Ground-based radar monitoring data compression method, medium and equipment

Abstract

The invention relates to the technical field of radar data processing, and provides a ground-based radar monitoring data compression method, which comprises the steps of S1, acquiring original data; converting the original data into an image of N channels; s2, constructing a neural network model; s3, inputting the image into the neural network model to obtain an output value; s3.1, obtaining the coordinates of each pixel in the image; s3.2, performing position coding on each pixel coordinate to obtain a multi-dimensional position coding tensor; s3.3, performing normalization processing on the real deformation values to obtain deformation value tensors; s3.4, inputting the multi-dimensional position coding tensor into the neural network model to obtain an output value; s4, inputting the real deformation value and the output value into a loss function, calculating the loss of the neural network model, if the loss is smaller than a set threshold value, storing parameters of the neural network model, and obtaining compressed data; otherwise, adjusting the parameters of the neural network model according to the loss reverse gradient, and returning to S3.4. According to the invention, multiple data compression can be realized, the storage space is reduced, and query is facilitated.

Description

Technical Field

[0001] The present invention relates to the field of radar data processing technology, and in particular to a ground-based radar monitoring data compression method, medium and equipment. Background Art

[0002] Radar data is actually a two-dimensional image. The smaller the angular resolution and distance resolution of observation, the smaller the granularity of deformation monitoring and the better the monitoring and early warning effect. Therefore, the radar monitoring data is a two-dimensional image with very high resolution. This increases the time for data transmission and poses challenges to the storage and query of historical data.

[0003] The current technology directly stores radar image data, or stores it as a binary file according to a fixed data structure. A small number of technologies store radar data in a database. However, since ground-based synthetic aperture radar is often used in online monitoring scenarios, the monitoring data will increase rapidly as the monitoring cycle increases. Assuming that the radar resolution unit is 1000×1000, and each resolution unit stores deformation data (4 bytes), the size of a frame of data is 3.81MB. Assuming that monitoring is performed every 5 minutes, the amount of data for one day is 24×60÷5×3.81MB≈1.07GB. The rapid increase in monitoring data adds a lot of pressure to the storage space, and the query of data will become slower as the amount of data increases, which is not conducive to subsequent data processing and early warning analysis. Summary of the invention

[0004] The present invention aims to provide a ground-based radar monitoring data compression method, aiming to solve the problem of large storage space and slow data query of existing radar data. The specific technical scheme is as follows:

[0005] The present invention provides a ground-based radar monitoring data compression method, comprising the following steps:

[0006] S1: Obtain N consecutive ground-based radar monitoring data to obtain raw data; and convert the raw data into an N-channel image, where N is a natural number from 2 to 10; the ground-based radar monitoring data includes information of all grid cells of the ground-based radar, and the grid cells contain distance coordinates, angle coordinates and true deformation values, and the grid cells correspond to pixels of the image one by one;

[0007] S2: Build a neural network model;

[0008] S3: Input the N-channel image into the neural network model to obtain the output value of the N-channel;

[0009] S3.1, obtaining the coordinates of each pixel in the N-channel image;

[0010] S3.2, using two symmetrical triangular wave functions to respectively perform position encoding on the coordinates of each pixel to obtain a multi-dimensional position encoding tensor;

[0011] S3.3, respectively normalizing the true deformation value of the N channels corresponding to each pixel to obtain the deformation value tensor of the N channels;

[0012] S3.4, input the multi-dimensional position encoding tensor into the neural network model for training to obtain the output value O of N channels x ;

[0013] S4: The actual deformation value of the N channel and the output value of the N channel O x Input the loss function and calculate the loss of the neural network model. If the loss is less than the set threshold, save the parameters of the neural network model and obtain the compressed data; otherwise, adjust the parameters of the neural network model according to the reverse gradient of the loss, set x=x+1 and return to S3.4, where: x is a natural number greater than or equal to 1.

[0014] Optionally, in S3.1, the pixel coordinates are (r, θ), where: r is the range coordinate of the grid unit, r is a natural number from 1 to row, and row is the resolution of the range in the ground-based radar monitoring data; θ is the angular coordinate of the grid unit, θ is a natural number from 1 to col, and col is the resolution of the angular in the ground-based radar monitoring data.

[0015] Optionally, in S3.2, the specific formulas of the two symmetrical triangular wave functions are:

[0016]

[0017] F 1 (x,i) and F 2 (x,i) is a symmetrical triangular wave function with a value range of [-1,1]; x is the input value, which can be r or θ; T is the period, i is the setting coefficient, i is [3,20];

[0018] In S3.2, the specific formula of the multi-dimensional position encoding tensor γ(X) is:

[0019] γ(X)=|F 1 (r,i), F 2 (r,i), F 1 (θ,i), F 2 (θ,i)|.

[0020] Optionally, in S3.3, the specific formula for normalization processing is:

[0021]

[0022] Where: D k is the deformation value of the kth channel after normalization of a certain grid unit; d k is the deformation value of the kth channel of a grid unit, k is a natural number from 1 to N; λ is the wavelength of the electromagnetic wave of the ground-based synthetic aperture radar.

[0023] Optionally, in S4, the specific formula of the loss function L is:

[0024] L=αL d +βL sum ;

[0025]

[0026] L sum =||S O -S d || 2 ;

[0027] Where: L d Represents the error between the output value of a pixel and the true deformation value; α is L d The weight coefficient is 0<α<1; β is L sum The weight coefficient, β = 1-α; S O is the sum of the output values ​​of the N channels of a pixel, O k is the output value of the kth channel; S d It is the sum of the true deformation values ​​of N channels of a certain pixel. L sum Indicates S O With S d The error between .

[0028] Optionally, the ground-based radar monitoring data compression method further includes:

[0029] S5: restore the original data according to the pixel coordinates, specifically:

[0030] S5.1. Input the multi-dimensional position encoding tensor and calculate the output O corresponding to each pixel in the compressed data through the neural network model. x ;

[0031] S5.2, according to the normalization process, the output value O of each pixel N channel is x They are converted into the true deformation values ​​corresponding to each grid unit to obtain the original data.

[0032] Optionally, the neural network model constructed in S2 is a four-layer fully connected neural network, and each layer of the fully connected neural network passes through an activation layer, and the activation function of the activation layer adopts a ReLU function; the set threshold in S4 is ε, e-5 ≤ε≤5e -4

[0033] The present invention also provides a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the ground-based radar monitoring data compression method as described above is implemented.

[0034] The present invention also provides an electronic device, comprising: at least one processor, at least one memory and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, the ground-based radar monitoring data compression method as described above is performed.

[0035] The technical solution of the present invention adopts a symmetric triangular wave function to perform position encoding on the coordinates of each pixel of the image, and then through fully connected neural network training, a function that implicitly expresses the correspondence between the position encoding tensor and the deformation value tensor is used through the neural network model, thereby realizing the compression of multiple ground-based radar monitoring data and effectively reducing the storage space of ground-based radar monitoring data, which is convenient for data query.

[0036] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0038] Figure 1 is a schematic diagram of the structure of a neural network model in an embodiment of the present invention;

[0039] Figure 2 is a schematic diagram of an image of ground-based radar monitoring data in an embodiment of the present invention;

[0040] Figure 3 is an image schematic diagram of ground-based radar monitoring data when L>ε in S4 in an embodiment of the present invention;

[0041] Figure 4 It is a schematic diagram of an image of ground-based radar monitoring data when L<ε in S4 in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0043] In one embodiment, a ground-based radar monitoring data compression method comprises the following steps:

[0044] S1: Obtain N consecutive ground-based radar monitoring data and obtain the following Figure 2 The raw data shown in FIG. 1 and the raw data are converted into an N-channel image, where N is a natural number from 2 to 10. The ground-based radar monitoring data includes information of all grid cells of the ground-based radar, and the grid cells contain distance coordinates, angle coordinates and true deformation values, and the grid cells correspond to pixels of the image one by one.

[0045] S2: Construct a neural network model; the neural network model constructed in S2 is a four-layer fully connected neural network, and each layer of the fully connected neural network passes through an activation layer, and the activation function of the activation layer adopts the ReLU function; Figure 1 shown.

[0046] S3: Input the N-channel image into the neural network model to obtain the output value of the N-channel;

[0047] S3.1, obtaining the coordinates of each pixel in the N-channel image;

[0048] In S3.1, the pixel coordinates are (r, θ), where: r is the range coordinate of the grid unit, r takes a natural number from 1 to row, and row is the resolution of the range in the ground-based radar monitoring data; θ is the angle coordinate of the grid unit, θ takes a natural number from 1 to col, and col is the resolution of the angle in the ground-based radar monitoring data.

[0049] S3.2, using two symmetrical triangular wave functions to respectively perform position encoding on the coordinates of each pixel to obtain a multi-dimensional position encoding tensor;

[0050] In S3.2, the specific formulas of the two symmetrical triangular wave functions are:

[0051]

[0052] F 1 (x,i) and F 2 (x,i) is a symmetrical triangular wave function, taking [-1,1]; x is the input value, taking r or θ; T is the period, i is the setting coefficient, i is [3,20];

[0053] In S3.2, the specific formula of the multi-dimensional position encoding tensor γ(X) is:

[0054] γ(X)=|F 1 (r,i), F 2 (r,i), F 1 (θ,i), F 2 (θ,i)|;

[0055] S3.3, respectively normalizing the true deformation value of the N channels corresponding to each pixel to obtain the deformation value tensor of the N channels;

[0056] In S3.3, the specific formula for normalization is:

[0057]

[0058] Where: D k is the deformation value of the kth channel after normalization of a certain grid unit; d k is the deformation value of the kth channel of a grid unit, k is a natural number from 1 to N; λ is the wavelength of the electromagnetic wave of the ground-based synthetic aperture radar.

[0059] S3.4, input the multi-dimensional position encoding tensor into the neural network model for training to obtain the output value O of N channels x ;

[0060] S4: The actual deformation value of the N channel and the output value of the N channel O x Input the loss function and calculate the loss of the neural network model. If the loss is less than the set threshold, save the parameters of the neural network model and get the compressed data; otherwise, adjust the parameters of the neural network model according to the reverse gradient of the loss, set x = x + 1 and return to S3.4, where x is a natural number greater than or equal to 1, set the threshold to ε, e -5 ≤ε≤5e -4 . See Figure 3 and Figure 4 .

[0061] In S4, the specific formula of the loss function L is:

[0062] L=αL d +βL sum ;

[0063]

[0064] L sum =||S O -S d || 2 ;

[0065] Where: L d Represents the error between the output value of a pixel and the true deformation value; L d The weight coefficient is 0<α<1; β is L sum The weight coefficient, β = 1-α; S O is the sum of the output values ​​of the N channels of a pixel, O k is the output value of the kth channel; S dIt is the sum of the true deformation values ​​of N channels of a certain pixel. L sum Indicates S O With S d The error between .

[0066] The ground-based radar monitoring data compression method also includes:

[0067] S5: restore the original data according to the pixel coordinates, specifically:

[0068] S5.1. Input the multi-dimensional position encoding tensor and calculate the output O corresponding to each pixel in the compressed data through the neural network model. x ;

[0069] S5.2, according to the normalization process, the output value O of each pixel N channel is x They are converted into the true deformation values ​​corresponding to each grid unit to obtain the original data.

[0070] In this embodiment, three ground-based radar monitoring data are obtained, that is, N=3, the radar's row=684, col=358, the radar's deformation data is stored using float32 type data, the size of a single data is 0.93MB, and the size of three frames of data is 2.8MB. Through the method of the present invention, the model position encoding i=8, the input is a 32-dimensional tensor, the number of parameters used by the model is 32×512+512+512×128+128+128×128+128+128×3+3=99459, the storage space is, and the data compression effect is very significant.

[0071] The technical solution applied in this embodiment uses a symmetric triangular wave function to perform position encoding on the coordinates of each pixel of the image, and then through fully connected neural network training, a function that implicitly expresses the correspondence between the position encoding tensor and the deformation value tensor is used through the neural network model, thereby achieving compression of multiple ground-based radar monitoring data and effectively reducing the storage space of ground-based radar monitoring data, which is convenient for data query.

[0072] This embodiment also includes a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the ground-based radar monitoring data compression method described above is implemented.

[0073] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0074] This embodiment also includes an electronic device, including: at least one processor, at least one memory, and computer program instructions stored in the memory. When the computer program instructions are executed by the processor, the ground-based radar monitoring data compression method as described above is performed.

[0075] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0076] The electronic device may be a computing device such as a mobile phone, a desktop computer, a notebook, a PDA, a cloud server, etc. The electronic device may include, but is not limited to, a processor and a memory. For example, the electronic device may also include an input / output device, a network access device, a bus, etc.

[0077] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0078] The memory can be used to store the computer program and / or module, and the processor implements the computer program by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0079] Wherein, if the module / unit integrated in the electronic device 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 such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for compressing ground-based radar monitoring data, characterized in that: The steps include: S1: Obtain N consecutive ground-based radar monitoring data to obtain raw data; and convert the raw data into an N-channel image, where N is a natural number from 2 to 10; the ground-based radar monitoring data includes information of all grid cells of the ground-based radar, and the grid cells contain distance coordinates, angle coordinates and true deformation values, and the grid cells correspond to pixels of the image one by one; S2: Build a neural network model; S3: Input the N-channel image into the neural network model to obtain the output value of the N-channel; S3.1, obtaining the coordinates of each pixel in the N-channel image; S3.2, using two symmetrical triangular wave functions to respectively perform position encoding on the coordinates of each pixel to obtain a multi-dimensional position encoding tensor; S3.3, respectively normalizing the true deformation value of the N channels corresponding to each pixel to obtain the deformation value tensor of the N channels; S3.4, input the multi-dimensional position encoding tensor into the neural network model for training to obtain the output value O of N channels x ; S4: The actual deformation value of the N channel and the output value of the N channel O x Input the loss function and calculate the loss of the neural network model. If the loss is less than the set threshold, save the parameters of the neural network model and obtain the compressed data; otherwise, adjust the parameters of the neural network model according to the reverse gradient of the loss, set x=x+1 and return to S3.4, where: x is a natural number greater than or equal to 1.

2. The ground-based radar monitoring data compression method according to claim 1, characterized in that: In S3.1, the pixel coordinates are (r, θ), where: r is the range coordinate of the grid unit, r takes a natural number from 1 to row, and row is the resolution of the range in the ground-based radar monitoring data; θ is the angle coordinate of the grid unit, θ takes a natural number from 1 to col, and col is the resolution of the angle in the ground-based radar monitoring data.

3. The ground-based radar monitoring data compression method according to claim 2, characterized in that: In S3.2, the specific formulas of the two symmetrical triangular wave functions are: F1(x,i) and F2(x,i) are symmetrical triangular wave functions with a value range of [-1,1]; x is the input value, which can be r or θ; T is the period, i is the setting coefficient, i is [3,20]; In S3.2, the specific formula of the multi-dimensional position encoding tensor γ(X) is: γ(X)=|F1(r,i),F2(r,i),F1(θ,i),F2(θ,i)|.

4. The method for compressing ground-based radar monitoring data according to claim 3, characterized in that: In S3.3, the specific formula for normalization processing is: Where: D k is the deformation value of the kth channel after normalization of a certain grid unit; d k is the deformation value of the kth channel of a grid unit, k is a natural number from 1 to N; λ is the wavelength of the electromagnetic wave of the ground-based synthetic aperture radar.

5. The method for compressing ground-based radar monitoring data according to claim 4, characterized in that: In S4, the specific formula of the loss function L is: L=αL d +βL sum ; L sum =||S O -S d || 2 ; Where: L d Represents the error between the output value of a pixel and the true deformation value; α is L d The weight coefficient is 0<α<1; β is L sum The weight coefficient, β = 1-α; S O is the sum of the output values ​​of the N channels of a pixel, O k is the output value of the kth channel; S d It is the sum of the true deformation values ​​of N channels of a certain pixel. L sum Indicates S O With S d The error between .

6. The method for compressing ground-based radar monitoring data according to claim 5, characterized in that: The ground-based radar monitoring data compression method also includes: S5: restore the original data according to the pixel coordinates, specifically: S5.

1. Input the multi-dimensional position encoding tensor and calculate the output O corresponding to each pixel in the compressed data through the neural network model. x ; S5.2, according to the normalization process, the output value O of each pixel N channel is x They are converted into the true deformation values ​​corresponding to each grid unit to obtain the original data.

7. The method for compressing ground-based radar monitoring data according to any one of claims 1 to 6, characterized in that: The neural network model constructed in S2 is a four-layer fully connected neural network, and each layer of the fully connected neural network passes through an activation layer, and the activation function of the activation layer adopts the ReLU function; the set threshold in S4 is ε, e -5 ≤ε≤5e -4 .

8. A computer-readable storage medium, characterized in that: Computer program instructions are stored thereon, and when the computer program instructions are executed by a processor, the ground-based radar monitoring data compression method as described in any one of claims 1 to 6 is implemented.

9. An electronic device, characterized in that: include: At least one processor, at least one memory and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, the ground-based radar monitoring data compression method as described in any one of claims 1 to 6.

Images