A two-dimensional rainfall field construction method and system, and a storage medium

Abstract

The application discloses a two-dimensional rainfall field construction method and system and a storage medium, and comprises the following steps: determining the position and quantity of millimeter wave radars according to the spatial resolution required for constructing a two-dimensional rainfall field in a target area; performing signal processing on the radar signals emitted omnidirectionally by the millimeter wave radars and the reflected signals received by the millimeter wave radars, obtaining the position and signal attenuation of a temporary link; constructing the corresponding relationship between the signal attenuation and the rainfall intensity by using a deep residual network; the sample set of the deep residual network is composed of the signal attenuation of each millimeter wave radar and the rainfall intensity data of a corresponding synchronous rain gauge; combining the corresponding relationship between the signal attenuation and the rainfall intensity, using microwave rain attenuation to retrieve rainfall intensity, obtaining the rainfall intensity results retrieved by each millimeter wave radar setting point, and collecting the two-dimensional rainfall field of the target area. The application makes full use of the characteristics of omnidirectional signal emission of the millimeter wave radars, combines rain attenuation characteristics to retrieve rainfall intensity, and makes up for the shortcomings of the existing single-transmitting and single-receiving link.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorological factor detection, and particularly relates to a two-dimensional rainfall field construction method and system. BACKGROUND

[0002] The existing method for measuring rainfall is mostly through erecting directional ultra-high frequency wireless microwave equipment on a mobile communication base station to transmit ultra-high frequency microwave signals in a point-to-point manner; then the intensity of the microwave signals is recorded by the receiving end, transmitted to the back-end cloud computing center platform through the Internet of Things data acquisition terminal; finally, the cloud computing center platform provides massive computing resources. The microwave data collected by the front end is analyzed and calculated through the platform, and a two-dimensional rainfall spatial field and time series with high spatial and temporal resolution are fitted out by means of intelligent algorithms. The existing method for inverting rainfall intensity by using the signal attenuation of a single link obtains the average rainfall intensity of the link, which is difficult to reflect the unevenness of the spatial distribution of rainfall on the link, and is difficult to correct the inversion results by using the rainfall station data due to the limitation of the location and number of mobile base stations and the long distance from the rainfall station in most cases. At the same time, the microwave link density of most cities at present cannot meet the requirement of constructing a relatively accurate two-dimensional rainfall field. SUMMARY

[0003] The present application aims to overcome the deficiencies in the prior art and provide a two-dimensional rainfall field construction method and system, which uses the omni-directional signal transmission characteristics of a millimeter wave radar and combines rain attenuation characteristics to invert rainfall intensity, thereby solving the technical problem of the prior art that the measurement result of rainfall intensity by using a single transmitting and single receiving link is not accurate enough.

[0004] To solve the above technical problems, the present application adopts the following technical solutions:

[0005] In a first aspect, the present application provides a two-dimensional rainfall field construction method, which comprises:

[0006] Step S1: determining the position and number of millimeter wave radars according to the spatial resolution required for constructing a two-dimensional rainfall field in a target area;

[0007] Step S2: performing signal processing on the omni-directionally transmitted radar signals and the received reflected signals of the millimeter wave radars to obtain the position and signal attenuation of the temporary link;

[0008] Step S3: using a deep residual network to construct the corresponding relationship between the signal attenuation and the rainfall intensity, and training the sample set of the deep residual network, which is composed of the signal attenuation of each millimeter wave radar and the rainfall intensity data of the corresponding rainfall station at the same period;

[0009] Step S4: combining the corresponding relationship between the signal attenuation amount and the rainfall intensity, using the microwave rain attenuation to retrieve the rainfall intensity, obtaining the rainfall intensity results retrieved by each millimeter wave radar setting point, and collecting to obtain a two-dimensional rainfall field of the target area.

[0010] In combination with the first aspect, further, in step S2, the signal processing process of the millimeter wave radar comprises:

[0011] The sawtooth wave modulation signal generated by the signal generator is input to the voltage-controlled oscillator to generate a frequency-modulated continuous wave signal;

[0012] The frequency-modulated continuous wave signal is divided into two paths after passing through the power divider, one path of the signal is radiated to the free space by the transmitter through the radio frequency front end, and the other path of the signal is adjusted in phase error through the phase shifter;

[0013] The signal received by the receiver is mixed with the signal adjusted in phase error after low-noise amplification processing to obtain a mixed signal;

[0014] The difference frequency signal is obtained after the mixed signal is filtered by the low-pass filter, and the difference frequency signal is converted into a digital signal by the analog-to-digital converter A / D;

[0015] After the digital signal is processed by the signal processing machine, the length, midpoint position and azimuth angle of the corresponding temporary link are obtained.

[0016] In combination with the first aspect, further, in step S2, the signal attenuation amount of the temporary link is the frequency difference between the radar transmission signal and the difference frequency signal.

[0017] In combination with the first aspect, further, in step S3, the deep residual network is set as a ResNet152 residual network model, and the model comprises a convolution layer, a residual block and a fully connected layer;

[0018] The residual block is a pre-activation residual block with a ReLu activation function, and the fully connected layer comprises an average pooling layer and a softmax classifier.

[0019] In combination with the first aspect, further, in step S3, training the deep residual network comprises:

[0020] The network parameters are initialized in a manner of 0 mean value and 0.02 standard deviation;

[0021] The sample set is randomly divided into a training set and a test set in a ratio of 6:4;

[0022] The deep residual network is trained on the training set and the network parameters are updated using the Adam optimization algorithm;

[0023] The updated deep residual network is tested for denoising effect by using the test set, and the network parameters are adjusted according to the accuracy update;

[0024] Based on the trained deep residual network, the corresponding relationship between signal attenuation and rainfall intensity is obtained.

[0025] In combination with the first aspect, further, in step S4:

[0026] The signal attenuation of all temporary links of each millimeter wave radar is input into the trained deep residual network to obtain the average rainfall intensity of each link;

[0027] The average rainfall intensity of each link is taken as the point rainfall intensity at the midpoint position of each corresponding temporary link;

[0028] According to the specified time resolution, the point rainfall intensity data is processed to obtain the point rainfall amount;

[0029] According to the required spatial resolution, the target area is divided into corresponding grids, and the rainfall amount of each grid is calculated using the inverse square distance method, and the calculation formula is as follows:

[0030]

[0031] In the formula: P i represents the rainfall amount of the i-th grid, i=1, 2…Z; wherein Z represents the number of grids of the target area; P h represents the point rainfall amount in the target area; d hi represents the distance from the center of the i-th grid to the point rainfall amount of the h-th point; a represents the number of point rainfall amount positions in the target area.

[0032] In the second aspect, the application provides a two-dimensional rainfall field construction system, which comprises:

[0033] A setting module is configured to determine the position and number of millimeter wave radars according to the spatial resolution required for constructing a two-dimensional rainfall field in a target area;

[0034] A signal processing module is configured to process the radar signals transmitted and reflected by the millimeter wave radars in all directions to obtain the position and signal attenuation of temporary links;

[0035] A corresponding relationship construction module is configured to construct the corresponding relationship between signal attenuation and rainfall intensity by using a deep residual network, and the sample set of the deep residual network is composed of signal attenuation of each millimeter wave radar and rainfall intensity data of a corresponding synchronous rain gauge station;

[0036] An inversion rain intensity module is used to combine the corresponding relationship between the signal attenuation amount and the rain intensity, to obtain the rain intensity result of each millimeter wave radar setting point, and to obtain the two-dimensional rain field of the target area.

[0037] In a third aspect, the present application provides a computer readable storage medium for storing the two-dimensional rain field construction method of any one of the first aspect.

[0038] Compared with the prior art, the present application has the following beneficial effects:

[0039] The present application uses omnidirectional and short-interval millimeter wave radar signals to form a temporary link with a wide coverage range, and the signal transmission is radiated, which has a wide coverage range compared with the existing microwave link. The deep residual network does not have the problems of gradient disappearance and gradient explosion compared with the traditional convolutional neural network, and the pre-activation residual module is used to regularize the model, which is more easily optimized. The deep residual network is denoised, the rain-induced attenuation is accurately identified, and the large error of the traditional method is compensated. The two-dimensional rain field is constructed by using the microwave rain attenuation inversion rain intensity and the synchronous data of the rain gauge station, which is more efficient and more accurate than the traditional link. In addition, the method of the present application can be used to build equipment according to the development of the target area, which is not limited by the location of the existing commercial microwave transmission tower, has strong applicability, small investment, high efficiency, and is suitable for popularization and use. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flow chart of a two-dimensional rain field construction method provided by an embodiment of the present application;

[0041] Figure 2 is a structural principle block diagram of a two-dimensional rain field construction system provided by an embodiment of the present application;

[0042] Figure 3 is a schematic diagram of a millimeter wave radar structure provided by an embodiment of the present application;

[0043] Figure 4 is a schematic diagram of a target area grid division provided by an embodiment of the present application. DETAILED DESCRIPTION

[0044] The technical solutions of the present application will be described in detail below with reference to the drawings and specific embodiments. It should be understood that the specific features in the embodiments and the specific features in the embodiments are detailed descriptions of the technical solutions of the present application, and are not limitations of the technical solutions of the present application. In the case of no conflict, the technical features in the embodiments and the embodiments can be combined with each other.

[0045] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0046] Example 1:

[0047] like Figure 1 As shown in the figure, this embodiment introduces a method for constructing a two-dimensional rainfall field, which specifically includes the following steps:

[0048] Step S1: Determine the location and number of millimeter-wave radars based on the spatial resolution required to construct a two-dimensional rainfall field for the target area;

[0049] In this step, a comprehensive analysis is conducted using rainfall data, flood data, hydrological and topographical data, land use data, and socio-economic data of the target area to determine the required spatiotemporal resolution and accuracy for designing the two-dimensional rainfall field.

[0050] Step S2: The millimeter-wave radar performs signal processing on its omnidirectional transmitted radar signal and received reflected signal to obtain the location of the temporary link and the signal attenuation.

[0051] Reference Figure 3 The signal processing process of millimeter-wave radar includes:

[0052] The sawtooth wave modulation signal generated by the signal generator is input to the voltage-controlled oscillator to generate a frequency-modulated continuous wave signal.

[0053] After passing through a power divider, the frequency-modulated continuous wave signal is split into two paths. One path passes through the radio frequency front end and the transmitter radiates electromagnetic energy into free space. The other path passes through a phase shifter for phase error adjustment.

[0054] The signal received by the receiver is processed by a low-noise amplifier and then mixed with the phase error adjusted signal to obtain a mixed signal.

[0055] After the low-pass filter removes the multiplication factor of the mixing signal, the difference frequency signal is obtained. The difference frequency signal is then converted into a digital signal by an analog-to-digital converter (A / D).

[0056] After the digital signal is processed by the signal processor, the length, midpoint position, and azimuth of the corresponding temporary link are obtained.

[0057] Specifically, the digital signal received by the signal processor is represented as:

[0058]

[0059] In the formula: AT A represents the amplitude of the signal transmitted by the transmitter. R f represents the amplitude of the signal received by the receiver. c T represents the initial frequency of the carrier, α represents the slope of the frequency modulation, and T represents the initial frequency of the carrier. m R0 represents the duration of the repetition period, R0 represents the distance of the target as a function of time when it is stationary, l represents the antenna number (i.e., the signal data received from antenna number l), the spatial interval between antennas is d, θ represents the angle between the target and the radar normal, v represents the velocity of the target object, and c represents the propagation speed of electromagnetic waves in free space, taken as 3 × 10-1. 8 m / s, K is the number of receiving antennas, and j is the exponent.

[0060] According to formula (1), the digital signal is processed at a sample frequency f s We perform sampling and list the independent variables for each dimension in parentheses on the left side of the equals sign as shown below:

[0061]

[0062] In the formula: n represents the sampling sequence number within a single frequency-modulated continuous wave cycle, n = 1, 2...N, and the time interval is t; m represents the sampling point in the m-th cycle, m = 1, 2...M, and the time interval is T. m The sampled data forms an M×N matrix.

[0063] Based on formula (2), the distance and velocity are jointly estimated, and the signal matrices in the fast and slow time domains are obtained as follows:

[0064]

[0065] The target distance and the target speed can be calculated by performing a fast Fourier transform on the signal matrix.

[0066] Similarly, the signal matrix for estimating the position and distance is as follows:

[0067]

[0068] By performing a joint angle and range estimation algorithm on this matrix, the parameters of the distance and azimuth between the target object and the radar can be estimated.

[0069] The target distance is used as the link length, and the midpoint of the link is determined using the target azimuth and distance.

[0070] As one embodiment of the present invention, the signal attenuation of the link is obtained by subtracting the receiving frequency from the transmission frequency of the radar signal.

[0071] After completing the radar setting, stably run for a period of time, collect the signal attenuation amount and the data of the nearby rainfall station in the same period (at least one month).

[0072] Step S3: using a deep residual network, constructing the corresponding relationship between the signal attenuation amount and the rainfall intensity, and training the sample set of the deep residual network, which is composed of the signal attenuation amount of each millimeter wave radar and the corresponding rainfall intensity data of the rainfall station in the same period;

[0073] As an embodiment of the present application, constructing the deep residual network comprises: applying a convolutional neural network ResNet_152 residual network model, setting a convolutional layer, a residual block and a fully connected layer; in the residual unit, introducing a shortcut connection from the input end to the output end of each convolutional layer, using an identity mapping as the shortcut connection, and reducing the complexity of the residual network; the residual block of the present application uses a pre-activated residual block, the activation function selects ReLu, the batch sample normalization operation is performed before each layer of convolution, and the activation function is used to increase the non-linear factor, solving the defect of insufficient expression ability of the linear model, strengthening the model regularization, meanwhile, the activation function after addition is deleted, so that the network model is more easily optimized, and the fully connected layer includes an average pooling layer and a softmax classifier.

[0074] In step S3 of the embodiment of the present application, the step of training the deep residual network comprises:

[0075] Step S3.1: initializing the network parameters in a manner of 0 mean value and 0.02 standard deviation;

[0076] Step S3.2: randomly dividing the sample set into a training set and a test set according to a proportion of 6:4;

[0077] Wherein, when the sample set is made, the signal attenuation amount of the temporary link is taken as the input data of the deep residual network, and the corresponding rainfall intensity data of the rainfall station in the same period is taken as the output data of the deep residual network;

[0078] Step S3.3: training the deep residual network on the training set and updating the network parameters by using the Adam optimization algorithm;

[0079] As an embodiment of the present application, the training number is set to 400 times, wherein the Adam optimization algorithm iterates the first moment estimation and the second moment estimation of the gradient, so as to update the neural network weight; the first moment estimation of the gradient m t As follows:

[0080] m t =β1m t-1 +(1-β1)g t (5)

[0081] In the formula, m tThe first moment estimation of the network gradient of the tth layer; m t-1 The first moment estimation of the network gradient of the t-1th layer, β1 is the exponential decay rate of the first moment estimation, and β1 is 0.9; β2 is the exponential decay rate of the second moment estimation, and β2 is 0.999;

[0082] The second moment estimation v t As follows:

[0083] v t = β2v t-1 + (1- β2) g t 2 (6)

[0084] In the formula, v t The second moment estimation of the network gradient of the tth layer, v t-1 The first moment estimation of the network gradient of the t-1th layer;

[0085] The gradient g t of the weight parameter of the network of the tth layer is:

[0086]

[0087] In the formula, β1 is the exponential decay rate of the first moment estimation, and β2 is the exponential decay rate of the second moment estimation, and β2 is 0.999;

[0088] The bias correction of m t , v t is as follows:

[0089]

[0090]

[0091] Based on the bias-corrected m t , v t , the weight parameters of the network model are updated by formula (10) as follows:

[0092]

[0093] In the formula, θ t The weight coefficient of the tth layer network, θ t-1 The weight coefficient of the t-1th layer network, t=1,2…152; a is the learning rate, and a is 0.001; ε is set to 10 to avoid the divisor being 0 -18 .

[0094] In the embodiment of the application, the loss function of the training network adopts a cross-entropy loss function, and L2 regularization is used to punish the weight parameters, thereby reducing the overfitting phenomenon.

[0095] Step S3.4: test the denoising effect of the updated deep residual network with the test set, and update and adjust the network parameters according to the accuracy;

[0096] Step S4: combining the corresponding relationship between the signal attenuation and the rainfall intensity, using the microwave rain attenuation to retrieve the rain intensity, obtaining the rain intensity results retrieved by each millimeter wave radar setting point, and collecting to obtain the two-dimensional rainfall field of the target area.

[0097] As an embodiment of the present application, step S4 specifically comprises:

[0098] Step S4.1: input the signal attenuation of all temporary links of each millimeter radar into the trained deep residual network to retrieve the average rain intensity of each link;

[0099] Step S4.2: taking each said link average rain intensity as the point rainfall intensity at the midpoint position of each corresponding temporary link;

[0100] Step S4.3: processing each point rainfall intensity data according to the specified time resolution to obtain each point rainfall;

[0101] Step S4.4: dividing the target area into corresponding grids according to the required spatial resolution, as shown in Figure 4 , each grid collects the point rainfall at the same time on the target area, and the rainfall of each grid is calculated by using the inverse square distance method according to formula (11):

[0102]

[0103] In the formula: P i represents the rainfall of the i-th grid, i=1, 2…Z; wherein Z represents the number of grids of the target area; P h represents the point rainfall in the target area; d hi represents the distance from the center of the i-th grid to the point rainfall of the h-th point; a represents the number of point rainfall positions in the target area.

[0104] The two-dimensional rainfall field construction method provided by the embodiment of the present application uses the characteristics of omnidirectional transmission signal and rain attenuation of millimeter wave radar to retrieve rain intensity, which makes up for the deficiency of single transmitting and single receiving link in the prior art. Compared with the traditional method, the present application has strong applicability, small investment, high efficiency, and is suitable for popularization and use.

[0105] Embodiment two:

[0106] As shown in Figure 2 , the embodiment of the present application provides a two-dimensional rainfall field construction system, which can be used to implement the method described in embodiment one, and specifically comprises:

[0107] The setting module is configured to determine the position and quantity of the millimeter wave radar according to the spatial resolution required for constructing the two-dimensional rainfall field in the target area;

[0108] The signal processing module is configured to perform signal processing on the radar signal transmitted omnidirectionally by the millimeter wave radar and the received reflected signal, to obtain the position and signal attenuation of the temporary link;

[0109] The constructing correspondence module is configured to construct the correspondence between the signal attenuation and the rainfall intensity by using a deep residual network, and the sample set of the deep residual network is composed of the signal attenuation of each millimeter wave radar and the rainfall intensity data of the corresponding synchronous rain gauge station.

[0110] The inversion rainfall intensity module is configured to combine the correspondence between the signal attenuation and the rainfall intensity, and use the microwave rain attenuation to invert the rainfall intensity, to obtain the rainfall intensity result inverted by each millimeter wave radar setting point, and to obtain the two-dimensional rainfall field of the target area by summarizing.

[0111] The two-dimensional rainfall field constructing system provided by the embodiment of the present application is based on the same technical concept as the two-dimensional rainfall field constructing method provided by the first embodiment, and can produce the beneficial effects as described in the first embodiment, and the contents not described in detail in the present embodiment can be referred to the first embodiment.

[0112] Embodiment three:

[0113] The embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of any one of the methods in the first embodiment.

[0114] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0115] The present application is described with reference to the flowcharts and / or block diagrams according to the methods, devices (systems), and computer program products of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a machine that implements the functions described in the flowcharts and / or block diagrams. Figure 1one or more processes and / or blocks Figure 1 an apparatus for performing the functions specified in the flowchart or multiple flows and / or blocks.

[0116] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flowchart or multiple flows and / or blocks. Figure 1 one or more processes and / or blocks Figure 1 an apparatus for performing the functions specified in the flowchart or multiple flows and / or blocks.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart or multiple flows and / or blocks. Figure 1 one or more processes and / or blocks Figure 1 an apparatus for performing the functions specified in the flowchart or multiple flows and / or blocks.

[0118] The above description is merely the preferred embodiments of the present application, and it should be pointed out that for those skilled in the art, some improvements and modifications can be made without departing from the technical principles of the present application, and these improvements and modifications should also be considered as falling within the scope of protection of the present application.

Claims

1. A two-dimensional rain field construction method, characterized by, The method comprises: Step S1: determining the position and number of millimeter wave radars according to the spatial resolution required for constructing a two-dimensional rainfall field in a target area; Step S2: the millimeter wave radars perform signal processing on the radar signals transmitted in all directions and the received reflected signals to obtain the position and signal attenuation of a temporary link; Step S3: a deep residual network is used to construct the correspondence between the signal attenuation and the rainfall intensity, and the sample set of the deep residual network is composed of the signal attenuation of each millimeter wave radar and the rainfall intensity data of a corresponding synchronous rain gauge station; Step S4: combining the correspondence between the signal attenuation and the rainfall intensity, the rainfall intensity is inversed using the microwave rain attenuation, the rainfall intensity results obtained by each millimeter wave radar setting point are obtained, and the two-dimensional rainfall field of the target area is obtained by summarizing; In step S4: the signal attenuation of all temporary links of each millimeter wave radar is input into the trained deep residual network to obtain the average rainfall intensity of each link; the average rainfall intensity of each link is taken as the point rainfall intensity at the midpoint position of each corresponding temporary link; the point rainfall intensity data is processed according to the specified time resolution to obtain the point rainfall; the target area is divided into corresponding grids according to the required spatial resolution, and the rainfall of each grid is calculated using the inverse square distance method, and the calculation formula is as follows: ， In the formula: represents the rainfall of the th grid, ; wherein represents the number of grids of the target area; represents the rainfall of the hth point in the target area; represents the distance from the th grid center to the rainfall of the hth point; and a represents the number of point rainfall locations in the target area.

2. The two-dimensional rainfall field construction method according to claim 1, characterized by, In step S2, the signal processing process of the millimeter wave radar comprises: a sawtooth wave modulation signal generated by a signal generator is input into a voltage-controlled oscillator to generate a frequency-modulated continuous wave signal; the frequency-modulated continuous wave signal is divided into two paths after passing through a power divider, one path of the signal passes through a radio frequency front end to be radiated into free space by a transmitter, and the other path of the signal passes through a phase shifter to adjust the phase error; the signal received by the receiver is mixed with the signal after the phase error adjustment after low-noise amplification processing to obtain a mixed signal; a low-pass filter filters out the frequency multiplication part of the mixed signal to obtain a difference frequency signal, and the difference frequency signal is converted into a digital signal by an analog-to-digital converter A / D; after the digital signal is processed by a signal processing machine, the length, midpoint position and azimuth angle of the corresponding temporary link are obtained.

3. The two-dimensional rainfall field construction method according to claim 2, characterized by, In step S2, the signal attenuation of the temporary link is the frequency difference between the radar transmitted signal and the difference frequency signal.

4. The two-dimensional rainfall field construction method according to claim 1, characterized by, In step S3, the deep residual network is set as a ResNet152 residual network model, and the model comprises a convolution layer, a residual block and a full connection layer; the residual block is a pre-activation residual block with a ReLu activation function, and the full connection layer comprises an average pooling layer and a softmax classifier.

5. The two-dimensional rainfall field construction method according to claim 4, characterized by, In step S3, training the deep residual network comprises: initializing the network parameters in the manner of 0 mean value and 0.02 standard deviation; randomly dividing the sample set into a training set and a test set in a proportion of 6:4; training the deep residual network on the training set and updating the network parameters using the Adam optimization algorithm; testing the denoising effect of the updated deep residual network using the test set, and adjusting the network parameters according to the accuracy; obtaining the correspondence between the signal attenuation and the rainfall intensity based on the trained deep residual network.

6. A two-dimensional rain field construction system characterized by comprising: The system comprises: The setting module is configured to determine the position and quantity of the millimeter wave radar according to the spatial resolution required for constructing the two-dimensional rainfall field in the target area; The signal processing module is configured to perform signal processing on the radar signals transmitted and the reflected signals received by the millimeter wave radar in an all-directional manner, to obtain the position and signal attenuation of the temporary link; The constructing corresponding relationship module is configured to construct the corresponding relationship between the signal attenuation and the rainfall intensity by using a deep residual network, and to train the sample set of the deep residual network, which is composed of the signal attenuation of each millimeter wave radar and the rainfall intensity data of the corresponding synchronous rain gauge station; The inversion rainfall intensity module is configured to combine the corresponding relationship between the signal attenuation and the rainfall intensity, to use the microwave rain attenuation to invert the rainfall intensity, to obtain the rainfall intensity results obtained by inverting the setting points of each millimeter wave radar, and to obtain the two-dimensional rainfall field of the target area by summarizing. In the inversion rainfall intensity module: The signal attenuation of all temporary links of each millimeter wave radar is input into the trained deep residual network to obtain the average rainfall intensity of each link; The average rainfall intensity of each link is taken as the point rainfall intensity at the midpoint position of each corresponding temporary link; The point rainfall intensity data is processed according to the specified time resolution to obtain the point rainfall of each point; According to the required spatial resolution, the target area is divided into corresponding grids, and the distance square reciprocal method is used to calculate the rainfall of each grid, and the calculation formula is as follows: ， In the formula: represents the rainfall of the th grid, ; wherein represents the number of grids of the target area; represents the rainfall of the hth point in the target area; represents the distance from the th grid center to the rainfall of the hth point; and a represents the number of point rainfall locations in the target area.

7. A computer readable storage medium for storing the two-dimensional rainfall field construction method of any one of claims 1-5.

Images