Laser ranging signal processing method and system based on multi-echo suppression

Abstract

The invention discloses a laser ranging signal processing method and system based on multi-echo suppression, relates to the technical field of photoelectric detection and signal processing, and is particularly suitable for high-precision distance measurement in a multi-echo interference environment. The method is used for solving the problems of pulse width measurement distortion and threshold model failure caused by waveform broadening of a waveform threshold under multiple echoes, performing high-frequency sampling and mathematical diagnosis on an acquired original waveform, and sensing a multi-echo distortion type through peak detection and skewness analysis; according to a diagnosis result, performing physical stripping and suppression of interference echoes by adopting a multi-Gaussian waveform decomposition or multiple scattering analysis model to obtain a purified waveform; extracting characteristic parameters of the purified waveform to obtain a real root-mean-square pulse width; based on the optimal normalized waveform threshold value of the purified pulse width dynamic mapping, high-precision distance calculation is realized in combination with precise time interpolation and migration error compensation; distance measurement errors caused by multiple echoes are overcome, and the distance measurement precision in a complex environment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photoelectric detection and signal processing, and more particularly to a laser ranging signal processing method and system based on multi-echo suppression. BACKGROUND

[0002] In the existing high-precision laser ranging, especially in the observation of the ground and the sea by the spaceborne laser radar, the threshold method based on the received pulse waveform characteristics is widely used for ranging. One advanced method relies on a dynamic optimization normalized waveform threshold model. The model maps the root mean square pulse width of the received waveform to a best threshold ratio to minimize the ranging error caused by the target slope and system noise. The core is to accurately measure the root mean square pulse width of the received waveform, and to assume that the echo waveform is approximately symmetric Gaussian shape, so as to establish a deterministic mapping relationship between the pulse width and the best threshold.

[0003] The prior art has the following disadvantages: However, in actual application, when the laser radar detects rough and large slope ground, forest canopy or penetrates through cloudy medium such as sea water, the multi-echo phenomenon caused by multiple scattering and multi-path effect will cause the received waveform to be seriously widened, distorted or overlapped with multiple peaks, the waveform loses the symmetric Gaussian feature, the measured root mean square pulse width will be significantly larger and distorted, and if the distorted pulse width is directly substituted into the existing dynamic threshold mapping model, the best threshold calculated will be catastrophically shifted, and then a huge systematic ranging error will be generated, which can be up to hundreds of meters, so that the advanced technology completely fails in complex scenes. Therefore, there is an urgent need for a pre-processing method that can effectively suppress interference and restore the real features of the waveform in a multi-echo environment.

[0004] In view of the above problems, the present application provides a solution. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a laser ranging signal processing method and system based on multi-echo suppression to solve the problems raised in the background art.

[0006] To achieve the above object, the present application provides the following technical scheme: A laser ranging signal processing method based on multi-echo suppression, comprising the steps of: Step S1: high-frequency sampling of the collected original waveform, estimation of the background mean and background standard deviation of the background noise, baseline correction and setting of the denoising threshold, obtaining of the preprocessed waveform, multi-echo diagnosis, judgment of whether it belongs to the multi-peak superposition type by detecting the number of local peaks, judgment of whether it belongs to the tail scattering type by calculating the waveform skewness, and determination of the setting of the multi-echo flag and the category according to the diagnosis result. Step S2: If the diagnosis is a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. The multi-Gaussian waveform decomposition method is adopted, and the waveform is decomposed into multiple Gaussian components through nonlinear least squares fitting. The main echo component is selected according to the application target and reconstructed into a purified waveform. For the tail scattering type, the exponential tail suppression method is adopted to identify the tail interval after the main peak and perform logarithmic linear fitting to estimate the tail attenuation model. The purified waveform is obtained by weighted subtraction. Step S3: Set an effective threshold based on the noise standard deviation, determine the effective range of the purified waveform, calculate the energy, time centroid, and root mean square pulse width of the purified waveform, compare the extracted root mean square pulse width with the single echo reference width, determine whether it exceeds the preset lower and upper bound coefficient range, calculate the energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range, if the self-check fails, feed back to step S2 to adjust the suppression weight, component selection, or model parameters; Step S4: Based on the root mean square pulse width of the purified waveform and the single echo reference width, calculate the normalized width, interpolate to obtain the normalized threshold coefficient, multiply it by the peak value to generate the absolute threshold level, search for the threshold crossing point on the rising edge of the purified waveform, calculate the round-trip flight time in combination with the transmission time stamp, convert it into distance, compensate for the flight time, and obtain the final distance.

[0007] In a preferred embodiment, step S1 includes the following: Estimating the background mean and background standard deviation of background noise includes: selecting continuous sampling points that do not contain target echoes as the background estimation set at the beginning of the gated sampling window, and calculating the background mean and background standard deviation based on the background estimation set; The detection of the number of local peaks includes: estimating the first derivative of the preprocessed waveform; counting the number of zero-crossing points where the derivative changes from positive to negative within the effective signal range above the signal candidate threshold; and using this number as the criterion for the number of local peaks. The time centroid and root mean square width of the preprocessed waveform are calculated. Based on the time centroid, root mean square width and waveform energy, the normalized third central moment of the waveform is calculated as the skewness. The skewness is compared with the skewness threshold obtained through calibration to determine whether it belongs to the trailing scattering type.

[0008] In a preferred embodiment, step S2 includes the following: The waveform is decomposed into multiple Gaussian components by nonlinear least squares fitting, which includes: determining the initial number of Gaussian components based on the number of detected local peaks, using the time and amplitude corresponding to each local peak as the initial center time and initial amplitude of the corresponding Gaussian component, using the single echo reference width or the empirically allocated value of the overall waveform width as the initial width, and performing iterative fitting to solve the parameters of each Gaussian component. Selecting the main echo component based on the application objective includes: calculating the signal-to-noise ratio of each Gaussian component, and selecting the component with the earliest or latest arrival time from the components that meet the signal-to-noise ratio threshold as the main echo component based on the ranging application objective. The purified waveform is obtained by weighted subtraction, which involves defining a suppression weight that gradually changes from the tailing point using a smooth window function, multiplying the tailing attenuation model by the suppression weight and subtracting it from the preprocessed waveform, and then truncating the result with a non-negative truncation to obtain the purified waveform.

[0009] In a preferred embodiment, step S3 includes the following: The calculation of the energy sum, time centroid, and root mean square pulse width of the purified waveform includes: determining the effective signal range of the purified waveform based on the effective threshold; within the effective range, calculating the sum of waveform sample values ​​as the energy sum; calculating the weighted average of the timestamps with energy as the weight as the time centroid; and calculating the root mean square value of the deviation of the timestamps from the time centroid as the root mean square pulse width.

[0010] In a preferred embodiment, step S4 includes the following: The normalized threshold coefficient obtained by interpolation includes: based on the calculated normalized width, querying the optimal threshold coefficient table generated for different pulse widths during the offline calibration stage, and obtaining the normalized threshold coefficient by interpolation; Compensating for flight time involves calculating a distance walk time correction using a linear correction model based on the root mean square pulse width of the purified waveform, and then using this correction to adjust the round-trip flight time.

[0011] A laser ranging signal processing system based on multi-echo suppression includes: a waveform preprocessing and diagnostic module, a multi-echo suppression and purification module, a feature extraction and self-testing module, and a dynamic threshold ranging module, with signal connections between the modules; Waveform preprocessing and diagnostic module: High-frequency sampling is performed on the acquired raw waveform, the background mean and background standard deviation of the background noise are estimated, baseline correction is performed and a noise reduction threshold is set to obtain the preprocessed waveform, and multi-echo diagnosis is performed. The number of local peaks is detected to determine whether it belongs to the multi-peak superposition type, and the waveform skewness is calculated to determine whether it belongs to the trailing scattering type. Based on the diagnostic results, the multi-echo flag and category are determined. Multi-echo suppression and purification module: If diagnosed as a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. The multi-Gaussian waveform decomposition method is adopted, and the waveform is decomposed into multiple Gaussian components through nonlinear least squares fitting. The main echo component is selected according to the application target and reconstructed into a purified waveform. For the tail scattering type, the exponential tail suppression method is adopted to identify the tail interval after the main peak and perform logarithmic linear fitting to estimate the tail attenuation model. The purified waveform is obtained by weighted subtraction. Feature extraction self-test module: Based on the noise standard deviation, an effective threshold is set to determine the effective range of the purified waveform. The energy, time centroid, and root mean square pulse width of the purified waveform are calculated. The extracted root mean square pulse width is compared with the single echo reference width to determine whether it exceeds the preset lower and upper bound coefficient range. The energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range is calculated. If the self-test fails, it is fed back to the multi-echo suppression purification module to adjust the suppression weight, component selection, or model parameters. Dynamic threshold ranging module: Based on the root mean square pulse width of the purified waveform and the single echo reference width, the normalized width is calculated, the normalized threshold coefficient is obtained by interpolation, and the absolute threshold level is generated by multiplying the peak value. The threshold crossing point is searched at the rising edge of the purified waveform, the round-trip flight time is calculated by combining the transmission time stamp, and converted into distance. The flight time is compensated to obtain the final distance.

[0012] The technical effects and advantages of the laser ranging signal processing method based on multi-echo suppression of this invention are as follows: By physically stripping away multi-echo interference and restoring the true waveform pulse width, combined with dynamic threshold mapping and distance walk compensation, the meter-level or even hundred-meter-level errors that traditional methods may produce in complex environments are reduced to centimeter-level, greatly improving ranging accuracy. It effectively handles different types of multi-echoes caused by spatial geometric overlap and multiple scattering by the medium, ensuring stable and reliable operation of the system in complex terrain, vegetation cover, and turbid media. It possesses the ability to automatically diagnose multi-echo types, adaptively select suppression algorithms, and verify the physical rationality of parameters, reducing reliance on manual intervention and prior knowledge, and improving the system's intelligence level and long-term on-orbit maintainability. Based on a clear mathematical and physical model and mature digital signal processing algorithms, it can be efficiently implemented on existing DSP or FPGA hardware platforms and is easily integrated into existing spaceborne or airborne lidar systems. Attached Figure Description

[0013] Fig. 1 This is a schematic diagram of the laser ranging signal processing method based on multi-echo suppression according to the present invention.

[0014] Fig. 2 This is a schematic diagram of a laser ranging signal processing system module based on multi-echo suppression according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example Please see Figs. 1-2 As shown, this invention discloses a laser ranging signal processing method based on multi-echo suppression, including the following steps: Step S1: High-frequency sampling is performed on the acquired raw waveform to estimate the background mean and background standard deviation of the background noise. Baseline correction is performed and a noise reduction threshold is set to obtain the preprocessed waveform. Multi-echo diagnosis is performed. The number of local peaks is detected to determine whether it belongs to the multi-peak superposition type. The waveform skewness is calculated to determine whether it belongs to the trailing scattering type. Based on the diagnosis results, the multi-echo flag and category are determined. Step S2: If the diagnosis is a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. The multi-Gaussian waveform decomposition method is adopted, and the waveform is decomposed into multiple Gaussian components through nonlinear least squares fitting. The main echo component is selected according to the application target and reconstructed into a purified waveform. For the tail scattering type, the exponential tail suppression method is adopted to identify the tail interval after the main peak and perform logarithmic linear fitting to estimate the tail attenuation model. The purified waveform is obtained by weighted subtraction. Step S3: Set an effective threshold based on the noise standard deviation, determine the effective range of the purified waveform, calculate the energy, time centroid, and root mean square pulse width of the purified waveform, compare the extracted root mean square pulse width with the single echo reference width, determine whether it exceeds the preset lower and upper bound coefficient range, calculate the energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range, if the self-check fails, feed back to step S2 to adjust the suppression weight, component selection, or model parameters; Step S4: Based on the root mean square pulse width of the purified waveform and the single echo reference width, calculate the normalized width, interpolate to obtain the normalized threshold coefficient, multiply it by the peak value to generate the absolute threshold level, search for the threshold crossing point on the rising edge of the purified waveform, calculate the round-trip flight time in combination with the transmission time stamp, convert it into distance, compensate for the flight time, and obtain the final distance.

[0017] In step S1, the acquired raw waveform is sampled at high frequency to estimate the background noise mean and standard deviation. Baseline correction is performed, and a noise reduction threshold is set to obtain a preprocessed waveform. Multi-echo diagnosis is then performed. The number of local peaks is detected to determine whether the waveform is of the multi-peak superposition type, and the waveform skewness is calculated to determine whether it is of the trailing scattering type. Based on the diagnostic results, the multi-echo flag and category are determined. Specific details include: Having completed one laser pulse emission and entered gated sampling, the analog front-end output is converted into a digital waveform that can be used for subsequent suppression and ranging calculations. The presence of multiple echoes and their main morphology, such as multi-peak superposition or trailing scattering, can be calculably diagnosed to determine whether to proceed to the subsequent multiple echo stripping step. Multiple scattering and multi-target reflection can significantly change the echo pattern. For example, atmospheric, cloud, and aerosol layer echoes are affected by multiple scattering, and their importance is related to factors such as optical thickness, distance, receiving field of view, and RFOV. In particular, before proceeding to threshold time extraction, distortion diagnosis and necessary multi-echo suppression must be completed in this step; otherwise, the threshold time is easily affected by peak shape and tail distortion, resulting in distance wandering and accuracy errors. Record the launch timestamp when a launch event occurs. The transmission time stamp can be derived from: the rising edge trigger of the transmission drive current, the digital pulse generated by the comparator from the output of the transmission monitoring photodiode, and the timestamp of the transmission channel by the TDC; Enable gating sampling, gating start point With gate duration It is obtained by converting from the maximum and minimum ranging ranges, for example sampling frequency The sampling interval is set and readable by the ADC clock. The number of sampling points was calculated to be... Once sampling is complete, the ADC output sequence is the original waveform. ; To clarify the timeline mapping, after obtaining Then, define the time corresponding to each sampling point: , ;in, For the first The relative time of each sampling point; The relative time to the start of the gated window is obtained from the gated logic configuration. For sampling point index; The sampling interval; The total number of sampling points is determined by the gate duration and... Calculate and round down; Before the echo arrives, the first part of the gating window contains only electronic noise and ambient light noise. A continuous signal is selected within the gating window. 1 sampling point as the background estimation set ;in, , These are configurable parameters for the implementer. For example, they account for 10% of the total sampling points. The selection principle is that this interval should not contain any possible target echoes, so as to avoid estimating the signal as noise. After confirmation Then, the background mean and standard deviation are calculated and used as the baseline level and noise fluctuation scale, respectively. Background mean It can be represented as: ;in, This represents the number of sampling points in the background interval. For background sampling index set; Output the raw sample to the ADC; Furthermore, the background standard deviation is calculated in the form of sample variance to avoid bias: ;in, The background standard deviation; , , , Same meaning as above; For squaring operations; After completing the noise statistics, baseline correction and hard thresholding denoising are performed, and confidence coefficients are defined. A value of 5 is preferred to balance false alarm rate and fidelity. The denoising threshold is defined as follows: ;in, This is the noise reduction threshold; and Calculated from the above two formulas; Confidence coefficient; Baseline correction and threshold truncation are performed on all sampling points to obtain the preprocessed sequence. : ; Optionally, threshold suppression can be further implemented; implementers may choose one or both simultaneously: ;in, This is the preprocessed non-negative waveform; This is a function to find the maximum value. The mean value is the background value. This represents the threshold value corresponding to the coordinates after the baseline has been removed; In full-waveform LiDAR processing, an echo may contain superimposed components of multiple targets. The echo can be decomposed into multiple components or echo components to characterize different targets along the optical path. Therefore, the first step is to determine whether there are multiple local peaks in the waveform in order to identify geometric multiple echoes, such as multiple echoes caused by canopy, ground, building facade, etc. To reduce noise-induced spurious peaks, in the already obtained Then, its first derivative is estimated, and the sampling interval required for derivative estimation is... As already stated at the beginning of this step The calculations show that either a five-point central difference or a three-point central difference can be used; here, the discrete derivative in five-point form is given to improve accuracy: , ;in, for The discrete first derivative estimate; For preprocessed waveforms; The sampling interval is determined by... calculate; This represents the total number of sampling points; The set of indices that satisfy the valid signal range The number of zero-crossings where the internal detection derivative changes from positive to negative, to clarify The method for obtaining this is to first use a coarse threshold to determine the candidate signal region: for example, taking... , The default value is 5~10, and it is set to... Then, in Internal statistics satisfy and The number of index points, denoted as ; involved As calculated from the aforementioned noise statistics, The threshold coefficient can be configured for implementers; when When EchoType is set to multi-peak superposition and MultiEchoFlag is set to True, the multi-component decomposition and stripping step is then performed. This criterion is consistent with the engineering fact that full-waveform LiDAR has multiple peaks and multiple components. When multiple scattering is significant, the waveform may not show clear multiple peaks, but rather exhibit an asymmetric shape with a stretched trailing edge of the main peak and a significant energy tail. The contribution of multiple scattering to the LiDAR signal is affected by the medium and geometric conditions, and under certain conditions should not be ignored. To identify such single-peaked but tailed multiple echoes, a skewness is introduced. As an asymmetric criterion, to ensure that the source of the parameters is clear, the time centroid and the width scale are defined first; In obtaining Next, calculate the zeroth moment, approximate the energy with the first moment, use time-weighted summation, and define the discrete energy sum: ;in, For discrete energy sum; For preprocessed waveforms; This represents the total number of sampling points; Define the time center of gravity, arrival time statistics center : ;in, As the focus of time; For energy and; For sampling timestamps; For preprocessed waveforms; Define the root mean square width corresponding to the second-order central moment as the waveform scale. : ;in, The root mean square time width; In already obtained Then, the skewness is calculated in the form of the standardized third-order central moments. This form belongs to the commonly used definition of skewness: ;in, Skewness; For energy and; For sampling timestamps; As the focus of time; The root mean square width; For preprocessed waveforms; The larger the absolute value of the skewness, the more asymmetrical the waveform. A skewness threshold is set. The value can be taken as 0.1~0.3; its specific value can be obtained by calibrating a single-echo standard target: when collecting standard echoes in an environment without multiple scattering and without multiple target superposition, its statistical value can be obtained. Distribution and taking the 95th percentile as ; when and When EchoType is set to trailing scattering type, MultiEchoFlag is set to True; when and At that time, it was considered that the subsequent parameter extraction and threshold ranging steps could be performed by approximating the single peak. The key output quantity is: the preprocessed waveform. Noise statistics Number of multi-peaks skewness MultiEchoFlag and EchoType: Multi-peak superposition, trailing scattering, and near-single echo; This directly determines the execution branch for the next step: if MultiEchoFlag=True, then proceed to multi-echo stripping and suppression; if False, the stripping step can be skipped and the purification parameter extraction can be directly initiated. In this case, purification can be considered as... In this special case, the diagnostic-then-suppress link ensures that the input parameters of the subsequent formulas have a clear source and physical meaning, thereby avoiding the amplification of the error caused by directly using the distortion width and distortion centroid for threshold calculation under multi-echo conditions.

[0018] In step S2, if the diagnosis is a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. A multi-Gaussian waveform decomposition method is used, employing nonlinear least squares fitting to decompose the waveform into multiple Gaussian components. Based on the application objective, the main echo component is selected and reconstructed into a purified waveform. For the tailing scattering type, an exponential tailing suppression method is used to identify the tailing interval after the main peak and perform logarithmic linear fitting to estimate the tailing attenuation model. The purified waveform is obtained through weighted subtraction. Specific details include: Following the output of the previous step, MultiEchoFlag and EchoType: If the previous step clearly indicates the existence of multiple echoes, regardless of whether they manifest as multi-peak superposition or trailing scattering, the echo components harmful to ranging, non-target peaks in multi-target superposition, and trailing or broadening terms caused by multiple scattering will be removed from the digital domain. By stripping away the impurities, a purified waveform that is as close as possible to the reflection of a single effective target is obtained. Its core idea is consistent with the method in full-waveform LiDAR processing, which decomposes the waveform into several echo components to extract their respective parameters, and outputs... This will serve as the sole input waveform for the next step of extracting and self-testing true feature parameters, thereby ensuring that the pulse width, centroid, and threshold mapping in subsequent calculations are all based on the main echo rather than the distorted superimposed waveform. Will Represented as The Gaussian components are superimposed with residual noise, among which The number of peaks obtained from the previous step Decision, for example, to Or make ;in, The maximum number of components set for implementers is usually 2 to 6 to control the amount of computation. Nonlinear least squares fitting is sensitive to initial values, therefore the source of the initial values ​​must be explicitly given, which was obtained in the previous step when detecting the zero-crossing point of the derivative. A set of candidate peak position indices If the value passes through zero or a local maximum, then the initial center time of each component can be set to... What it requires As given in step one, the initial amplitude of each component can be set to the amplitude at the candidate peak point. Initial width of each component The root mean square width of a single echo can be obtained from the pulse during experimental calibration. The echo width is given, for example, the echo width measured on a standard Lambertian target or a planar hard target, or the overall width obtained from the previous step. Assign based on experience, for example, let ; Definition of the first The Gaussian component at time... The value is: ;in, For the first The component is related to the first... The contribution of each sampling point; For the first Peak amplitude of each component; For the first The center time of each component; For the first The time standard deviation of each component; For sampling timestamps; The overall fitted signal is then: ;in, To fit and reconstruct the waveform; The number of components, by and upper limit Decide; It is a component function; Define residual: The fitting objective is to minimize the sum of squared residuals: ;in, The objective function is... The vector of parameters to be estimated; To fit the set of indices, we can take... The range above a certain threshold is used to reduce the dominant noise points; By setting solver selection and termination conditions, Levenberg-Marquardt and L-M class methods can be used to solve nonlinear least squares problems. Mature implementation paths exist in general numerical libraries and classic MINPACKs, such as MINPACK's LMDER1, which calculates Jacobian or LMDIF. The Jacobian can be approximated by difference, and minimization is performed under a modified Levenberg-Marquardt method. Sum of squares of nonlinear functions; Specifically, the processing unit in each iteration Perform the following calculations: Calculate the residual vector ; Calculate the Jacobian matrix If the implementer is unwilling to explicitly derive Jacobian, a differential approximation can be used to obtain Jacobian, corresponding to the LMDIF path; Solving for the increment And update parameters ; The process terminates when convergence conditions are met, such as the relative decrease of the objective function being less than a threshold, the relative change of parameters being less than a threshold, or the maximum number of iterations being reached. The MINPACK documentation provides engineering practices for controlling termination using tolerances such as TOL, FTOL, and XTOL. The default convergence configuration can be set to: maximum number of iterations. ; Threshold for relative change in objective function ; Threshold for relative change of parameters These thresholds are adjustable parameters for the implementer, but giving default values ​​can reduce implementation uncertainty; After completing the fitting, we get The optimal estimate is determined, and the effective main echo is identified. An implementable decision-making logic is provided, ensuring its consistency with the application objectives. If the application involves determining ground distance through vegetation, the component with the latest arrival time that meets the signal-to-noise ratio threshold is typically selected as the ground echo, i.e., the maximum. The rest are echoes from the canopy or intermediate targets; If the application is for nearest obstacle detection, the one with the earliest arrival time and whose amplitude meets the threshold is usually selected, i.e., the smallest. As a barrier echo; If the application is for water depth measurement, two components, water surface and water bottom, can be selected and output separately; The calculation from the previous step can be used. As a noise metric, the component signal-to-noise ratio is defined approximately as follows: ;in, To prevent zero trace amounts and to set a threshold For example, 5~10, this threshold filters out too small components and reduces noise fitting spurious peaks; Assuming the final selected main echo component index is The purified waveform can then be reconstructed from the main echo. Alternatively, other components can be subtracted from the original waveform to preserve the main echo and residual details: ;in, The purified waveform after multi-peak stripping; For preprocessed waveforms; For the first Component fitting values; Main echo index; In turbid media, especially water, multiple scattering can cause waveform time broadening and tailing, affecting detection accuracy. Related research on underwater LiDAR points out that modeling multiple scattering events can change echo detection performance and time broadening pattern, suggesting that the scattering effect needs to be explicitly considered. Therefore, a piecewise fitting exponential tailing suppression scheme is provided for tailing scattering, and deconvolution can be superimposed when needed to compress the broadening. To fit the tailing pattern, the position of the main peak must first be determined. This location can be determined by... The search for the global maximum value yields: Define the trailing start index ;in, This is a post-peak protection interval used to avoid mistaking the initial descent of the main peak as a pure tail. For example, take... ;in, The single echo width calibration value is used; the tail end point can be taken as the point where the waveform falls back to near the noise threshold. ,satisfy And if a certain number of consecutive points are true, then a trailing index set is obtained. ; Within the tailing interval, the tail is approximated using an exponential model: , ;in, For trailing model values; This is the initial amplitude of the trailing effect; The attenuation coefficient; For timestamps; The timestamp of the start of the trail is provided by and Relationship obtained; For estimation , We employ feasible log-linear regression, first performing a safety measure on the tailing samples before taking the logarithm: to avoid zero logarithmic values, we... Below a certain small positive number And take the natural logarithm of both sides of the model: , ;in, and It can be obtained through least-squares linear fitting; get , Then, a tail estimation is constructed on the entire waveform. and by weight The purified waveform is obtained by performing subtraction. : ;in, The purified waveform after tail suppression; For preprocessed waveforms; The tailing estimate is calculated using an exponential model. To suppress weights; To avoid Hard cutting at the point of view causes spectral leakage and waveform discontinuity. A smooth window transformation can be used, such as a cosine window that gradually transitions from 0 to 1. An explicit implementation definition is given: let the smoothing length be... ,For example ; when hour ; when hour ; when hour ; When EchoType is a trailing scattering type and has extreme broadening, for example Exceeding the rated single echo width Several times larger, deconvolution can be used to further compress and broaden. The rationale for deconvolution is that the observed signal can be regarded as the convolution superposition of real echo and point spread, impulse response noise. The Richardson–Lucy, R–L algorithm has been used in optics and signal processing as an iterative deconvolution method for a long time, and there are variants that have been developed under noisy conditions. Specify the point spread function, PSF, or impulse response. Method of obtaining: During factory or periodic calibration, echoes are collected from a standard target, such as a planar hard target or a target at a known distance, and then normalized. It is stored in non-volatile memory and then read at runtime. ,right Or after the trailing part has been removed Performing R-L iterations, since the R-L formula has various discrete forms in different fields, we provide an implementable discrete form, where... For discrete convolution, For point-by-point multiplication and division, iteration : ;in, For the first The purified signal is estimated in the next iteration; For input observation signals; These are the impulse response and point spread function. for Time reversal; To prevent zero constant; This represents the number of iterations. Using the diagnostic data from step one as input: For multi-peak superposition type, the output is obtained by multi-component fitting and stripping. For trailing scattering types, the output is suppressed by exponential trailing and can be optionally deconvolved. Output The common goal is to recover the main echo of a single effective reflection interface as much as possible in terms of mathematical morphology, so that subsequent calculations of pulse width, centroid, and threshold mapping are no longer dominated by non-target components or scattering tails. This is consistent with the basic processing idea of ​​waveform decomposition to extract echo parameters in full-waveform LiDAR processing. At the same time, the use of an engineering-implementable LM-type solver, such as the MINPACK path, ensures that the fitting process and termination conditions can be directly implemented and used as inputs for the next step to extract true feature parameters and perform self-checking feedback.

[0019] In step S3, an effective threshold is set based on the noise standard deviation to determine the effective range of the purified waveform. The energy, time centroid, and root mean square pulse width of the purified waveform are calculated. The extracted root mean square pulse width is compared with the single-echo reference width to determine whether it exceeds the preset lower and upper bound coefficient range. The energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range is calculated. If the self-check fails, feedback is sent to step S2 to adjust the suppression weights, component selection, or model parameters. Specific details include: from Extract all the feature parameters required for subsequent ranging calculations, such as peak value, time centroid, root mean square pulse width, and morphological parameters required for threshold mapping. Then, use boundary verification and consistency checks to determine whether the suppression is excessive or insufficient. If necessary, feed back to the previous step to adjust the suppression intensity or component selection. This parameter extraction + self-checking feedback structure can reduce the sensitivity of the algorithm chain to single fitting and peeling off accidental errors, making it more robust under engineering conditions. Before parameter extraction, the effective waveform range is determined to reduce the impact of noise tails on moment estimation. Since the noise scale has already been obtained in step one... Define a valid threshold ;in, Dimensionless, in Find satisfaction on earliest index With the latest index ,Will As a valid interval, if If the value is empty or the length is less than the minimum number of points, the current echo is determined to be unusable, and the status of no valid echo can be output. Peak amplitude is used for threshold generation. The threshold is often obtained as a proportion or mapping of the peak value. To avoid the peak value being affected by the quantization fence effect, it can be... To ensure feasibility and low computational cost, three-point parabolic interpolation or cubic spline interpolation should be performed nearby. Three-point parabolic interpolation can be taken , , Three points are used to fit a quadratic curve and find its vertex, thereby obtaining the subsampling peak time. With peak Among them, the required timestamp Depend on get; A moment-based parameter extraction method is adopted. To calculate the moments, the effective interval energy and sum are first calculated. : ;in, For the effective interval energy sum; For a valid range index set; For purification waveform; In obtaining Post-calculation time centroid : ;in, The time centroid of the main echo; For energy and; For timestamps; For purification waveform; The effective interval; Then calculate the root mean square width of the vein. : ;in, The root mean square time width of the main echo; After suppressing the trailing edge, the main echo should be closer to a symmetrical shape. To determine whether excessive clipping caused the waveform's trailing edge energy to be incorrectly removed, in Recalculate the skewness Since the skewness formula has already been disclosed in step one, it is only necessary to... Replace with and will Replace with That's it; no new parameters are introduced. The skewness normalization form is the commonly used definition. On the contrary Significantly increased, for example, increased by more than a certain percentage. , If the default value is 1.5~2.0, it may indicate that the suppression weight is too large or the main echo is selected incorrectly. You should go back to step two to reduce the suppression weight or reselect the component. Two types of self-tests are given: pulse width lower and upper bound self-tests, and energy fidelity self-tests. Pulse width lower and upper bound self-check: Single echo reference width has been obtained during calibration. Measured under standard target conditions and without multiple echoes, the actual main echo width should generally not be significantly smaller than the reference width; otherwise, it means that the energy is truncated or the interpolation peak is incorrect. Therefore, a lower bound is set: ;in, The default value is 0.7~0.9. If this is violated, feedback will be sent to step two to reduce the suppression intensity or expand the effective range. Regarding the upper boundary, if Still much larger ,For example , The default value is 3~10, depending on the application. If this value is set too low, it may indicate insufficient multi-echo suppression. In this case, you should follow up with step two to increase the number of components. Alternatively, increase the tail modeling range or enable deconvolution enhancement; Energy fidelity self-check: Compare the energy of the effective range of the preprocessed waveform with the energy of the purified waveform: Let Then the energy fidelity ratio is ,like Too low a value, such as less than 0.3-0.5, may indicate excessive inhibition; if... If the value is close to 1 but still has multiple peaks, then the suppression may be insufficient. The weights for step two can be adjusted. Weighting and decision threshold Make adaptive adjustments; Output a set of true characteristic parameters of the main echo used for ranging calculation: peak value Peak time Time focus Mean square root width and the skewness used for self-testing. Energy ratio This process, which inputs the data into the next step of dynamic threshold generation and distance calculation, tightly links multi-echo suppression with threshold ranging: the threshold mapping is no longer based on the spurious width of the distorted waveform, but on the main echo. and The true form of the distance measurement is obtained, which fundamentally reduces the ranging error under multi-echo conditions.

[0020] In step S4, based on the root mean square pulse width of the purified waveform and the single echo reference width, the normalized width is calculated, the normalized threshold coefficient is obtained by interpolation, and the absolute threshold level is generated by multiplying the peak values. The threshold crossing point is searched at the rising edge of the purified waveform, and the round-trip flight time is calculated in combination with the transmission time stamp and converted into distance. The flight time is compensated to obtain the final distance. The specific contents include: Continuing from the main echo characteristic parameters output in the previous step, especially , , It generates a dynamic threshold that adapts to the current main echo pattern, extracts the threshold crossing time and converts the flight time and distance, and compensates or reduces amplitude-related deviations, time and distance walks. The rising edge detection of the fixed threshold will generate a rangewalk due to the change in echo amplitude, and can be corrected according to the amplitude or equivalent amplitude proxy. The threshold setting is extended from a fixed ratio to an adaptive ratio driven by pulse width and morphological parameters. During factory calibration or offline simulation, different main echo widths are considered. or its normalized form Generate the optimal threshold coefficient table ;in Runtime reading And interpolate values ​​in the table to get And normalize the threshold coefficients. This refers to the proportion of the threshold level to the peak value of the main echo. In obtaining Then, the absolute threshold level is generated. The threshold is defined as: ;in, The threshold level; The normalized threshold coefficient; The main echo peak value; Extracting the threshold crossing time requires determining whether to search on the rising edge, defining the rising edge search interval: indexed by peak value. Tracing back from the center, find the earliest satisfied... and index pairs ,in This search process can be implemented as a linear scan or a binary search in DSPs and FPGAs; After finding the crossing point, linear interpolation is used to obtain the subsampling threshold time. : ;in, The threshold crossing time; For the first Timestamp of each sampling point; The sampling interval; The threshold level; Values ​​of adjacent samples; To prevent zero constant; To convert the threshold crossing time into distance, time stamps need to be emitted. With the received time stamp Defineable ;in, The time travel relative to the launch timescale; Round trip is defined as: ;in, Round-trip flight time; For receiving the judgment time stamp; For launch time markers; Distance, calculated one way, can be expressed as: ;in, Distance to target; It is the speed of light in a vacuum. The refractive index of the medium is approximately 1 for the atmosphere and is set according to the calibration or model for water bodies. This represents the round-trip flight time; the coefficient 2 indicates the round-trip conversion to a one-way trip. A fixed or inappropriate threshold can cause threshold crossing time shift due to amplitude variations. When two echoes differ only in amplitude and the threshold is fixed, the weaker echo will cross the threshold closer to the pulse center, resulting in formative error. Furthermore, amplitude surrogate parameters, such as timeoverthreshold, can be used for correction and reduction, through multi-echo suppression and... Driven Adaptive design reduces erroneous threshold selection caused by waveform distortion at the source; Using pulse width Define a time correction amount for the correction method of the proxy quantity: ;in, This is a distance travel time correction factor; The mean square root pulse width of the main echo; Let be the correction factor; then the corrected flight time is: The distance is ; By adaptively adjusting the dynamic threshold, the ranging accuracy of the system in complex environments is greatly improved. Combined with the aforementioned physical echo stripping and waveform purification, it can not only effectively suppress the errors caused by multiple echoes, but also adjust the ranging threshold in real time according to the actual reflection characteristics of the target, ensuring high-precision distance calculation. The entire system forms a complete and feasible technical closed loop from the original echo data acquisition and interference stripping to the final distance calculation, which significantly improves ranging accuracy, especially in environments with large slopes, complex media, or multi-path reflections.

[0021] This invention discloses a laser ranging signal processing system based on multi-echo suppression, comprising: a waveform preprocessing and diagnostic module, a multi-echo suppression and purification module, a feature extraction and self-testing module, and a dynamic threshold ranging module, wherein the modules are interconnected by signals; Waveform preprocessing and diagnostic module: High-frequency sampling is performed on the acquired raw waveform, the background mean and background standard deviation of the background noise are estimated, baseline correction is performed and a noise reduction threshold is set to obtain the preprocessed waveform, and multi-echo diagnosis is performed. The number of local peaks is detected to determine whether it belongs to the multi-peak superposition type, and the waveform skewness is calculated to determine whether it belongs to the trailing scattering type. Based on the diagnostic results, the multi-echo flag and category are determined. Multi-echo suppression and purification module: If diagnosed as a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. The multi-Gaussian waveform decomposition method is adopted, and the waveform is decomposed into multiple Gaussian components through nonlinear least squares fitting. The main echo component is selected according to the application target and reconstructed into a purified waveform. For the tail scattering type, the exponential tail suppression method is adopted to identify the tail interval after the main peak and perform logarithmic linear fitting to estimate the tail attenuation model. The purified waveform is obtained by weighted subtraction. Feature extraction self-test module: Based on the noise standard deviation, an effective threshold is set to determine the effective range of the purified waveform. The energy, time centroid, and root mean square pulse width of the purified waveform are calculated. The extracted root mean square pulse width is compared with the single echo reference width to determine whether it exceeds the preset lower and upper bound coefficient range. The energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range is calculated. If the self-test fails, it is fed back to the multi-echo suppression purification module to adjust the suppression weight, component selection, or model parameters. Dynamic threshold ranging module: Based on the root mean square pulse width of the purified waveform and the single echo reference width, the normalized width is calculated, the normalized threshold coefficient is obtained by interpolation, and the absolute threshold level is generated by multiplying the peak value. The threshold crossing point is searched at the rising edge of the purified waveform, the round-trip flight time is calculated by combining the transmission time stamp, and converted into distance. The flight time is compensated to obtain the final distance.

[0022] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0023] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0024] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0025] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0026] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0027] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A laser ranging signal processing method based on multi-echo suppression, characterized in that, Includes steps; Step S1: High-frequency sampling is performed on the acquired raw waveform to estimate the background mean and background standard deviation of the background noise. Baseline correction is performed and a noise reduction threshold is set to obtain the preprocessed waveform. Multi-echo diagnosis is performed. The number of local peaks is detected to determine whether it belongs to the multi-peak superposition type. The waveform skewness is calculated to determine whether it belongs to the trailing scattering type. Based on the diagnosis results, the multi-echo flag and category are determined. Step S2: If the diagnosis is a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. The multi-Gaussian waveform decomposition method is adopted, and the waveform is decomposed into multiple Gaussian components through nonlinear least squares fitting. The main echo component is selected according to the application target and reconstructed into a purified waveform. For the tail scattering type, the exponential tail suppression method is adopted to identify the tail interval after the main peak and perform logarithmic linear fitting to estimate the tail attenuation model. The purified waveform is obtained by weighted subtraction. Step S3: Set an effective threshold based on the noise standard deviation, determine the effective range of the purified waveform, calculate the energy, time centroid, and root mean square pulse width of the purified waveform, compare the extracted root mean square pulse width with the single echo reference width, determine whether it exceeds the preset lower and upper bound coefficient range, calculate the energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range, if the self-check fails, feed back to step S2 to adjust the suppression weight, component selection, or model parameters; Step S4: Based on the root mean square pulse width of the purified waveform and the single echo reference width, calculate the normalized width, interpolate to obtain the normalized threshold coefficient, multiply it by the peak value to generate the absolute threshold level, search for the threshold crossing point on the rising edge of the purified waveform, calculate the round-trip flight time in combination with the transmission time stamp, convert it into distance, compensate for the flight time, and obtain the final distance.

2. The laser ranging signal processing method based on multi-echo suppression according to claim 1, characterized in that, Estimating the background mean and background standard deviation of background noise involves selecting continuous sampling points that do not contain target echoes at the beginning of the gated sampling window as the background estimation set, and calculating the background mean and background standard deviation based on this background estimation set.

3. The laser ranging signal processing method based on multi-echo suppression according to claim 2, characterized in that, The detection of the number of local peaks includes: estimating the first derivative of the preprocessed waveform; counting the number of zero-crossing points where the derivative changes from positive to negative within the effective signal range above the candidate signal threshold; and using this number as the criterion for the number of local peaks.

4. The laser ranging signal processing method based on multi-echo suppression according to claim 2, characterized in that, The calculation of waveform skewness includes: calculating the time centroid and root mean square width of the preprocessed waveform; calculating the normalized third central moment of the waveform as the skewness based on the time centroid, root mean square width and waveform energy; and determining whether it belongs to the trailing scattering type by comparing the skewness with the skewness threshold obtained through calibration.

5. The laser ranging signal processing method based on multi-echo suppression according to claim 1, characterized in that, The waveform is decomposed into multiple Gaussian components by nonlinear least squares fitting, which includes: determining the initial number of Gaussian components based on the number of detected local peaks; using the time and amplitude corresponding to each local peak as the initial center time and initial amplitude of the corresponding Gaussian component; using the single echo reference width or the empirically assigned value of the overall waveform width as the initial width; and iteratively fitting to solve for the parameters of each Gaussian component.

6. The laser ranging signal processing method based on multi-echo suppression according to claim 5, characterized in that, Selecting the main echo component based on the application objective includes: calculating the signal-to-noise ratio of each Gaussian component, and selecting the component with the earliest or latest arrival time from the components that meet the signal-to-noise ratio threshold as the main echo component based on the ranging application objective.

7. The laser ranging signal processing method based on multi-echo suppression according to claim 5, characterized in that, The purified waveform is obtained by weighted subtraction, which involves defining a suppression weight that gradually changes from the tailing point using a smooth window function, multiplying the tailing attenuation model by the suppression weight and subtracting it from the preprocessed waveform, and then truncating the result with a non-negative truncation to obtain the purified waveform.

8. The laser ranging signal processing method based on multi-echo suppression according to claim 1, characterized in that, The calculation of the energy sum, time centroid, and root mean square pulse width of the purified waveform includes: determining the effective signal range of the purified waveform based on the effective threshold; within the effective range, calculating the sum of waveform sample values ​​as the energy sum; calculating the weighted average of the timestamps with energy as the weight as the time centroid; and calculating the root mean square value of the deviation of the timestamps from the time centroid as the root mean square pulse width.

9. The laser ranging signal processing method based on multi-echo suppression according to claim 1, characterized in that, The normalized threshold coefficient obtained by interpolation includes: based on the calculated normalized width, querying the optimal threshold coefficient table generated for different pulse widths during the offline calibration stage, and obtaining the normalized threshold coefficient by interpolation; Compensating for flight time involves calculating a distance walk time correction using a linear correction model based on the root mean square pulse width of the purified waveform, and then using this correction to adjust the round-trip flight time.

10. A laser ranging signal processing system based on multi-echo suppression, comprising: a waveform preprocessing and diagnostic module, a multi-echo suppression and purification module, a feature extraction and self-testing module, and a dynamic threshold ranging module, used to implement the laser ranging signal processing method based on multi-echo suppression as described in any one of claims 1-9, characterized in that... ； Waveform preprocessing and diagnostic module: High-frequency sampling is performed on the acquired raw waveform, the background mean and background standard deviation of the background noise are estimated, baseline correction is performed and a noise reduction threshold is set to obtain the preprocessed waveform, and multi-echo diagnosis is performed. The number of local peaks is detected to determine whether it belongs to the multi-peak superposition type, and the waveform skewness is calculated to determine whether it belongs to the trailing scattering type. Based on the diagnostic results, the multi-echo flag and category are determined. Multi-echo suppression and purification module: If diagnosed as a multi-peak superposition type, the preprocessed waveform is modeled as a superposition of multiple Gaussian components. The multi-Gaussian waveform decomposition method is adopted, and the waveform is decomposed into multiple Gaussian components through nonlinear least squares fitting. The main echo component is selected according to the application target and reconstructed into a purified waveform. For the tail scattering type, the exponential tail suppression method is adopted to identify the tail interval after the main peak and perform logarithmic linear fitting to estimate the tail attenuation model. The purified waveform is obtained by weighted subtraction. Feature extraction self-test module: Based on the noise standard deviation, an effective threshold is set to determine the effective range of the purified waveform. The energy, time centroid, and root mean square pulse width of the purified waveform are calculated. The extracted root mean square pulse width is compared with the single echo reference width to determine whether it exceeds the preset lower and upper bound coefficient range. The energy fidelity ratio of the purified waveform and the preprocessed waveform within the effective range is calculated. If the self-test fails, it is fed back to the multi-echo suppression purification module to adjust the suppression weight, component selection, or model parameters. Dynamic threshold ranging module: Based on the root mean square pulse width of the purified waveform and the single echo reference width, the normalized width is calculated, the normalized threshold coefficient is obtained by interpolation, and the absolute threshold level is generated by multiplying the peak value. The threshold crossing point is searched at the rising edge of the purified waveform, the round-trip flight time is calculated by combining the transmission time stamp, and converted into distance. The flight time is compensated to obtain the final distance.

Images