Ground wave radar data quality control method based on CNN-LSTM model

Abstract

The invention discloses a ground wave radar data quality control method based on a CNN-LSTM model, and the method comprises the steps: determining a radar detection data range affected by the terrain according to the accuracy difference of radar echo data; according to the radar detection data range, training a CNN-LSTM neural network to obtain a target model; and correcting the radar data of the terrain-affected area according to the target model to obtain corrected radar data. According to the embodiment of the invention, the radar detection data influenced by coastlines, islands and other terrains can be corrected in combination with actual marine environment information, the ocean current direction accuracy synthesized by the radar detection data is improved, and the method can be widely applied to the technical field of computers.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a ground wave radar data quality control method based on a CNN-LSTM model. Background Art

[0002] At present, the acquisition of ocean current data mainly comes from numerical simulation and actual observation. Although numerical simulation can simulate the ocean current conditions of the entire sea area in real time, its accuracy is low. Although the actual fixed-point observation data is accurate, it can often only obtain data from one point or a vertical profile. It is difficult to obtain observation data for a large sea area and is easily affected by the environment. High-frequency ground wave radar has the advantages of beyond-horizon, large range, all-weather and low cost, and it is less affected by weather. In the field of using ground wave radar to invert and calculate ocean currents, it is now constantly improving and gradually becoming mature. Although the existing technology can obtain high-precision ocean current information in the high-precision area of ​​ground wave radar observation, since the traditional calculation method is based on radar echo information and does not consider the impact of the actual marine environment, the range of this high-precision area is much smaller than the range of sea area data actually obtained by the radar.

[0003] In the field of using high-frequency ground wave radar data to obtain ocean current data, there are mainly the following problems:

[0004] 1. Most of the existing inversion algorithms are based on electronic information theory, which has certain limitations. In coastal areas, radar electromagnetic waves are easily disturbed by coastlines and islands, the seabed topography is relatively complex, and factors such as tides, runoff, and wind also have an impact. Radar electromagnetic echoes cannot effectively analyze the interaction between these physical factors, so the accuracy of ocean current information obtained using electronic information theory is relatively low.

[0005] 2. The current technology of using neural networks to correct radar data, most of the neural network input layer only uses the echo parameters of the radar high-precision area, and does not add the marine environment information that affects the physical process of ocean currents. A small number of them add tides and wind fields, but when correcting radar data in areas affected by terrain, since neither the influence of terrain is considered nor the spatial features of the data are extracted, when correcting radar detection data affected by terrain, the correction accuracy is slightly lacking. At the same time, each data point needs to be trained separately. A single trained neural network cannot be applied to the radar data correction of the entire affected area, which reduces practicality. Summary of the invention

[0006] The main purpose of the embodiment of the present invention is to propose a ground wave radar data quality control method based on the CNN-LSTM model, which can improve the accuracy of the ocean current direction synthesized by radar detection data.

[0007] To achieve the above object, an embodiment of the present invention proposes a method for controlling ground wave radar data quality based on a CNN-LSTM model, comprising the following steps:

[0008] According to the accuracy difference of radar echo data, the range of radar detection data affected by terrain is determined;

[0009] According to the radar detection data range, the CNN-LSTM neural network is trained to obtain a target model;

[0010] The radar data of the area affected by the terrain is corrected according to the target model to obtain corrected radar data.

[0011] In some embodiments, determining the range of radar detection data affected by terrain according to the accuracy difference of radar echo data comprises the following steps:

[0012] Based on the accuracy differences of radar echo data, data is detected within the radar detection area near the coastline and islands;

[0013] The area where the ellipse direction angle error of the empirical orthogonal function of the synthetic ocean current is greater than 40° is determined as the radar detection data range affected by the terrain.

[0014] In some embodiments, training a CNN-LSTM neural network according to the radar detection data range to obtain a target model includes the following steps:

[0015] Establish an initial model based on the CNN-LSTM neural network;

[0016] The 9×9 matrix with the data point to be corrected as the center point is used as the input data of the CNN-LSTM neural network;

[0017] The training cycle of the CNN-LSTM neural network is set to 200 rounds, the weight parameters of the network model are updated using the Adam optimization algorithm, and the initial learning rate is set to 0.001;

[0018] During the training of the network model, the number of convolutional layers, the size of the convolution kernel, and the number of hidden layers of the LSTM module are adjusted;

[0019] The final network structure is determined, all input training data are used as training sets, the neural network is trained, the training is terminated when the target minimum error is achieved and remains relatively stable, and a radar data correction model is generated as the target model.

[0020] In some embodiments, the process of training the CNN-LSTM neural network further includes the following steps:

[0021] According to the EOF ellipse directional angle error, the mean flow directional angle error and the EOF ellipse major and minor axis errors, an evaluation index is determined to verify the target model; this step is specifically as follows:

[0022] A data matrix is ​​established, where the first and second columns are u component and v component respectively, and then the velocity matrix is ​​decomposed into the first mode and the second mode by EOF analysis; wherein the major axis of the ellipse represents the first mode of the eigenvalue of the maximum standard deviation of the velocity, and the minor axis represents the second mode of the eigenvalue;

[0023] The direction angle of the ellipse uses the formula Calculated, where v1 and v2 are the eigenvectors of the first and second modes respectively:

[0024] The calculation formulas for EOF ellipse direction angle error and mean flow direction angle error are:

[0025] Δθ=|(θ1-θ2+180°)mod360°-180°|, where θ1 and θ2 represent the angles of the EOF ellipse direction or the mean flow direction of the observed value and the corrected value, respectively, mod is the modulus operation, and Δθ is the error angle;

[0026] The calculation formula of the major and minor axis error E of the EOF ellipse is:

[0027] Among them, Lmaj radar Represents the major axis of the radar data EOF ellipse; Lmaj observation Represents the major axis of the EOF ellipse of the observed data; Lmin radar Represents the minor axis of the radar data EOF ellipse; Lmin observation Represents the minor axis of the observed data EF ellipse.

[0028] In some embodiments, establishing an initial model according to a CNN-LSTM neural network comprises the following steps:

[0029] The features in the convolutional neural network CNN are extracted through the convolution operation. The convolution operation gradually extracts local features by sliding a convolution kernel on the input data. The calculation formula for feature extraction is: Among them, Y(x, y) represents the value of the position (x, y) in the feature matrix extracted by the convolution neural network; r is the ReLU activation function; the size of the convolution kernel is (m×n); I(x+i, y+j) is the value of the corresponding convolution kernel position (i, j) in the input data matrix; K(i, j) is the value of the position (i, j) in the convolution kernel; b is the bias term;

[0030] The calculation formula for average pooling is as follows:

[0031]

[0032] Among them, O(x, y) represents the value of the position (x, y) in the output feature matrix; the size of the pooling window is (m×n); s is the step size, which represents the step size of each sliding of the pooling window; Y(x·s+i, y·s+j) represents the element in the input feature map covered by the pooling window;

[0033] The calculation formula of the network parameters of the LSTM neural network is:

[0034] f t =σ(W f ·[h t-1 , x t ]+b f )

[0035] i t =σ(W i ·[h t-1 , x t ]+b i )

[0036]

[0037] o t =σ(W o ·[h t-1 , x t ]+b o )

[0038] h t =o t *tanh(C t )

[0039] Where σ represents the sigmoid activation function; tanh represents the hyperbolic tangent activation function; · represents the dot product; * represents the Hadamard product; t is the time step; x t is the input data at time t; h t is the hidden state at time t; Indicates the candidate unit state; C t is the cell state at time t; f t Is the forget gate, which determines the t What information is discarded in t It is the input gate, which controls how much information of the candidate state at the current moment needs to be saved in the unit state; t It is the output gate, which controls the current state C t How much information is output to the hidden state h t ; W f , W i, W c and W o is the weight of the corresponding gate and unit; b f 、b i 、b c and b o is the bias term corresponding to the gate and unit;

[0040] The calculation formula of the loss function used in neural network training is:

[0041]

[0042] Where N is the number of samples; y i is the value of the observation on the i-th sample; is the value of the predicted value on the i-th sample.

[0043] In some embodiments, the 9×9 matrix with the data point to be corrected as the center point is used as the input data of the CNN-LSTM neural network, including the following steps:

[0044] According to the observed composite speed of ocean current u and v components, tidal velocity u component, tidal velocity v component, wind speed u component, wind speed v component and altitude data, the input data matrix of the CNN module is constructed, where each batch of input data contains matrix data of 10 time steps, t, t+1, ..., t+9; after two convolution and pooling operations, the input matrix data is flattened and input to the LSTM module, the output of the LSTM module is combined with the flattened altitude data, and finally the fully connected layer FC outputs the corrected u and v components of 10 time steps;

[0045] The input data is normalized, and the expression of the normalization is:

[0046]

[0047] Among them, X i represents the normalized result; X ioriginal Represents the original data in the input item; X max Represents the maximum value of the original data; X min Represents the minimum value of the original data;

[0048] The normalized input training data is randomly divided into two parts, 80% and 20%, as the training set and test set, respectively, for training network parameters and testing network performance.

[0049] In some embodiments, the radar data of the area affected by the terrain is corrected according to the target model to obtain the corrected radar data, specifically:

[0050] The radar data of the area affected by the terrain is corrected according to the target model. The input data matrix during correction includes the composite velocity of the ocean current u and v components, the tidal velocity u component, the tidal velocity v component, the wind speed u component, the wind speed v component and the altitude data obtained by radar detection, and finally the corrected radar data is obtained.

[0051] Another aspect of the embodiment of the present invention further provides a ground wave radar data quality control system based on a CNN-LSTM model, comprising:

[0052] The first module is used to determine the range of radar detection data affected by terrain according to the accuracy difference of radar echo data;

[0053] The second module is used to train the CNN-LSTM neural network according to the radar detection data range to obtain a target model;

[0054] The third module is used to correct the radar data of the area affected by the terrain according to the target model to obtain corrected radar data.

[0055] To achieve the above objective, another aspect of an embodiment of the present invention provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above-mentioned method when executing the computer program.

[0056] To achieve the above objective, another aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned method is implemented.

[0057] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0058] The embodiments of the present invention include at least the following beneficial effects: The present invention provides a method for controlling the quality of ground wave radar data based on a CNN-LSTM model, which determines the range of radar detection data affected by terrain according to the accuracy difference of radar echo data; trains the CNN-LSTM neural network according to the range of radar detection data to obtain a target model; and corrects the radar data in the area affected by terrain according to the target model to obtain corrected radar data. The embodiments of the present invention can correct the radar detection data affected by terrain such as coastlines and islands in combination with actual marine environmental information, thereby improving the accuracy of the ocean current direction synthesized by the radar detection data. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of an implementation environment provided by an embodiment of the present invention;

[0060] Figure 2 is a flow chart of the overall steps provided by an embodiment of the present invention;

[0061] Figure 3 It is a flowchart of specific implementation steps provided by an embodiment of the present invention;

[0062] Figure 4 is a flowchart of the steps for training a CNN-LSTM neural network provided by an embodiment of the present invention;

[0063] Figure 5 It is a schematic diagram of the framework structure of the network model provided by an embodiment of the present invention;

[0064] Figure 6 is a schematic diagram of an LSTM neural network structure provided by an embodiment of the present invention;

[0065] Figure 7 It is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present invention. They are only examples of devices and methods consistent with some aspects of the embodiments of the present invention as detailed in the attached claims.

[0067] It is understood that the terms "first", "second", etc. used in the present invention may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0068] The terms "at least one", "multiple", "each", "any", etc. used in the present invention, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0069] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0070] Before describing the embodiments of the present invention in detail, some related technologies involved in the embodiments of the present invention are first described as follows:

[0071] CNN-LSTM neural network model: The CNN-LSTM neural network model combines convolutional neural networks (CNN) and long short-term memory (LSTM) networks. CNN is a feedforward neural network that includes convolution calculations. It uses convolution and pooling operations to effectively extract local features from input data and is mainly used to process data with spatial structures (such as images and videos). LSTM is a special recurrent neural network (RNN) that is mainly used to process time series data. It is used to solve the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequence data. The CNN-LSTM neural network model combines the advantages of CNN in extracting spatial features and LSTM in capturing temporal features, and is suitable for processing complex spatiotemporal data.

[0072] Deep learning: Deep learning is a new research direction in the field of machine learning. It combines low-level features to form more abstract high-level representation attribute categories or features to discover distributed feature representations of data.

[0073] Ground wave radar: Ground wave radar is a major means of sea detection. Its detection principle is to use the characteristic of small diffraction propagation attenuation of the conductive ocean surface to emit high-frequency radio waves, which can break through the horizon and detect targets 300 kilometers away with high detection accuracy.

[0074] The ground wave radar data quality control method based on the CNN-LSTM model provided in an embodiment of the present invention relates to the field of computer technology. The ground wave radar data quality control method based on the CNN-LSTM model provided in an embodiment of the present invention can be applied to a terminal, can also be applied to a server, and can also be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a vehicle-mounted terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements the ground wave radar data quality control method based on the CNN-LSTM model, etc., but is not limited to the above forms.

[0075] The present invention can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0076] like Figure 1 FIG. 1 is a schematic diagram of an implementation environment provided by an embodiment of the present invention. Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected to a network wirelessly or wired to complete data transmission and exchange.

[0077] Server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0078] In addition, the server 101 can also be a node server in the blockchain network. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm.

[0079] The terminal 102 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal 102 may also be a vehicle-mounted terminal of various device types described above, but is not limited thereto. The terminal 102 and the server 101 may be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment of the present invention.

[0080] Based on the example Figure 1 In the implementation environment shown, an embodiment of the present invention provides a ground wave radar data quality control method based on a CNN-LSTM model. The following is an example of applying the ground wave radar data quality control method based on a CNN-LSTM model to a server 101. It can be understood that the method can also be applied to a terminal 102.

[0081] Reference Figure 2 , Figure 2 The flowchart of the ground wave radar data quality control method based on the CNN-LSTM model applied to the server provided in the embodiment of the present invention, the execution subject of the method can be any of the aforementioned computer devices (including servers or terminals). Figure 2 , the method may include the following steps:

[0082] According to the accuracy difference of radar echo data, the range of radar detection data affected by terrain is determined;

[0083] According to the radar detection data range, the CNN-LSTM neural network is trained to obtain a target model;

[0084] The radar data of the area affected by the terrain is corrected according to the target model to obtain corrected radar data.

[0085] In some embodiments, determining the range of radar detection data affected by terrain according to the accuracy difference of radar echo data comprises the following steps:

[0086] Based on the accuracy differences of radar echo data, data is detected within the radar detection area near the coastline and islands;

[0087] The area where the ellipse direction angle error of the empirical orthogonal function of the synthetic ocean current is greater than 40° is determined as the radar detection data range affected by the terrain.

[0088] In some embodiments, training a CNN-LSTM neural network according to the radar detection data range to obtain a target model includes the following steps:

[0089] Establish an initial model based on the CNN-LSTM neural network;

[0090] The 9×9 matrix with the data point to be corrected as the center point is used as the input data of the CNN-LSTM neural network;

[0091] The training cycle of the CNN-LSTM neural network is set to 200 rounds, the weight parameters of the network model are updated using the Adam optimization algorithm, and the initial learning rate is set to 0.001;

[0092] During the training of the network model, the number of convolutional layers, the size of the convolution kernel, and the number of hidden layers of the LSTM module are adjusted;

[0093] The final network structure is determined, all input training data are used as training sets, the neural network is trained, the training is terminated when the target minimum error is achieved and remains relatively stable, and a radar data correction model is generated as the target model.

[0094] In some embodiments, the process of training the CNN-LSTM neural network further includes the following steps:

[0095] According to the EOF ellipse directional angle error, the mean flow directional angle error and the EOF ellipse major and minor axis errors, an evaluation index is determined to verify the target model; this step is specifically as follows:

[0096] A data matrix is ​​established, where the first and second columns are u component and v component respectively, and then the velocity matrix is ​​decomposed into the first mode and the second mode by EOF analysis; wherein the major axis of the ellipse represents the first mode of the eigenvalue of the maximum standard deviation of the velocity, and the minor axis represents the second mode of the eigenvalue;

[0097] The direction angle of the ellipse uses the formula Calculated, where v1 and v2 are the eigenvectors of the first and second modes respectively:

[0098] The calculation formulas for EOF ellipse direction angle error and mean flow direction angle error are:

[0099] Δθ=|(θ1-θ2+180°)mod360°-180°|, where θ1 and θ2 represent the angles of the EOF ellipse direction or the mean flow direction of the observed value and the corrected value, respectively, mod is the modulus operation, and Δθ is the error angle;

[0100] The calculation formula of the major and minor axis error E of the EOF ellipse is:

[0101] Among them, Lmaj radar Represents the major axis of the radar data EOF ellipse; Lmaj observvation Represents the major axis of the EOF ellipse of the observed data; Lmin radar Represents the minor axis of the radar data EOF ellipse; Lmin observation Represents the minor axis of the observed data EOF ellipse.

[0102] In some embodiments, establishing an initial model according to a CNN-LSTM neural network comprises the following steps:

[0103] The features in the convolutional neural network CNN are extracted through the convolution operation. The convolution operation gradually extracts local features by sliding a convolution kernel on the input data. The calculation formula for feature extraction is: Among them, Y(x, y) represents the value of the position (x, y) in the feature matrix extracted by the convolution neural network; r is the ReLU activation function; the size of the convolution kernel is (m×n); I(x+i, y+j) is the value of the corresponding convolution kernel position (i, j) in the input data matrix; K(i, j) is the value of the position (i, j) in the convolution kernel; b is the bias term;

[0104] The calculation formula for average pooling is as follows:

[0105]

[0106] Among them, O(x, y) represents the value of the position (x, y) in the output feature matrix; the size of the pooling window is (m×n); s is the step size, which represents the step size of each sliding of the pooling window; Y(x·s+i, y·s+j) represents the element in the input feature map covered by the pooling window;

[0107] The calculation formula of the network parameters of the LSTM neural network is:

[0108] f t =σ(W f ·[h t-1 , x t ]+b f )

[0109] i t =σ(Wi ·[h t-1 , x t ]+b i )

[0110]

[0111] o t =σ(W o ·[h t-1 , x t ] +b o )

[0112] h t =o t *tanh(C t )

[0113] Where σ represents the sigmoid activation function; tanh represents the hyperbolic tangent activation function; · represents the dot product; * represents the Hadamard product; t is the time step; x t is the input data at time t; h t is the hidden state at time t; Indicates the candidate unit state; C t is the cell state at time t; f t Is the forget gate, which determines the t What information is discarded in t It is the input gate, which controls how much information of the candidate state at the current moment needs to be saved in the unit state; t It is the output gate, which controls the current state C t How much information is output to the hidden state h t ; W f , W i , W c and W o is the weight of the corresponding gate and unit; b f 、b i 、b c and b o is the bias term corresponding to the gate and unit;

[0114] The calculation formula of the loss function used in neural network training is:

[0115]

[0116] Where N is the number of samples; y i is the value of the observation on the i-th sample; is the value of the predicted value on the i-th sample.

[0117] In some embodiments, the 9×9 matrix with the data point to be corrected as the center point is used as the input data of the CNN-LSTM neural network, including the following steps:

[0118] According to the observed composite speed of ocean current u and v components, tidal velocity u component, tidal velocity v component, wind speed u component, wind speed v component and altitude data, the input data matrix of the CNN module is constructed, where each batch of input data contains matrix data of 10 time steps, t, t+1, ..., t+9; after two convolution and pooling operations, the input matrix data is flattened and input to the LSTM module, the output of the LSTM module is combined with the flattened altitude data, and finally the fully connected layer FC outputs the corrected u and v components of 10 time steps;

[0119] The input data is normalized, and the expression of the normalization is:

[0120]

[0121] Among them, X i represents the normalized result; X ioriginal Represents the original data in the input item; X max Represents the maximum value of the original data; X min Represents the minimum value of the original data;

[0122] The normalized input training data is randomly divided into two parts, 80% and 20%, as the training set and test set, respectively, for training network parameters and testing network performance.

[0123] In some embodiments, the radar data of the area affected by the terrain is corrected according to the target model to obtain the corrected radar data, specifically:

[0124] The radar data of the area affected by the terrain is corrected according to the target model. The input data matrix during correction includes the composite velocity of the ocean current u and v components, the tidal velocity u component, the tidal velocity v component, the wind speed u component, the wind speed v component and the altitude data obtained by radar detection, and finally the corrected radar data is obtained.

[0125] The following describes the specific implementation process of the embodiment of the present invention in detail by taking a specific application scenario as an example:

[0126] refer to Figure 3 The specific implementation steps of the present invention include the following steps:

[0127] Step 1: Based on the accuracy difference of radar echo data, in the radar detection area near the coastline and islands, the area where the angular error of the empirical orthogonal function (EOF) ellipse of the detection data synthesized ocean current is greater than 40° is determined as the radar detection data range affected by the terrain.

[0128] Step 2: Train the CNN-LSTM neural network, refer to Figure 4 and Figure 5 , the specific detailed steps are as follows:

[0129] Step 0: Modeling of CNN-LSTM neural network. The feature extraction in convolutional neural network CNN mainly depends on the convolution operation. The convolution operation extracts local features step by step by sliding a convolution kernel on the input data. The calculation formula of feature extraction is as follows:

[0130]

[0131] In the formula, Y( x， y) represents the feature matrix extracted by the product neural network ( x， y) position; r is the ReLU activation function; the size of the convolution kernel is (m×n); I(x+i, y+j) is the value of the corresponding convolution kernel position (i, j) in the input data matrix; K(i, j) is the value of the convolution kernel position (i, j); b is the bias term.

[0132] The calculation formula for average pooling is as follows:

[0133]

[0134] In the formula, O( x， y) represents the output feature matrix ( x， y) position; the size of the pooling window is (m×n); s is the step size, which indicates the step size of each sliding of the pooling window; Y(x·s+i, y·s+j) represents the element in the input feature map covered by the pooling window.

[0135] LSTM neural network is a special recurrent neural network (RNN). In order to solve the problem of gradient vanishing or gradient exploding when traditional RNN processes long time series, LSTM introduces two mechanisms, namely memory cell and gated control structure. Among them, the memory cell is responsible for storing and updating the long-term dependency information in the time series. The gated control structure refers to the use of gate mechanism to control the use of state information. It contains three gates, namely forget gate, input gate and output gate. These gates jointly determine the retention, update and final output of information. The structure of LSTM neural network is as follows Figure 6 The main calculation formula is as follows:

[0136] f t =σ(W f ·[h t-1 , x t ]+b f )

[0137] i t =σ(W i ·[h t-1 , x t ]+b i )

[0138]

[0139] o t =σ(W o ·[h t-1 , x t ]+b o )

[0140] h t =o t *tanh(C t )

[0141] In the above formula, σ represents the sigmoid activation function; tanh represents the hyperbolic tangent activation function; · represents the dot product; * represents the Hadamard product; t is the time step; x t is the input data at time t; h t is the hidden state at time t; Indicates the candidate unit state; C t is the cell state at time t; ft is the forget gate, which determines the cell state C t What information is discarded in tIt is the input gate, which controls how much information of the candidate state at the current moment needs to be saved in the unit state; t It is the output gate, which controls the current state C t How much information is output to the hidden state h t ; W f , W i , W c and W o is the weight of the corresponding gate and unit; b f 、b i 、b c and b o are the bias terms for the corresponding gates and cells.

[0142] The calculation formula of the loss function used in neural network training is:

[0143]

[0144] Where N is the number of samples, that is, the total number of data points; yi is the value of the observation on the i-th sample; is the value of the predicted value on the i-th sample.

[0145] Step 1: The input data of the neural network are all 9×9 matrices with the data point to be corrected as the center point. The input data matrix of the CNN module includes the observed composite speed of the ocean current u and v components, the tidal velocity u component, the tidal velocity v component, the wind speed u component, the wind speed v component and the altitude data. Each batch of input data contains matrix data of 10 time steps, t, t+1, ..., t+9. After two convolution and pooling operations, the input matrix data is flattened and input to the LSTM module. The output of the LSTM module is combined with the flattened altitude data, and finally the corrected u and v components of 10 time steps are output through the fully connected layer FC. To speed up the operation, the data can be normalized:

[0146]

[0147] Among them, X i represents the normalized result; X ioriginal Represents the original data in the input item; X max Represents the maximum value of the original data; X min Represents the minimum value of the original data. The normalized input training data is randomly divided into two parts, 80% and 20%, as the training set and test set, respectively, for training network parameters and testing network performance.

[0148] Step 2: The training cycle of the CNN-LSTM network model is set to 200 rounds. The weight parameters of the network model are updated using the Adam optimization algorithm. The initial learning rate is set to 0.001. In order to prevent overfitting of the model and enhance the generalization ability of the model, L2 regularization is used during model training, and the weight decay coefficient is set to 0.00001.

[0149] Step 3: During the model training, the number of convolutional layers, the size of the convolution kernel, and the number of hidden layers of the LSTM module are adjusted, and the radar data correction effect of the model under different parameters is compared to minimize the error.

[0150] Step 4: After the above steps, the final network structure is determined, and all input training data are used as training sets to train the neural network. The training is terminated when the target minimum error is achieved and remains relatively stable, and the radar data correction model is generated.

[0151] Step 5: Evaluation indicators: EOF ellipse angular error, mean flow angular error, and EOF ellipse major and minor axis errors.

[0152] The calculation method of EOF ellipse is as follows: firstly, a data matrix is ​​established, where the first and second columns are u component and v component respectively, and then the velocity matrix is ​​decomposed into the first mode and the second mode by EOF analysis. The major axis of the ellipse represents the first mode of the eigenvalue of the maximum standard deviation of the velocity, and the minor axis is the second mode of the eigenvalue. The direction angle of the ellipse is calculated using the following formula, where v1 and v2 are the eigenvectors of the first mode and the second mode respectively.

[0153]

[0154] The calculation formulas for EOF ellipse direction angle error and mean flow direction angle error are as follows:

[0155] Δθ=|(θ1-θ2+180°)mod360°-180°|

[0156] In the above formula, θ1 and θ2 represent the angles of the EOF ellipse direction or the mean flow direction of the observed value and the corrected value, respectively, mod is the modulo operation, and △θ is the error angle.

[0157] The calculation formula for the major and minor axis errors of the EOF ellipse is as follows:

[0158] Among them, Lmaj radar Represents the major axis of the radar data EOF ellipse; Lmaj observation Represents the major axis of the EOF ellipse of the observed data; Lmin radar Represents the minor axis of the radar data EOF ellipse; Lmin observationRepresents the minor axis of the EOF ellipse of the observed data. Step 6: Save the model: Save the trained and verified model.

[0159] Step 3: The model trained in step 2 is used to correct radar data in areas affected by terrain. The input data matrix during correction includes the composite velocity of ocean current u and v components, tidal velocity u component, tidal velocity v component, wind speed u component, wind speed v component and altitude data obtained by radar detection, and finally the corrected radar data is obtained.

[0160] In summary, the embodiment of the present invention obtains the nonlinear correlation between the ocean flow field and radar echoes, terrain and marine environmental information based on the deep learning method, corrects the radar detection data affected by the terrain such as coastline and islands, and improves the accuracy of the ocean current direction synthesized by the radar detection data.

[0161] The embodiments of the present invention have the following characteristics:

[0162] 1. To improve the inversion calculation method of ocean currents, this technology uses both ground wave radar echo data and actual ocean environment data.

[0163] 2. Compared with the traditional method of inverting ocean currents through the first-order peak information of ground wave radar, this technology adopts a deep learning method.

[0164] 3. This technology uses the deep learning method of the CNN-LSTM neural network to obtain the optimal correlation coefficient between ocean current direction and radar data and marine environment data.

[0165] Compared with the prior art, the present invention has the following advantages:

[0166] 1. High accuracy: The most direct physical process affecting the ocean flow field is ocean dynamic factors, such as tides, wind fields, runoff, terrain distribution, etc. The traditional ocean current inversion method is an indirect calculation method based on the ground wave radar echo signal. The present invention combines the radar echo signal with the ocean environment information through the CNN-LSTM neural network, integrating the advantages of CNN in extracting spatial features and LSTM in capturing temporal features, thereby improving the accuracy of radar detection of ocean current direction.

[0167] 2. Good generalization ability: When correcting radar data in areas affected by terrain, there is no need to build a neural network model for each radar data point. A CNN-LSTM neural network model can achieve good correction results. At the same time, as long as the time contained in the data used for training is long enough, the trained model can also be used for data correction in any time period in the area. There is no need to retrain the model every time data is corrected at different times. This can greatly reduce time costs and has more practical application value.

[0168] Another aspect of the embodiment of the present invention further provides a ground wave radar data quality control system based on a CNN-LSTM model, comprising:

[0169] The first module is used to determine the range of radar detection data affected by terrain according to the accuracy difference of radar echo data;

[0170] The second module is used to train the CNN-LSTM neural network according to the radar detection data range to obtain a target model;

[0171] The third module is used to correct the radar data of the area affected by the terrain according to the target model to obtain corrected radar data.

[0172] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0173] The embodiment of the present invention further provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned ground wave radar data quality control method based on the CNN-LSTM model when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0174] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0175] See also Figure 7 , Figure 7 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0176] The processor 701 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention.

[0177] The memory 702 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 702, and the processor 701 calls and executes the ground wave radar data quality control method based on the CNN-LSTM model of the embodiment of the present invention;

[0178] Input / output interface 703, used to implement information input and output;

[0179] Communication interface 704, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0180] A bus 705 that transmits information between the various components of the device (e.g., the processor 701, the memory 702, the input / output interface 703, and the communication interface 704);

[0181] The processor 701 , the memory 702 , the input / output interface 703 and the communication interface 704 are connected to each other in communication within the device via a bus 705 .

[0182] An embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned ground wave radar data quality control method based on the CNN-LSTM model is implemented.

[0183] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0184] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0185] It should be noted that in various specific embodiments of the present invention, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present invention needs to obtain the user's sensitive personal information, it will obtain the user's separate permission or consent through a pop-up window or jump to a confirmation page, and after clearly obtaining the user's separate permission or consent, it will obtain the necessary user-related data for the normal operation of the embodiment of the present invention.

[0186] The embodiments described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. Those skilled in the art can appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0187] Those skilled in the art will appreciate that the technical solutions shown in the figures do not limit the embodiments of the present invention and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0188] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0189] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0190] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0191] It should be understood that in the present invention, "at least one (item)" refers to one or more, and "plurality" refers to two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can represent: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0192] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0193] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0194] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0195] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including multiple instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store programs.

[0196] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the embodiments of the present invention is not limited thereby. Any modification, equivalent substitution and improvement made by those skilled in the art without departing from the scope and essence of the embodiments of the present invention shall be within the scope of the rights of the embodiments of the present invention.

Claims

1. A ground wave radar data quality control method based on CNN-LSTM model, characterized in that: The following steps are involved: According to the accuracy difference of radar echo data, the range of radar detection data affected by terrain is determined; According to the radar detection data range, the CNN-LSTM neural network is trained to obtain a target model; The radar data of the area affected by the terrain is corrected according to the target model to obtain corrected radar data.

2. According to a method for controlling ground wave radar data quality based on a CNN-LSTM model according to claim 1, it is characterized in that: Determining the range of radar detection data affected by terrain according to the accuracy difference of radar echo data includes the following steps: Based on the accuracy differences of radar echo data, data is detected within the radar detection area near the coastline and islands; The area where the ellipse direction angle error of the empirical orthogonal function of the synthetic ocean current is greater than 40° is determined as the radar detection data range affected by the terrain.

3. The method for controlling ground wave radar data quality based on the CNN-LSTM model according to claim 1, characterized in that: The step of training the CNN-LSTM neural network according to the radar detection data range to obtain a target model includes the following steps: Establish an initial model based on the CNN-LSTM neural network; The 9×9 matrix with the data point to be corrected as the center point is used as the input data of the CNN-LSTM neural network; The training cycle of the CNN-LSTM neural network was set to 200 rounds, the weight parameters of the network model were updated using the Adam optimization algorithm, and the initial learning rate was set to 0.001; During the training of the network model, the number of convolutional layers, the size of the convolution kernel, and the number of hidden layers of the LSTM module are adjusted; The final network structure is determined, all input training data are used as training sets, the neural network is trained, and the training is terminated when the target minimum error is achieved and remains relatively stable, and a radar data correction model is generated as the target model.

4. The method for controlling ground wave radar data quality based on the CNN-LSTM model according to claim 3, characterized in that: The process of training the CNN-LSTM neural network also includes the following steps: According to the EOF ellipse directional angle error, the mean flow directional angle error and the EOF ellipse major and minor axis errors, an evaluation index is determined to verify the target model; this step is specifically as follows: A data matrix is ​​established, where the first and second columns are u component and v component respectively, and then the velocity matrix is ​​decomposed into the first mode and the second mode by EOF analysis; wherein the major axis of the ellipse represents the first mode of the eigenvalue of the maximum standard deviation of the velocity, and the minor axis represents the second mode of the eigenvalue; The direction angle of the ellipse uses the formula Calculated, where v1 and v2 are the eigenvectors of the first and second modes respectively: The calculation formulas for EOF ellipse direction angle error and mean flow direction angle error are: Δθ=|(θ1-θ2+180°)mod360°-180°|, where θ1 and θ2 represent the angles of the EOF ellipse direction or the mean flow direction of the observed value and the corrected value, respectively, mod is the modulus operation, and Δθ is the error angle; The calculation formula of the major and minor axis error E of the EOF ellipse is: Among them, Lmaj radar Represents the major axis of the radar data EOF ellipse; Lmaj observation Represents the major axis of the EOF ellipse of the observed data; Lmin radar Represents the minor axis of the radar data EOF ellipse; Lmin observation Represents the minor axis of the observed data EOF ellipse.

5. The method for controlling ground wave radar data quality based on the CNN-LSTM model according to claim 3, characterized in that: The initial model is established according to the CNN-LSTM neural network, including the following steps: The features in the convolutional neural network CNN are extracted through the convolution operation. The convolution operation gradually extracts local features by sliding a convolution kernel on the input data. The calculation formula for feature extraction is: Among them, Y(x,y) represents the value of the position (x,y) in the feature matrix extracted by the convolution neural network; r is the ReLU activation function; the size of the convolution kernel is (m×n); I(x+i,y+j) is the value of the corresponding convolution kernel position (i,j) in the input data matrix; K(i,j) is the value of the position (i,j) in the convolution kernel; b is the bias term; The calculation formula for average pooling is as follows: Among them, O(x,y) represents the value of the (x,y) position in the output feature matrix; the size of the pooling window is (m×n); s is the step size, which represents the step size of each sliding of the pooling window; Y(x·s+i,y·s+j) represents the element in the input feature map covered by the pooling window; The calculation formula of the network parameters of the LSTM neural network is: f t =σ(W f ·[h t-1 ,x t ]+b f ) i t =σ(W i ·[h t-1 ,x t ]+b i ) the t =σ(W o ·[h t-1 ,x i ]+b o ) h t =o t *tanh(C t ) Where σ represents the sigmoid activation function; tanh represents the hyperbolic tangent activation function; · represents the dot product; * represents the Hadamard product; t is the time step; x t is the input data at time t; h t is the hidden state at time t; Indicates the candidate unit state; C t is the cell state at time t; f t is the forget gate, which determines the state of the cell C t What information is discarded in t It is the input gate, which controls how much information of the candidate state at the current moment needs to be saved in the unit state; t It is the output gate, which controls the current state C t How much information is output to the hidden state h t ; W f , W i , W c and W o is the weight of the corresponding gate and unit; b f , b i , b c and b o is the bias term corresponding to the gate and unit; The calculation formula of the loss function used in neural network training is: Where N is the number of samples; y i is the value of the observation on the i-th sample; is the value of the predicted value on the i-th sample.

6. The method for controlling ground wave radar data quality based on the CNN-LSTM model according to claim 3, characterized in that: The 9×9 matrix with the data point to be corrected as the center point is used as the input data of the CNN-LSTM neural network, including the following steps: According to the observed ocean current u and v component composite speeds, tidal velocity u component, tidal velocity v component, wind speed u component, wind speed v component and altitude data, the input data matrix of the CNN module is constructed, where each batch of input data contains matrix data of 10 time steps, t, t+1, ..., t+9; after two convolution and pooling operations, the input matrix data is flattened and input to the LSTM module, the output of the LSTM module is combined with the flattened altitude data, and finally the fully connected layer FC outputs the corrected u and v components of 10 time steps; The input data is normalized, and the expression of the normalization is: Among them, X i represents the normalized result; X ioriginal Represents the original data in the input item; X max Represents the maximum value of the original data; X min Represents the minimum value of the original data; The normalized input training data is randomly divided into two parts, 80% and 20%, as the training set and test set, respectively, for training network parameters and testing network performance.

7. The method for controlling ground wave radar data quality based on the CNN-LSTM model according to claim 1, characterized in that: The radar data of the area affected by the terrain is corrected according to the target model to obtain the corrected radar data, specifically: The radar data of the area affected by the terrain is corrected according to the target model. The input data matrix during correction includes the composite velocity of the ocean current u and v components, the tidal velocity u component, the tidal velocity v component, the wind speed u component, the wind speed v component and the altitude data obtained by radar detection, and finally the corrected radar data is obtained.

8. A ground wave radar data quality control system based on CNN-LSTM model, characterized in that: include: The first module is used to determine the range of radar detection data affected by terrain according to the accuracy difference of radar echo data; The second module is used to train the CNN-LSTM neural network according to the radar detection data range to obtain a target model; The third module is used to correct the radar data of the area affected by the terrain according to the target model to obtain corrected radar data.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the method according to any one of claims 1 to 7.

Images