A self - calibration method for the non - linearity of a lidar detector

Abstract

The present invention discloses a self - calibration method for the non - linearity of a lidar detector, comprising the following steps: (1) Collect lidar echo signal data, control the laser to emit pulse energy to increase step by step from 0% to 100%, set multiple energy levels, and collect atmospheric echo signals at different distances at each energy level to obtain an observation signal matrix; (2) Establish a theoretical signal model of lidar atmospheric echo and define a non - linear response model to deduct background noise; (3) Perform singular value decomposition on the observation signal matrix, extract the principal components to construct a rank - one approximation matrix; (4) Calibrate the non - linear response curve; The present invention does not require prior calibration of laser energy or atmospheric parameters, only depends on the characteristics of the echo signal itself, forms an intensity gradient naturally based on distance - energy scanning, and synchronously calibrates the linear region and the saturation region; effectively improving the measurement accuracy and stability of the lidar system.

Description

Technical Field

[0001] The present invention relates to the technical field of lidar, and particularly to a self - calibration method for the nonlinearity of a lidar detector. Background Technique

[0002] Lidar needs to process echo signals with a span of several orders of magnitude in fields such as atmospheric aerosol detection. The detector needs to maintain a linear response for weak signals in the far field, while nonlinear distortion is likely to occur due to saturation under strong signals in the near field. Traditional single - point calibration methods usually rely on off - line calibration in a laboratory environment, which requires complex calibration equipment and a precise reference signal with known intensity, and it is difficult to cover the full dynamic range. Existing atmospheric detection lidar technologies rely on accurately calibrating laser energy or atmospheric parameters (such as backscattering coefficient) when retrieving atmospheric parameters. In practical applications, the calibrated parameters become inaccurate due to environmental changes and system aging. When calibrating, it is usually required to disassemble the detector from the lidar system for calibration, which greatly increases the maintenance cost and affects data continuity. At the same time, preset polynomial or exponential models are difficult to match the real nonlinear characteristics, and Poisson noise and circuit noise will significantly interfere with the accuracy of parameter inversion. Summary of the Invention

[0003] Object of the Invention: The object of the present invention is to provide a self - calibration method for the nonlinearity of a lidar detector. Through the joint inversion of dynamic scanning pulse energy and signals at different distances, in - situ nonlinear calibration is directly realized by using the echo signals of the lidar system itself, so as to solve the problems existing in the background technique.

[0004] Technical Solution: A self - calibration method for the nonlinearity of a lidar detector according to the present invention includes the following steps:

[0005] (1) Collect lidar echo signal data, control the laser to emit pulse energy increasing step by step from 0% to 100%, set multiple energy levels, and collect atmospheric echo signals at different distances at each energy level to obtain an observation signal matrix;

[0006] (2) Establish a theoretical signal model of lidar atmospheric echo and define a nonlinear response model to subtract background noise;

[0007] (3) Perform singular value decomposition on the observation signal matrix, extract the principal components to construct a rank - one approximation matrix;

[0008] (4) Construct a cost function, and alternately update the rank - one matrix and the response function through iterative optimization to minimize the reconstruction error, and finally calibrate the nonlinear response curve.

[0009] Further, in step (1), the energy level is set to 10 - 30 levels, and the signal acquisition time interval at each energy level is 1 - 10 seconds, ensuring that the signal dynamic range covers the linear region to the non-linear saturation region of the detector.

[0010] Further, in step (2), the lidar atmospheric echo theoretical signal model is expressed as:

[0011] ;

[0012] where, is the pulse energy, t is the signal acquisition moment corresponding to the pulse energy, is the backscattering coefficient, is the atmospheric extinction coefficient, R is the distance of the echo signal, and C is a constant term related to the system efficiency.

[0013] The non-linear response model is defined as:

[0014] ;

[0015] where, is the actual observed signal matrix, is the non-linear response function to be calibrated, is the noise term.

[0016] Further, step (3) is specifically as follows: Let be decomposed into the outer product of the distance-related backscattering attenuation term and the emission pulse energy-related term :

[0017] ;

[0018] where, P is the backscattering attenuation term represented in vector form related to the distance R; T is the emission pulse energy term represented in vector form , related to the pulse energy E; the superscript * represents the transpose symbol;

[0019] Perform singular value decomposition on the observed signal matrix , and obtain:

[0020] ;

[0021] where, is the constant coefficient obtained by decomposition, is the distance-related vector obtained by decomposition, is the pulse energy-related vector obtained by decomposition, r is the order of decomposition; m is the index corresponding to the order; Extract the principal components to construct a rank-one approximation matrix:

[0022] ;

[0023] wherein, is an estimated distance-related term, is an estimated energy-related term, is a constant term.

[0024] Furthermore, step (4) includes the following steps:

[0025] (41) Perform SVD decomposition on the observed signal to obtain an initial rank-one approximation ;

[0026] (42) Construct a set of data pairs , where i and j are valid data indices; k is the number of update iterations;

[0027] (43) Use the set of data pairs to fit the non-linear function ;

[0028] (44) Use the non-linear function to update the signal; the formula is as follows:

[0029] ;

[0030] wherein, is the inverse function of the function ;

[0031] (45) Re-perform SVD decomposition on the updated to extract a new rank-one approximation;

[0032] (46) Construct a cost function:

[0033] ;

[0034] wherein, is the observation weight, is the total variation regularization term used to suppress the oscillation of the response curve caused by noise; is the weight of the corresponding regularization term;

[0035] Repeat steps S41 - S46 in S47 for iterative optimization, and minimize the cost function J by alternately updating the rank-one matrix and the response function , and finally obtain the functional form of the non-linear response curve.

[0036] Further, step (43) is specifically as follows: The data pairs are curve-smoothly fitted by using a parameterization method with a known function form or a non-parameterization method with an unknown function form.

[0037] Further, in step (46), weights are introduced into the cost function to achieve adaptive optimization of the signal-to-noise ratio.

[0038] Further, in step (46), signals below the detection threshold are mask-labeled to dynamically exclude invalid data during iteration.

[0039] An electronic device according to the present invention includes a memory, a processor, and a computer program stored on the memory. When the processor executes the program, the steps of any one of the above methods are implemented.

[0040] A computer-readable storage medium according to the present invention stores a computer program. When the program is executed by a processor, the steps of any one of the above methods are implemented.

[0041] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: The present invention calibrates the non-linear response curve of the lidar photodetector without pre-calibrating the laser energy or atmospheric parameters, relying only on the characteristics of the echo signal itself. Using the distance-energy scan to naturally form an intensity gradient, synchronously calibrating the linear region and the saturation region. It has the advantages of fully automatic calibration and wide dynamic coverage. The present invention does not require a complex external calibration device for separate calibration. It is compatible with complex interferences such as Poisson noise and circuit noise, and supports real-time online calibration. It has the advantages of high speed, high flexibility, and high precision. The present invention is not affected by the aerosol concentration distribution, can adapt to internal and external environmental changes, can quickly and accurately calibrate the non-linear response curve, and significantly improves the measurement accuracy and stability of the lidar system. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flowchart of the present invention;

[0043] Figure 2 is a flowchart of the dynamic iterative optimization method of the present invention;

[0044] Figure 3 is an exemplary diagram of the lidar echo signal curve of the present invention;

[0045] Figure 4 is an exemplary diagram of the normalized echo signal curve of the present invention;

[0046] Figure 5 is an exemplary diagram of the result of the non-linear response curve of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0048] As Figure 1 shown, an embodiment of the present invention provides a zoom self-calibration method for the focal length of a coherent wind-measuring lidar telescope, including the following steps:

[0049] S1. Collect lidar echo signal data, set the lidar to maintain an effective collection of atmospheric echo signals in a fixed direction, control the laser pulse energy to increase step by step from 0% to 100% (usually 10 - 30 energy levels can be set), and at each set energy, continuously collect the atmospheric echo signals within a set time interval (generally 1 s can be set). The echo signals at different distances R at different times t are obtained matrix. As Figure 3 shown, the echo signals under different pulse energy setting conditions are given curve. It can be seen that the dynamic range of the lidar echo signal varies greatly with distance. The weak far-field signals cover the linear region of the detector, due to large noise fluctuations. The strong near-field signals cover the non-linear saturation region of the detector. When the pulse energy increases, the intensity of the near-field signals deviates from proportional increase due to non-linear effects.

[0050] Step S2. Physical modeling of atmospheric echo signals. The theoretical signal model of lidar atmospheric echo can be expressed as:

[0051] ;

[0052] where is the pulse energy, t is the signal acquisition time corresponding to the pulse energy, is the backscattering coefficient, is the atmospheric extinction coefficient, R is the distance of the echo signal, and C is a constant term related to the system efficiency.

[0053] The non-linear response model of the detector establishes the relationship between the observed echo signal and the true echo signal and can be expressed as:

[0054] ;

[0055] where is the actual observed signal matrix, is the non-linear response function to be calibrated, is the noise term.

[0056] Use the background noise region of the echo signal at each time t to estimate the background noise and subtract it. As Figure 4As shown, after deducting the background noise, the normalized echo signal curve obtained by normalizing the maximum pulse energy echo curve shows that the curve is flat but fluctuates greatly at long distances, while the signal-to-noise ratio is high but the curve is bent at short distances.

[0057] S3. Perform rank-one matrix decomposition (SVD) on the echo signal matrix. According to the physical modeling in step S2, it can be decomposed into the outer product of the distance-dependent backscattering attenuation term and the transmit pulse energy-related term :

[0058] ;

[0059] where P is the backscattering attenuation term represented in vector form and is related to the distance R; T is the transmit pulse energy term represented in vector form and is related to the pulse energy E; the superscript * represents the transpose symbol;

[0060] and the non-linear effect makes the rank of the observed signal matrix destroy the original rank-one structure. Therefore, perform singular value decomposition on the observed signal matrix to obtain:

[0061] ;

[0062] where is the constant coefficient obtained by decomposition, is the distance-related vector obtained by decomposition, is the pulse energy-related vector obtained by decomposition, r is the order of decomposition; m is the index of the corresponding order; extract the principal components to construct a rank-one approximation matrix:

[0063] ;

[0064] where is the estimated distance-related term, is the estimated energy-related term, is the constant term.

[0065] S4. Non-linear response curve estimation. Use the iterative optimization method to estimate the non-linear response curve. As Figure 2 shown, the specific steps include:

[0066] S41. Perform SVD decomposition on the observed signal to obtain the initial rank-one approximation ;

[0067] S42. Construct the data pair set , where i and j are valid data indices;

[0068] S43. Use the data pair set to fit a non - linear function ;

[0069] In this embodiment, a parametric method with a known function form is adopted to perform curve smoothing fitting on the data, and a non - parametric representation of is obtained.

[0070] The specific function form is defined according to the general detector non - linear curve model as:

[0071] ;

[0072] where a is the parameter to be optimized; x represents the input signal;

[0073] Or a non - parametric method with an unknown function form is adopted to perform curve smoothing fitting on the data, and a non - parametric representation of is obtained.

[0074] S44. Use the non - linear function to update the signal; the formula is as follows:

[0075] ;

[0076] where is the inverse function of the function ;

[0077] S45. Re - perform SVD decomposition on the updated to extract a new rank - one approximation;

[0078] S46. Construct a cost function:

[0079] ;

[0080] where is the observation weight, is the total variation regularization term used to suppress the oscillation of the response curve caused by noise; is the weight of the corresponding regularization term;

[0081] S47. Repeat steps S41 - S46 for iterative optimization. By alternately updating the rank - one matrix and the response function , minimize the cost function J, and finally obtain the function form of the non - linear response curve .

[0082] When the error of the cost function is less than the set allowable error, finally obtain the function form of the non - linear response curve As shown in Figure 5 the figure, an example diagram of the results of the non-linear response curve of the lidar detector is given. In the case where there is noise in the echo signal, the inversion result is in perfect agreement with the true value.

Claims

1. A nonlinear self-calibration method for a laser radar detector, characterized in that: The following steps are involved: S1 collects laser radar echo signal data, sets multiple energy levels, collects atmospheric echo signals at different distances at each energy level, and obtains an observation signal matrix; S2 establishes the theoretical signal model of the laser radar atmospheric echo and defines the nonlinear response model to deduct the background noise; the theoretical signal model of the laser radar atmospheric echo is expressed as: ； in, is the pulse energy, t is the signal acquisition time corresponding to the pulse energy, is the backscattering coefficient, is the atmospheric extinction coefficient, R is the distance of the echo signal, and C is a constant term related to the system efficiency; The nonlinear response model is defined as: ； in, is the actual observed signal matrix, is the nonlinear response function to be calibrated, is the noise term; S3 performs singular value decomposition on the observed signal matrix, extracts the principal components and constructs a rank-one approximate matrix; S4 constructs a cost function, updates the rank-one matrix and the response function alternately through iterative optimization, minimizes the reconstruction error, and finally calibrates the nonlinear response curve.

2. The nonlinear self-calibration method of a laser radar detector according to claim 1, characterized in that: In step S1, the energy level is set to 10-30 levels, and the signal acquisition time interval at each energy level is 1-10 seconds, ensuring that the signal dynamic range covers the linear region to the nonlinear saturation region of the detector.

3. The nonlinear self-calibration method of a laser radar detector according to claim 1, characterized in that: Step S3 is as follows: Decomposed into distance-dependent backscatter attenuation terms and the transmitted pulse energy The outer product of: ； Where P is the backscatter attenuation term expressed in vector form Related to the distance R; T is the transmitted pulse energy term expressed in vector form , which is related to the pulse energy E; the superscript * indicates the transposed symbol; The observed signal matrix Perform singular value decomposition and get: ； in, is the constant coefficient obtained by decomposition, To decompose the distance correlation vector, is the pulse energy correlation vector obtained by decomposition, r is the decomposition order; m is the index of the corresponding order; extract the principal component to construct a rank-one approximate matrix: ； in, is the estimated distance related term, is the estimated energy-related term, is a constant term.

4. The nonlinear self-calibration method of a laser radar detector according to claim 3, characterized in that: Step S4 includes the following steps: S41 uses observation signals Perform SVD decomposition to obtain the initial rank-one approximation ; S42 builds a data pair set , where i, j are valid data indexes; k is the number of update iterations; S43 Using data to collect , fitting nonlinear function ; S44 uses nonlinear functions , update signal; the formula is as follows: ； in, For function The inverse function of S45 for the updated Re-perform SVD decomposition and extract a new rank-one approximation; S46 constructs the cost function: ； in, is the observation weight, is the total variation regularization term, which is used to suppress the oscillation of the response curve caused by noise; is the weight of the corresponding regularization term; S47 repeats steps S41 to S46 to perform iterative optimization by alternately updating the rank-one matrix With the response function , minimize the cost function J, and finally obtain the functional form of the nonlinear response curve .

5. The nonlinear self-calibration method of a laser radar detector according to claim 4, characterized in that: Step S43 is specifically as follows: a parametric method in a known function form or a non-parametric method in an unknown function form is used to perform curve smoothing fitting on the data pair.

6. The nonlinear self-calibration method of a laser radar detector according to claim 4, characterized in that: In step S46, weights are introduced into the cost function , to achieve adaptive optimization of signal-to-noise ratio.

7. The nonlinear self-calibration method of a laser radar detector according to claim 4, characterized in that: In step S46, signals below the detection threshold are masked and invalid data are dynamically excluded during iteration.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images