Method for decomposing full waveform with oscillation signal in airborne laser radar

Abstract

The invention discloses a method for decomposing a full waveform with an oscillation signal in an airborne laser radar, which belongs to the technical field of laser radar data processing, is used for waveform decomposition, and comprises the following steps: reading and converting original data, carrying out data partitioning, selecting a wavelet threshold for each piece of data for denoising, constructing a hybrid model, and selecting initial parameters of the hybrid model. And performing particle swarm optimization to obtain optimized parameters, constructing output data of each piece of data, and splicing all the output data to complete full-waveform decomposition. According to the method, a large amount of high-frequency noise is almost completely removed, and the retention degree of signals is good; full-waveform decomposition is directly carried out, oscillation signals are fitted, the influence caused by the oscillation signals is reduced, and distance data of the airborne laser radar are extracted; the multi-point search and randomness of the particle swarm can more possibly find a global optimal solution, the particle update is natural and parallelizable, the method is suitable for distributed calculation, and the calculation speed is increased.

Description

Technical Field

[0001] The invention discloses a full waveform decomposition method with an oscillation signal in an airborne laser radar, belonging to the technical field of laser radar data processing. Background Art

[0002] In an airborne lidar system, laser light is first generated by a laser. After passing through a telescope and lenses to filter out unwanted light, the received laser signal is converted into an electrical signal by a photomultiplier tube. When the laser light strikes a target, it generates a backscattered signal that is received by the airborne lidar receiving system. This signal appears as a peak in the full waveform data. Due to impedance mismatch, the entire circuit generates low-frequency oscillations due to the input variation of the echo signal. When the acquisition system collects the electrical echo signal, this oscillation is also detected and superimposed on the echo signal, forming a longer tail. In full waveform decomposition, an approximate function is used to fit the entire echo signal hitting the target, and its extreme points are extracted as the measured distance. This oscillation not only affects the shape of the entire waveform, causing deviations when the function optimizes the echo signal fit, but also shifts the extreme points of the echo signal, causing variations in distance and affecting the final data accuracy. Summary of the Invention

[0003] The purpose of the present invention is to provide a full waveform decomposition method with an oscillating signal in an airborne laser radar, so as to solve the problem in the prior art that the oscillating signal affects the waveform decomposition of the laser radar data.

[0004] A full waveform decomposition method for an airborne laser radar with an oscillating signal includes reading and converting raw data, dividing the data into blocks, selecting a wavelet threshold for denoising each block of data, constructing a hybrid model, selecting initial parameters of the hybrid model, performing particle swarm optimization to obtain optimized parameters, constructing output data for each block of data, and splicing all output data to complete full waveform decomposition.

[0005] The airborne lidar includes an optical path system and an acquisition system. The optical path system and the acquisition system share the same square wave with a frequency of 5kHz. The square wave controls the laser emission of the optical path system and the frequency acquisition of the acquisition system. The optical path system and the acquisition system are triggered at the edge of the square wave so that the echo signal corresponds one-to-one with the collected data. The optical path system receives the echo signal, and the collected data is a quantized voltage value, which is a negative number.

[0006] Reading in and converting the raw data involves reading in binary numbers in groups of 16 bits and converting them to decimal numbers. After reading in, the decimal numbers are inverted bit by bit, negative numbers are converted to positive numbers, and all data are stored in tab-separated form. All data are divided into blocks, different blocks are numbered, and assigned to various processes of the central processing unit for processing. After each block of data is processed, each processed block of data is sorted by number. Finally, the data is spliced ​​and stored, and distance data is extracted from the original data and used as the full waveform data of the airborne lidar.

[0007] Wavelet threshold denoising includes selecting the number of adaptive decomposition layers, discrete wavelet transform, determining the wavelet threshold that meets the system, and inverse wavelet transform; Determining the number of decomposition layers includes cutting and extracting the part with data at the front end of the data to determine the length of the collected data. , decomposition level for: ; Where, is the laser pulse width, Acquisition frequency for airborne lidar systems; Discrete wavelet transform involves taking waveform data as a given signal and performing discrete wavelet transform: ; Where, is the detail coefficient obtained by discrete wavelet transform, is the scale that controls the expansion and contraction of the wavelet. is the wavelet basis function, To control the translation of the wavelet basis position, is a given signal, It's time.

[0008] Determining the wavelet threshold involves calculating the energy density of the detail coefficients of each layer in the decomposition layer. : ; ; Where, It is The detail factor of the layer, Indicates the first The detail coefficient, It is the length of the layer; set up , find the density difference that meets the following judgment conditions : ; Where, is the experience value, represents the median, It is Energy density of the layer; For those that meet the judgment conditions , from the corresponding Estimating noise soft threshold in subbands of discrete wavelet transform of layers : ; Wavelet Thresholding for: ; The detail coefficients above the wavelet threshold are subjected to inverse wavelet transform and reconstructed into a given signal after noise is filtered out.

[0009] Constructing a hybrid coupling model that meets the requirements of airborne lidar systems ,include: ; Where, is a natural constant, is the Gaussian function amplitude, is the angular frequency, is the amplitude of the damped oscillatory function, is the attenuation factor, is the Gaussian mean, is the signal standard deviation, is the time scale; Selecting the initial parameters of the hybrid model includes selecting the initial parameters of the estimated Gaussian function and the initial parameters of the damped oscillation function, performing multi-level wavelet decomposition on the given signal after filtering out the noise, and extracting sub-signals in different frequency bands. The multi-level wavelet decomposition is: ; Where, is a low-pass filter in a given signal after filtering out noise, is a high-pass filter in a given signal after filtering out noise, represents the scale factor, is the approximate coefficient of the low-pass filter after filtering, For scale The approximate signal under is the detail coefficient after filtering by the low-pass filter; is the discrete time index, To sum the index, is a set of integers; right Perform low-frequency convolution to obtain low-frequency information. Perform high-frequency convolution to obtain high-frequency information, and continue to perform low-frequency convolution on the low-frequency information, while the high-frequency information remains unchanged. After the decomposition, the detail coefficients of each sub-band from high frequency to low frequency are obtained, and the low-frequency component and the intermediate-frequency component are selected. After the inverse wavelet transform reconstruction, the component where the Gaussian peak is located and the component where the oscillation signal is located are obtained; The area around the end point of the Gaussian peak is selected as the starting point of the oscillation signal, and the initial value of the Gaussian term is used as a parameter to extract the phase at the starting moment of the oscillation. Perform initial parameter correction and use the corrected initial parameters as the initial particles of the particle swarm optimization algorithm.

[0010] The initial parameters of the Gaussian function include: for: ; Where, is the independent variable of the Gaussian function, which is equivalent to the component where the Gaussian peak is located; Use the strength of the low-frequency signal as , the center point of the low-frequency signal is , the half-width of the low-frequency signal As : .

[0011] The initial value of the damped oscillation function includes 、 、 and initial phase , damped oscillation function for: ; Get the analysis signal : ; Where, is the original signal, which is equivalent to the component where the oscillation signal is located; Frequency domain after Euler transform for: ; Where, is a symbolic function, is the original signal in the frequency domain; set up , In the time domain: ; Where, is the intermediate variable in the frequency domain, is the intermediate variable in the time domain, represents the convolution operation, is the original signal in the time domain; Find : ; Where, To represent the limit symbol, Indicates that the integral is subjected to Cauchy principal value processing.

[0012] Find Then, we bring it into the analytical signal formula and get ; for: ; Where, is the complex argument function; right The envelope signal of is logarithmically transformed and then linearly fitted: ; Get the slope of the linear fit result and intercept , 、 for: ; ; for: .

[0013] Particle swarm optimization involves randomly generating a group of particles, initializing particle positions and particle velocities, and using mean square error to calculate The fitting effect of When the mean square error of the fitting result is the smallest, the current particle swarm optimization solution is the optimal solution, and the individual optimal solution of the particle and the global optimal solution at this time are recorded; calculate When the fitting result is , the particle is updated as: ; ; Where, For the The particle velocity of the sub-particle update, is the inertia factor, and is the acceleration constant, Representation interval The random number in For the Particle No. The individual optimal solution of the sub-particle update, Indicates the The global optimal solution for sub-particle update, For the Particle position of sub-particle update; After completing the particle swarm optimization, the Gaussian function parameters and the damped oscillation function parameters are converged through multiple iterative learning to complete the full waveform decomposition.

[0014] Compared with the existing technology, the present invention has the following beneficial effects: a large amount of high-frequency noise is completely removed, and the signal is better preserved; full waveform decomposition is directly performed to fit the oscillation signal, reducing its impact and extracting the distance data of the airborne lidar; the multi-point search and randomness of the particle swarm make it more likely to find the global optimal solution, and the particle update can be naturally parallelized, which is suitable for distributed computing and speeds up the calculation speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flowchart of the present invention; Figure 2 Optimize waveform for fitting; Figure 3 is the original waveform; Figure 4 is the first waveform component after wavelet transformation; Figure 5 is the second waveform component after wavelet transformation; Figure 6 is the third waveform component after wavelet transformation; Figure 7 is the fourth waveform component after wavelet transformation; Figure 8 is the fifth waveform component after wavelet transformation; Figure 9 is the sixth waveform component after wavelet transformation; Figure 10 is the seventh waveform component after wavelet transformation; Figure 11 is the eighth waveform component after wavelet transformation; Figure 12 is the wavelet threshold denoised waveform. DETAILED DESCRIPTION

[0016] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0017] A full waveform decomposition method for an airborne laser radar with an oscillating signal includes reading and converting raw data, dividing the data into blocks, selecting a wavelet threshold for denoising each block of data, constructing a hybrid model, selecting initial parameters of the hybrid model, performing particle swarm optimization to obtain optimized parameters, constructing output data for each block of data, and splicing all output data to complete full waveform decomposition.

[0018] The airborne lidar includes an optical path system and an acquisition system. The optical path system and the acquisition system share the same square wave with a frequency of 5kHz. The square wave controls the laser emission of the optical path system and the frequency acquisition of the acquisition system. The optical path system and the acquisition system are triggered at the edge of the square wave so that the echo signal corresponds one-to-one with the collected data. The optical path system receives the echo signal, and the collected data is a quantized voltage value, which is a negative number.

[0019] Reading in and converting the raw data involves reading in binary numbers in groups of 16 bits and converting them to decimal numbers. After reading in, the decimal numbers are inverted bit by bit, negative numbers are converted to positive numbers, and all data are stored in tab-separated form. All data are divided into blocks, different blocks are numbered, and assigned to various processes of the central processing unit for processing. After each block of data is processed, each processed block of data is sorted by number. Finally, the data is spliced ​​and stored, and distance data is extracted from the original data and used as the full waveform data of the airborne lidar.

[0020] Wavelet threshold denoising includes selecting the number of adaptive decomposition layers, discrete wavelet transform, determining the wavelet threshold that meets the system, and inverse wavelet transform; Determining the number of decomposition layers includes cutting and extracting the part with data at the front end of the data to determine the length of the collected data. , decomposition level for: ; Where, is the laser pulse width, Acquisition frequency for airborne lidar systems; Discrete wavelet transform involves taking waveform data as a given signal and performing discrete wavelet transform: ; Where, is the detail coefficient obtained by discrete wavelet transform, is the scale that controls the expansion and contraction of the wavelet. is the wavelet basis function, To control the translation of the wavelet basis position, is a given signal, It's time.

[0021] Determining the wavelet threshold involves calculating the energy density of the detail coefficients of each layer in the decomposition layer. : ; ; Where, It is The detail factor of the layer, Indicates the first The detail coefficient, It is the length of the layer; set up , find the density difference that meets the following judgment conditions : ; Where, is the experience value, represents the median, It is Energy density of the layer; For those that meet the judgment conditions , from the corresponding Estimating noise soft threshold in subbands of discrete wavelet transform of layers : ; Wavelet Thresholding for: ; The detail coefficients above the wavelet threshold are subjected to inverse wavelet transform and reconstructed into a given signal after noise is filtered out.

[0022] Constructing a hybrid coupling model that meets the requirements of airborne lidar systems ,include: ; Where, is a natural constant, is the Gaussian function amplitude, is the angular frequency, is the amplitude of the damped oscillatory function, is the attenuation factor, is the Gaussian mean, is the signal standard deviation, is the time scale; Selecting the initial parameters of the hybrid model includes selecting the initial parameters of the estimated Gaussian function and the initial parameters of the damped oscillation function, performing multi-level wavelet decomposition on the given signal after filtering out the noise, and extracting sub-signals in different frequency bands. The multi-level wavelet decomposition is: ; Where, is a low-pass filter in a given signal after filtering out noise, is a high-pass filter in a given signal after filtering out noise, represents the scale factor, is the approximate coefficient of the low-pass filter after filtering, For scale The approximate signal under is the detail coefficient after filtering by the low-pass filter; is the discrete time index, To sum the index, is a set of integers; right Perform low-frequency convolution to obtain low-frequency information. Perform high-frequency convolution to obtain high-frequency information, and continue to perform low-frequency convolution on the low-frequency information, while the high-frequency information remains unchanged. After the decomposition, the detail coefficients of each sub-band from high frequency to low frequency are obtained, and the low-frequency component and the intermediate-frequency component are selected. After the inverse wavelet transform reconstruction, the component where the Gaussian peak is located and the component where the oscillation signal is located are obtained; The area around the end point of the Gaussian peak is selected as the starting point of the oscillation signal, and the initial value of the Gaussian term is used as a parameter to extract the phase at the starting moment of the oscillation. Perform initial parameter correction and use the corrected initial parameters as the initial particles of the particle swarm optimization algorithm.

[0023] The initial parameters of the Gaussian function include: for: ; Where, is the independent variable of the Gaussian function, which is equivalent to the component where the Gaussian peak is located; Use the strength of the low-frequency signal as , the center point of the low-frequency signal is , the half-width of the low-frequency signal As : .

[0024] The initial value of the damped oscillation function includes 、 、 and initial phase , damped oscillation function for: ; Get the analysis signal : ; Where, is the original signal, which is equivalent to the component where the oscillation signal is located; Frequency domain after Euler transform for: ; Where, is a symbolic function, is the original signal in the frequency domain; set up , In the time domain: ; Where, is the intermediate variable in the frequency domain, is the intermediate variable in the time domain, represents the convolution operation, is the original signal in the time domain; Find : ; Where, To represent the limit symbol, Indicates that the integral is subjected to Cauchy principal value processing.

[0025] Find Then, we bring it into the analytical signal formula and get ; for: ; Where, is the complex argument function; right The envelope signal of is logarithmically transformed and then linearly fitted: ; Get the slope of the linear fit result and intercept , 、 for: ; ; for: .

[0026] Particle swarm optimization involves randomly generating a group of particles, initializing particle positions and particle velocities, and using mean square error to calculate The fitting effect of When the mean square error of the fitting result is the smallest, the current particle swarm optimization solution is the optimal solution, and the individual optimal solution of the particle and the global optimal solution at this time are recorded; calculate When the fitting result is , the particle is updated as: ; ; Where, For the The particle velocity of the sub-particle update, is the inertia factor, and is the acceleration constant, Representation interval The random number in For the Particle No. The individual optimal solution of the sub-particle update, Indicates the The global optimal solution for sub-particle update, For the Particle position of sub-particle update; After completing the particle swarm optimization, the Gaussian function parameters and the damped oscillation function parameters are converged through multiple iterative learning to complete the full waveform decomposition.

[0027] Power systems are often affected by low-frequency oscillation signals. In the airborne lidar system used in this invention, the photomultiplier tube (PMT) is also affected by these signals when converting light to electricity, resulting in reduced data accuracy. Commonly used algorithms for processing low-frequency oscillation signals include the Prony algorithm, Gaussian filtering for denoising, mean smoothing filtering, and other filtering methods. Traditional Gaussian decomposition within full waveform decomposition methods is also used, as are optimization methods such as the least squares method and the LM algorithm.

[0028] Gaussian filtering is a linear smoothing filter based on the Gaussian function. It is widely used in two-dimensional image processing. When processing one-dimensional time series, it is used to smooth the waveform. This method takes a continuous number of odd points in the one-dimensional time series data, weights them to obtain the center value, and then slides this method over the data. The core idea is to suppress noise through weighted averaging. The weights are determined by the Gaussian function, with points closer to the center pixel receiving larger weights. This can be used to filter out high-frequency noise. In the full waveform decomposition step, Gaussian decomposition is a process of fitting the original echo signal to the superposition of multiple Gaussian function curves. Laser altimetry systems typically emit Gaussian-distributed pulse signals. The full waveform echo signal, after reflection from complex objects, is composed of multiple echo signals and can be considered a superposition of multiple Gaussian signals. Full waveform decomposition can be simply summarized as fitting a Gaussian function model to the data to ultimately obtain the actual data points.

[0029] The LM algorithm is an optimization algorithm for solving nonlinear least squares problems, particularly suitable for curve fitting and parameter estimation. It combines the advantages of gradient descent and the Gauss-Newton method, balancing the characteristics of these two methods by dynamically adjusting the damping parameter. The LM algorithm updates the parameters by iteratively adjusting the parameter vector to minimize the objective function.

[0030] The Fourier transform is a classic tool for signal analysis. It converts signals from the time domain to the frequency domain, revealing the signal's global frequency components through linear combinations of sinusoidal basis functions. However, the Fourier transform has significant limitations, and the wavelet transform was proposed to address this issue. By introducing wavelet basis functions with finite support, the wavelet transform enables multiresolution analysis of signals at different scales and locations, thereby taking into account both global trends and local details. This property makes it particularly suitable for processing full-waveform lidar data, which often contains a mixture of non-stationary noise and target reflected pulses. The core idea of ​​denoising is to decompose the signal into subbands of different scales and frequencies using the wavelet transform. By exploiting the differences in signal and noise characteristics in the wavelet domain, thresholding is used to suppress the noise coefficient while retaining the valid signal. This is particularly effective for removing non-stationary noise such as Gaussian white noise and impulse noise. The main steps are wavelet transform and threshold selection.

[0031] The discrete wavelet transform is used to decompose a signal into wavelet signals. Wavelets are functions that are small, highly localized oscillations in time. The Fourier transform decomposes the signal into infinitely long sines and cosines, thus losing all temporal position information, while the wavelet transform is a time-localized scaled and shifted version that well preserves temporal position information. Unlike the Fourier transform, which always uses complex exponential basis functions, the wavelet transform uses time-localized oscillation functions as basis functions. This basis function can be scaled and translated, with different scales of scaling corresponding to the low and high frequencies of the signal, while the translation incorporates temporal information. Wavelet analysis is more complex than Fourier analysis because it must specify the basis functions from which it is constructed.

[0032] Wavelet transforms can extract local signal features at different frequencies and times. Strong reflected pulses in the full LiDAR waveform correspond to large values, while noise appears as widely distributed, low-amplitude coefficients. The wavelet basis function used is the Symlets wavelet. Symlets have compact support and near symmetry, making them suitable for processing phase-sensitive signals. Therefore, this wavelet is used to process the full waveform data of airborne LiDAR.

[0033] Since the oscillation tail is caused by the drastic voltage change generated by the photomultiplier tube receiving the signal, it is an active oscillation. The voltage generated by the full-waveform echo signal rises rapidly as an excitation, affecting the output signal. The final output can be regarded as a dynamic coupling process between the laser pulse and the system response. The damped sine term of the hybrid model represents the impulse response of the system or environment, and the Gaussian term represents the instantaneous energy distribution of the laser emission. The convolution operation describes the dynamic coupling process between the laser pulse and the system response. The convolution model can more realistically reflect the delay effect and frequency modulation in signal propagation. The propagation of the Gaussian pulse excites the system oscillation, and the amplitude and phase of the oscillation change dynamically with time. Due to the nonlinear characteristics of the convolution model, the initial values ​​of the Gaussian term and the damped sine term need to be estimated step by step.

[0034] When estimating initial parameter values, key subbands are first selected. After wavelet transform, the full waveform echo peak is typically located in the low-frequency subband, while damped oscillations are typically located in the mid-frequency subband. After selection and verification, the low-frequency subband and a portion of the mid-frequency subband are selected for inverse wavelet transform to obtain the reconstructed signal components. The reconstructed low-frequency subband contains Gaussian peak signals, while the reconstructed portion of the mid-frequency subband contains most oscillatory signals. Initial parameter estimation of the hybrid model is performed on the reconstructed low-frequency and mid-frequency signal components. This approach optimizes resource utilization and simplifies the initial parameter estimation method. The hybrid model contains both a Gaussian function and a damped oscillation function, both of which are estimated separately.

[0035] The initial parameter amplitude of a damped sinusoidal signal decays exponentially over time, the oscillation frequency is related to the signal's periodicity, and the instantaneous phase should be linearly related to time. During estimation, the real signal is first converted into an analytical signal via the Hilbert transform, from which the instantaneous amplitude and phase are extracted. The Hilbert transform is a linear time-invariant transform that converts a real-valued signal into a signal orthogonal to its original signal. However, the current data waveform is not easy to analyze, and obtaining the initial parameters of the current data waveform is difficult. Therefore, its complex signal, also known as the analytical signal, is required. The analytical signal of a signal is a complex signal composed of the original signal and its Hilbert transform.

[0036] The embodiment of the present invention uses an airborne laser radar to conduct data experiments. Some of the binary raw data collected by the airborne laser radar is in hexadecimal format, and the data converted from the airborne laser radar raw file is in decimal format. Figure 3 This is one of the waveforms collected during the airborne lidar fixed-point test, corresponding to the full waveform data collected during a single laser emission.

[0037] Figure 4 、 Figure 5 、 Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 、 Figure 11 The series of pictures are the waveform components after wavelet transformation. Figure 4 is the approximate coefficient component of the layer 0 waveform, Figure 5 、 Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 、 Figure 11 They are the detail coefficient components of the waveforms from layer 1 to layer 7.

[0038] Figure 12 This is a wavelet threshold denoised waveform. By adaptively selecting the number of decomposition layers and denoising thresholds, the algorithm can filter out most of the noise. Figure 2 To optimize the waveform, the full waveform data is fitted and function parameters are extracted to obtain a stable and more accurate representation of the target distance. The partially calibrated distance consists of three columns: the first column is the target position extracted from the waveform, the second column is the intensity of the target echo, and the third column is the measured distance. Figure 1 The technical process of the present invention includes reading in binary data, performing data conversion, and processing three data blocks separately. Each data block undergoes the same processing, including wavelet threshold denoising and extracting the intermediate value of the wavelet coefficient. Then, it is divided into two branches, one of which inputs the wavelet transform layer 1 to obtain the overall characteristics and the initial value of the Gaussian function, and the other inputs the wavelet transform layer 2 to perform initial fitting and obtain the initial value of the damping function. Then, a hybrid model is constructed for the two initial values, particle swarm global optimization is performed, target parameters are obtained, three data are output, data are spliced ​​after sorting, and the data is output.

[0039] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A full waveform decomposition method for an oscillating signal in an airborne laser radar, characterized in that: It includes reading and converting the original data, dividing the data into blocks, selecting wavelet threshold denoising for each block of data, building a hybrid model, selecting the initial parameters of the hybrid model, performing particle swarm optimization, obtaining the optimized parameters, building the output data of each block of data, splicing all the output data, and completing the full waveform decomposition.

2. The full waveform decomposition method for an airborne laser radar with an oscillating signal according to claim 1, characterized in that: The airborne lidar includes an optical path system and an acquisition system. The optical path system and the acquisition system share the same square wave with a frequency of 5kHz. The square wave controls the laser emission of the optical path system and the frequency acquisition of the acquisition system. The optical path system and the acquisition system are triggered at the edge of the square wave so that the echo signal corresponds one-to-one with the collected data. The optical path system receives the echo signal, and the collected data is a quantized voltage value, which is a negative number.

3. The full waveform decomposition method for an airborne laser radar with an oscillating signal according to claim 2, characterized in that: Reading in and converting the raw data involves reading in binary numbers in groups of 16 bits and converting them to decimal numbers. After reading in, the decimal numbers are inverted bit by bit, negative numbers are converted to positive numbers, and all data are stored in tab-separated form. All data are divided into blocks, different blocks are numbered, and assigned to various processes of the central processing unit for processing. After each block of data is processed, each processed block of data is sorted by number. Finally, the data is spliced ​​and stored, and distance data is extracted from the original data and used as the full waveform data of the airborne lidar.

4. The method for full waveform decomposition of an oscillating signal in an airborne laser radar according to claim 3, characterized in that: Wavelet threshold denoising includes selecting the number of adaptive decomposition layers, discrete wavelet transform, determining the wavelet threshold that meets the system, and inverse wavelet transform; Determining the number of decomposition layers includes cutting and extracting the part with data at the front end of the data to determine the length of the collected data. , decomposition level for: ； Where, is the laser pulse width, Acquisition frequency for airborne lidar systems; Discrete wavelet transform involves taking waveform data as a given signal and performing discrete wavelet transform: ； Where, is the detail coefficient obtained by discrete wavelet transform, is the scale that controls the expansion and contraction of the wavelet. is the wavelet basis function, To control the translation of the wavelet basis position, is a given signal, It's time.

5. The full waveform decomposition method for an airborne laser radar with an oscillating signal according to claim 4, characterized in that: Determining the wavelet threshold involves calculating the energy density of the detail coefficients of each layer in the decomposition layer. : ； ； Where, It is The detail factor of the layer, Indicates the first The detail coefficient, It is the length of the layer; set up , find the density difference that meets the following judgment conditions : ； Where, is the experience value, represents the median, It is Energy density of the layer; For those that meet the judgment conditions , from the corresponding Estimating noise soft threshold in subbands of discrete wavelet transform of layers : ； Wavelet Thresholding for: ； The detail coefficients above the wavelet threshold are subjected to inverse wavelet transform and reconstructed into a given signal after noise is filtered out.

6. The method for full waveform decomposition of an oscillating signal in an airborne laser radar according to claim 5, characterized in that: Constructing a hybrid coupling model that meets the requirements of airborne lidar systems ,include: ； Where, is a natural constant, is the Gaussian function amplitude, is the angular frequency, is the amplitude of the damped oscillatory function, is the attenuation factor, is the Gaussian mean, is the signal standard deviation, is the time scale; Selecting the initial parameters of the hybrid model includes selecting the initial parameters of the estimated Gaussian function and the initial parameters of the damped oscillation function, performing multi-level wavelet decomposition on the given signal after filtering out the noise, and extracting sub-signals in different frequency bands. The multi-level wavelet decomposition is: ； Where, is a low-pass filter in a given signal after filtering out noise, is a high-pass filter in a given signal after filtering out noise, represents the scale factor, is the approximate coefficient of the low-pass filter after filtering, For scale The approximate signal under is the detail coefficient after filtering by the low-pass filter; is the discrete time index, To sum the index, is a set of integers; right Perform low-frequency convolution to obtain low-frequency information. Perform high-frequency convolution to obtain high-frequency information, and continue to perform low-frequency convolution on the low-frequency information, while the high-frequency information remains unchanged. After the decomposition, the detail coefficients of each sub-band from high frequency to low frequency are obtained, and the low-frequency component and the intermediate-frequency component are selected. After the inverse wavelet transform reconstruction, the component where the Gaussian peak is located and the component where the oscillation signal is located are obtained; The area around the end point of the Gaussian peak is selected as the starting point of the oscillation signal, and the initial value of the Gaussian term is used as a parameter to extract the phase at the starting moment of the oscillation. Perform initial parameter correction and use the corrected initial parameters as the initial particles of the particle swarm optimization algorithm.

7. The method for full waveform decomposition of an oscillating signal in an airborne laser radar according to claim 6, characterized in that: The initial parameters of the Gaussian function include: for: ； Where, is the independent variable of the Gaussian function, which is equivalent to the component where the Gaussian peak is located; Use the strength of the low-frequency signal as , the center point of the low-frequency signal is , the half-width of the low-frequency signal As : 。 8. The method for full waveform decomposition of an oscillating signal in an airborne laser radar according to claim 7, characterized in that: The initial value of the damped oscillation function includes 、 、 and initial phase , damped oscillation function for: ； Get the analysis signal : ； Where, is the original signal, which is equivalent to the component where the oscillation signal is located; Frequency domain after Euler transform for: ； Where, is a symbolic function, is the original signal in the frequency domain; set up , In the time domain: ； Where, is the intermediate variable in the frequency domain, is the intermediate variable in the time domain, represents the convolution operation, is the original signal in the time domain; Find : ； Where, To represent the limit symbol, Indicates that the integral is subjected to Cauchy principal value processing.

9. The method for full waveform decomposition of an oscillating signal in an airborne laser radar according to claim 8, characterized in that: Find Then, we bring it into the analytical signal formula and get ; for: ； Where, is the complex argument function; right The envelope signal of is logarithmically transformed and then linearly fitted: ； Get the slope of the linear fit result and intercept , 、 for: ； ； for: 。 10. The method for full waveform decomposition of an oscillating signal in an airborne laser radar according to claim 9, characterized in that: Particle swarm optimization involves randomly generating a group of particles, initializing particle positions and particle velocities, and using mean square error to calculate The fitting effect of When the mean square error of the fitting result is the smallest, the current particle swarm optimization solution is the optimal solution, and the individual optimal solution of the particle and the global optimal solution at this time are recorded; calculate When the fitting result is , the particle is updated as: ； ； Where, For the The particle velocity of the sub-particle update, is the inertia factor, and is the acceleration constant, Representation interval The random number in For the Particle No. The individual optimal solution of the sub-particle update, Indicates the The global optimal solution for sub-particle update, For the Particle position of sub-particle update; After completing the particle swarm optimization, the Gaussian function parameters and the damped oscillation function parameters are converged through multiple iterative learning to complete the full waveform decomposition.

Images