyzes the relationship between the existing model complexity and matching accuracy under noisy conditions.A denoising method based on asymmetric generalized Gaussian model is proposed,which achieves a comprehensive optimal design in three evaluation indicators: model complexity,model fitting accuracy,and model anti-noise ability.In addition,for the multi-parameter global optimization problem existing in the model fitting process,this dissertation adopts the theory of spotted hyena swarm optimization to deal with it.The combination of the two is the whole content of the denoising of the group-optimal asymmetric generalized Gaussian method.The experimental results show that compared with the existing methods such as the Gaussian distributed echo model,the improved Gaussian distributed echo model and the piecewise function echo model,the proposed method is more effective in the fitting accuracy of the ideal echo signal and the anti-interference of the model under the condition of strong noise.It is optimal in terms of performance and model stability under different noise intensities.(3)To solve the problem of the superposition of multiple noises faced by the photon counting LiDAR system,a photon counting signal denoising method based on elastic variational mode extraction is proposed.Based on the existing variational modal decomposition algorithm,this method uses elastic net regression to reconstruct the solution model based on Tikhonov regularization in the frequency domain for the existing mathematical model for solving ill-posed inverse problems.Then a Wiener filter is constructed around the target extraction frequency,and the penalty function of the target signal is synthesized to form a joint constraint equation system in the timefrequency domain,and its variational decomposition is calculated.Finally,the alternating direction method of multipliers method is used to iteratively solve the formed variational problem.The goal of extracting signals from clusters of noisy signals at a specified center frequency is achieved.The experimental results show that,compared with the existing methods such as Haar wavelet transform,empirical mode decomposition,variational mode decomposition,and Hilbert envelope,the method has better signal denoising ability under strong noise conditions and better performance under different noise intensity conditions.The performance indicators such as denoising stability and operation speed are comprehensively optimal.(4)The photon counting LiDAR using pseudo-random coding will generate bit errors when it is subjected to environmental noise.The traditional method can remove the signal noise,but it cannot solve the ranging error caused by the bit error.Aiming at this problem,a photon counting signal denoising method based on N-recoded pulse is proposed.This method is based on the idea of multi-level coding.On the basis of the existing pseudo-random sequence,an error-correcting code mechanism is introduced,and the xoroshiro128 algorithm with better performance is used as the underlying pseudo-random sequence generator.The experimental results show that,compared with the existing longest linear feedback shift register sequence and Mersenne rotation sequence,this method not only has better theoretical autocorrelation performance,but also has a good error correction effect on the bit error caused by noise.