Method for constructing jamming situation based on radar seeker active and passive information fusion

Abstract

The application discloses a jamming situation construction method based on radar seeker active and passive information fusion, constructs a jamming situation environment, and forms an active and passive information fusion sample by calculating active information of all platforms and passive information of all devices in the environment, so that the deficiency that the prior art is limited by a known radar feature library, limited fusion information can be acquired, and feature information of a target cannot be effectively extracted is solved. The training set of the active and passive information fusion is used to train the constructed convolutional neural network, so that the trained convolutional neural network has good nonlinear mapping capability, can quickly extract platform and device feature parameters, and obtains corresponding fusion evaluation grades, so that the deficiency that electromagnetic feature calculation error of the target is large and there is false fusion is solved. The active and passive information acquired by the radar seeker in the application has complementarity, and more comprehensive and accurate platform and device information is mastered.

Description

Technical Field

[0001] This invention belongs to the field of radar communication technology, and further relates to a method for constructing an electromagnetic situation based on the fusion of active and passive information from a radar seeker in the field of radar jamming technology. This invention can be used to reflect the situation in complex electromagnetic environments in real time, merging and fusing electromagnetic environment and equipment information to construct an electromagnetic situation, providing a basis for subsequent anti-jamming strategy scheduling. Background Technology

[0002] With the widespread application of electromagnetic technology and frequency-using equipment, the increasingly complex electromagnetic interference environment seriously affects the survival of radar systems. The primary goal of interference situation awareness is to efficiently extract and fuse electromagnetic characteristics from the interference environment in real time to provide timely and accurate intelligence to users. Interference situation construction, as a key technology in interference situation awareness, can use mathematical models to calculate and fuse electromagnetic environment and equipment information that is difficult for humans to understand into easily understood and acceptable numerical values. These values ​​can objectively and in real time reflect the electromagnetic situation and can be used for subsequent anti-jamming strategy scheduling. However, existing interference situation construction methods suffer from problems such as inaccurate information fusion between radar and reconnaissance equipment, lack of alignment with the same target, and weak correlation between situational elements.

[0003] In his paper "Construction of a Knowledge-Based Maritime Battlefield Situation Assessment Auxiliary Decision-Making System" (Command Information Systems and Technology, 2020, 11(4):15-19), Sun Yuxiang, based on the analysis of the main characteristics of electromagnetic situation elements, provides a method for constructing interference situation based on knowledge discovery technology. The implementation steps of this method are as follows: First, using the electromagnetic information of various radiation source targets and combining it with the electromagnetic spectrum information in the radar feature information database to generate electromagnetic situation; Second, using reconnaissance equipment to detect target information and forming intelligence information for situation consistency processing; Third, through situation element extraction and correlation processing, verifying and deduplicating information of different control areas, levels and granularities to form unified situation information, including multi-target event activity information and knowledge from information fusion model analysis; Fourth, using knowledge discovery technology to analyze situation information and comprehensively applying data, knowledge and models to construct the situation. The shortcomings of this method are that it is limited by the known radar feature information database when acquiring situation elements, and the fusion information that can be acquired is limited. Therefore, this method cannot effectively extract the feature information of targets, affecting the accuracy of the fusion results.

[0004] The Rocket Force Engineering University of the Chinese People's Liberation Army disclosed a method for constructing a jamming situation using a diffuse reflector false target in its patent application, "A System and Method for Generating Jamming Situations Using a Diffuse Reflector False Target" (Application No. 202111313009.4, Publication No.: CN114035175A). The method involves the following steps: First, an energy acquisition and measurement subsystem acquires laser energy and measures the laser energy density in front of the mirror; second, an image acquisition subsystem displays the position of the diffuse reflector false target image in the field of view in real time, and adjusts the azimuth and pitch angles of the gimbal based on the position of the diffuse reflector false target image in the field of view to establish a spatial relationship between the diffuse reflector and the target; third, a jamming situation generation subsystem determines the target distance and attitude angle using corner detection and monocular visual pose measurement based on the laser energy density in front of the mirror and the target image, thereby constructing the jamming situation using the false target. However, this method still has shortcomings. It only fuses the diffuse reflector and target information from spatial position information, resulting in significant errors in calculating the electromagnetic characteristics of the target and a problem of incorrect fusion.

[0005] In summary, although many researchers have studied methods for constructing jamming situations, there is still a lack of easily implemented methods for constructing jamming situations that integrate information from both spatial location and electromagnetic characteristics to reflect the correlation features of situational elements in the current complex and ever-changing radar countermeasures process. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of the existing technology by providing a method for constructing an interference situation by fusing active and passive information of a radar seeker. This method solves the problems of existing interference situation construction methods, which are limited by known radar feature libraries, resulting in the lack of acquisition of active and passive information, inability to effectively extract target feature information, large errors in the calculation of target electromagnetic features, and erroneous associations.

[0007] The specific approach to achieving the objective of this invention is as follows: This invention constructs an interference situation environment and calculates an active-passive information fusion sample by combining the active information of all platforms and the passive information of all devices in the environment. This overcomes the shortcomings of existing technologies, which are limited by known radar feature libraries, resulting in limited fusion information and an inability to effectively extract target feature information. This invention constructs a convolutional neural network (CNN). Since CNNs have the ability to handle classification problems, and fusion effect evaluation is essentially a classification problem, the fusion effect evaluation can be performed using the classification capabilities of CNNs. The constructed CNN is trained using the training set of the active-passive information fusion sample. This results in a well-trained CNN with good nonlinear mapping capabilities, enabling it to quickly extract platform and device feature parameters and obtain the corresponding fusion evaluation level. This addresses the shortcomings of large calculation errors in the electromagnetic features of targets, leading to erroneous fusion.

[0008] To achieve the above objectives, the technical solution adopted by the invention includes the following steps:

[0009] Step 1, Constructing the interference situation environment:

[0010] A platform, B radar devices, and C jamming devices are combined to form a jamming situation environment, where A ≥ 10, and the values ​​of B and C are equal to those of A.

[0011] Step 2, generate active and passive information fusion samples:

[0012] Step 2.1: Calculate the radial velocity of each platform, the distance between the seeker and the platform, and the azimuth angle of each platform.

[0013] Step 2.2: Calculate the pulse width, carrier frequency, and pulse repetition period of the signal for each radar device.

[0014] Step 2.3: Calculate the azimuth angle and power of the signal from each interfering device.

[0015] Step 2.4: Combine the active information from all platforms and the passive information from all devices into a fusion sample of active and passive information.

[0016] Step 3, Generate the training set:

[0017] Step 3.1: Select at least 1000 active and passive information fusion samples and normalize the feature parameters of each sample;

[0018] Step 3.2: Calculate the target fusion evaluation level for each sample and label each sample with the corresponding fusion evaluation level label;

[0019] Step 3.3: Compose a training set by combining all normalized samples and their corresponding sample level labels.

[0020] Step 4, construct the convolutional neural network:

[0021] Step 4.1: Build a 12-layer convolutional neural network, whose structure is connected in sequence as follows: first convolutional layer, first pooling layer, first deconvolutional layer, second convolutional layer, second pooling layer, second deconvolutional layer, third convolutional layer, third deconvolutional layer, fourth convolutional layer, fifth convolutional layer, sixth convolutional layer, and fully connected layer.

[0022] Step 4.2: Set the number of convolutional kernels in the first to sixth convolutional layers to 128, 64, 32, 16, 16, and 8 respectively, and set the kernel size to 1×3 for each layer; use max pooling for the first to second pooling layers, set the kernel size to 1×2 for each layer, and set the pooling stride to 1×2 for each layer; set the number of convolutional kernels in the first to third deconvolutional layers to 8, 8, and 16 respectively, and set the kernel size to 1×3 for each layer.

[0023] Step 5, train the convolutional neural network:

[0024] The training set is input into the convolutional neural network, and the parameters of each layer of the convolutional neural network are iteratively updated using the backpropagation gradient descent method until the loss function of the convolutional neural network converges, thus obtaining the trained convolutional neural network.

[0025] Step 6: Assess the degree of integration:

[0026] Using the same normalization method as in step 3.3, the normalized sample for each sample to be evaluated for fusion level is input into the trained convolutional neural network, which outputs the evaluation level of the fusion level of the sample.

[0027] Compared with the prior art, the present invention has the following advantages:

[0028] First, this invention constructs an interference situation environment and, by calculating the active information of all platforms and the passive information of all devices in the environment, forms an active-passive information fusion sample. This overcomes the shortcomings of existing technologies, which are limited by known radar feature libraries, resulting in limited fusion information and an inability to effectively extract target feature information. This allows the invention to support the higher accuracy requirements for establishing fusion relationships between platforms and devices in complex interference situations, and broadens its application scenarios.

[0029] Secondly, this invention utilizes a training set combining active and passive information fusion to train a constructed convolutional neural network. This results in a well-trained convolutional neural network with excellent nonlinear mapping capabilities, enabling it to quickly extract platform and equipment feature parameters and obtain corresponding fusion evaluation levels. This addresses the shortcomings of large errors in the calculation of target electromagnetic features and the resulting erroneous fusion. Furthermore, the active and passive information acquired by the radar seeker in this invention is complementary, providing more comprehensive and accurate platform and equipment information and improving the utilization efficiency of reconnaissance equipment resources. Attached Figure Description

[0030] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0031] Reference Figure 1 The specific implementation steps of the present invention will be further described in detail below with reference to the embodiments.

[0032] Step 1: Construct the interference situation environment.

[0033] A platform, B radar devices, and C jamming devices are combined to form a jamming situation environment, where A ≥ 10, and the values ​​of B and C are equal to those of A.

[0034] Step 2: Generate active and passive information fusion samples.

[0035] Step 2.1: Calculate the platform radial velocity, the distance between the seeker and the platform, and the platform azimuth angle, respectively. These parameters are obtained using the following formulas:

[0036]

[0037]

[0038]

[0039] Among them, v r Let f represent the radial velocity of the platform, λ represent the radar wavelength, and f represent the radial velocity of the platform. d R represents the Doppler frequency shift of the platform, R represents the distance between the seeker and the platform, c represents the speed of light, and t represents the Doppler frequency shift of the platform. R θ represents the platform echo pulse delay time, θ represents the platform azimuth angle, arcsin(·) represents the arcsine function, φ represents the phase difference between the two platform signals received by the seeker, π represents pi, and d represents the antenna spacing.

[0040] Step 2.2: Calculate the pulse width, carrier frequency, and pulse repetition period of the radar equipment signal, which are obtained by the following formulas:

[0041] PW = TOE - TOA

[0042]

[0043] PRI = t k -t k-1

[0044] Where PW represents the pulse width of the radar signal, TOE represents the pulse end time, TOA represents the pulse arrival time, RF represents the carrier frequency of the radar signal, k represents the digital output channel number, and f s The sampling rate is represented by , n represents the number of sampling points, PRI represents the pulse repetition period of the radar device signal, and t represents the sampling rate. k t represents the pulse emission time at time k. k-1 This represents the pulse emission time at time k-1.

[0045] Step 2.3: Calculate the azimuth angle and power of the interfering device signal, which are obtained by the following formulas:

[0046]

[0047]

[0048] Where, θ j φ represents the azimuth angle of the interfering device signal. j P represents the phase difference between the two interference signals received by the seeker. r P represents the power of the interfering device's signal. j G represents the transmission power of the jamming device. j G represents the gain of the transmitting antenna of the jamming device. r L represents the gain of the seeker's receiving antenna. r Indicates the seeker head receiving loss, γ j R represents the seeker polarization loss. j f represents the distance between the seeker and the jamming device. r This indicates the operating frequency of the interfering device.

[0049] Step 3: Generate the training set.

[0050] Step 3.1: Normalize the feature parameters of each sample according to the following formula:

[0051]

[0052] Where, x ij ' represents the j-th target feature parameter of the i-th sample after normalization, j represents the index of the target feature parameter, j=1 represents the radial velocity of the platform, j=2 represents the distance between the seeker and the platform, j=3 represents the azimuth angle of the platform, j=4 represents the pulse width of the radar equipment signal, j=5 represents the carrier frequency of the radar equipment signal, j=6 represents the pulse repetition period of the radar equipment signal, j=7 represents the azimuth angle of the jamming equipment signal, j=8 represents the power of the jamming equipment signal, x ij Let x represent the j-th target feature parameter in the i-th sample. jmax Let x represent the maximum value of the j-th target feature parameter. jmin This represents the minimum value of the j-th target feature parameter.

[0053] Step 3.2, calculate the fusion evaluation level for each sample target according to the following formula:

[0054]

[0055] Among them, G i This represents the fusion evaluation level of the i-th sample, if 0 ≤ Gi A fusion assessment level of ≤0.3 is "poor"; if it is 0.3... <G i A fusion evaluation level of ≤0.7 is "average"; if 0.7 < G, the fusion evaluation level is "general". i When the value is ≤1, the fusion assessment level is "very good", where i represents the sample number, i = 1, 2...N, and N represents the total number of samples. i1 Let u represent the radial velocity of the platform for the i-th sample. i2 u represents the distance between the seeker head and the platform for the i-th sample. i3 Let u represent the azimuth angle of the platform for the i-th sample. i4 u represents the pulse width of the radar device signal for the i-th sample. i5 u represents the carrier frequency of the radar device signal for the i-th sample. i6 u represents the pulse repetition period of the radar device signal for the i-th sample. i7 U represents the azimuth angle of the interfering device signal of the i-th sample. i8 This represents the power of the interfering device signal in the i-th sample.

[0056] Each sample was labeled with a fusion assessment level label, with the label for "poor" fusion assessment level set to 0, the label for "average" fusion assessment level set to 1, and the label for "very good" fusion assessment level set to 2.

[0057] Step 3.3: Compose a training dataset by combining all normalized samples and their corresponding grade labels.

[0058] Step 4: Construct a convolutional neural network.

[0059] Step 4.1: Build a 12-layer convolutional neural network, whose structure is connected in series as follows: first convolutional layer, first pooling layer, first deconvolutional layer, second convolutional layer, second pooling layer, second deconvolutional layer, third convolutional layer, third deconvolutional layer, fourth convolutional layer, fifth convolutional layer, sixth convolutional layer, and fully connected layer.

[0060] Step 4.2: Set the number of convolutional kernels in the first to sixth convolutional layers to 128, 64, 32, 16, 16, and 8 respectively, and set the kernel size to 1×3 for each layer; use max pooling for the first to second pooling layers, set the kernel size to 1×2 for each layer, and set the pooling stride to 1×2 for each layer; set the number of convolutional kernels in the first to third deconvolutional layers to 8, 8, and 16 respectively, and set the kernel size to 1×3 for each layer.

[0061] Step 5: Train the convolutional neural network.

[0062] The training set is input into the convolutional neural network, and the parameters of each layer of the convolutional neural network are iteratively updated using the backpropagation gradient descent method until the loss function of the convolutional neural network converges, thus obtaining the trained convolutional neural network.

[0063] The formula for calculating the loss function is as follows:

[0064]

[0065] Where Loss represents the loss function of the convolutional neural network, N represents the total number of samples, ∑· represents the summation operation, g = 0, 1, ..., m, m represents the total number of fusion evaluation levels, and y ig This represents the relationship between the predicted and true values ​​of the fusion evaluation level g for the i-th sample in the training dataset. When the true and predicted values ​​are equal, then y... ig =1, when the actual value and the predicted value are not equal. ig =0, p ig Let g represent the predicted probability of the fusion evaluation level g of the i-th sample in the training dataset.

[0066] Step 6: Assess the degree of integration.

[0067] Using the same normalization method as in step 3.3, the normalized sample for each sample to be evaluated for fusion level is input into the trained convolutional neural network, which outputs the evaluation level of the fusion level of the sample.

[0068] The effects of this invention will be further illustrated below with simulation experiments:

[0069] 1. Simulation experimental conditions:

[0070] The hardware platform for the simulation experiment of this invention is: Intel i58300H CPU with a main frequency of 2.3GHz and 8GB of memory.

[0071] The software platform for the simulation experiment of this invention is: Windows 10 operating system and Python 3.8.

[0072] The parameters of the training and test sets used in the simulation experiments of this invention are set as follows: the radial velocity of the platform is 16-19 m / s, the distance between the seeker and the platform is 10-15 km, the azimuth angle of the platform is 0-180°, the pulse width (PW) of the radar equipment signal is 10-200 μs, the carrier frequency (RF) of the radar equipment signal is 2.9-3.1 GHz, the pulse repetition period (PRI) of the radar equipment signal is 100-150 μs, the azimuth angle of the jamming equipment signal is 0-180°, and the power of the jamming equipment signal is 80-120 W.

[0073] The training set contains 3000 samples and corresponding fusion evaluation level values, and the test set contains 1000 samples and corresponding fusion evaluation level values. The fusion evaluation results of the present invention are evaluated using the test set.

[0074] The simulation experiment of the present invention uses the following configuration parameters for the convolutional neural network: training batch size is set to 100, learning rate is 0.0005, loss rate is 0.5, and total number of training epochs is 500. The network parameters of the convolutional neural network trained for 100, 200, 300, 400 and 500 epochs using the training set are recorded respectively.

[0075] 2. Simulation content and result analysis:

[0076] The simulation experiment of this invention involves normalizing 1000 samples in the test set using the method of this invention, then randomly dropping data, and inputting the test set after random data loss into a trained convolutional neural network to obtain the fusion evaluation level value of this invention.

[0077] The method of this invention is used to obtain the fusion evaluation level of each sample in the test set. The fusion evaluation level is divided into three categories: "poor", "average", and "very good". Convolutional neural networks trained for 100, 200, 300, 400, and 500 epochs were used to test the test set. The accuracy of the fusion evaluation level "poor" is calculated by dividing the number of correctly predicted samples in the test set by the total number of samples with the "poor" fusion evaluation level; the accuracy of the fusion evaluation level "average" is calculated by dividing the number of correctly predicted samples in the test set by the total number of samples with the "average" fusion evaluation level; and the accuracy of the fusion evaluation level "very good" is calculated by dividing the number of correctly predicted samples in the test set by the total number of samples with the "very good" fusion evaluation level. The final fusion evaluation level prediction results are shown in Table 1.

[0078] Table 1. Predicted Integration Assessment Level (Unit: Percentage)

[0079] Number of training rounds 100 rounds 200 rounds 300 rounds 400 rounds 500 rounds Integration assessment rating: "Poor" 95.32 95.35 96.25 96.32 96.52 Integration assessment level: "Average" 96.73 96.91 97.06 97.19 97.36 The integration assessment level was "very good". 95.46 95.89 96.28 96.47 96.73

[0080] As shown in Table 1, after 500 training cycles of the convolutional neural network, the accuracy rate of the present invention in predicting the fusion evaluation level of "poor" is 96.52%, the accuracy rate in predicting the fusion evaluation level of "average" is 97.36%, and the accuracy rate in predicting the fusion evaluation level of "very good" is 96.63%, which proves that the present invention is relatively accurate in evaluating the fusion of active and passive information under interference conditions.

Claims

1. A method for constructing an interference situation based on radar seeker active and passive information fusion, characterized in that, A fusion sample of active and passive information, consisting of active information from all platforms and passive information from all devices, is generated. A convolutional neural network is trained using this fusion training set. The steps of this interference situation construction method include the following: Step 1, Constructing the interference situation environment: A platform, B radar devices, and C jamming devices are combined to form a jamming situation environment, where A ≥ 10, and the values ​​of B and C are equal to those of A. Step 2, generate active and passive information fusion samples: Step 2.1: Calculate the radial velocity of each platform, the distance between the seeker and the platform, and the azimuth angle of each platform. Step 2.2: Calculate the pulse width, carrier frequency, and pulse repetition period of the signal for each radar device. Step 2.3: Calculate the azimuth angle and power of the signal from each interfering device. Step 2.4: Combine the active information from all platforms and the passive information from all devices into a fusion sample of active and passive information. Step 3, Generate the training set: Step 3.1: Select at least 1000 active and passive information fusion samples and normalize the feature parameters of each sample; Step 3.2: Calculate the target fusion evaluation level for each sample and label each sample with the corresponding fusion evaluation level label; Step 3.3: Compose a training set by combining all normalized samples and their corresponding sample level labels. Step 4, construct the convolutional neural network: Step 4.1: Build a 12-layer convolutional neural network, whose structure is connected in sequence as follows: first convolutional layer, first pooling layer, first deconvolutional layer, second convolutional layer, second pooling layer, second deconvolutional layer, third convolutional layer, third deconvolutional layer, fourth convolutional layer, fifth convolutional layer, sixth convolutional layer, and fully connected layer. Step 4.2: Set the number of convolutional kernels in the first to sixth convolutional layers to 128, 64, 32, 16, 16, and 8 respectively, and set the kernel size to 1×3 for each layer; use max pooling for both the first and second pooling layers, and set the pooling kernel size to 1. ×2, the pooling stride is set to 1×2; the number of convolution kernels in the first to third deconvolution layers are set to 8, 8 and 16 respectively, and the size of the convolution kernels is set to 1×3. Step 5, train the convolutional neural network: The training set is input into the convolutional neural network, and the parameters of each layer of the convolutional neural network are iteratively updated using the backpropagation gradient descent method until the loss function of the convolutional neural network converges, thus obtaining the trained convolutional neural network. Step 6: Assess the degree of integration: Using the same normalization method as in step 3.3, the normalized sample for each sample to be evaluated for fusion level is input into the trained convolutional neural network, which outputs the evaluation level of the fusion level of the sample.

2. The radar seeker active-passive information fusion based jamming situation construction method according to claim 1, characterized in that, The platform radial velocity, the distance between the seeker and the platform, and the platform azimuth angle mentioned in step 2.1 are obtained from the following formulas: Among them, v r Let f represent the radial velocity of the platform, λ represent the radar wavelength, and f represent the radial velocity of the platform. d R represents the Doppler frequency shift of the platform, R represents the distance between the seeker and the platform, c represents the speed of light, and t represents the Doppler frequency shift of the platform. R θ represents the platform echo pulse delay time, θ represents the platform azimuth angle, arcsin(·) represents the arcsine function, φ represents the phase difference between the two platform signals received by the seeker, π represents pi, and d represents the antenna spacing.

3. The method of claim 1, wherein the method further comprises: determining a radar cross section (RCS) of the target; and determining a radar cross section (RCS) of the jammer. The pulse width, carrier frequency, and pulse repetition period of the radar equipment signal mentioned in step 2.2 are obtained by the following formulas: PW = TOE - TOA PRI = t k -t k-1 Where PW represents the pulse width of the radar signal, TOE represents the pulse end time, TOA represents the pulse arrival time, RF represents the carrier frequency of the radar signal, k represents the digital output channel number, and f s The sampling rate is represented by , n represents the number of sampling points, PRI represents the pulse repetition period of the radar device signal, and t represents the sampling rate. k t represents the pulse emission time at time k. k-1 This represents the pulse emission time at time k-1.

4. The radar seeker active-passive information fusion based jamming situation construction method according to claim 1, characterized in that, The azimuth angle and power of the interference device signal mentioned in step 2.3 are obtained by the following formula: Where, θ j φ represents the azimuth angle of the interfering device signal. j P represents the phase difference between the two interference signals received by the seeker. r P represents the power of the interfering device's signal. j G represents the transmission power of the jamming device. j G represents the gain of the transmitting antenna of the jamming device. r L represents the gain of the seeker's receiving antenna. r Indicates the seeker head receiving loss, γ j R represents the seeker polarization loss. j f represents the distance between the seeker and the jamming device. r This indicates the operating frequency of the interfering device.

5. The radar seeker active-passive information fusion based jamming situation construction method according to claim 1, characterized in that, The normalization operation described in step 3.1 is implemented by the following formula: Where, x ij ' represents the j-th target feature parameter of the i-th sample after normalization, j represents the index of the target feature parameter, j=1 represents the radial velocity of the platform, j=2 represents the distance between the seeker and the platform, j=3 represents the azimuth angle of the platform, j=4 represents the pulse width of the radar equipment signal, j=5 represents the carrier frequency of the radar equipment signal, j=6 represents the pulse repetition period of the radar equipment signal, j=7 represents the azimuth angle of the jamming equipment signal, j=8 represents the power of the jamming equipment signal, x ij Let x represent the j-th target feature parameter in the i-th sample. jmax Let x represent the maximum value of the j-th target feature parameter. jmin This represents the minimum value of the j-th target feature parameter.

6. The method of claim 1, wherein, The fusion evaluation level of each sample target mentioned in step 3.2 is calculated by the following formula: Among them, G i This represents the fusion evaluation level of the i-th sample, if 0 ≤ G i A fusion assessment level of ≤0.3 is "poor"; if it is 0.3... <G i A fusion evaluation level of ≤0.7 is "average"; if 0.7 < G, the fusion evaluation level is "general". i When the value is ≤1, the fusion assessment level is "very good", where i represents the sample number, i = 1, 2...N, and N represents the total number of samples. i1 Let u represent the radial velocity of the platform for the i-th sample. i2 u represents the distance between the seeker head and the platform for the i-th sample. i3 Let u represent the azimuth angle of the platform for the i-th sample. i4 u represents the pulse width of the radar device signal for the i-th sample. i5 u represents the carrier frequency of the radar device signal for the i-th sample. i6 u represents the pulse repetition period of the radar device signal for the i-th sample. i7 U represents the azimuth angle of the interfering device signal of the i-th sample. i8 This represents the power of the interfering device signal in the i-th sample.

7. The method of claim 1, wherein, The fusion assessment level labels mentioned in step 3.2 refer to the following: the label for a fusion assessment level of "poor" is set to 0, the label for a fusion assessment level of "average" is set to 1, and the label for a fusion assessment level of "very good" is set to 2.

8. The radar seeker active-passive information fusion based jamming situation construction method according to claim 1, characterized in that, The loss function of the convolutional neural network described in step 5 is as follows: Where Loss represents the loss function of the convolutional neural network, ∑· represents the summation operation, g = 0, 1, ..., m, m represents the total number of fusion evaluation levels, and y ig This represents the relationship between the true value and the predicted value of the fusion evaluation level g for the i-th sample in the training dataset. When the true value and the predicted value are equal, then y... ig =1, when the actual value and the predicted value are not equal. ig =0, p ig Let g represent the predicted probability of the fusion evaluation level g of the i-th sample in the training dataset.

Images