Laser radar aliasing waveform decomposition method based on waveform characteristics

Abstract

The invention provides a laser radar aliasing waveform decomposition method based on waveform characteristics, which constructs a waveform mathematical model according to seed light characteristics of a laser radar, and decomposes aliasing echoes into a plurality of single echo data in combination with an LM curve fitting algorithm, thereby achieving the purpose of accurately extracting multi-target ranging information. The technical basis of the invention is a laser radar full waveform detection technology, that is, a laser radar detection system can sample an analog electric signal converted from an optical signal into a digital waveform signal, and the processing object of the invention is a waveform sampling point sequence of seed light and return light detected by a laser radar.

Description

Technical Field

[0001] The invention relates to the field of visibility meters, and in particular to a laser radar aliasing waveform decomposition method based on waveform features. Background Art

[0002] LiDAR is an active detection technology that is widely used in remote sensing, surveying and mapping. With the integration and miniaturization of laser light sources and the continuous advancement of detection technology, LiDAR technology has made great progress in recent years. In the fields of assisted driving, robotics, security monitoring, 3D measurement, etc., LiDAR has become one of the general technical means.

[0003] The performance requirements of the surveying and mapping industry for LiDAR are different from those in other technical fields. It has higher requirements for core technical indicators such as point frequency, ranging, accuracy, and penetration. In the face of long-range and complex target scenes, the return light signal of LiDAR is often complex. Taking pulsed LiDAR as an example, the seed light pulse signal is a relatively ideal pulse waveform. The return light signal is affected by factors such as the target's spatial distribution characteristics, geometric characteristics, and reflection characteristics. Its form is often complex and diverse, and it is easy to form special echo signals such as multiple echoes, aliased echoes, and irregular echoes. Such signals will affect the calculation of LiDAR ranging and intensity information, thereby causing noise in the LiDAR point cloud data and affecting the measurement accuracy. At the same time, some useful signals will be submerged by aliased echoes and irregular echoes, resulting in incomplete detection of targets and affecting the "penetration" of LiDAR. Summary of the invention

[0004] The main purpose of the present invention is to provide a laser radar aliasing waveform decomposition method based on waveform characteristics, so as to solve the problem that the seed light pulse signal is a relatively ideal pulse waveform, and the echo signal is affected by factors such as the target spatial distribution characteristics, geometric characteristics, and reflection characteristics. Its form is often complex and diverse, and it is easy to form special echo signals such as multiple echoes, aliased echoes, and irregular echoes. Such signals will affect the calculation of the laser radar ranging and intensity information, thereby causing noise in the laser radar point cloud data, affecting the measurement accuracy.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a laser radar aliasing waveform decomposition method based on waveform characteristics, according to the seed light characteristics of the laser radar, a waveform mathematical model is constructed, and the aliasing echo is decomposed into multiple single echo data in combination with the LM curve fitting algorithm, so as to achieve the purpose of accurately extracting multi-target ranging information. The method includes: S1. Obtain waveform sampling sequence: Obtain the waveform sampling sequence and sampling point time sequence of the laser radar seed light / return light after AD sampling, and the time interval between adjacent sampling points is a known fixed value; S2. Calculate the background noise value: Calculate the background noise value based on the signal in a certain interval around the waveform data, and take the average signal in the noise interval as the background noise value, so that it changes dynamically with the waveform data; S3, waveform data smoothing: use mean filtering and other algorithms to smooth the original waveform data; S4. Determine the valid waveform range: take the smoothed waveform data, the point where the value is greater than the background noise value and the value increases for n consecutive times as the starting point, and the point where the value decreases for n consecutive times as the ending point, where n is related to the characteristics of the laser radar, and the waveform sequence involved in the subsequent calculation is the valid waveform range; S5. Setting the basic function model of waveform decomposition: The single pulse waveform function model is determined by the seed light emission characteristics and the response characteristics of the laser radar photoelectric detection system. The practical model of the present invention is in a specific form, and the parameters to be fitted characterize the pulse waveform intensity, peak position, and pulse width; S6. Initial determination of the number of pulse waveforms: the number of starting waveforms is determined according to the pulse waveform sequence, and the aliasing waveform function is set to the superposition of the corresponding number of basic functions; S7, setting the initial value of the fitting parameter and the search range: determine the initial value of the variable according to the waveform peak value, position, and seed light pulse width, and set the initial value of the pulse width parameter and the search range of the fitting parameter; S8. Fitting using LM algorithm: Use LM algorithm to perform curve fitting and calculate the optimal estimate of parameters; S9, verify fitting accuracy: substitute the fitting parameters into the model to calculate the fitting waveform sequence, and make a difference calculation deviation with the original waveform sequence. If the predetermined index is not reached, return to step S6 to increase the number of pulse functions to be fitted. If the index is reached, output the fitting result and determine the distance measurement and reflection intensity information of the reflecting target.

[0006] In the preferred embodiment, steps S1-S2 are specifically as follows: S11, let a pulse signal be sampled by AD and the output data sequence be: ; The corresponding sampling point time series is: ; The time interval between adjacent sampling points is the system AD sampling period , is a known fixed value: ; S21. Calculate the noise floor value corresponding to the waveform data based on the signal value within a certain interval around the waveform data. Although for a fixed laser radar detection system, the noise floor of the echo signal is generally at a fixed level, the noise floor of the waveform sequence is generally dynamically changing due to the influence of factors such as the ambient light source and the temperature characteristics of the hardware system itself. The noise floor calculation method described in the present invention can be used to estimate the noise floor of the current waveform data in real time, so that the noise floor value changes dynamically with the waveform data.

[0007] In the preferred embodiment, the specific steps of steps S3-S4 are: S31, waveform data smoothing: Smoothing the original waveform data. The smoothing method may adopt filtering algorithms such as mean filtering. The waveform data sequence after smoothing is: ; S41. Determine the effective waveform range: The effective waveform is a data sequence involved in the subsequent waveform decomposition calculation. The determination method is as follows: take the smoothed waveform data, and the point whose value is greater than the background noise value and has n consecutive values ​​increasing is the starting point of the effective waveform; the point whose value is greater than the background noise value and has n consecutive values ​​decreasing is the ending point of the effective waveform; the value of n is related to the characteristics of the laser radar, and the present invention is described by taking n=4 as an example in the claims description; Starting point k Sure: ; End point j Sure: ; The waveform sequence involved in subsequent calculations is: .

[0008] In the preferred embodiment, the specific steps of step S5 are: S51. Setting a waveform decomposition basic function model: According to the ranging principle of pulse laser radar, the seed light pulse waveform still maintains the waveform characteristics of the seed light after being reflected by the target. However, due to the influence of the reflection of the adjacent target, the return light received by the laser radar is usually a superposition of multiple pulse signals. The waveform decomposition is to extract the multiple superimposed pulse waveforms, and then obtain the ranging and intensity information of multiple adjacent targets. The waveform function model of a single pulse is the basic function model of waveform decomposition, which can be determined by the emission characteristics of the seed light and the response characteristics of the laser radar photoelectric detection system. The Gaussian function is an ideal laser radar seed light pulse model and a symmetrical function model. The actual laser radar seed light pulse waveform is affected by the laser characteristics and the laser radar photoelectric response characteristics, and is deformed on the basis of the Gaussian model, and the rising and falling edges of the waveform are asymmetric. The basic model of waveform decomposition function is: .

[0009] In the preferred embodiment, the specific steps of step S6 are: S61. Initial determination of the number of pulse waveforms: The aliased waveform is usually formed by the superposition of multiple basic pulse waveforms. Determine the number of starting waveforms to facilitate rapid convergence of waveform fitting calculation. The aliased waveform function to be fitted is the superposition of N basic functions. Let: ; The initial form of the aliasing waveform function is: .

[0010] In the preferred embodiment, the specific steps of step S7 are: S71, setting the initial value and search range of the waveform decomposition fitting parameter: As described in step S61, if the number of pulse waveforms initially determined is N, it is necessary to determine the initial values ​​and search ranges of multiple variables, which are expressed as follows: ; The initial value of the variable is determined by the peak value, peak position, and seed light pulse width of the waveform. c The initial value can be set to the same pulse width parameter value as the seed light : ; The search range for the fitting parameters is: .

[0011] In the preferred embodiment, the specific steps of step S8 are: S81. Use LM algorithm to fit the aliased waveform curve: According to the function model of the aliased waveform , variable initial values, variable search range, use the (LM) Levenberg-Marquardt algorithm for curve fitting, and calculate the optimal estimate of the parameters under the current state.

[0012] In the preferred embodiment, the specific steps of step S9 are: S91, verify the fitting accuracy: substitute the optimal estimated value of the parameter obtained by the fitting algorithm described in step 81 into the aliasing waveform function model , calculate the fitted waveform sequence: ; The original aliased waveform sequence Perform difference calculation to obtain the deviation between the fitting function and the original waveform sequence : ; like If the predetermined index is not reached, return to step 6 and increase the number of pulse functions to be fitted, that is: ; Repeat steps S6-S9; like If the predetermined index is reached, the fitting result is output, and the distance measurement and reflection intensity information of the reflecting target can be determined based on the multiple pulse waveforms obtained by decomposition.

[0013] In the preferred solution, the specific steps of using the LM algorithm in step S8 are as follows: According to the aliasing waveform function model, variable initial values ​​(and variable search range) determined in the previous step S71, the initial parameter vector is set , the number is determined according to the number of pulse waveforms initially determined; For a given waveform sampling point sequence And the corresponding sampling point time series , according to the current parameter vector Calculate model predictions ; Calculate the residual vector ,in Indicates the sampling point number; Compute the Jacobian matrix , whose elements is the model function about The parameters in The partial derivatives at the sampling points; For the basic model of the waveform decomposition function in the present invention, respectively calculate the The partial derivative of Calculate the gradient vector ,in is the transpose of the Jacobian matrix; Compute the approximate Hessian matrix ; Calculate the parameter update amount according to the LM algorithm formula : First calculate the damping factor ; calculate ,in is the identity matrix; Then calculate ; Update parameter vector ; Calculate the updated residual vector and the new objective function value; Compare the new objective function value with the objective function value of the previous iteration. If the convergence condition is met, the iteration is stopped and the current parameter vector This is the optimal estimate of the parameters; Otherwise, return to the above steps to continue iterating. In subsequent iterations, the damping factor may need to be adjusted according to the situation. .

[0014] The present invention provides a laser radar aliasing waveform decomposition method based on waveform features. According to the seed light features of the laser radar, a waveform mathematical model is constructed. Combined with the LM curve fitting algorithm, the aliasing echo is decomposed into multiple single echo data, so as to achieve the purpose of accurately extracting multi-target ranging information.

[0015] The technical basis of the present invention is the laser radar full waveform detection technology, that is, the laser radar detection system can sample the analog electrical signal converted from the optical signal into a digital waveform signal. The processing object of the present invention is the waveform sampling point sequence of the seed light and the return light detected by the laser radar.

[0016] By constructing a waveform mathematical model and combining it with the LM curve fitting algorithm, complex aliased echoes can be decomposed into multiple single echo data. When faced with special echo signals such as multiple echoes and aliased echoes generated in long-range and complex target scenarios, multi-target ranging information can be effectively extracted. For example, in actual surveying and mapping scenarios, echo signals reflected by multiple targets are often superimposed together. This method can accurately separate the echoes of each target, thereby accurately obtaining the distance of each target and avoiding ranging errors caused by aliasing.

[0017] The complex morphology of the laser radar return signal is affected by many factors, which will cause noise in the point cloud data. This decomposition method can effectively process such signals and reduce the negative impact of noise on measurement accuracy. For example, in a complex environment, the noise signal generated by factors such as ambient light interference and changes in target reflection characteristics can be processed by this method to more accurately reflect the true position of the target and improve the overall measurement accuracy.

[0018] When some useful signals are submerged by aliased echoes or shaped echoes, the target detection will be incomplete. This technology can decompose the aliased waveforms and restore the useful signals that were originally submerged, so as to detect the target completely and improve the LiDAR's ability to detect targets in complex scenes, such as obtaining target information more comprehensively and accurately when penetrating complex obstacles.

[0019] By calculating the noise floor value of the waveform sequence, the dynamic changes in the noise floor caused by factors such as the influence of the ambient light source and the temperature characteristics of the hardware system itself are taken into account, so that the noise floor value can be adjusted in real time with the waveform data. This helps to more accurately identify effective signals under different environmental conditions and further improve the accuracy of calculations such as waveform decomposition and ranging. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a schematic diagram of the laser radar waveform sampling sequence of the present invention; Figure 2 is a schematic diagram of the laser radar waveform noise floor of the present invention; Figure 3 Schematic diagram of a typical laser radar seed light pulse waveform of the present invention; Figure 4 This is a schematic diagram of the initial determination of the number of aliased waveforms of the present invention; Figure 5 Schematic diagram of selecting initial values ​​of parameters to be fitted for the aliased waveform of the present invention; DETAILED DESCRIPTION Example 1 like Figures 1 to 5 As shown, a laser radar aliasing waveform decomposition method based on waveform characteristics is constructed according to the seed light characteristics of the laser radar. The waveform mathematical model is combined with the LM curve fitting algorithm to decompose the aliasing echo into multiple single echo data to achieve the purpose of accurately extracting multi-target ranging information. The method includes: S1. Obtain waveform sampling sequence: Obtain the waveform sampling sequence and sampling point time sequence of the laser radar seed light / return light after AD sampling, and the time interval between adjacent sampling points is a known fixed value; S2. Calculate the background noise value: Calculate the background noise value based on the signal in a certain interval around the waveform data, and take the average signal in the noise interval as the background noise value, so that it changes dynamically with the waveform data; S3, waveform data smoothing: use mean filtering and other algorithms to smooth the original waveform data; S4. Determine the valid waveform range: take the smoothed waveform data, the point where the value is greater than the background noise value and the value increases for n consecutive times as the starting point, and the point where the value decreases for n consecutive times as the ending point, where n is related to the characteristics of the laser radar, and the waveform sequence involved in the subsequent calculation is the valid waveform range; S5. Setting the basic function model of waveform decomposition: The single pulse waveform function model is determined by the seed light emission characteristics and the response characteristics of the laser radar photoelectric detection system. The practical model of the present invention is in a specific form, and the parameters to be fitted characterize the pulse waveform intensity, peak position, and pulse width; S6. Initial determination of the number of pulse waveforms: the number of starting waveforms is determined according to the pulse waveform sequence, and the aliasing waveform function is set to the superposition of the corresponding number of basic functions; S7, setting the initial value of the fitting parameter and the search range: determine the initial value of the variable according to the waveform peak value, position, and seed light pulse width, and set the initial value of the pulse width parameter and the search range of the fitting parameter; S8. Fitting using LM algorithm: Use LM algorithm to perform curve fitting and calculate the optimal estimate of parameters; S9, verify fitting accuracy: substitute the fitting parameters into the model to calculate the fitting waveform sequence, and make a difference calculation deviation with the original waveform sequence. If the predetermined index is not reached, return to step S6 to increase the number of pulse functions to be fitted. If the index is reached, output the fitting result and determine the distance measurement and reflection intensity information of the reflecting target.

[0021] Example 2 Further illustrate with reference to Example 1, Figure 1-5 The structure shown, the specific steps of steps S1-S2 are: S11, let a pulse signal be sampled by AD and the output data sequence be: ; The corresponding sampling point time series is: ; The time interval between adjacent sampling points is the system AD sampling period , is a known fixed value:

[0022] as follows Figure 1 The figure shows a schematic diagram of the waveform sequence obtained after the laser radar samples a typical echo signal.

[0023] S21. Calculate the noise floor value corresponding to the waveform data based on the signal value within a certain interval around the waveform data. Although for a fixed laser radar detection system, the noise floor of the echo signal is generally at a fixed level, the noise floor of the waveform sequence is generally dynamically changing due to the influence of factors such as the ambient light source and the temperature characteristics of the hardware system itself. The noise floor calculation method described in the present invention can be used to estimate the noise floor of the current waveform data in real time, so that the noise floor value changes dynamically with the waveform data.

[0024] as follows Figure 2 The figure shows a waveform data with background noise. The signal within a certain range before and after the waveform interval is taken as the noise interval, and the mean value of the signal in the noise interval is taken as the background noise value, which is recorded as .

[0025] Example 3 Further illustrate with reference to Example 1, Figure 1-5 The structure shown, the specific steps of steps S3-S4 are: S31, waveform data smoothing: Smoothing the original waveform data. The smoothing method may adopt filtering algorithms such as mean filtering. The waveform data sequence after smoothing is: ; S41. Determine the effective waveform range: The effective waveform is a data sequence involved in the subsequent waveform decomposition calculation. The determination method is as follows: take the smoothed waveform data, and the point whose value is greater than the background noise value and has n consecutive values ​​increasing is the starting point of the effective waveform; the point whose value is greater than the background noise value and has n consecutive values ​​decreasing is the ending point of the effective waveform; the value of n is related to the characteristics of the laser radar, and the present invention is described by taking n=4 as an example in the claims description; Starting point k Sure: ; End point j Sure: ; The waveform sequence involved in subsequent calculations is: .

[0026] In the preferred embodiment, the specific steps of step S5 are: S51. Setting a waveform decomposition basic function model: According to the ranging principle of pulse laser radar, the seed light pulse waveform still maintains the waveform characteristics of the seed light after being reflected by the target. However, due to the influence of the reflection of the adjacent target, the return light received by the laser radar is usually a superposition of multiple pulse signals. The waveform decomposition is to extract the multiple superimposed pulse waveforms, and then obtain the ranging and intensity information of multiple adjacent targets. The waveform function model of a single pulse is the basic function model of waveform decomposition, which can be determined by the emission characteristics of the seed light and the response characteristics of the laser radar photoelectric detection system. The Gaussian function is an ideal laser radar seed light pulse model and a symmetrical function model. The actual laser radar seed light pulse waveform is affected by the laser characteristics and the laser radar photoelectric response characteristics, and is deformed on the basis of the Gaussian model. The rising edge and falling edge of the waveform are asymmetric. As shown in the figure below, the rising edge of the pulse waveform is steeper and the falling edge is gentler.

[0027] The basic model of the waveform decomposition function used in the present invention is: .

[0028] Example 3 Further illustrate with reference to Example 1, Figure 1-5 The structure shown, steps S3-S4 are specifically steps 4 and step S6 is specifically steps: S61. Aliasing waveforms are usually formed by the superposition of multiple basic pulse waveforms. Determine the number of starting waveforms to facilitate rapid convergence of waveform fitting calculations, as follows Figure 4The waveform sequence shown can quickly detect three local peaks, so the initial number of waveforms is set to, and the aliased waveform function to be fitted is the superposition of three basic functions, let: ; The initial form of the aliasing waveform function is: .

[0029] In the preferred embodiment, the specific steps of step S7 are: S71, setting the initial value and search range of the waveform decomposition fitting parameter: As described in step S61, if the number of pulse waveforms initially determined is N, it is necessary to determine the initial values ​​and search ranges of multiple variables; if the number of pulse waveforms initially determined is 3, it is necessary to determine the initial values ​​and search ranges of 9 variables: the formula is expressed as: ; The initial value of the variable is determined by the peak value, peak position, and seed light pulse width of the waveform. As described in step S6, the initial value of the variable of the aliased waveform is selected as follows: Figure 5 As shown: The initial value of the variable is determined by the peak value, peak position, and seed light pulse width of the waveform. c The initial value can be set to the same pulse width parameter value as the seed light : ; The search range for the fitting parameters is: .

[0030] In the preferred embodiment, the specific steps of step S8 are: S81. Use LM algorithm to fit the aliased waveform curve: According to the function model of the aliased waveform , variable initial values, variable search range, use the (LM) Levenberg-Marquardt algorithm for curve fitting, and calculate the optimal estimate of the parameters under the current state.

[0031] In the preferred embodiment, the specific steps of step S9 are: S91, verify the fitting accuracy: substitute the optimal estimated value of the parameter obtained by the fitting algorithm described in step 81 into the aliasing waveform function model , calculate the fitted waveform sequence: ; The original aliased waveform sequence Perform difference calculation to obtain the deviation between the fitting function and the original waveform sequence : ; like If the predetermined index is not reached, return to step 6 and increase the number of pulse functions to be fitted, that is: ; Repeat steps S6-S9; like If the predetermined index is reached, the fitting result is output, and the distance measurement and reflection intensity information of the reflecting target can be determined based on the multiple pulse waveforms obtained by decomposition.

[0032] In the preferred solution, the specific steps of using the LM algorithm in step S8 are as follows: According to the aliasing waveform function model, variable initial values ​​(and variable search range) determined in the previous step S71, the initial parameter vector is set , the number is determined according to the number of pulse waveforms initially determined; For a given waveform sampling point sequence And the corresponding sampling point time series , according to the current parameter vector Calculate model predictions ; Calculate the residual vector ,in Indicates the sampling point number; Compute the Jacobian matrix , whose elements is the model function about The parameters in The partial derivatives at the sampling points; For the basic model of the waveform decomposition function in the present invention, respectively calculate the The partial derivative of Calculate the gradient vector ,in is the transpose of the Jacobian matrix; Compute the approximate Hessian matrix ; Calculate the parameter update amount according to the LM algorithm formula : First calculate the damping factor ; calculate ,in is the identity matrix; Then calculate ; Update parameter vector ; Calculate the updated residual vector and the new objective function value; Compare the new objective function value with the objective function value of the previous iteration. If the convergence condition is met, the iteration is stopped and the current parameter vector This is the optimal estimate of the parameters; Otherwise, return to the above steps to continue iterating. In subsequent iterations, the damping factor may need to be adjusted according to the situation. Through the above steps, the optimization is iterated continuously, and finally the optimal estimated value of the parameter that meets the conditions is calculated to achieve the curve fitting of the aliased waveform.

[0033] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A laser radar aliasing waveform decomposition method based on waveform features, characterized by: According to the seed light characteristics of the lidar, a waveform mathematical model is constructed, and combined with the LM curve fitting algorithm, the aliased echo is decomposed into multiple single echo data to achieve the purpose of accurately extracting multi-target ranging information. The method includes: S1. Obtain waveform sampling sequence: Obtain the waveform sampling sequence and sampling point time sequence of the laser radar seed light / return light after AD sampling, and the time interval between adjacent sampling points is a known fixed value; S2. Calculate the background noise value: Calculate the background noise value based on the signal in a certain interval around the waveform data, and take the average signal in the noise interval as the background noise value, so that it changes dynamically with the waveform data; S3, waveform data smoothing: use mean filtering and other algorithms to smooth the original waveform data; S4. Determine the valid waveform range: take the smoothed waveform data, the point where the value is greater than the background noise value and the value increases for n consecutive times as the starting point, and the point where the value decreases for n consecutive times as the ending point, where n is related to the characteristics of the laser radar, and the waveform sequence involved in the subsequent calculation is the valid waveform range; S5. Setting the basic function model of waveform decomposition: The single pulse waveform function model is determined by the seed light emission characteristics and the response characteristics of the laser radar photoelectric detection system. The practical model of the present invention is in a specific form, and the parameters to be fitted characterize the pulse waveform intensity, peak position, and pulse width; S6. Initial determination of the number of pulse waveforms: the number of starting waveforms is determined according to the pulse waveform sequence, and the aliasing waveform function is set to the superposition of the corresponding number of basic functions; S7, setting the initial value of the fitting parameter and the search range: determine the initial value of the variable according to the waveform peak value, position, and seed light pulse width, and set the initial value of the pulse width parameter and the search range of the fitting parameter; S8. Fitting using LM algorithm: Use LM algorithm to perform curve fitting and calculate the optimal estimate of parameters; S9, verify fitting accuracy: substitute the fitting parameters into the model to calculate the fitting waveform sequence, and make a difference calculation deviation with the original waveform sequence. If the predetermined index is not reached, return to step S6 to increase the number of pulse functions to be fitted. If the index is reached, output the fitting result and determine the distance measurement and reflection intensity information of the reflecting target.

2. According to the method for decomposing laser radar aliased waveforms based on waveform features as described in claim 1, it is characterized by: step The specific steps of S1-S2 are: S11, let a pulse signal be sampled by AD and the output data sequence be: ； The corresponding sampling point time series is: ； The time interval between adjacent sampling points is the system AD sampling period , is a known fixed value: ； S21. Calculate the noise floor value corresponding to the waveform data based on the signal value within a certain interval around the waveform data. Although for a fixed laser radar detection system, the noise floor of the echo signal is generally at a fixed level, the noise floor of the waveform sequence is generally dynamically changing due to the influence of factors such as the ambient light source and the temperature characteristics of the hardware system itself. The noise floor calculation method described in the present invention can be used to estimate the noise floor of the current waveform data in real time, so that the noise floor value changes dynamically with the waveform data.

3. According to the method for decomposing laser radar aliased waveforms based on waveform features as claimed in claim 1, it is characterized by: The specific steps of steps S3-S4 are: S31, waveform data smoothing: Smoothing the original waveform data. The smoothing method may adopt filtering algorithms such as mean filtering. The waveform data sequence after smoothing is: ； S41. Determine the effective waveform range: The effective waveform is a data sequence involved in the subsequent waveform decomposition calculation. The determination method is as follows: take the smoothed waveform data, and the point whose value is greater than the background noise value and has n consecutive values ​​increasing is the starting point of the effective waveform; the point whose value is greater than the background noise value and has n consecutive values ​​decreasing is the ending point of the effective waveform; the value of n is related to the characteristics of the laser radar, and the present invention is described by taking n=4 as an example in the claims description; Starting point k Sure: ； End point j Sure: ； The waveform sequence involved in subsequent calculations is: 。 4. The laser radar aliasing waveform decomposition method based on waveform features according to claim 1, characterized in that: The specific steps of step S5 are: S51. Setting a waveform decomposition basic function model: According to the ranging principle of pulse laser radar, the seed light pulse waveform still maintains the waveform characteristics of the seed light after being reflected by the target. However, due to the influence of the reflection of the adjacent target, the return light received by the laser radar is usually a superposition of multiple pulse signals. The waveform decomposition is to extract the multiple superimposed pulse waveforms, and then obtain the ranging and intensity information of multiple adjacent targets. The waveform function model of a single pulse is the basic function model of waveform decomposition, which can be determined by the emission characteristics of the seed light and the response characteristics of the laser radar photoelectric detection system. The Gaussian function is an ideal laser radar seed light pulse model and a symmetrical function model. The actual laser radar seed light pulse waveform is affected by the laser characteristics and the laser radar photoelectric response characteristics, and is deformed on the basis of the Gaussian model, and the rising and falling edges of the waveform are asymmetric. The basic model of waveform decomposition function is: 。 5. The laser radar aliasing waveform decomposition method based on waveform features according to claim 1, characterized in that: The specific steps of step S6 are: S61. Initial determination of the number of pulse waveforms: The aliased waveform is usually formed by the superposition of multiple basic pulse waveforms. Determine the number of starting waveforms to facilitate rapid convergence of waveform fitting calculation. The aliased waveform function to be fitted is the superposition of N basic functions. Let: ； The initial form of the aliasing waveform function is: 。 6. The laser radar aliasing waveform decomposition method based on waveform features according to claim 5 is characterized by: The specific steps of step S7 are: S71, setting the initial value and search range of the waveform decomposition fitting parameter: As described in step S61, if the number of pulse waveforms initially determined is N, it is necessary to determine the initial values ​​and search ranges of multiple variables, which are expressed as follows: ； The initial value of the variable is determined by the peak value, peak position, and seed light pulse width of the waveform. c The initial value can be set to the same pulse width parameter value as the seed light : ； The search range for the fitting parameters is: 。 7. The laser radar aliasing waveform decomposition method based on waveform features according to claim 1, characterized in that: The specific steps of step S8 are: S81. Use LM algorithm to fit the aliased waveform curve: According to the function model of the aliased waveform , variable initial values, variable search range, use the (LM) Levenberg-Marquardt algorithm for curve fitting, and calculate the optimal estimate of the parameters under the current state.

8. The laser radar aliasing waveform decomposition method based on waveform features according to claim 7, characterized in that: The specific steps of step S9 are: S91, verify the fitting accuracy: substitute the optimal estimated value of the parameter obtained by the fitting algorithm described in step 81 into the aliasing waveform function model , calculate the fitted waveform sequence: ； The original aliased waveform sequence Perform difference calculation to obtain the deviation between the fitting function and the original waveform sequence : ； like If the predetermined index is not reached, return to step 6 and increase the number of pulse functions to be fitted, that is: ； Repeat steps S6-S9; like If the predetermined index is reached, the fitting result is output, and the distance measurement and reflection intensity information of the reflecting target can be determined based on the multiple pulse waveforms obtained by decomposition.

9. The laser radar aliasing waveform decomposition method based on waveform features according to claim 7, characterized in that: Step S8 uses the LM algorithm. The specific steps are: According to the aliasing waveform function model, variable initial values ​​(and variable search range) determined in the previous step S71, the initial parameter vector is set , the number is determined according to the number of pulse waveforms initially determined; For a given waveform sampling point sequence And the corresponding sampling point time series , according to the current parameter vector Calculate model predictions ; Calculate the residual vector ,in Indicates the sampling point number; Compute the Jacobian matrix , whose elements is the model function about The parameters in The partial derivatives at the sampling points; For the basic model of the waveform decomposition function in the present invention, respectively calculate the The partial derivative of Calculate the gradient vector ,in is the transpose of the Jacobian matrix; Compute the approximate Hessian matrix ; Calculate the parameter update amount according to the LM algorithm formula : First calculate the damping factor ; calculate ,in is the identity matrix; Then calculate ; Update parameter vector ; Calculate the updated residual vector and the new objective function value; Compare the new objective function value with the objective function value of the previous iteration. If the convergence condition is met, the iteration is stopped and the current parameter vector This is the optimal estimate of the parameters; Otherwise, return to the above steps to continue iterating. In subsequent iterations, the damping factor may need to be adjusted according to the situation. .

Images