An online ranging method for lidar saturation signals based on waveform compensation

Abstract

The present invention discloses an online ranging method for a laser radar saturation signal based on waveform compensation, comprising a waveform state detection module, a waveform compensation module, and a moment identification module. The waveform state detection module detects the signal state by comparing the amplitude of the original echo signal with the acquisition range; the waveform compensation module determines a mathematical basic model based on the skewness characteristics of the non-saturated signal and determines the model parameters using a fitting optimization method; the saturated signal is compensated to a reconstructed non-saturated signal by attenuating and delaying the basic model, wherein the attenuation factor is determined by estimating the saturation length by fitting the rising and falling edge sampling points of the saturated signal, and the delay factor is determined based on the principle of minimizing the matching error between the attenuated signal and the saturated signal; the moment identification module is used to extract the waveform parameters of the reconstructed signal. The present invention can invert high-precision distance and intensity values ​​from highly saturated signals, thereby expanding the dynamic range of laser radar detection.

Description

Technical Field

[0001] The present invention relates to the field of laser radar (LiDAR) measurement, and in particular to an online ranging method for LiDAR saturation signals based on waveform compensation. This method directly processes the input saturation signal using a programmable processor to execute a waveform compensation algorithm and a moment discrimination algorithm, achieving online measurement of target distance and echo intensity. Background Art

[0002] As an active detection technology, lidar (LiDAR) is widely used in fields such as ocean exploration, topographic mapping, and autonomous driving due to its compact size, high accuracy, and long range. LiDAR actively emits laser pulses toward a target and records the return pulses using an analog-to-digital converter. Using a time-of-day discrimination algorithm, it extracts the emission and arrival times of the laser pulses to measure the target's distance. Combined with the laser pulse's angle of departure, it generates a three-dimensional point cloud of the target. For example, in detection applications such as autonomous driving, lidar typically requires a detection range of several meters to several hundred meters. However, due to variations in target reflectivity and detection range, the amplitude of the return signal can experience significant dynamic variations. When long-range detection is required, the amplitude of the backscattered return signal from a close-range target typically exceeds the maximum range of the receiving system, causing the return signal to become distorted and saturated. This saturated signal loses waveform information above the maximum range of the receiving system, exhibiting characteristics such as peak cancellation and pulse broadening. Directly processing this saturated signal using traditional time-of-day discrimination algorithms results in poor ranging accuracy and an inability to extract echo intensity information. Therefore, during the data processing process, traditional lidars are generally unable to process the saturated signals, thereby limiting the increase in the density of close-range target point clouds and shortening the lidar detection range.

[0003] To overcome the impact of saturation signals on lidar ranging performance, two common approaches are currently used. One is hardware-based saturation signal ranging technology, which adjusts the gain of the lidar's transceiver channels to avoid the generation of saturation signals. This hardware-based approach primarily includes automatic gain control systems and multi-channel transceiver systems. By avoiding saturation signals, this hardware-based approach effectively expands the lidar's detection range, but it also increases the lidar's design cost and complexity. Furthermore, variations in the receive channel gain lead to discontinuous variations in echo intensity at different distances from the same target, making it difficult to utilize this echo intensity information. The other is algorithm-based saturation signal ranging technology, which includes waveform non-compensation and waveform compensation algorithms. The waveform non-compensation algorithm uses partial features of the saturation signal as the arrival time of the echo pulse, such as the rising edge or center of gravity of the saturation signal. However, this approach fails to fully utilize the characteristic information of the saturation signal, resulting in ranging accuracy limited by the sampling rate of the analog-to-digital converter, and generally unable to extract echo intensity. The waveform compensation algorithm compensates for the saturated region of the saturated signal, transforming it into a reconstructed unsaturated signal. A conventional time-of-day discrimination algorithm is then used to extract characteristic parameters from the reconstructed unsaturated signal, enabling distance inversion and echo intensity estimation from the saturated signal. The waveform compensation algorithm processes the reconstructed unsaturated signal, enabling the time-of-day discrimination algorithm to fully utilize the characteristic information of the saturated signal, improving the accuracy of distance and intensity estimation. For example, Gaussian decomposition can be used to extract target distance based on the reconstructed unsaturated signal, achieving sub-sampling ranging accuracy.

[0004] Depending on the algorithm processing platform, algorithm-based ranging methods can be divided into post-processing ranging technology and online ranging technology. Post-processing ranging technology requires recording and storing echo signals, and then processing the stored echo signals on a host computer, making it impossible to extract target distance in real time. Online ranging technology uses a programmable processor and digital waveform processing technology to achieve online measurement of target distance and echo intensity. This invention belongs to an online ranging technology based on a waveform compensation algorithm, which is a sub-alternative algorithm-based saturation signal ranging technology.

[0005] According to patent reviews provided by existing patent offices, existing patent applications utilize the least squares method to fit the undistorted sampling points of the saturated signal, estimating the amplitude and shape parameters of the restored unsaturated signal based on the waveform compensation algorithm. The restored unsaturated signal, with a determined amplitude and shape, is then truncated according to the saturation level and matched with the saturated signal to estimate the position parameters of the restored unsaturated signal, thereby completing the restoration of the saturated signal. This patent encompasses the "LIDAR Saturation Waveform Restoration Method Based on the Least Squares Method" disclosed in Chinese Patent CN112230197A. This patent does not address the issue of saturated signal broadening. As the saturation level of the saturated signal increases, the pulse broadening of the saturated signal increases. At this point, the error in estimating the model parameters by directly fitting the undistorted saturated sampling points increases, rendering the patent's ranging capability ineffective. The present invention differs from the aforementioned patent in the following ways: First, the compensation method differs. The aforementioned patent estimates model parameters by fitting unsaturated sampling points and implements saturated signal compensation based on a fixed parameter model. The present invention implements a model combination by attenuating and translating the basic model of the system echo signal to compensate for the saturated signal. Second, the associated parameter in the compensation function differs (the saturation length). The aforementioned patent does not define the concept of saturation length, while the present invention defines the concept of saturation length and designs a saturation length correction method to reduce model parameter estimation errors caused by pulse broadening. Third, the amplitude and shape parameter estimation methods differ. The aforementioned patent simultaneously estimates amplitude and shape parameters by fitting unsaturated sampling points. The present invention first establishes the basic model combination based on the time domain signal representation of the unsaturated signal, designs a segmented fitting algorithm to estimate the original saturation length, and then estimates the amplitude parameters based on the corrected original saturation length. Fourth, the parameter estimation steps differ (rough estimation and optimized estimation). The aforementioned patent only performs a rough estimation of model parameters based on unsaturated sampling points. The present invention first performs a rough estimation of model parameters based on the saturated signal characteristics and then optimizes the model parameters in the time identification algorithm.

[0006] Another patent application uses a deep learning-based method to restore the saturated signal. This patent collects saturated and unsaturated signals as training data, training LSTM and LSTM-CNN models respectively. The trained LSTM model then restores the saturated signal to an unsaturated signal, and then uses the LSTM-CNN model to extract the characteristic parameters of the restored unsaturated signal to measure the target distance. This patent encompasses the "Single-Pulse LiDAR Echo Feature Extraction Method Based on Deep Learning" disclosed in Chinese patent CN112986950A. This patent trains the network model solely based on echo signal characteristics, but the collected echo signal itself has broadening distortion. As the saturation level of the saturated signal increases, the ranging performance of this patent decreases. The differences between the present application and the above-mentioned patents include: first, the compensation method is different. The above-mentioned patent restores the saturated signal based on the LSTM model trained with the saturated signal, while the present application realizes model combination by attenuating and translating the basic model of the system echo signal to compensate for the saturated signal; second, the characteristic parameter estimation method is different. The above-mentioned patent estimates the characteristic parameters of the restored non-saturated signal based on the LSTM-CNN model trained with the non-saturated signal, while the present patent first performs a rough estimation of the model parameters based on the characteristics of the saturated signal, and then optimizes and estimates the model parameters in the moment identification algorithm to extract the characteristic parameters of the restored non-saturated signal.

[0007] The two aforementioned patents cover the waveform compensation algorithm within the algorithmic approach, and also include a patent for the non-waveform compensation algorithm within the algorithmic approach. This patent determines the arrival time of the echo signal based on the rising edge of the saturation signal. This patent then compensates for the determined arrival time based on the sampling length during which the saturation signal sample amplitude exceeds a set threshold, thereby enabling distance measurement of the saturation signal. These patents encompass the "Laser Ranging Method and System Based on Pulse Width Compensation" disclosed in Chinese Patent CN115343693A. This patent only compensates for the distance parameter determined based on the rising edge of the saturation signal and cannot estimate the amplitude parameter. The differences between the present application and the above-mentioned patents include: first, the ranging method based on the saturated signal is different. The above-mentioned patent compensates the distance parameter determined directly based on the rising edge of the saturated signal according to the sampling length at which the sampling amplitude of the saturated signal is higher than the set threshold, thereby realizing the inversion of the target distance. The present application belongs to the waveform compensation algorithm in the algorithm method, firstly restores the saturated signal, and then uses the moment identification algorithm to extract the distance parameter of the restored non-saturated signal; second, the method for determining the associated parameters in the compensation function is different (the saturation length). The above-mentioned patent uses the sampling length higher than the set threshold as the saturation length. The present application designs a segmented fitting algorithm to estimate the original saturation length, and designs the saturation length correction method, and uses the corrected original saturation length as the saturation length; third, the amplitude parameter estimation method is different. The above-mentioned patent cannot estimate the amplitude parameter. The present application first makes a rough estimation of the amplitude parameter according to the saturated signal length, and then optimizes the estimation of the amplitude parameter in the moment identification algorithm.

[0008] To increase the dynamic range of laser radar detection, reduce the cost and complexity of hardware-based designs, and achieve real-time, algorithmic inversion of high-precision distance and intensity information from saturated signals, the present invention discloses a laser radar saturated signal online ranging method based on waveform compensation. This saturated signal online ranging method uses an analog-to-digital converter to acquire saturated signals, executes the waveform compensation algorithm and the moment identification algorithm on a programmable processor, and achieves online compensation of the saturated signal and online extraction of characteristic parameters, completing the inversion of target distance and estimation of echo intensity. The present invention has broad application prospects in the field of laser radar measurement. Summary of the Invention

[0009] The present invention discloses a laser radar saturation signal online ranging method based on waveform compensation, which is characterized in that the saturation signal online ranging method is executed based on a programmable processor, and the input saturation signal is directly processed to achieve online measurement of target distance and echo intensity; the saturation signal online ranging method includes a waveform compensation algorithm and a moment identification algorithm; the waveform compensation algorithm establishes a basic model of the system echo signal for the non-saturated signal, and complements the saturated signal into a reconstructed non-saturated signal by attenuating and translating the basic model of the system echo signal; the reconstructed non-saturated signal is processed using the moment identification algorithm to extract characteristic parameters of the target echo;

[0010] The saturation signal is defined as the echo signal amplitude exceeding the maximum range A of the receiving system. th , resulting in distortion, flat top and other phenomena in the middle part of the output signal; the saturation signal is divided into three parts: saturation interval, rising edge and falling edge, and the length of the saturation interval is defined as the original saturation length l raw , that is, the starting point s1 of the saturation interval and the ending point s n To overcome the original saturation length l raw The broadening effect requires the original saturation length l raw Correction, and the corrected original saturation length l raw Named as saturation length l; after the saturation signal is digitized by the analog-to-digital converter, the saturation interval sampling point T s Defined as: T s ={t s |(f raw (t)≥A th )∩(s1<t<s n )}, the starting sampling point of the saturation interval is The saturation interval cutoff sampling point is f raw (t) is the original echo signal recorded; the rising edge sampling point R s Defined as: Falling edge sampling point F s Defined as: Where σ is the original echo signal f raw The noise standard deviation of (t);

[0011] The saturation signal online ranging method comprises the following steps:

[0012] (1) Establishing a compensation model: Select a mathematical model based on the different time domain signal representations of the non-saturated signal of the laser radar; Based on the mathematical model, perform fitting optimization on the non-saturated signal and determine the mathematical model parameters as the basic model f of the system echo signal.e (t); By the basic model f of the system echo signal e (t) Perform attenuation and translation to establish a compensation model f for the saturation signal r (t), that is, f r (t) = k a ×f e (tk τ ), k a is the attenuation factor, k τ is the delay factor;

[0013] (2) Determine the attenuation factor k a : According to the compensation model f r (t), establish the saturation length l and the attenuation factor k a The relational model k a =g(l); the saturation length l needs to be the original saturation length l of the saturated signal raw Correction is performed by correcting the model λ(l raw ) for the original saturation length l raw Correction is performed, and the correction output is the saturation length l; based on the relationship model k a =g(l), determine the attenuation factor k a ;

[0014] (3) Determine the delay factor k τ : Determine the compensation model f r The attenuation factor k of (t) a Then, set the delay factor k τ The search interval S τ and the search step Δτ, according to the original echo signal f raw (t) and the degree of matching of the compensation model, adjusting the delay factor k τ The size of the rising edge sampling point R s The delay factor k is determined based on the principle of minimum matching error at τ ;

[0015] (4) Reconstructing the saturated signal: Determining the attenuation factor k a and the delay factor k τ After that, according to the compensation model f r (t) = k a ×f e (tk τ ), calculate the saturation interval sampling point T s The compensation point is used to complete the saturated signal into the reconstructed non-saturated signal;

[0016] (5) extracting characteristic parameters of the reconstructed non-saturated signal: the characteristic parameters of the reconstructed non-saturated signal include pulse amplitude a, pulse position τ, and pulse width w. A time discrimination algorithm is used to extract the characteristic parameters of the reconstructed non-saturated signal to achieve distance inversion and echo intensity estimation of the saturated signal;

[0017] (6) Realizing the online execution of the saturation signal ranging method: The architecture design of the saturation signal online ranging method in the programmable processor includes a waveform state detection module, a waveform compensation module and a time identification module; in the waveform state detection module, the original echo signal f is raw The sampling amplitude of (t) is 1-δ times the receiving system range A th Compare and judge the original echo signal f raw (t) is in saturation state, the δ coefficient is defined as the value of the amplitude of the sampling point in the saturation interval within the receiving system range A th The fluctuation coefficient near the echo signal can be used to prevent misjudgment of the echo signal state; when the original echo signal f raw The sampling amplitude of (t) is higher than (1-δ)·A th When the original echo signal f raw (t) is in saturation state, the original echo signal f raw (t) is sequentially input to the waveform compensation module and the time identification module to realize the distance inversion and echo intensity estimation of the saturated signal; otherwise, the original echo signal f raw (t) is a non-saturated state, and the original echo signal f raw (t) is directly input to the time identification module to extract the target distance and echo intensity.

[0018] The laser radar saturation signal online ranging method based on waveform compensation is characterized in that the basic model f of the system echo signal e (t) A Gaussian model, and / or a Gaussian mixture model, and / or a B-spline model may be selected; a reasonable selection of the basic model may be achieved based on the time domain characteristics of the system's transmitted signal.

[0019] The laser radar saturation signal online ranging method based on waveform compensation is characterized in that the original saturation length l raw The segmented fitting algorithm can be used to determine the rising edge sampling point R of the saturation signal. s Perform fitting to obtain the rising edge sampling point fitting curve r c (t); the falling edge sampling point F of the saturation signal s Perform fitting to obtain the falling edge sampling point fitting curve f c (t); Let:

[0020]

[0021] Solve the time starting point s1 of the saturation interval and the time ending point s of the saturation interval of the saturation signal n , and then calculate the original saturation length l raw =s n -s1.

[0022] The laser radar saturation signal online ranging method based on waveform compensation is characterized in that the saturation length l and the attenuation factor k a The relational model k a =g(l), can be established as follows:

[0023] 1) Adjust the compensation model f r The attenuation factor k of (t) a ,Right now To generate n groups of modulated non-saturated signals;

[0024] 2) For each group of the modulated non-saturated signals, set the receiving system range A according to the system parameters. th , the modulated non-saturated signal is higher than the receiving system range A th The amplitude of is truncated to generate n groups of modulated saturation signals;

[0025] 3) For the modulated saturation signal of group i, the corresponding attenuation factor is The original saturation length is determined by using a piecewise fitting algorithm Since the modulated saturation signal has no broadening effect, the saturation length is consistent with the original saturation length, and thus the attenuation factor can be obtained. and the corresponding saturation length l i ;

[0026] 4) After all the n groups of modulated saturation signals are processed according to step 3), the attenuation factor can be obtained. and the corresponding saturation length [l 1 ,l 2 ,…,l n ] data set; fitting and optimizing the acquired data set, calculating the relationship model k a =g(l).

[0027] The laser radar saturation signal online ranging method based on waveform compensation is characterized in that the delay factor k τ The non-precise line search algorithm can be used for adjustment; the search interval S τ The saturation interval [s1,s n], adjust the candidate delay factor with a fixed step size, i.e. The compensation model f r (t) and the original echo signal f raw (t), the error between the rising edge sampling points R s The smallest candidate delay factor is used as the final delay factor k τ , as shown in formula (2).

[0028]

[0029] The online ranging method for laser radar saturation signals based on waveform compensation is characterized in that the moment identification algorithm includes a peak method, and / or an interpolation method, and / or a Gaussian decomposition method.

[0030] The laser radar saturation signal online ranging method based on waveform compensation is characterized in that the original saturation length correction model λ(l raw ) can be established by following the steps below:

[0031] 1) Fix the distance between the laser radar and the target, adjust the echo intensity received by the laser radar system by blocking the light path, and collect the saturated signals of different original saturation lengths

[0032] 2) Calculate the original saturation length l of the saturation signal using a segmented fitting algorithm raw , the collected saturation signals are divided into n groups according to the size of the original saturation length, that is,

[0033] 3) For the i-th group, the saturation signal is collected, and the corresponding original saturation length is Since the original saturation length is greater than or equal to the saturation length, the original saturation length correction model λ(l raw ) has a value range of 0 to 1; in the original saturation length correction model λ(l raw ) is within the range of 0 to 1, and the original saturation length is adjusted according to the step size Δλ Corresponding correction factor Wherein, the correction coefficient is the original saturation length l raw When the original saturation length correction model λ(l raw ) value;

[0034] 4) For each waveform of the saturation signal collected in the i-th group, at each correction coefficient Next, calculate the saturation length l j Based on the saturation length l j, use the saturation signal online ranging algorithm to measure the target distance information; when the i-th group collects all the waveforms of the saturation signal and completes the processing, the correction coefficient is statistically When the distance measurement accuracy of the target distance of the i-th group is

[0035] 5) Repeat steps 3) to 4) until all correction coefficients in the interval 0 to 1 are traversed according to the step size Δλ, and the correction coefficient corresponding to the minimum ranging accuracy is used as the current original saturation length The corresponding correction coefficient λ i ;

[0036] 6) Repeat steps 3) to 5) until n groups of the saturation signals collected are processed to obtain the original saturation length and the corresponding calibration data set, that is, and [λ1,λ2,…,λ n ]; Based on the polynomial model, the acquired data set is fitted and optimized to establish the saturation length correction model λ(l raw ).

[0037] The online ranging method for laser radar saturation signals based on waveform compensation is characterized in that the waveform compensation module is implemented in the programmable processor, and the specific implementation process is as follows:

[0038] 1) Determine the saturation length: read the original echo signal f output by the waveform state detection module raw (t), the original echo signal f raw (t) The first one higher than (1-δ)·A th The sampling time corresponding to the sampling amplitude is the starting sampling point of the saturation interval. The last one is higher than (1-δ)·A th The sampling moment corresponding to the sampling amplitude of is the cutoff sampling point of the saturation interval. According to the sampling point and Determine the rising edge fitting curve r c (t); According to the sampling point and Determine the falling edge fitting curve f c (t), solve the starting point s1 of the saturation interval and the ending point s1 of the saturation interval according to formula (3) n :

[0039]

[0040] Then calculate the original saturation length l raw =s n-s1; According to the determined saturation length correction model λ(l raw ), for the original saturation length l raw Correction, calculation of the saturation length l;

[0041] 2) Determine the attenuation factor k a : Read the saturation length l calculated in step 1), and according to the established saturation length l and the attenuation factor k a The relational model k a =g(l), calculate the attenuation factor k a ;

[0042] 3) Determine the delay factor k τ : Read the attenuation factor k determined in step 2) a , the rising edge sampling point R s Fixed to N is the number of rising edge sampling points, which is determined by the pulse width of the laser radar non-saturated signal; the search step length Δτ is set to α times the saturation length, that is, Δτ = α·l, to ensure that the number of searches is the same when processing the saturated signals with different saturation lengths; the search interval S is set to τ Limited to the saturation interval [s1,s n ], wherein the search interval S τ It can be further limited according to the waveform time domain characteristics to improve the execution efficiency of the algorithm; the candidate delay factor is adjusted according to the search step size Δτ, that is, The candidate delay factor In this case, record the compensation model f r (t) and the original echo signal f raw (t) at the rising edge sampling point R s The mean absolute error of the compensation model f r (t) and the original echo signal f raw (t) Evaluation criteria for the matching error between delay factors; after the mean absolute value errors of all candidate delay factors are calculated, a comparator is used to select the minimum mean absolute value error, and the candidate delay factor corresponding to the minimum mean absolute error is used as the delay factor k τ ;

[0043] 4) Reconstruct the saturated signal: read the attenuation factor k determined in steps 2) and 3) a and the delay factor k τ , according to the compensation model f r (t) = k a ×f e (tk τ ), calculate the saturation interval sampling point Ts The corresponding compensation point is obtained by completing the saturated signal into the reconstructed non-saturated signal, and the reconstructed non-saturated signal is input into the time identification module.

[0044] The waveform compensation-based online ranging method for laser radar saturation signals is characterized in that the execution speed and resource usage of the online ranging method for saturation signals on a programmable processor are optimized:

[0045] 1) In the compensation model f r (t) determination link, the basic model f of the system echo signal e (t) The more complex the method is, the more parameters there are, the more resources the saturated signal online ranging method occupies in the programmable processor, the higher the computational complexity, and the longer the execution time;

[0046] 2) In the calculation link of the saturation length l, it is used to determine the fitting curve r c (t) and f c The rising edge sampling point R of (t) s and the falling edge sampling point F s The more points there are, the more resources the saturation signal online ranging method occupies in the programmable processor, the higher the computational complexity, and the longer the execution time;

[0047] 3) In the delay factor k τ The calculation link, the search interval S τ The larger the value is, the smaller the delay factor search step Δτ is, the more resources the saturated signal online ranging method occupies in the programmable processor, and the longer the execution time is;

[0048] 4) In the extraction step of reconstructing the characteristic parameters of the non-saturated signal, the more complex the selected moment identification algorithm is, the more resources the saturated signal online ranging method occupies in the programmable processor, the higher the computational complexity is, and the slower the execution speed is.

[0049] The waveform compensation-based online ranging method for laser radar saturation signals is characterized in that the programmable processor includes a field programmable gate array (FPGA), and / or a complex programmable logic device (CPLD), and / or an ARM processor (Advanced RISC processors), and / or a digital signal processor (DSP). According to the logic characteristics of the programmable processor, the saturation signal online ranging algorithm is reasonably designed to improve the online ranging calculation time and the online ranging accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The workflow diagram of the online ranging method for lidar saturation signals based on waveform compensation

[0051] Figure 2 Schematic diagram of saturation signal

[0052] Figure 3 Schematic diagram for estimating the original saturation length

[0053] Figure 4 Obtaining a schematic diagram for the decay factor and saturation length datasets

[0054] Figure 5 Schematic diagram for reconstructing a saturated signal

[0055] Figure 6 The structure diagram of the test system for the online ranging method of laser radar saturation signal based on waveform compensation DETAILED DESCRIPTION

[0056] The specific implementation of the present invention is further described in detail below with reference to the accompanying drawings.

[0057] The present invention discloses a laser radar saturation signal online ranging method based on waveform compensation. The working process of an embodiment of the method is as follows: Figure 1 This embodiment consists of three parts: a waveform state detection module, a waveform compensation module, and a moment identification module. The waveform state detection module is used to detect whether the original echo signal is the saturated signal. The waveform compensation module establishes a basic model of the system echo signal for the non-saturated signal, and completes the saturated signal into a reconstructed non-saturated signal by attenuating and translating the basic model of the system echo signal. The moment identification module processes the reconstructed non-saturated signal to extract the characteristic parameters of the target echo.

[0058] The saturation signal in this embodiment is divided into three parts: saturation interval, rising edge and falling edge. Figure 2 As shown. The saturation interval length is defined as the original saturation length l raw , that is, the starting point s1 of the saturation interval and the ending point s n To overcome the original saturation length l raw The broadening effect requires the original saturation length l raw Correction, and the corrected original saturation length l raw Named as saturation length l; after the saturation signal is digitized by the analog-to-digital converter, the saturation interval sampling point T s Defined as: T s ={t s |(f raw (t)≥Ath )∩(s1<t<s n )}, the starting sampling point of the saturation interval is The saturation interval cutoff sampling point is f raw (t) is the original echo signal recorded; the rising edge sampling point R s Defined as: Falling edge sampling point F s Defined as: Where σ is the original echo signal f raw The standard deviation of the noise of (t).

[0059] The implementation steps of this embodiment are as follows:

[0060] (1) Establishing a compensation model: Select a mathematical model based on the different time domain signal representations of the non-saturated signal of the laser radar; Based on the mathematical model, perform fitting optimization on the non-saturated signal and determine the mathematical model parameters as the basic model f of the system echo signal. e (t); By the basic model f of the system echo signal e (t) Perform attenuation and translation to establish a compensation model f for the saturation signal r (t), that is, f r (t) = k a ×f e (tk τ ), k a is the attenuation factor, k τ is the delay factor;

[0061] (2) Determine the attenuation factor k a : According to the compensation model f r (t), establish the saturation length l and the attenuation factor k a The relational model k a =g(l); the saturation length l needs to be the original saturation length l of the saturated signal raw Correction is performed by correcting the model λ(l raw ) for the original saturation length l raw Correction is performed, and the correction output is the saturation length l; based on the relationship model k a =g(l), determine the attenuation factor k a ;

[0062] (3) Determine the delay factor k τ : Determine the compensation model f r The attenuation factor k of (t) a Then, set the delay factor k τThe search interval S τ and the search step Δτ, according to the original echo signal f raw (t) and the degree of matching of the compensation model, adjusting the delay factor k τ The size of the rising edge sampling point R s The delay factor k is determined based on the principle of minimum matching error at τ ;

[0063] (4) Reconstructing the saturated signal: Determining the attenuation factor k a and the delay factor k τ After that, according to the compensation model f r (t) = k a ×f e (tk τ ), calculate the saturation interval sampling point T s The compensation point is used to complete the saturated signal into the reconstructed non-saturated signal;

[0064] (5) extracting characteristic parameters of the reconstructed non-saturated signal: the characteristic parameters of the reconstructed non-saturated signal include pulse amplitude a, pulse position τ, and pulse width w. A time discrimination algorithm is used to extract the characteristic parameters of the reconstructed non-saturated signal to achieve distance inversion and echo intensity estimation of the saturated signal;

[0065] (6) Realizing the online execution of the saturation signal ranging method: The architecture design of the saturation signal online ranging method in the programmable processor includes a waveform state detection module, a waveform compensation module and a time identification module; in the waveform state detection module, the original echo signal f is raw The sampling amplitude of (t) is 1-δ times the receiving system range A th Compare and judge the original echo signal f raw (t) is in saturation state, the δ coefficient is defined as the value of the amplitude of the sampling point in the saturation interval within the receiving system range A th The fluctuation coefficient near the echo signal can be used to prevent misjudgment of the echo signal state; when the original echo signal f raw The sampling amplitude of (t) is higher than (1-δ)·A th When the original echo signal f raw (t) is in saturation state, the original echo signal f raw (t) is sequentially input to the waveform compensation module and the time identification module to realize the distance inversion and echo intensity estimation of the saturated signal; otherwise, the original echo signal f raw (t) is a non-saturated state, and the original echo signal f raw (t) is directly input to the time identification module to extract the target distance and echo intensity.

[0066] In step (1) of establishing the compensation model, this embodiment selects a Gaussian mixture model as the basic model of the system echo signal, namely:

[0067]

[0068] Where, ξ e is the model parameter vector, defined as (a1, b1, c1, …, a M ,b M ,c M ,M,k dc ); M is the number of Gaussian models; a i ,b i and c i is the amplitude, position and standard deviation of the i-th Gaussian model; k dc The model parameter vector ξ in this embodiment is e The specific determination process is:

[0069] 1) Based on the laser radar system, multiple groups of non-saturated signals are collected at the same target. According to the multi-pulse coherent superposition algorithm, the collected multiple groups of non-saturated signals are superimposed to obtain the noise-reduced non-saturated signal;

[0070] 2) When M takes different values, the Levenberg-Marquardt algorithm is used to fit the noise-reduced non-saturated signal to determine the model parameter vector ξ under different values ​​of M. e ;

[0071] 3) Calculating the skewness of the Gaussian mixture model (4) and the noise-reduced non-saturated signal, and the mean square error between the Gaussian mixture model (4) and the noise-reduced non-saturated signal under different values ​​of M;

[0072] 4) The model parameter vector corresponding to the minimum M that makes the skewness and mean square error meet the error requirements is used as the final model parameter vector ξ e .

[0073] In step (2), the attenuation factor k is determined a In this embodiment, in order to speed up the execution of the algorithm, the sampling point and Determine the rising edge fitting curve r c (t); According to the sampling point and Determine the falling edge fitting curve f c (t), such as Figure 3 As shown; according to formula (3), the starting point s1 of the saturation interval and the ending point s of the saturation interval are solved. n :

[0074]

[0075] Then calculate the original saturation length l raw =s n -s1.

[0076] In step (2), the attenuation factor k is determined a In this embodiment, the saturation length l and the attenuation factor k are established according to the following steps: a The relational model k a =g(l):

[0077] 1) Adjust the compensation model f r The attenuation factor k of (t) a ,Right now To generate n groups of modulated non-saturated signals;

[0078] 2) For each group of the modulated non-saturated signals, set the receiving system range A according to the system parameters. th , the modulated non-saturated signal is higher than the receiving system range A th The amplitude of is truncated to generate n groups of modulated saturation signals;

[0079] 3) For the modulated saturation signal of group i, the corresponding attenuation factor is The original saturation length is determined by using a piecewise fitting algorithm Since the modulated saturation signal has no broadening effect, the saturation length is consistent with the original saturation length, and thus the attenuation factor can be obtained. and the corresponding saturation length l i ,like Figure 4 As shown;

[0080] 4) After all the n groups of modulated saturation signals are processed according to step 3), the attenuation factor can be obtained. and the corresponding saturation length [l 1 ,l 2 ,…,l n This embodiment optimizes the obtained data set according to model (6) and calculates the relationship model k a = parameters γ1 and γ2 in g(l).

[0081]

[0082] In step (3), the delay factor k is determined τ In this embodiment, the non-exact line search algorithm is used to adjust the delay factor k τ ; The search interval Sτ The saturation interval [s1,s n ], adjust the candidate delay factor with a fixed step size, i.e. The compensation model f r (t) and the original echo signal f raw (t), the error between the rising edge sampling points R s The smallest candidate delay factor is used as the final delay factor k τ , as shown in formula (2).

[0083]

[0084] The specific implementation steps are as follows:

[0085] 1) In order to ensure the efficiency of the algorithm, the rising edge sampling point R s Fixed to N is the number of rising edge sampling points, which is determined by the pulse width of the non-saturated laser radar signal;

[0086] 2) The search step length Δτ is set to α times the saturation length l, that is, Δτ=α·l, to ensure that the number of searches is the same when processing the saturated signals with different saturation lengths; the search interval S τ The saturation interval [s1,s n ];

[0087] 3) Adjust the candidate delay factor according to the search step size Δτ, that is, Record the candidate delay factor at the i-th Under the above mentioned compensation model f r (t) and the original echo signal f raw (t) at the rising edge sampling point R s The mean absolute error of

[0088] 4) After the mean absolute value errors of all candidate delay factors are calculated, a comparator is used to select the minimum mean absolute value error, and the candidate delay factor corresponding to the minimum mean absolute value error is used as the delay factor k τ .

[0089] In step (5) of extracting the characteristic parameters of the reconstructed non-saturated signal, this embodiment uses Gaussian decomposition method to extract the characteristic parameters of the reconstructed non-saturated signal, wherein the objective function of the fitting is set to:

[0090]

[0091] Where ξ is the waveform characteristic parameter vector, defined as (A r ,τr ,w r ) T ; A r ,τ r and w r are the pulse amplitude, pulse position and pulse width respectively; the Gauss-Newton method is used to solve the least squares problem of formula (9) to obtain high-precision characteristic parameters:

[0092]

[0093] Where, ξ * is the optimized waveform characteristic parameter vector, f s (t) is the reconstructed non-saturated signal. The specific execution process is as follows:

[0094] 1) Setting the initial vector of the characteristic parameters of the reconstructed unsaturated signal to Among them, τ e is the pulse position of the basic model of the system echo signal;

[0095] 2) Calculate the Jacobian matrix J(ξ k ) and the residual vector And update the waveform characteristic parameter vector according to formula (10):

[0096]

[0097] In the formula, the residual vector N c To cut the length of the effective waveform segment, it is related to the laser radar pulse width and the sampling rate of the receiving system;

[0098] 3) When the number of iterations reaches the set number of iterations K, the iteration stops and the optimized feature parameters are output.

[0099] In step (2), the attenuation factor k is determined a In this embodiment, the original saturation length correction model λ(l raw ):

[0100] 1) Fix the distance between the laser radar and the target, adjust the echo intensity received by the laser radar system by blocking the light path, and collect the saturated signals of different original saturation lengths

[0101] 2) Calculate the original saturation length l of the saturation signal using a segmented fitting algorithm raw , the collected saturation signals are divided into n groups according to the size of the original saturation length, that is,

[0102] 3) For the i-th group, the saturation signal is collected, and the corresponding original saturation length is Since the original saturation length is greater than or equal to the saturation length, the original saturation length correction model λ(l raw ) has a value range of 0 to 1; in the original saturation length correction model λ(l raw ) is within the range of 0 to 1, and the original saturation length is adjusted according to the step size Δλ Corresponding correction factor Wherein, the correction coefficient is the original saturation length l raw When the original saturation length correction model λ(l raw ) value;

[0103] 4) For each waveform of the saturation signal collected in the i-th group, at each correction coefficient Next, calculate the saturation length l j Based on the saturation length l j , calculate the attenuation factor k a =g(l j ), using an inexact line search algorithm to determine the delay factor k τ , obtain the reconstructed non-saturated signal, and extract the characteristic parameters of the reconstructed non-saturated signal based on the Gaussian decomposition method; after all the waveforms of the saturated signal collected by the i-th group are processed, the correction coefficient is statistically When the distance measurement accuracy of the target distance of the i-th group is

[0104] 5) Repeat steps 3) to 4) until all correction coefficients in the interval 0 to 1 are traversed according to the step size Δλ, and the correction coefficient corresponding to the minimum ranging accuracy is used as the current original saturation length The corresponding correction coefficient λ i ;

[0105] 6) Repeat steps 3) to 5) until n groups of the saturation signals collected are processed to obtain the original saturation length and the corresponding calibration data set, that is, and [λ1,λ2,…,λ n ]; To ensure the efficiency of the algorithm, this embodiment adopts a second-order polynomial model to perform fitting optimization on the acquired data set and establish the saturation length correction model λ(l raw ).

[0106] In step (6) of implementing the online execution of the saturated signal ranging method, this embodiment executes the waveform compensation-based laser radar saturated signal online ranging method based on a field programmable gate array, and the specific execution steps are as follows:

[0107] 1) Waveform state detection: By transforming the original echo signal f raw The sampling amplitude of (t) is 1-δ times the receiving system range A th Compare and judge the original echo signal f raw (t) Whether it is in a saturated state, the receiving system range A of the laser radar in this embodiment th =900mV, the sampling point amplitude is within the receiving system range A th The fluctuation value near the original echo signal f is less than 9mV, so the fluctuation coefficient δ is set to 0.01; raw The sampling amplitude of (t) is higher than (1-δ)·A th When the original echo signal f raw (t) is a saturation signal. In order to speed up the algorithm execution efficiency and ensure that the effective segment of the saturation signal can be completely intercepted, the saturation signal is in the interval The sampling point is intercepted and the intercepted sampling point is input to step 2); otherwise, the original echo signal f raw (t) is a non-saturated signal, and the non-saturated signal is intercepted in the interval [t τ -15,t τ +16], t τ is the maximum sampling amplitude of the non-saturated signal a τ The corresponding sampling moment, and the intercepted sampling point and the initial vector of the characteristic parameters of the non-saturated signal ξ0=(a τ ,t τ ,w) T Input to step 5), w is the pulse width of the laser radar. In this embodiment, the pulse width w of the laser radar is 10ns, and the sampling rate of the receiving system is 1GSa / s.

[0108] 2) Determine the saturation length: read the original echo signal f output by the waveform state detection module raw (t), the original echo signal f raw (t) The first one higher than (1-δ)·A th The sampling time corresponding to the sampling amplitude is the starting sampling point of the saturation interval. The last one is higher than (1-δ)·A th The sampling moment corresponding to the sampling amplitude of is the cutoff sampling point of the saturation interval. According to the sampling point and Determine the rising edge fitting curve r c (t); According to the sampling point and Determine the falling edge fitting curve f c(t), solve the starting point s1 of the saturation interval and the ending point s1 of the saturation interval according to formula (3) n :

[0109]

[0110] Then calculate the original saturation length l raw =s n -s1; According to the determined saturation length correction model λ(l raw ), for the original saturation length l raw Correction, calculation of the saturation length l;

[0111] 3) Determine the attenuation factor k a : Read the saturation length l calculated in step 2), and according to the established saturation length l and the attenuation factor k a The relational model k a =g(l), calculate the attenuation factor k a ;

[0112] 4) Determine the delay factor k τ : Read the attenuation factor k determined in step 3) a , the rising edge sampling point R s Fixed to N is the number of rising edge sampling points. In order to speed up the execution efficiency of the algorithm, this embodiment fixes the number of rising edge sampling points N to 16, sets the search step Δτ to α to 0.1, and sets the search interval S to τ Set to [s1+0.4l, s1+0.5l]; adjust the candidate delay factor according to the search step size Δτ, that is, The candidate delay factor In this case, record the compensation model f r (t) and the original echo signal f raw (t) at the rising edge sampling point R s The mean absolute error of the compensation model f r (t) and the original echo signal f raw (t) Evaluation criteria for the matching error between delay factors; after the mean absolute value errors of all candidate delay factors are calculated, a comparator is used to select the minimum mean absolute value error, and the candidate delay factor corresponding to the minimum mean absolute error is used as the delay factor k τ ;

[0113] 5) Reconstruct the saturated signal: read the attenuation factor k determined in steps 3) and 4) a and the delay factor k τ , according to the compensation model fr (t) = k a ×f e (tk τ ), calculate the saturation interval sampling point T s The corresponding compensation point is used to complete the saturated signal into the reconstructed non-saturated signal, such as Figure 5 As shown; In this embodiment, in order to ensure that the waveform segments of the saturated signal and the non-saturated signal input to the time identification module are consistent, the reconstructed non-saturated signal is intercepted in the interval [t c -15,t c +16] sampling points, where t c The sampling time corresponding to the maximum sampling amplitude of the reconstructed non-saturated signal is the sampling point after the interception and the initial vector of the characteristic parameters of the reconstructed non-saturated signal. Enter into step 5).

[0114] 6) Feature parameter extraction: Read the waveform segment output from step 1) or step 5) and the waveform feature parameter initial vector ξ0, and calculate the Jacobian matrix J(ξ) of the kth iteration according to model (8). k ) and the residual vector The inverse matrix [J(ξ k )J(ξ k )] -1 , and then solve the waveform characteristic parameter vector ξ of the k+1th time k+1 When the number of iterations is higher than the set number of iterations K, the waveform characteristic parameter vector after final optimization is output to realize online measurement of target distance and echo intensity. In this embodiment, the set number of iterations K is set to 2.

[0115] This embodiment implements the waveform-compensated laser radar saturation signal online ranging method based on the field programmable gate array (FPGA), and tests the ranging performance and runtime of the algorithm. During the test, the distance between the laser radar system and a diffuse reflector with a reflectivity of 90% is fixed, and the energy of the echo signal is adjusted by adjusting the optical attenuation plate in the optical unit to change the saturation of the saturated signal. In this embodiment, saturation is defined as the percentage of the original saturation length to the laser pulse width; the field programmable gate array receives and processes the original echo signal output by the optical unit, and transmits the extracted characteristic parameters to the host computer via Ethernet. The structural diagram of the test system is shown in the figure below. Figure 6As shown. In this embodiment, ranging accuracy and ranging precision are used to evaluate the ranging performance of the online ranging method of the saturated signal of the laser radar based on waveform compensation, wherein the ranging accuracy is defined as the deviation between the average value of the measured distance and the true distance, and the ranging precision is defined as the standard deviation of the measured distance. The test results of this embodiment on the saturated signal with different saturations are shown in Table 1. The test results show that when the saturation is in the range of 0 to 165%, the ranging accuracy of this embodiment fluctuates in the range of -1.89 to 1.74 cm, and the ranging precision is less than 2.53 cm. The running time of this embodiment for processing a single saturated signal based on the ZYNQ7000 type FPGA is 7.9 μs, wherein the working clock frequency of the FPGA is 250 MHz.

[0116] Table 1 Ranging results of this embodiment

[0117]

[0118]

[0119] The above is merely a basic scheme for implementing the present invention, but the scope of protection of the present invention is not limited thereto. Any changes and substitutions that can be imagined by anyone skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims. All changes that fall within the meaning and scope of equivalents of the claims are intended to be included within the scope of the claims.

Claims

1. A laser radar saturation signal online ranging method based on waveform compensation, characterized in that: A saturated signal online ranging method is executed based on a programmable processor, directly processing the input saturated signal to achieve online measurement of target distance and echo intensity. The saturated signal online ranging method includes a waveform compensation algorithm and a moment identification algorithm. The waveform compensation algorithm establishes a basic model of the system echo signal for the non-saturated signal, and completes the saturated signal into a reconstructed non-saturated signal by attenuating and translating the basic model of the system echo signal. The reconstructed non-saturated signal is processed using the moment identification algorithm to extract characteristic parameters of the target echo. The saturation signal is defined as the echo signal amplitude exceeding the maximum range A of the receiving system. th , resulting in distortion, flat top and other phenomena in the middle part of the output signal; the saturation signal is divided into three parts: saturation interval, rising edge and falling edge, and the length of the saturation interval is defined as the original saturation length l raw , that is, the starting point s1 of the saturation interval and the ending point s n To overcome the original saturation length l raw The broadening effect requires the original saturation length l raw Correction, and the corrected original saturation length l raw Named as saturation length l; after the saturation signal is digitized by the analog-to-digital converter, the saturation interval sampling point T s Defined as: T s ={t s |(f raw (t)≥A th )∩(s1<t<s n )}, the starting sampling point of the saturation interval is The saturation interval cutoff sampling point is f raw (t) is the original echo signal recorded; the rising edge sampling point R s Defined as: Falling edge sampling point F s Defined as: Where σ is the original echo signal f raw The standard deviation of noise (t); The saturation signal online ranging method comprises the following steps: (1) Establishing a compensation model: Select a mathematical model based on the different time domain signal representations of the non-saturated signal of the laser radar; Based on the mathematical model, perform fitting optimization on the non-saturated signal and determine the mathematical model parameters as the basic model f of the system echo signal. e (t); By the basic model f of the system echo signal e (t) Perform attenuation and translation to establish a compensation model f for the saturation signal r (t), that is, f r (t) = k a ×f e (tk τ ), k a is the attenuation factor, k τ is the delay factor; (2) Determine the attenuation factor k a : According to the compensation model f r (t), establish the saturation length l and the attenuation factor k a The relational model k a =g(l); the saturation length l needs to be the original saturation length l of the saturated signal raw Correction is performed by correcting the model λ(l raw ) for the original saturation length l raw Correction is performed, and the correction output is the saturation length l; based on the relationship model k a =g(l), determine the attenuation factor k a ; (3) Determine the delay factor k τ : Determine the compensation model f r The attenuation factor k of (t) a Then, set the delay factor k τ The search interval S τ and the search step Δτ, according to the original echo signal f raw (t) and the degree of matching of the compensation model, adjusting the delay factor k τ The size of the rising edge sampling point R s The delay factor k is determined based on the principle of minimum matching error at τ ; (4) Reconstructing the saturated signal: Determining the attenuation factor k a and the delay factor k τ After that, according to the compensation model f r (t) = k a ×f e (tk τ ), calculate the saturation interval sampling point T s The compensation point is used to complete the saturated signal into the reconstructed non-saturated signal; (5) extracting characteristic parameters of the reconstructed non-saturated signal: the characteristic parameters of the reconstructed non-saturated signal include pulse amplitude a, pulse position τ, and pulse width w. A time discrimination algorithm is used to extract the characteristic parameters of the reconstructed non-saturated signal to achieve distance inversion and echo intensity estimation of the saturated signal; (6) realizing online execution of the saturation signal ranging method: the architecture design of the saturation signal online ranging method in the programmable processor includes a waveform state detection module, a waveform compensation module and a time identification module; In the waveform state detection module, the original echo signal f raw The sampling amplitude of (t) is 1-δ times the receiving system range A th Compare and judge the original echo signal f raw (t) is in saturation state, the δ coefficient is defined as the value of the amplitude of the sampling point in the saturation interval within the receiving system range A th The fluctuation coefficient near the echo signal can be used to prevent misjudgment of the echo signal state; when the original echo signal f raw The sampling amplitude of (t) is higher than (1-δ)·A th When the original echo signal f raw (t) is in saturation state, the original echo signal f raw (t) is sequentially input to the waveform compensation module and the time identification module to realize the distance inversion and echo intensity estimation of the saturated signal; otherwise, the original echo signal f raw (t) is a non-saturated state, and the original echo signal f raw (t) is directly input to the time identification module to extract the target distance and echo intensity.

2. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The basic model f of the system echo signal e (t) A Gaussian model, and / or a Gaussian mixture model, and / or a B-spline model may be selected.

3. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The original saturation length l raw The segmented fitting algorithm can be used to determine the rising edge sampling point R of the saturation signal. s Perform fitting to obtain the rising edge sampling point fitting curve r c (t); the falling edge sampling point F of the saturation signal s Perform fitting to obtain the falling edge sampling point fitting curve f c (t); make: Solve the time starting point s1 of the saturation interval and the time ending point s of the saturation interval of the saturation signal n , and then calculate the original saturation length l raw =s n -s1.

4. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The saturation length l and the attenuation factor k a The relational model k a =g(l), can be established as follows: 1) Adjust the compensation model f r The attenuation factor k of (t) a ,Right now To generate n groups of modulated non-saturated signals; 2) For each group of the modulated non-saturated signals, set the receiving system range A according to the system parameters. th , the modulated non-saturated signal is higher than the receiving system range A th The amplitude of is truncated to generate n groups of modulated saturation signals; 3) For the modulated saturation signal of group i, the corresponding attenuation factor is The original saturation length is determined by using a piecewise fitting algorithm Since the modulated saturation signal has no broadening effect, the saturation length is consistent with the original saturation length, and thus the attenuation factor can be obtained. and the corresponding saturation length l i ; 4) After all the n groups of modulated saturation signals are processed according to step 3), the attenuation factor can be obtained. and the corresponding saturation length The obtained data set is fitted and optimized to calculate the relationship model k a =g(l).

5. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The delay factor k τ The non-precise line search algorithm can be used for adjustment; the search interval S τ The saturation interval [s1,s n ], adjust the candidate delay factor with a fixed step size, i.e. The compensation model f r (t) and the original echo signal f raw (t), the error between the rising edge sampling points R s The smallest candidate delay factor is used as the final delay factor k τ , as shown in formula (2).

6. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The time identification algorithm includes a peak method, and / or an interpolation method, and / or a Gaussian decomposition method.

7. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The original saturation length correction model λ(l raw ) can be established by following the steps below: 1) fixing the distance between the laser radar and the target, adjusting the echo intensity received by the laser radar system by blocking the light path, and collecting the saturated signals of different original saturation lengths; 2) Calculate the original saturation length l of the saturation signal using a segmented fitting algorithm raw , the collected saturation signals are divided into n groups according to the size of the original saturation length, that is, 3) For the i-th group, the saturation signal is collected, and the corresponding original saturation length is Since the original saturation length is greater than or equal to the saturation length, the original saturation length correction model λ(l raw ) has a value range of 0 to 1; in the original saturation length correction model λ(l raw ) is within the range of 0 to 1, and the original saturation length is adjusted according to the step size Δλ Corresponding correction factor Wherein, the correction coefficient is the original saturation length l raw When the original saturation length correction model λ(l raw ) value; 4) For each waveform of the saturation signal collected in the i-th group, at each correction coefficient Next, calculate the saturation length l j Based on the saturation length l j , use the saturation signal online ranging algorithm to measure the target distance information; when the i-th group collects all the waveforms of the saturation signal and completes the processing, the correction coefficient is statistically When the distance measurement accuracy of the target distance of the i-th group is 5) Repeat steps 3) to 4) until all correction coefficients in the interval 0 to 1 are traversed according to the step size Δλ, and the correction coefficient corresponding to the minimum ranging accuracy is used as the current original saturation length The corresponding correction coefficient λ i ; 6) Repeat steps 3) to 5) until n groups of the saturation signals collected are processed to obtain the original saturation length and the corresponding calibration data set, that is, and [λ1,λ2,…,λ n ]; Based on the polynomial model, the acquired data set is fitted and optimized to establish the saturation length correction model λ(l raw ).

8. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The waveform compensation module is implemented in the programmable processor, and the specific implementation process is as follows: 1) Determine the saturation length: read the original echo signal f output by the waveform state detection module raw (t), the original echo signal f raw (t)The first one higher than (1-δ)·A th The sampling time corresponding to the sampling amplitude is the starting sampling point of the saturation interval. The last one is higher than (1-δ)·A th The sampling moment corresponding to the sampling amplitude of is the cutoff sampling point of the saturation interval. According to the sampling point and Determine the rising edge fitting curve r c (t); According to the sampling point and Determine the falling edge fitting curve f c (t), solve the starting point s1 of the saturation interval and the ending point s1 of the saturation interval according to formula (3) n : Then calculate the original saturation length l raw =s n -s1; According to the determined saturation length correction model λ(l raw ), for the original saturation length l raw Correction, calculation of the saturation length l; 2) Determine the attenuation factor k a : Read the saturation length l calculated in step 2), and according to the established saturation length l and the attenuation factor k a The relational model k a =g(l), calculate the attenuation factor k a ; 3) Determine the delay factor k τ : Read the attenuation factor k determined in step 3) a , the rising edge sampling point R s Fixed to N is the number of rising edge sampling points, which is determined by the pulse width of the laser radar non-saturated signal; the search step length Δτ is set to α times the saturation length, that is, Δτ = α·l, to ensure that the number of searches is the same when processing the saturated signals with different saturation lengths; the search interval S is set to τ Limited to the saturation interval [s1,s n ], wherein the search interval S τ It can be further limited according to the waveform time domain characteristics to improve the execution efficiency of the algorithm; the candidate delay factor is adjusted according to the search step size Δτ, that is, The candidate delay factor In this case, record the compensation model f r (t) and the original echo signal f raw (t) at the rising edge sampling point R s The mean absolute error of the compensation model f r (t) and the original echo signal f raw (t) Evaluation criteria for the matching error between delay factors; after the mean absolute value errors of all candidate delay factors are calculated, a comparator is used to select the minimum mean absolute value error, and the candidate delay factor corresponding to the minimum mean absolute error is used as the delay factor k τ ; 4) Reconstruct the saturated signal: read the attenuation factor k determined in steps 2) and 3) a and the delay factor k τ , according to the compensation model f r (t) = k a ×f e (tk τ ), calculate the saturation interval sampling point T s The corresponding compensation point is obtained by completing the saturated signal into the reconstructed non-saturated signal, and the reconstructed non-saturated signal is input into the time identification module.

9. The online ranging method for laser radar saturation signals based on waveform compensation according to claim 1, characterized in that: The programmable processor includes a Field-Programmable Gate Array (FPGA), and / or a Complex Programmable Logic Device (CPLD), and / or an ARM processor (Advanced RISC Machines), and / or a Digital Signal Processor (DSP).

Images