一种基于差分进化算法的激光雷达波形分解算法  

A LiDAR Waveform Decomposition Method Based on Differential Evolution Algorithm

作　　者：黄泽东 朱少岚[1] 赵意意[1] 陶金有 杨建峰[1] HUANG Zedong;ZHU Shaolan;ZHAO Yiyi;TAO Jinyou;YANG Jianfeng(Lunar and Deep Space Exploration Technology Laboratory,Xi′an Institute of Optics and Precision Mechanics of CAS,Xi′an 710119,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院西安光学精密机械研究所月球与深空探测技术研究室,西安710119 [2]中国科学院大学,北京100049

出　　处：《光子学报》2025年第3期80-93,共14页Acta Photonica Sinica

基　　金：国家重点研发计划项目(No.2023YFF1303801)。

摘　　要：全波形激光雷达的探测精度一定程度上依赖于波形分解精度,而目前常用的波形分解方法存在对初值敏感,分解稳定性不足等问题。针对这一现象,提出了差分进化Levenberg-Marquardt波形分解优化算法:以高斯函数为分解模型,经预处理获取参数初值后,首先使用差分进化算法进行初步优化,其次使用Levenberg-Marquardt优化算法对优化结果进行改善。通过理论分析验证其分解精度高的特点,并采集实际激光雷达数据进行处理,结果证明,所提方法能在一定程度上改善波形分解对于初值敏感、分解精度不稳定等问题,有效提高激光雷达波形分解的精度和稳定性。The rapid development of LiDAR technology has made the accuracy and stability of waveform decomposition critical factors limiting its practical application.Traditional waveform decomposition optimization algorithms face challenges such as sensitivity to initial conditions and insufficient fitting stability,which hinder the in-depth application and widespread use of full-waveform LiDAR across various fields.This paper focuses on optimizing the waveform decomposition algorithm for full-waveform LiDAR.The primary objective is to address key issues,such as sensitivity to initial values and poor fitting stability,present in traditional algorithms,thereby significantly enhancing the accuracy and stability of waveform decomposition.Regarding the research methodology,a decomposition model based on the Gaussian function is initially constructed.For the collected LiDAR echo data,a multi-step preprocessing procedure is applied.A wavelet denoising algorithm is employed to eliminate background noise,and the five-point cubic smoothing method is used to enhance the smoothness of the data.Building on this,the Gaussian inflection point method is implemented to provide a preliminary estimation of the initial parameter values.Following this,the parameter optimization process begins.This paper introduces the Differential Evolution and Levenberg-Marquardt(DELM)optimization algorithm for waveform decomposition.The Differential Evolution(DE)algorithm is first used for preliminary optimization.After determining the population size and dimensionality,the DE algorithm randomly generates individuals in the decision space,and through iterative processes such as mutation,crossover,and selection,it converges toward the optimal solution.Subsequently,the Levenberg-Marquardt(LM)algorithm is applied for secondary optimization of the DE algorithm results.The LM algorithm calculates the iteration step size based on critical components such as the objective function,Jacobian matrix,and damping factor,accurately updating the parameters,and decides wh

关 键 词：全波形激光雷达 波形分解 差分进化 LEVENBERG-MARQUARDT算法 参数优化 

分 类 号：TN958.98[电子电信—信号与信息处理]