Laser radar waveform decomposition method assisted by fine denoising

Abstract

The invention discloses a laser radar waveform decomposition method assisted by fine denoising. The method comprises the following steps of acquiring waveform data of a transmitted pulse and an echo waveform of a full-waveform laser radar system, performing waveform decomposition on the waveform data of the transmitted pulse by using an EM algorithm, and estimating background noise and random noise based on waveform effective components obtained by waveform decomposition and the waveform data of the transmitted pulse, based on the background noise and the random noise, denoising the waveform data of the echo waveform in combination with the intensity relevance of adjacent sampling values of the waveform data of the echo waveform, and performing waveform decomposition of multiple Gaussian group fractions on the denoised waveform data of the echo waveform by using an EM algorithm, and determining optimal waveform decomposition of the echo waveform according to a Bayesian information criterion. According to the method, optimal decomposition of echo waveform data can be realized on the premise of completing auxiliary noise estimation of transmitted pulse waveform data.

Description

technical field [0001] The present disclosure relates to the technical field of radar detection, and in particular to a laser radar waveform decomposition method aided by fine denoising. Background technique [0002] In practical applications, the waveform data collected by full-waveform LiDAR is inevitably adversely affected by noise such as background noise, thermal noise, dark current noise, and digital sampling noise. Noise estimation and filter denoising processing of waveform data are very important for accurate extraction of subsequent waveform information. [0003] At present, most filtering and denoising algorithms combine the threshold method with filtering algorithms (such as Gaussian filtering, Savitzky-Golay filtering, etc.), and generally estimate the background noise by intercepting the mean value of the front or end noise data of a certain amount of echo waveform data. Although this type of method can estimate the background noise simply and quickly, it fail...

AI Technical Summary

Problems solved by technology

Due to the existence of superimposed waves and "false inflection points", missing components and false detections of false components are very easy to occur during component detection, causing problems such as incorrect component number estimation and excessive estimation deviation of initial value of model parameters, which in turn leads to accurate waveform decomposition. Low performance, poor robustness and other shortcomings

Images