Fusion method of unmanned ship laser fog measuring radar and global reanalysis data

Abstract

The invention discloses a fusion method of unmanned ship laser fog measuring radar and global reanalysis data, which belongs to the technical field of meteorological detection, and adopts a four-dimensional variational method to combine high-resolution laser fog measuring radar visibility data with low-resolution but wider-coverage numerical mode data. Therefore, more complete and more accurate three-dimensional meteorological element data can be generated; in order to solve the problems of insufficient marine observation data and difficult agglomerate fog detection, an unmanned ship is adopted to carry a laser fog measurement radar to carry out sea fog maneuvering observation, and then laser fog measurement radar PPI data obtained through maneuvering observation is converted from polar coordinates to a rectangular coordinate system. And then the converted laser fog measuring radar data are input into an LAPS mode to generate three-dimensional observation field data, and then a four-dimensional variational method is adopted to fuse the three-dimensional observation field data and global reanalysis data, so that the sea fog sensing precision can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological detection, and in particular to a method for fusing an unmanned boat laser fog detection radar with global reanalysis data. Background Art

[0002] At present, the fusion of laser fog radar and global reanalysis data from numerical models at home and abroad mainly uses fusion modes such as MM5 and GSI, among which MM5 uses genetic algorithm and GSI uses three-dimensional variational method.

[0003] Global reanalysis data is an important data for analyzing weather processes in the meteorological field, but the spatial resolution of the data is 0.25X0.25, which cannot meet the application of marine cloud and fog observation, such as the detection of fog clusters, because its scale is often less than 5km. The 0.25 resolution of global reanalysis data is converted to a distance of about 25km. Therefore, it is impossible to conduct a detailed analysis of fog clusters in sea fog using only global reanalysis data.

[0004] These fusion systems cannot integrate the data of ship / boat laser fog detection radar, that is, they cannot integrate the load data on the moving platform, which will lead to the following problems:

[0005] First, fog can cause a sudden drop in visibility, limiting the sight of ship drivers and making it difficult to maintain effective lookout.

[0006] Second: In waters covered by fog, ship drivers cannot detect surrounding ships and obstacles at a sufficient distance, making it difficult to judge the dynamics of other ships and the avoidance actions taken by other ships, which greatly increases the risk of water traffic accidents such as ship collisions and groundings.

[0007] Based on this, the present invention designs a fusion method of unmanned boat laser fog detection radar and global reanalysis data to solve the above problems. Summary of the invention

[0008] In view of the above-mentioned shortcomings of the prior art, the present invention provides a method for fusing unmanned boat laser fog detection radar with global reanalysis data.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0010] A method for fusing unmanned boat laser fog radar with global reanalysis data comprises the following steps:

[0011] Step 1: Perform dead time correction, smoothing filter and geometric overlap factor correction on the laser fog radar PPI data.

[0012] Then remove the background noise;

[0013] Step 2: Convert the polar coordinates of the laser fog radar PPI data based on the unmanned boat route into rectangular coordinate grid data;

[0014] Step 3: The laser fog radar data after coordinate conversion is input into the LAPS numerical model. The LAPS numerical model fuses and interpolates the laser fog radar data to generate three-dimensional observation field data with continuous time and space and consistent resolution;

[0015] Step 4: Use the four-dimensional variational method to fuse the three-dimensional observation field data with the global reanalysis data. By correcting the input of the model, the distance between the model output and the corresponding observation within the fusion period is minimized. A target functional is defined to represent this distance. The gradient of the target functional with respect to the control variable is calculated by forward integration of the adjoint model corresponding to the original model. Then, the minimization algorithm is used to make it reach the minimum point to complete the fusion.

[0016] Furthermore, in step 1, the calculation formula for dead time correction is:

[0017]

[0018] Where M is the real count value, N is the measured count value, and the dead time is t dead .

[0019] Furthermore, the smoothing filter is processed as follows:

[0020] The average echo value is obtained by averaging the echo signals at several adjacent distance points. At the same time, the distance corresponding to the average echo value is also the average distance of these distances. The data smoothing and filtering processing also changes the distance resolution of the fog detection radar data. The original distance resolution is Rres. After W-point smoothing, its distance resolution is W*Rres.

[0021] Furthermore, the operation mode of calculating the number of points from the mean value is 3×3 smoothing or 5×5 smoothing.

[0022] Furthermore, in step 1, the formula for geometric overlap factor correction is:

[0023]

[0024] Wherein, the geometric overlap factor is O(H), P(H, λ) is the original echo signal, P′(H, λ) is the signal corrected by the geometric overlap factor, H is the height from the radar in m, and λ is the laser emission wavelength in nm.

[0025] Furthermore, in step 1, the formula for background noise removal is:

[0026] P′(H,λ)=P(H,λ)-P bg(H, λ)

[0027] Among them, P(H, λ) is the original echo signal, P′(H, λ) is the signal after background noise subtraction, and P bg (H, λ) is the signal of background noise;

[0028] P bg The calculation formula of (H, λ) is:

[0029]

[0030] Among them, H1~H N is the background noise height, and λ is the wavelength.

[0031] Furthermore, in step 2;

[0032] The coordinates of the laser fog radar point p data (r, α, θ) can be expressed in rectangular coordinates in the rectangular coordinate system as (x p ,y p , z p ):

[0033] x p =rcos(θ)cos(α);

[0034] y p =rcos(θ)sin(α);

[0035] z p =rsin(θ);

[0036] The longitude and latitude of the current unmanned boat are (Lon_p, Lat_p), and the longitude and latitude grid coordinates of the rectangular coordinates of the laser fog radar data It is expressed as:

[0037]

[0038] Furthermore, in step 4, the calculation formula of the objective functional (i.e., value function) of the optimal interpolation fusion method is:

[0039]

[0040] gradient:

[0041] Among them, J G represents the value function of global reanalysis data, J L It represents the three-dimensional observation field output by the laser radar after being sent into the LAPS mode. The subscript i represents any time, distributed in the time period [0, t N ], X is the control vector, x Gis the state vector of global reanalysis data, G is the error covariance matrix of global reanalysis data, R i is the observation error covariance matrix, H i is the observation operator, y i is the observation forecast vector at time i, x i is the state prediction vector at time i, and T is used to represent the operation of exchanging the rows and columns of the matrix.

[0042] Preferably, the state prediction vector x i Satisfies the mode control equation:

[0043]

[0044] Among them, M 0→i (x) is the model prediction operator from the starting time 0 to the time i.

[0045] Preferably, the four-dimensional variational method for fusing three-dimensional observation field data with global reanalysis data comprises the following steps:

[0046] S1. Determine the initial guess value of each control variable in the LAPS model And the model is forward integrated in the fusion window to obtain the

[0047] S2. Substitute the conjugate equation of the atmospheric motion state equation and reversely integrate the corresponding adjoint mode equation;

[0048] S3. Calculate the value function according to the above objective functional formula and find its partial derivative with respect to each control variable;

[0049] S4. Use the gradient descent algorithm to determine the descent direction of the value function:

[0050]

[0051] Where β is the optimal step size, n is the number of iterations, j, k = 1, 2, ..., n, is the optimal descent direction function;

[0052] S5, when J n -J n-1 ≤δ, output That is, fusion is completed;

[0053] Among them, J n The value function of the nth iteration, J n-1 The value function of the n-1th iteration, δ represents a very small value.

[0054] Compared with the prior art, the present invention has the following beneficial effects: 1. High-precision perception: The laser fog radar provides real-time visibility data with a distance resolution of 30 meters. After being integrated with the global reanalysis data, it can significantly improve the perception accuracy of sea fog;

[0055] 2. Global coverage: Global reanalysis data has the characteristics of global coverage, which can make up for the shortcomings of laser fog radar in spatial coverage;

[0056] 3. Dynamic adaptability: The mobile fusion algorithm is based on the maneuvering navigation of the unmanned boat. The PPI data of the laser fog radar placed on the unmanned boat is dynamically and real-timely interpolated with the global reanalysis data in space and time, so as to obtain a high-resolution horizontal visibility of a certain area, such as 500 meters;

[0057] 4. The present invention uses a four-dimensional variational method to combine high-resolution laser fog radar visibility data with low-resolution but more widely covered numerical model data to generate a more complete and accurate three-dimensional meteorological element data;

[0058] 5. To address the problems of insufficient observation data at sea and difficulty in detecting fog clusters, the present invention uses an unmanned boat equipped with a laser fog radar to conduct mobile observation of sea fog, then converts the PPI data of the laser fog radar obtained from the mobile observation into a polar coordinate system to a rectangular coordinate system, then inputs the converted laser fog radar data into the LAPS mode to generate three-dimensional observation field data, and then uses the four-dimensional variational method to fuse the three-dimensional observation field data with the global reanalysis data. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0060] Figure 1 The present invention is a logic block diagram of a method for fusing an unmanned boat laser fog detection radar with global reanalysis data. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Embodiment 1: In some embodiments, please refer to the drawings of the specification Figure 1 ,A method for fusing unmanned boat laser fog radar with global reanalysis data, comprising the following steps;

[0063] Step 1: Perform quality control such as dead time correction, smoothing filtering and geometric overlap factor correction on the laser fog radar PPI (Plan Position Indicator) data, and then remove the background noise;

[0064] Preferably, the working principle of PPI is to obtain the position information of the target through the rotation scanning of the antenna.

[0065] Preferably, the laser fog detection radar is composed of a laser emission system, an optical receiving system, a photoelectric conversion system, a signal processing system, a data acquisition system, etc. In the avalanche diode in the photoelectric conversion link, the upper limit of the photon counting mode count rate depends on the pulse pair resolution, that is, the minimum time interval at which each pulse can be separated. The reciprocal of the pulse pair resolution is the maximum count value. However, since most events in the light counting area usually occur randomly, the counting pulses may overlap. Therefore, the actual maximum count value is about one tenth of the above calculated value.

[0066] The dead time correction is calculated as:

[0067]

[0068] Where M is the real count value, N is the measured count value, and the dead time is t dead (Usually provided by the manufacturer).

[0069] Preferably, since the atmospheric state changes at any time and is affected by the random error of the detection equipment, the original signal output by the laser fog radar acquisition card has jitter phenomenon, and the echo signal can be smoothed and filtered to reduce the signal jitter;

[0070] Data smoothing processing is to average the echo signals at several adjacent distance points to obtain the average echo value. At the same time, the distance corresponding to the average echo value is also the average of these distances. The number of points for calculating the average value depends on the actual signal quality. Usually 3×3 smoothing, 5×5 smoothing, etc. are performed. 3×3 smoothing and 5×5 smoothing usually refer to the operation method of smoothing data in the fields of image processing and data processing. The 3×3 and 5×5 here represent the filter, also known as the size of the convolution kernel.

[0071] Data smoothing also changes the distance resolution of fog radar data. For example, if the original distance resolution is Rres, after W-point smoothing, the distance resolution becomes W*Rres.

[0072] Preferably, in the blind zone, the beam emitted by the laser is usually not in the receiving field of the telescope; in the transition zone, the atmospheric echo signal cannot be completely received by the telescope, resulting in the received echo signal not being consistent with the actual echo signal. The ratio of the number of photons received by the telescope to the actual number of photons is called the geometric overlap factor. This situation needs to be corrected during data processing, that is, the geometric overlap factor correction;

[0073] The geometric overlap factor is a parameter that approaches 1 as the distance increases. It is generally provided by the manufacturer. If the geometric overlap factor is O(H), the corrected signal is:

[0074]

[0075] Wherein, P(H, λ) is the original echo signal, P′(H, λ) is the signal corrected by the geometric overlap factor, H is the height from the radar in meters, and λ is the laser emission wavelength in nanometers.

[0076] Preferably, the data measured by the laser fog radar contains not only signals such as aerosols and clouds, but also sky background noise signals. According to the vertical distribution characteristics of the atmosphere, the high-altitude atmosphere is generally clean and contains almost no aerosols and clouds. Therefore, the high-altitude atmospheric echo signal is generally used to remove the background noise, and the average value of the high-altitude 1-2km signal is selected as the background noise, and the real atmospheric echo signal is subtracted from the original echo;

[0077] For example, if the maximum range of the laser fog radar is 10km, the average value of the signal within any 2km within 10km can be selected as the background signal for elimination. The calculation method is shown in the following formula:

[0078] P′(H,λ)=P(H,λ)-P bg (H, λ)

[0079] Among them, P(H, λ) is the original echo signal, P′(H, λ) is the signal after background noise subtraction, and P bg (H, λ) is the signal of background noise;

[0080] P bg The calculation formula of (H, λ) is:

[0081]

[0082] Among them, H1~H N is the background noise height, and λ is the wavelength.

[0083] Step 2: Convert the polar coordinates of the laser fog radar PPI data based on the unmanned boat route into rectangular coordinate grid data;

[0084] Preferably, the coordinates (r, α, θ) of the laser fog radar point p can be expressed in rectangular coordinates in a rectangular coordinate system as (x p ,y p , z p ):

[0085] x p =rcos(θ)cos(α);

[0086] y p =rcos(θ)sin(α);

[0087] z p =rsin(θ);

[0088] The longitude and latitude of the current unmanned boat are (Lon_p, Lat_p), and the longitude and latitude grid coordinates of the rectangular coordinates of the laser fog radar data It is expressed as:

[0089]

[0090] Step 3: The laser fog radar data after coordinate conversion is input into the LAPS numerical model (Local Analysis and Prediction System). The LAPS numerical model fuses and interpolates the laser fog radar data to generate three-dimensional observation field data with continuous time and space and consistent resolution.

[0091] Step 4: Use the four-dimensional variational method to fuse the three-dimensional observation field data with the global reanalysis data (ERA5, MERRA2, NCEPNCAR Reanlysis1, JRA55).

[0092] The basic principle of fusion is to minimize the distance between the model output (or some of their functions) and the corresponding observation within the fusion period by correcting the model input, and define a target functional to represent this distance. This constitutes a functional minimization problem under constraints (control equations). The gradient of the target functional with respect to the control variable can be calculated by forward integration of the adjoint model corresponding to the original model (after linearization), and then a minimization algorithm (such as gradient descent algorithm, quasi-Newton method) is used to make it reach the minimum point.

[0093] Preferably, the calculation formula of the objective functional (i.e., the value function) of the optimal interpolation fusion method is:

[0094]

[0095] gradient:

[0096] Among them, JG represents the value function of global reanalysis data, J L It represents the three-dimensional observation field output by the laser radar after being sent into the LAPS mode. The subscript i represents any time, distributed in the time period [0, t N ], t N The final value is the result of the experiment, that is, it is determined by the actual number of iterations;

[0097] X is the control vector, x G is the state vector of global reanalysis data, G is the error covariance matrix of global reanalysis data, R i is the observation error covariance matrix, H i is the observation operator, y i is the observation forecast vector at time i, x i is the state prediction vector at time i, and T is used to represent the operation of exchanging the rows and columns of the matrix.

[0098] Preferably, the state prediction vector x i Satisfies the mode control equation:

[0099]

[0100] Among them, M 0→i (x) is the model prediction operator from the starting time 0 to the time i.

[0101] Preferably, the four-dimensional variational method for fusing three-dimensional observation field data with global reanalysis data comprises the following steps:

[0102] S1. Determine the initial guess value of each control variable in the LAPS model And the model is forward integrated in the fusion window to obtain the

[0103] S2. Substitute the conjugate equation of the atmospheric motion state equation and reversely integrate the corresponding adjoint mode equation;

[0104] S3. Calculate the value function according to the above objective functional formula and find its partial derivative with respect to each control variable;

[0105] S4. Use the gradient descent algorithm to determine the descent direction of the value function:

[0106]

[0107] Where β is the optimal step size, n is the number of iterations, j, k = 1, 2, ..., n, is the optimal descent direction function;

[0108] S5, when J n -Jn-1 ≤δ, output That is, fusion is completed;

[0109] Among them, J n The value function of the nth iteration, J n-1 The value function of the n-1th iteration, δ represents a very small value obtained by experiment.

[0110] The four-dimensional variational method is mainly used to study multi-dimensional problems. It solves the problems by converting them into minimization or maximization problems of functions. The four-dimensional variational algorithm (4DVar) is developed based on the three-dimensional variational fusion method. It considers the three-dimensional distribution of the state variable space and the one-dimensional time distribution. This method adjusts the analysis time interval of the model and the update time of the observation data to the same frequency, and compared with the three-dimensional variational algorithm (3DVar), the four-dimensional variational algorithm considers the changes of the background field over time more, so it can be applied to complex nonlinear models.

[0111] The four-dimensional variational method (4D-Var) is used in meteorological data fusion. It can fuse observational data from different sources with the numerical model background field to produce the best estimate.

[0112] This application uses a four-dimensional variational method to combine high-resolution laser fog radar visibility data with low-resolution but more comprehensive numerical model data to generate a more complete and accurate three-dimensional meteorological element data.

[0113] In order to solve the problems of insufficient observation data at sea and difficulty in detecting fog clusters, the present invention adopts an unmanned boat equipped with a laser fog radar to conduct mobile observation of sea fog, then converts the PPI data of the laser fog radar obtained by the mobile observation into a rectangular coordinate system, and then inputs the converted laser fog radar data into the LAPS mode to generate three-dimensional observation field data, and then adopts a four-dimensional variational method to fuse the three-dimensional observation field data with the global reanalysis data, which can significantly improve the perception accuracy of sea fog.

[0114] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fusing unmanned boat laser fog radar with global reanalysis data, characterized in that: The steps include: Step 1: Perform dead time correction, smoothing filter and geometric overlap factor correction on the laser fog radar PPI data, and then remove the background noise; Step 2: Convert the polar coordinates of the laser fog radar PPI data based on the unmanned boat route into rectangular coordinate grid data; Step 3: The laser fog radar data after coordinate conversion is input into the LAPS numerical model. The LAPS numerical model fuses and interpolates the laser fog radar data to generate three-dimensional observation field data with continuous time and space and consistent resolution; Step 4: Use the four-dimensional variational method to fuse the three-dimensional observation field data with the global reanalysis data. By correcting the input of the model, the distance between the model output and the corresponding observation within the fusion period is minimized. A target functional is defined to represent this distance. The gradient of the target functional with respect to the control variable is calculated by forward integration of the adjoint model corresponding to the original model. Then, the minimization algorithm is used to make it reach the minimum point to complete the fusion.

2. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 1 is characterized in that: In step 1, the dead time correction calculation formula is: Where M is the real count value, N is the measured count value, and the dead time is t dead .

3. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 1 is characterized in that: In step 1, the smoothing filter is processed as follows: The average echo value is obtained by averaging the echo signals at several adjacent distance points. At the same time, the distance corresponding to the average echo value is also the average distance of these distances. The data smoothing and filtering processing also changes the distance resolution of the fog detection radar data. The original distance resolution is Rres. After W-point smoothing, its distance resolution is W*Rres.

4. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 3 is characterized in that: The number of points to calculate the distance from the mean is calculated using either 3×3 smoothing or 5×5 smoothing.

5. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 1 is characterized in that: In step 1, the formula for geometric overlap factor correction is: Among them, the geometric overlap factor is O(H), P(H, λ) is the original echo signal, P ′ (H, λ) is the signal corrected by the geometric overlap factor, H is the height from the radar, in meters, and λ is the laser emission wavelength, in nanometers.

6. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 1, characterized in that: In step 1, the formula for background noise removal is: P ′ (H,λ)=P(H,λ)-P bg (H,λ) Among them, P(H, λ) is the original echo signal, P ′ (H, λ) is the signal after background noise subtraction, P bg (H, λ) is the signal of background noise; P bg The calculation formula of (H, λ) is: Among them, H1~H N is the background noise height, and λ is the wavelength.

7. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 1, characterized in that: In step 2; The coordinates of the laser fog radar point p data (r, α, θ) can be expressed in rectangular coordinates in the rectangular coordinate system as (x p ,y p , z p ): x p =rcos(θ)cos(α); y p =rcos(θ)sin(α); z p =rsin(θ); The longitude and latitude of the current unmanned boat are (Lon_p, Lat_p), and the longitude and latitude grid coordinates of the rectangular coordinates of the laser fog radar data It is expressed as:

8. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 1, characterized in that: In step 4, the calculation formula of the target functional of the optimal interpolation fusion method is: gradient: Among them, J G represents the value function of global reanalysis data, J L It represents the three-dimensional observation field output by the laser radar after entering the LAPS mode. The subscript i represents any time, distributed in the time period [0, t N ], X is the control vector, x G is the state vector of global reanalysis data, G is the error covariance matrix of global reanalysis data, R i is the observation error covariance matrix, H i is the observation operator, y i is the observation forecast vector at time i, x i is the state prediction vector at time i, and T is used to represent the operation of exchanging the rows and columns of the matrix.

9. The method for fusing unmanned boat laser fog radar with global reanalysis data according to claim 8, characterized in that: State prediction vector x i Satisfies the mode control equation: Among them, M 0→i (x) is the model prediction operator from the starting time 0 to the time i.

10. The method for fusing unmanned boat laser fog radar and global reanalysis data according to claims 1-9, characterized in that: The four-dimensional variational method integrates three-dimensional observation field data with global reanalysis data, including the following steps: S1. Determine the initial guess value of each control variable in the LAPS model And the model is forward integrated in the fusion window to obtain the S2. Substitute the conjugate equation of the atmospheric motion state equation and reversely integrate the corresponding adjoint mode equation; S3. Calculate the value function according to the above objective functional formula and find its partial derivative with respect to each control variable; S4. Use the gradient descent algorithm to determine the descent direction of the value function: Where β is the optimal step size, n is the number of iterations, j, k = 1, 2, ..., n, is the optimal descent direction function; S5, when J n -J n-1 ≤δ, output That is, fusion is completed; Among them, J n The value function of the nth iteration, J n-1 The value function of the n-1th iteration, δ represents a very small value.

Images