A Cognitive Radar Waveform Design Method and Device for Maneuvering Target Path Prediction

Abstract

The present invention discloses a cognitive radar waveform design method and device for maneuvering target path prediction. The method includes: detecting a moving target by the radar transmitting a narrowband waveform; training the obtained echo through an LSTM network to construct a time series prediction model; predicting the filtered trajectory of the moving target to obtain the target trajectory; determining the spatial position information of the moving target relative to the radar through the detection pulses transmitted by the radar; determining the elevation angle information and azimuth angle information of the moving target according to the result of the filtered trajectory prediction and the spatial position information; calculating the RCS information of the moving target according to the elevation angle information and azimuth angle information; determining the inter-pulse DFC coding information according to the RCS information; and determining the transmission frequency of each pulse in each group of pulses transmitted by the radar according to the inter-pulse DFC coding information. The present invention can improve the low intercept and anti-jamming performance of the radar and can be widely applied to the field of radar data processing technology.

Description

Technical Field

[0001] The present invention relates to the technical field of radar data processing, and in particular to a cognitive radar waveform design method and device for maneuvering target path prediction. Background Art

[0002] A radar is a radio detection and ranging device, which has excellent performance such as a long detection range, long working hours, and strong penetration ability, and is widely used in related fields such as national defense and civilian use. How to obtain target characteristics, motion and other information from radar scattered echoes is the fundamental means for radar to meet the needs of related fields. In order to adapt to the rapidly developing technology and interact with environmental state information to achieve the purpose of changing the radar's own transmission parameters according to changes in the environmental state, the concept of cognitive radar system has been proposed in the radar field and has become a hot issue in the research of radar technology. Cognition was originally a biological concept, referring to a series of mental activities that a person has when in a certain environment, including observation, thinking, judgment, reasoning, etc. Therefore, introducing this concept into radar technology means that the radar can perceive and learn environmental information according to the constantly changing complex working scenarios, and can timely adjust its own relevant parameters, so as to achieve technical operations such as detection, recognition, and classification, and provide highly reliable data and technical references for signal processing.

[0003] In order to reduce the risk of radar signals being intercepted and detected by interceptors, radio frequency stealth technology has been vigorously developed in recent years. The purpose of stealth is to put the other party in an environment of continuous speculation, making it difficult to interpret its active feature information, or submerging its information in a noise environment so that it cannot be effectively accumulated for a long time. The method of reducing active features is called LPI technology. In order to improve the LPI performance of radar, the main current approach is complex radar waveform design technology and radar cognitive frequency emission and power distribution technology.

[0004] When using radar for the task of tracking air targets, in order to improve target tracking performance and have excellent LPI capabilities, having advanced fire control radar technology is a key factor to achieve this goal. During the process of radar tracking a target, the radar transmitter usually first uses a high-power transmission waveform to irradiate the target to achieve the purpose of detecting and tracking the target. However, this will cause the received echo signal to be mixed with high-power clutter, increasing the difficulty of signal processing, and the high-power signal is easily intercepted, reducing the radar's LPI performance.

[0005] It can be known from theoretical analysis that the size of the target RCS affects the probability of the target being detected by the radar. The larger the RCS, the greater the possibility of being detected by the radar. At present, electromagnetic stealth of related equipment usually adopts various methods to minimize the effective scattering cross-section of the target for electromagnetic waves, so as to reduce the probability of being detected by the opponent's radar. With the rapid development of technologies such as stealth and anti-stealth, radar target recognition technology, and dynamic precise navigation, the performance requirements for estimating the RCS of complex targets have also increased. It is not only necessary to make specific optimizations and designs for conventional targets to minimize the RCS value of the target to avoid radar detection. At the same time, the radar signal processing system also needs to adaptively adjust the detection algorithm to ensure the calculation accuracy and estimation accuracy of the target RCS, thereby reducing missed detections.

[0006] Since Kalman proposed the Kalman filter, radar target tracking has been greatly developed. It uses the target measurement echoes obtained during the target detection process to continuously update the target state information. Once the initial state information of the target is determined, the track of the target can be formed and the state information of the target at different times can be predicted, making the radar play a powerful role in both military and civilian fields. At the same time, the Kalam filter (Kalman filter) is an optimal linear filter. Due to the continuous changes in application scenarios such as modern battlefields and cities, the radar detection objects and environmental states are complex and changeable. Therefore, the radar emission parameters during target tracking also need to change with the changes of the target and the environment. In traditional radar target tracking, the detection and tracking stages are two relatively independent parts. Even for a target tracking system with an adaptive filtering function, it only adjusts the algorithm parameters during the filtering process and only achieves stable target tracking within the scope of data processing. When the external environment changes and the target's own motion state is abnormal, such as when the target is affected by clutter interference or makes a rapid maneuver, due to the fixed radar detection and filtering parameters and the deterioration of the measurement quality, the effect of target tracking will still deteriorate accordingly, and even a loss-of-track phenomenon will occur over time. Therefore, at present, there is an urgent need for a software and hardware adaptive radar that can adapt to changes in external environmental conditions and the target's own state. By perceiving the target state information during target tracking, it predicts the state of the target at subsequent moments and guides the adjustment of the software and hardware parameters of the transmitter, thereby ensuring the quality of subsequent measurements and improving the target tracking performance. However, due to structural limitations, traditional mechanical scanning radars generally use the rotation of their own arrays to detect targets, and the scanning frequency cannot be changed. At the same time, it is also difficult to effectively change parameters such as the wavelength and transmission power of the radar emission signal by using target and environmental information. Summary of the Invention

[0007] In view of this, the embodiments of the present invention provide a cognitive radar waveform design method and device for maneuvering target path prediction to improve the low intercept and anti-interference performance of the radar.

[0008] On the one hand, an embodiment of the present invention provides a cognitive radar waveform design method for maneuvering target path prediction, including:

[0009] The radar emits a narrowband waveform to detect a moving target, and obtains an echo containing target state information;

[0010] Train the echo through an LSTM network to construct a time series prediction model;

[0011] Perform filtered trajectory prediction on the moving target according to the time series prediction model to obtain the target trajectory;

[0012] According to the target trajectory, determine the spatial position information of the moving target relative to the radar through the detection pulses emitted by the radar;

[0013] According to the result of the filtered trajectory prediction and the spatial position information, determine the elevation angle information and azimuth angle information of the moving target;

[0014] Calculate the RCS information of the moving target according to the elevation angle information and the azimuth angle information;

[0015] Determine the inter-pulse DFC coding information according to the RCS information;

[0016] Determine the transmission frequency of each pulse in each group of pulses emitted by the radar according to the inter-pulse DFC coding information.

[0017] Optionally, the radar emits a narrowband waveform to detect a moving target, and the obtained echo containing target state information includes:

[0018] By performing pulse compression on the echo of the transmitted LFM waveform pulse group, obtain the change in the distance of the target relative to the radar during the radar detection stage;

[0019] Process the radar echo information at different times to obtain the target motion trajectory data set during the radar detection stage.

[0020] Optionally, according to the target trajectory, determining the spatial position information of the moving target relative to the radar through the detection pulses emitted by the radar includes:

[0021] According to different detection pulses emitted by the radar, obtain different RCS frequency response functions reflected by the moving target;

[0022] According to the target information echo obtained by the radar in different postures, predict the position information of the moving target at different times.

[0023] Optionally, determining the pitch angle information and azimuth angle information of the moving target according to the result predicted by the filtering trajectory and the spatial position information includes:

[0024] According to the result predicted by the filtering trajectory and the spatial position information, through the calculation of trigonometric functions, the pitch angle and azimuth angle of the moving target relative to the radar are obtained.

[0025] Optionally, calculating the RCS information of the moving target according to the pitch angle information and the azimuth angle information includes:

[0026] Obtain the number of discrete strong scattering centers on the moving target;

[0027] Obtain the scattering coefficients of each strong scattering point and the phase of each scattering point relative to the scattering center;

[0028] Decompose the moving target into multiple strong scattering points;

[0029] According to the number of strong scattering centers, the scattering coefficients and the phases, calculate the complex target RCS information of multiple strong scattering points.

[0030] Optionally, in the process of calculating the RCS information, it further includes:

[0031] According to the transmitted signal of the radar and the scattering coefficients of each scattering point, calculate the echo information at the radar receiver end and the projection distance of the scattering point on the radar line of sight from the radar receiving end.

[0032] Optionally, the method further includes:

[0033] According to the maximum RCS frequency response function and the corresponding frequency at the current angle of the moving target azimuth, use the radar equation to calculate the optimal transmission power used by the radar at the current moment;

[0034] Perform power allocation according to the optimal transmission power.

[0035] Another aspect of the embodiments of the present invention further provides a cognitive radar waveform design device for maneuvering target path prediction, including:

[0036] The first module is used to detect the moving target by the radar transmitting a narrowband waveform to obtain an echo containing target state information;

[0037] The second module is used to train the echo through an LSTM network to construct a time series prediction model;

[0038] The third module is used to perform filtering trajectory prediction on the moving target according to the time series prediction model to obtain a target trajectory;

[0039] A fourth module, configured to determine spatial position information of the moving target relative to the radar according to the target trajectory and detection pulses emitted by the radar;

[0040] A fifth module, configured to determine elevation angle information and azimuth angle information of the moving target according to the result predicted by the filtered trajectory and the spatial position information;

[0041] A sixth module, configured to calculate RCS information of the moving target according to the elevation angle information and the azimuth angle information;

[0042] A seventh module, configured to determine inter-pulse DFC coding information according to the RCS information;

[0043] An eighth module, configured to determine the transmission frequency of each pulse in each group of pulses emitted by the radar according to the inter-pulse DFC coding information.

[0044] Another aspect of the embodiments of the present invention further provides an electronic device, including a processor and a memory;

[0045] The memory is used to store a program;

[0046] The processor executes the program to implement the method as described above.

[0047] Another aspect of the embodiments of the present invention further provides a computer-readable storage medium, where the storage medium stores a program, and the program is executed by a processor to implement the method as described above.

[0048] The embodiments of the present invention also disclose a computer program product or a computer program, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the method as described above.

[0049] Embodiments of the present invention detect moving targets by a radar transmitting a narrowband waveform, and obtain echoes containing target state information; train the echoes through an LSTM network to construct a time series prediction model; predict the filtered trajectory of the moving target according to the time series prediction model to obtain the target trajectory; determine the spatial position information of the moving target relative to the radar according to the target trajectory through the detection pulses transmitted by the radar; determine the elevation angle information and azimuth angle information of the moving target according to the result of the filtered trajectory prediction and the spatial position information; calculate the RCS information of the moving target according to the elevation angle information and the azimuth angle information; determine the inter-pulse DFC coding information according to the RCS information; and determine the transmission frequency of each pulse in each group of pulses transmitted by the radar according to the inter-pulse DFC coding information. The present invention first uses an LSTM network to predict the path of a target detected and accumulated for a period of time, and at the same time combines the RCS frequency response characteristics of the target to perform inter-pulse DFC of the radar transmitting waveform during the path prediction stage. This can effectively reduce the energy consumption of the radar during target detection. At the same time, since different combinations of inter-pulse frequency coding are used according to the state information of the target during the radar pulse transmission process, the low intercept and anti-jamming performance of the radar can be improved. Description of the Drawings

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.

[0051] Figure 1 It is the overall step flow chart provided by the embodiments of the present invention;

[0052] Figure 2 It is a schematic diagram of radar narrowband target detection;

[0053] Figure 3 It is a schematic diagram of simulating the target motion trajectory;

[0054] Figure 4 It is a schematic diagram of the recurrent neural network structure;

[0055] Figure 5 It is a schematic diagram of the recurrent neural network unfolded according to the time series;

[0056] Figure 6 It is a schematic diagram of the LSTM network system structure;

[0057] Figure 7 It is a schematic diagram of the LSTM network system structure unfolded;

[0058] Figure 8 Schematic diagram of the LSTM forget gate structure;

[0059] Figure 9 Schematic diagram of the LSTM input gate structure;

[0060] Figure 10 Schematic diagram of the LSTM output gate structure;

[0061] Figure 11 Schematic diagram of the LSTM target motion trajectory prediction;

[0062] Figure 12 Schematic diagram of the maneuvering target attitude angle and radar elevation angle;

[0063] Figure 13 Schematic diagram for studying the use of the target RCS fluctuation model;

[0064] Figure 14 Schematic diagram of the target detection capability of the LFM waveform;

[0065] Figure 15 Schematic diagram of the target detection capability of the cognitive frequency agile waveform. Detailed implementation manners

[0066] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0067] In recent years, deep neural networks have been widely applied due to the greatly enhanced computing power of computers. There are many different network frameworks in deep learning to address various problems. For feed-forward convolutional neural network frameworks like convolutional neural networks, they are commonly used in applications such as classification, recognition, and decision-making. However, feed-forward networks cannot preserve past information, and their performance in tasks like sequence prediction is often poor. By introducing recurrent networks, the information at the current moment can be connected to the information at the previous moment. This enables the input of the current network layer to not only obtain the output result of the previous layer but also the output of the hidden layer nodes at those previous moments. This endows the recurrent neural network with the characteristic of having short-term memory to preserve information. However, recurrent neural networks can only describe short-term dynamic behaviors and cannot capture long-term associations between data. For target motion estimation, multiple point-trajectory samples over a period of time are often required to exhibit strong motion characteristics, and the trained model will have strong robustness characteristics. Using only recurrent neural networks for model training will lead to problems such as a decline in the system's robustness. Moreover, as the iteration progresses, its weight coefficients may grow or decline exponentially, eventually resulting in gradient explosion or gradient vanishing, which causes the entire model to diverge. At the same time, the proposed LSTM network can just solve this problem.

[0068] The scattered echo of a target contains the position, motion speed, characteristic size, etc. of the detected target, which is the basis of radar signal processing. RCS is a physical quantity that measures the strength of the electromagnetic waves scattered by a target and affects the data quality of the target echo to a certain extent. Therefore, high-quality accurate acquisition and processing of radar echoes are inseparable from the study of RCS. At the same time, the study of RCS also has extremely important research significance for waveform optimization, resource management, target detection, target imaging, target recognition, etc. of cognitive radars. Accurately judging and estimating the changing trend of target RCS can more effectively utilize the parameters of the radar transmission waveform, more reasonably allocate radar resources, ensure the data quality of the target echo, and ensure the measurement accuracy of the radar. In a low signal-to-noise ratio environment, if the target RCS can be effectively estimated, the parameters of the transmission waveform can be selectively adjusted accordingly, enabling the radar to extract target information from the echo in the best way, thereby improving the radar's working performance. At the same time, due to the different electromagnetic scattering characteristics caused by the differences in the physical structures of targets, the uniqueness of a target can be reflected in its RCS. Therefore, accurately acquiring and analyzing the target RCS helps to estimate information such as the size, shape, and attitude of the radar target, and thus the radar parameters can be changed by leveraging the prior information of radar target detection to achieve the cognitive radar waveform design method.

[0069] Therefore, in view of the problems existing in the prior art, the present invention proposes a cognitive radar waveform design method for maneuvering target path prediction, including:

[0070] The radar emits a narrowband waveform to detect a moving target, and obtains an echo containing target state information;

[0071] Train the echo through an LSTM network to construct a time series prediction model;

[0072] Perform filtered trajectory prediction on the moving target according to the time series prediction model to obtain the target trajectory;

[0073] According to the target trajectory, determine the spatial position information of the moving target relative to the radar through the detection pulses emitted by the radar;

[0074] According to the result of the filtered trajectory prediction and the spatial position information, determine the elevation angle information and azimuth angle information of the moving target;

[0075] Calculate the RCS information of the moving target according to the elevation angle information and the azimuth angle information;

[0076] Determine the inter-pulse DFC coding information according to the RCS information;

[0077] Determine the transmission frequency of each pulse in each group of pulses emitted by the radar according to the inter-pulse DFC coding information.

[0078] Optionally, the echo containing target state information obtained by the radar emitting a narrowband waveform to detect a moving target includes:

[0079] By performing pulse compression on the echo of the transmitted LFM waveform pulse group, obtain the change in the distance of the target relative to the radar during the radar detection stage;

[0080] Process the radar echo information at different times to obtain the target motion trajectory data set during the radar detection stage.

[0081] Optionally, the determining the spatial position information of the moving target relative to the radar through the detection pulses emitted by the radar according to the target trajectory includes:

[0082] According to different detection pulses emitted by the radar, obtain different RCS frequency response functions reflected by the moving target;

[0083] According to the target information echo obtained by the radar at different attitudes, predict the position information of the moving target at different times.

[0084] Optionally, determining the pitch angle information and azimuth angle information of the moving target according to the result predicted by the filtering trajectory and the spatial position information includes:

[0085] According to the result predicted by the filtering trajectory and the spatial position information, through the calculation of trigonometric functions, the pitch angle and azimuth angle of the moving target relative to the radar are obtained.

[0086] Optionally, calculating the RCS information of the moving target according to the pitch angle information and the azimuth angle information includes:

[0087] Obtain the number of discrete strong scattering centers on the moving target;

[0088] Obtain the scattering coefficients of each strong scattering point and the phases of each scattering point relative to the scattering center;

[0089] Decompose the moving target into multiple strong scattering points;

[0090] According to the number of strong scattering centers, the scattering coefficients, and the phases, calculate the complex target RCS information of multiple strong scattering points.

[0091] Optionally, in the process of calculating the RCS information, it further includes:

[0092] According to the transmitted signal of the radar and the scattering coefficients of each scattering point, calculate the echo information at the radar receiver end and the projection distance of the scattering point on the radar line of sight from the radar receiving end.

[0093] Optionally, the method further includes:

[0094] According to the maximum RCS frequency response function and the corresponding frequency at the current angle of the moving target azimuth, use the radar equation to calculate the optimal transmit power used by the radar at the current moment;

[0095] Perform power allocation according to the optimal transmit power.

[0096] Another aspect of the embodiments of the present invention further provides a cognitive radar waveform design device for predicting the path of a maneuvering target, including:

[0097] The first module is used to detect a moving target by the radar transmitting a narrowband waveform and obtain an echo containing target state information;

[0098] The second module is used to train the echo through an LSTM network to construct a time series prediction model;

[0099] The third module is used to perform filtering trajectory prediction on the moving target according to the time series prediction model to obtain a target trajectory;

[0100] The fourth module is configured to determine the spatial position information of the moving target relative to the radar according to the detection pulses transmitted by the radar based on the target trajectory;

[0101] The fifth module is configured to determine the elevation angle information and azimuth angle information of the moving target according to the result of the filtered trajectory prediction and the spatial position information;

[0102] The sixth module is configured to calculate the RCS information of the moving target according to the elevation angle information and the azimuth angle information;

[0103] The seventh module is configured to determine the inter-pulse DFC coding information according to the RCS information;

[0104] The eighth module is configured to determine the transmission frequency of each pulse in each group of pulses transmitted by the radar according to the inter-pulse DFC coding information.

[0105] Another aspect of the embodiments of the present invention further provides an electronic device, including a processor and a memory;

[0106] The memory is used to store programs;

[0107] The processor executes the program to implement the method as described above.

[0108] Another aspect of the embodiments of the present invention further provides a computer-readable storage medium, where the storage medium stores a program, and the program is executed by a processor to implement the method as described above.

[0109] The embodiments of the present invention also disclose a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the method as described above.

[0110] The following will describe the specific implementation process of the present invention in detail with reference to the accompanying drawings of the specification:

[0111] The object of the present invention is to propose a cognitive radar waveform design method based on the prediction of fast maneuvering target path information. This method predicts the target path to obtain the azimuth angle and elevation angle of the target relative to the radar, determines the target RCS value at the current illumination moment of the radar according to the azimuth and elevation angles, and at the same time, according to the target scattering model, determines the frequency response function information corresponding to the maximum RCS at the current moment. According to the radar equation, the radar transmission parameters are adjusted to achieve cognitive radar waveform design.

[0112] The present invention can change the radar transmission parameters according to the target state information, achieving the purpose of cognitive waveform design. Compared with the traditional radar transmission waveform, the waveform designed by the present invention achieves the purpose of saving the use resources of radar equipment through the prediction of target state information. At the same time, the cognitive radar waveform design method can effectively improve the LPI performance of the radar by transmitting pulses with different frequencies within a pulse group, similar to achieving the effect of inter-pulse DFC.

[0113] As Figure 1 shown, the overall implementation process of the present invention includes the following steps:

[0114] Step 1, the radar transmits a narrowband waveform for target detection;

[0115] Step 2, training of the LSTM network model;

[0116] Step 3, using the LSTM network for target estimation and prediction;

[0117] Step 4, obtaining the spatial position of the target relative to the radar;

[0118] Step 5, obtaining the elevation angle of the target relative to the radar;

[0119] Step 6, obtaining the azimuth angle of the target relative to the radar;

[0120] Step 7, obtaining the RCS information of the target at the current moment;

[0121] Step 8, finding the maximum RCS of the target at the current moment;

[0122] Step 9, finding the frequency information corresponding to the maximum RCS;

[0123] Step 10, performing inter-pulse DFC according to the predicted target RCS at different moments;

[0124] Step 11, completing the cognitive radar waveform design for target path prediction.

[0125] Specifically, the specific implementation process of each step is described in detail below:

[0126] Step 1: The radar transmits a narrowband waveform for target detection:

[0127] In the initial stage, the target makes a rapid maneuvering movement within the radar detection range. The radar discovers the target and detects the target by transmitting a narrowband waveform. The purpose of detecting the target is to effectively accumulate the motion state information of the target and obtain a complete target motion model, laying a good foundation for subsequent target motion trajectory prediction. Its narrowband detection model is as Figure 2 shown.

[0128] In the initial target detection stage, the radar mainly uses narrowband LFM waveforms. By analyzing the ambiguity function of the LFM waveform, its ambiguity function shows an inclined knife-edge shape, indicating that the LFM waveform has excellent range-Doppler coupling. This also means that for fast targets, the LFM waveform has excellent target detection capabilities. At the same time, as a good pulse compression waveform, the LFM waveform can effectively obtain the target motion characteristics. By performing pulse compression on the echoes of the LFM waveform pulse group, the change in the target's distance relative to the radar during the radar detection stage can be obtained. Processing the radar echo information at different times can obtain the target motion trajectory data set during the radar detection stage.

[0129] The data set is very important for LSTM sequence prediction. A good data set not only requires a large amount of data but also needs to cover as many possible flight states as possible and reasonably allocate the time for each motion state, so that the trained network model can have better robustness and achieve ideal prediction accuracy. The data of the present invention is obtained through simulation modeling of maneuvering targets, which includes various common flight motion states and maneuvering modes. During the simulation of the flight trajectory, when the target's motion range is large, it is often necessary to calculate the curvature of the earth. In this article, it is assumed that the target's motion range is small, and the earth's curvature is not considered when modeling the target's flight trajectory.

[0130] Define the state vector of the target at a certain moment during flight as which are the longitude, latitude, altitude, speed, azimuth angle, and pitch angle of the target at this moment respectively. When the azimuth angle is defaulted to 0, the target's flight azimuth is towards the due north direction. When the azimuth angle is greater than 0, it rotates clockwise, and when it is less than 0, it rotates counterclockwise. The default maximum recovery angle of the target is (-90, +90), with the upward direction along the height being positive and the downward direction being negative.

[0131] When the target's flight speed is V, the azimuth angle is θ, and the pitch angle is during the process of the target's uniform linear motion, within the time of ΔT, its distance change is ΔS = V·ΔT. Then, the projections of ΔS on the vertical plane and the horizontal plane are ΔS V and ΔS H respectively, and their transformation formulas are:

[0132]

[0133]

[0134] And within the time of ΔT, the change amounts ΔL, ΔB, and ΔH of the target's longitude-latitude-altitude coordinates satisfy the following relationships:

[0135]

[0136] M i and N i are the radius of curvature of the meridian and the radius of curvature of the prime vertical respectively, and the calculation formulas can be expressed as:

[0137]

[0138]

[0139] where a is the first eccentricity.

[0140] After transformation by formula (3), the offset of the target in the geodetic coordinates within ΔT time can be obtained as:

[0141]

[0142] For the calculation of velocity, it is assumed that the target is in uniformly accelerated motion within ΔT time, and the acceleration pair is A t , then the velocity calculation formula of the target is:

[0143] V i+1 = V i + A i ·ΔT (7)

[0144] The above illustrates the change process of the position information and velocity information of the target during linear motion. It is also necessary to consider the motion changes when the target makes non-linear turning maneuvers. Compared with the target's linear motion, the calculation processes of its position information and velocity information when the target makes other motions are the same as those during linear motion, except that the azimuth angle and elevation angle of the target will change, and their calculation formulas are:

[0145]

[0146]

[0147] where A H and A V are the acceleration of the target in the horizontal direction and the acceleration in the vertical direction respectively.

[0148] In summary, during the entire process of the target's flight, its state transformation process is:

[0149]

[0150] After modeling the target motion, different motion trajectories of the target can be obtained by adjusting each parameter in Equation (10). By combining multiple different motion trajectories, estimation data including various motion states such as linear motion, turning maneuvers, dives, and hovering can be obtained. The simulation trajectories used in the present invention simulate the linear motion, accelerating motion, hovering motion, climbing motion, diving motion, etc. of the target, as Figure 3 shown, which is the target motion trajectory used in the present invention.

[0151] Through Figure 3 it can be seen that the target trajectory effectively simulates the changes in the motion postures of a maneuvering target such as linear motion, hovering, and climbing, and has strong practical application combination ability.

[0152] Step 2: Training of the LSTM network model;

[0153] The present invention mainly uses an LSTM network to establish a time series prediction model for predicting the prior information used in the tracking process of the model. It can solve problems such as the traditional method not being able to make full use of the motion state information provided by the target. Considering that the prediction process of prior information is actually a prediction of a time series, the biggest feature of time series prediction is that the predicted value at the current moment is not only related to the data at the previous moment. For the aircraft trajectory, its previous flight state actually indicates the change law of its future flight trajectory, and this law includes trendiness, periodicity, and irregularity. For short-term point track prediction, it also ensures the stationarity of the time series, indicating that the characteristics of the current time series can extend into the future. These conditions determine the theoretical feasibility of obtaining the target motion motor information at the next moment through time series prediction.

[0154] In recent years, deep neural networks have been widely used due to the greatly enhanced computing power of computers. There are many different network frameworks in deep learning to deal with various different problems. For a feedforward convolutional neural network framework like a convolutional neural network, it is commonly used in applications such as classification, recognition, and decision-making. However, a feedforward network cannot preserve past information, and its performance in tasks such as sequence prediction is often poor. By introducing a recurrent network, the information at the current moment can be connected to the previous information, which makes the input of the current network layer not only able to obtain the output result of the previous layer, but also able to obtain the output of the hidden layer nodes at those moments. This determines that the recurrent neural network has the characteristic of having short-term memory and being able to preserve information, as Figure 4 shown as the topological structure of the recurrent neural network, and Figure 5 is the schematic diagram of the network structure after it is unfolded according to the time series.

[0155] Figure 4 and Figure 5They are respectively a recurrent neural network structure and an equivalent network unfolded in time series, as Figure 4 and Figure 5 shown. Its basic structure can be split into an input layer, a hidden layer, and an output layer. Let I represent the input of the network, H represent the activation value of the hidden layer, and O represent the output of the network. The forward propagation process of the entire network can be described by the following four formulas:

[0156]

[0157] H t = f(μ t ) (12)

[0158] z t = W HO H f + b o (13)

[0159] O t = g(z t ) (14)

[0160] During the entire propagation process, where μ t represents the weighted input of the hidden layer, f is the activation function of the hidden layer, z t represents the weighted input of the output layer, and g is the activation function of the output layer. The main role of the activation function is to transform a single linear relationship into a non-linear relationship, enabling the training of the entire system to more accurately approximate the target value required by the present invention. W IH , W HH , W HO are respectively the weight matrices from the input layer to the hidden layer, the weight matrix between the hidden layers, and the weight matrix from the hidden layer to the output layer. Each layer is interconnected through W IH , W HH , W HO .

[0161] Backpropagation is actually a feedback process. By defining a cost function to judge the quality of the training results and using the gradient descent method to gradually find the optimal values of the weights, the training results can be improved. This is achieved by defining an error function as shown in Equation (15). The larger its value, the worse the training effect will be.

[0162]

[0163] Among them, L t is the target value, O t is the output value, and k is the scale constant. The weight update process is Equation (16):

[0164]

[0165] Among them, is the weight at time t after update, w t-1 is the weight at time t - 1 before update, γ is the learning rate of the weight. The faster the γ weight is updated, the slower the reproduction is.

[0166] Using Equation (15) for the in Equation (16) to perform gradient operation, the following error term is obtained:

[0167]

[0168]

[0169] During the network training at the last moment, the error term at the output can be calculated as:

[0170]

[0171] For the error term in the hidden layer, it is calculated as:

[0172]

[0173] When in the training network at other moments, the error term is calculated as:

[0174]

[0175] Through iteration, the errors of the final output unit and the hidden layer unit are calculated as:

[0176]

[0177] When all the error terms are calculated, the next step is to update the weights of each layer to optimize the entire network. For the output layer, the updated weight can be expressed as:

[0178]

[0179] For the input layer, compared with Equation (23), the updated weight is:

[0180]

[0181] For the recurrent layer, the updated weight is:

[0182]

[0183] As can be seen from the above introduction, the recurrent neural network can learn and describe dynamic temporal behaviors. However, the recurrent neural network can only describe short-term dynamic behaviors and cannot capture long-term associations between data. For target motion estimation, it often requires multiple point measurements over a period of time to exhibit strong motion characteristics, and the trained model will have strong robustness. Using only the recurrent neural network for model training will lead to problems such as a decrease in the robustness of the system. Moreover, as the iteration progresses, its weight coefficients may increase or decrease exponentially, eventually resulting in the gradient explosion or gradient vanishing problem, which causes the entire model to diverge. The proposed LSTM can just solve this problem.

[0184] Step 3: Use the LSTM network for target estimation and prediction;

[0185] The target trajectory prediction in step 3 of the present invention can obtain the corresponding trajectory prediction result, which is then used to confirm the spatial position of the radar in step 4.

[0186] Specifically, LSTM is proposed to solve the problems of long-term dependencies that the recurrent neural network cannot handle and issues such as gradient explosion or gradient vanishing. LSTM has three main gate structures to control and affect the relevant functions of the network. Among them, the forget gate determines the new input information received by the neuron, and the output gate determines the output state at the current moment. The system structure of LSTM is as Figure 6 and Figure 7 shown:

[0187] Figure 6 and Figure 7 are the system structure diagram and the unfolded diagram of the LSTM network system structure respectively. Figure 6 and Figure 7 respectively show the main structure of the LSTM network and its internal structure after unfolding. It can be seen that the internal structure of the LSTM network is more complex, and different gate structures are introduced to process and update the long-term memory C t-1 information.

[0188] LSTM introduces a more complex gate structure on the basis of the recurrent neural network. By making relevant derivations for several different gate units of LSTM. For the forget gate structure as Figure 8 shown, it can be seen that the output f t of the previous gate and the input x t at time t, and the state h t-1 of the hidden layer at time t - 1 are related, as shown in Equation (26).

[0189] Figure 8 is the LSTM forget gate structure. It can be seen that the forget gate is based on ht-1 and X t-1 to calculate the forgetting factor f t , thus controlling the part to be forgotten in long-term memory.

[0190] f t = σ(W f ·[h t-1 , x t +b f ) (26)

[0191] where W f is the weight matrix associated with f t relating h t-1 and x t , b f is the bias matrix, and σ is the sigmoid function.

[0192] The structure of the input gate is as Figure 9 shown. The input of the input gate will respectively pass through sigmoid and ranh to obtain the input gate forgetting factor i t and the current memory two values. By multiplying i t and , the candidate value of the current memory is obtained

[0193] Figure 9 is the LSTM input gate structure. As shown in the figure, the input gate determines the part of the current memory t to be forgotten through i , thus retaining useful information and generating the candidate value

[0194] i t = σ(W i ·[h t-1 , x t +b i ) (27)

[0195]

[0196] where W i , W c are respectively the weight matrix associated with i t relating h t-1 and x t and the weight matrix associated with relating h t-1 and x t , b i and b c are the bias matrices. Therefore, from formulas (27) and (28), it can be known that the input of the input gate is related to i t and related, combined with Figure 9 , the input of the input gate can be obtained as shown in Equation (29):

[0197]

[0198] After the state of the neuron at time t is updated as shown in Equation (30):

[0199] C t = C t-1 e f t + IC t (30)

[0200] After the state of the neuron at time t is updated to be:

[0201] C t = C t-1 e f t + IC t (31)

[0202] For the structure of the output gate as Figure 10 shown, it determines the value of the finally required output part, and defines Equations (32) and (33):

[0203] Figure 10 is the LSTM output gate structure. It can be seen that O t is the output gate forgetting factor, and it and the updated neuron state information jointly determine the final output h t at the current moment.

[0204] o t = σ(W o [h t-1 , x t + b o ) (32)

[0205] h t = o t e tanh(C t ) (33)

[0206] In the above formula, W o is the weight matrix that associates o t with h t-1 and x t , and b o is the bias matrix. Through LSTM, the motion trajectory of the target can be effectively estimated, as Figure 11 shown.

[0207] Through Figure 11It can be seen that in order to ensure that the generated data set can achieve effective and good training results, the generated flight tracks include multiple motions such as straight-line motion, accelerating motion, climbing motion, diving motion, decelerating motion, and hovering motion, or to ensure the robustness of the training waveforms.

[0208] Since having more than three layers of LSTM is likely to lead to a deterioration in the model training effect, the model used in the present invention is a two-layer LSTM network, the optimization model is "adam", the loss value is "msle", and the activation function is "relu". The target motion trajectory models a target including motions such as uniform motion, accelerating motion, rapid climbing motion, and diving motion for filtering trajectory prediction.

[0209] Step 4: Obtain the spatial position of the target relative to the radar;

[0210] For the radar, the elevation angle of the target relative to the radar is different at different times. At the same time, the attitude position of the target relative to the radar is different at different times. Therefore, under different detection pulses of the radar, the RCS frequency response function reflected by the target is different, as Figure 12 shown. It can be seen that the cognitive radar can predict the position information of the target at different times through the target information echoes in different postures.

[0211] In the initial stage, the radar emits a narrowband LFM pulse waveform for target detection. At the same time, the LSTM network is used to learn the target motion trajectory. Through the learned target path information, the path of the target in the prediction stage can be accurately estimated, and the radar transmission frequency can be changed using the state information of the target to achieve the purpose of radar cognitive transmission.

[0212] Step 5: Obtain the elevation angle of the target relative to the radar;

[0213] As Figure 12 shown, according to the prediction of the LSTM network and the spatial position of the target relative to the radar, the height information of the target relative to the radar and the target motion attitude information can be returned based on the position information of the target. Through the calculation of trigonometric functions, the elevation angle of the target relative to the radar can be obtained.

[0214] Step 6: Obtain the azimuth angle of the target relative to the radar;

[0215] Similarly, according to Figure 12 shown, the azimuth angle of the target relative to the radar can be equivalent to the motion attitude angle of the target at the current moment. Similar to Step 5, according to the trigonometric function relationship, the target attitude angle of the target relative to the radar coordinate system at the current moment can be obtained, which can be expressed as the included angle between the extension line of the main motion direction of the target and the radar plane.

[0216] It should be noted that the embodiments of the present invention are based on trigonometric functions andFigure 12 As shown, the pitch angle and azimuth angle can be characterized, which is just a spatial relative position information.

[0217] Step 7: Obtain the target RCS information at the current moment;

[0218] In free space, the scattering points are independently distributed. For simple targets, there are already complete analytical methods for RCS. However, for relatively complex large targets, their scattering characteristics are also relatively complex. Generally, a large target is decomposed into the scattering of multiple independent strong scattering points combined with each other, and the RCS of the complex target with multiple scattering points is calculated. The mathematical expression of the complex target RCS is expressed as:

[0219]

[0220] In Equation (34), N is the number of discrete strong scattering centers on the target, σ n is n the scattering coefficient of the th n strong scattering point, and

[0221] is the phase of the t th n scattering point relative to the scattering center. In the actual scenario, generally, a large target is decomposed into multiple strong scattering points. Each strong scattering point will generate a scattered wave for the incident electromagnetic wave. Due to the different radial distances of each scattering point relative to the radar receiver, their time delays and phases are also different. The echoes of all strong scattering points are superimposed in space to form the overall echo and reach the receiver.

[0221] If the transmitted signal is S t (t), and the scattering coefficients of each scattering point are σ n , then the echo at the radar receiver end is S r (t), that is:

[0222]

[0223] d n = R n ·cosθ n (36)

[0224] In Formulas (35) and (36), N is the number of strong scattering centers on the target, σ n is n the RCS value of the n th n strong scattering point, d n is the projection distance of the scattering point on the radar line of sight from the radar receiving end, R n is the radial distance of the scattering point from the radar, and θ n is the radar illumination angle.

[0225]

[0226] As can be seen from Equation (37), for a large target with multiple scattering points, the variation of its RCS value is jointly determined by the radar illumination angle of view and the scattering intensity of each scattering point. The RCS fluctuation model used in this embodiment is as Figure 13 shown. Figure 13 Figure Figure 13 shows the variation curve of the target RCS frequency response function at different azimuth angles. Since the electromagnetic reflection cross-section area of the target changes according to the relative radar pitch angle, for different azimuth angles, its RCS variation is different. At the same time, it can be seen from Figure 13 that at different radar transmission frequencies, the RCS fluctuation models of the target at different attitude angles are different.

[0227] Step 8: Find the maximum RCS of the target at the current moment;

[0228] As can be seen from Figure 13 , the RCS of the target at different angles shows different forms of the corresponding frequency response functions. Using the predicted RCS frequency response function at the same frequency f m , and under the condition of the target azimuth angle θ at time t, find the maximum RCS frequency response function:

[0229]

[0230] where, o θmax is the maximum target RCS frequency response function at the angle θ, is the m th RCS frequency response function corresponding to the frequency f m at the angle θ. So far, the algorithm has completed the radar cognitive frequency selection work and obtained the frequency f θmax corresponding to the maximum RCS frequency response function of the target.

[0231] Step 9: Find the frequency information corresponding to the maximum RCS;

[0232] Through the maximum RCS frequency response function o θmax at the current angle θ of the target azimuth and its corresponding frequency f θmax , the optimal transmission power used by the radar at this moment can be calculated using the radar equation, and the purpose of power allocation can be achieved.

[0233] It should be noted that in the target prediction stage, the radar also irradiates the target in the form of electromagnetic pulses. The purpose is to enable the radar to emit an inter-pulse frequency-coded waveform that satisfies the RCS frequency response characteristics of the target in the target prediction stage, achieving the purpose of cognitive transmission.

[0234] Step 10: Perform inter-pulse DFC according to the predicted target RCS at different times;

[0235] In the stage of predicting the motion path of a fast - maneuvering target, the radar selects multiple pulses that meet the emission performance to form a pulse group. The composition of the pulse group is based on a PRI reference time requirement in the time series, and multiple PRIs form a pulse group. Each PRI is mapped to the corresponding predicted target state information, meeting the requirements of inter - pulse DFC coding.

[0236] Step 11: Complete the cognitive radar waveform design for target path prediction;

[0237] It should be noted that the radar emits a group of pulses. In this group of pulses, the emission frequency of each pulse is determined by the DFC sequence, and the determination of DFC is determined by the RCS corresponding to the relative position of the target to the radar determined by the estimated position (pitch, azimuth) of the target estimation.

[0238] In this embodiment, 16 different pulses are taken as the main research object. By exploring the different frequency response functions of the target RCS in the echo during the target path prediction stage to achieve radar cognitive frequency emission, and taking the radar target detection power as the evaluation index, a comparison is made with the fixed - parameter LFM waveform. The target appears at 7000 meters and 8750 meters away from the radar. For low - slow - small targets, the RCS of the two targets set in the experiment is 0.01, and the ratio of the radar transmission power to the clutter power is - 16 dB. The target detection performances of the fixed - parameter LFM waveform and the cognitive frequency agile waveform are respectively Figure 14 and Figure 15 as shown.

[0239] In Figure 14 and Figure 15 , the experiment uses the 16 - pulse echo for long - time accumulation. It can be seen that for the fixed - parameter LFM waveform, there will be many false targets, and the target at 8750 meters has been completely submerged in the noise interference. At the same time, the false targets will cause great interference to target detection. For the cognitive frequency agile waveform, it can effectively reduce the interference of false targets generated by noise, and can detect the targets at 7000 meters and 8750 meters away from the radar. It can make the effective detection range of the radar under the clutter background increase by 25% compared with the fixed - parameter LFM, effectively realizing the cognitive emission method for the path prediction of fast - maneuvering targets.

[0240] To sum up, the present invention has the following characteristics:

[0241] 1. For fast - maneuvering targets, the present invention proposes a cognitive radar waveform design method for maneuvering target path prediction.

[0242] 2. In the process of maneuvering target path prediction, a method combining the LSTM network and the target RCS is used. By finding the maximum RCS of the target, the radar emission frequency is changed to achieve the function of inter - pulse DFC.

[0243] 3. The method proposed by the present invention can not only meet the premise of resource conservation, but also improve the low intercept and anti-jamming performance of the radar.

[0244] Compared with the prior art, the present invention proposes a cognitive radar waveform design method based on the path prediction of fast maneuvering targets. First, the present invention uses an LSTM network to predict the path of the target detected and accumulated for a period of time, and at the same time combines the target RCS frequency response characteristics to perform inter-pulse DFC of the radar transmission waveform during the path prediction stage. This can effectively reduce the energy consumption of the radar during target detection. At the same time, since different combinations of inter-pulse frequency coding are used according to the state information of the target during the radar pulse transmission, the low intercept and anti-jamming performance of the radar can be improved.

[0245] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order mentioned in the operating diagrams. For example, depending on the functions / operations involved, two consecutive blocks shown may actually be executed substantially simultaneously or the blocks can sometimes be executed in the reverse order. In addition, the embodiments presented and described in the flowcharts of the present invention are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated, where the order of various operations is changed and where sub-operations described as part of a larger operation are executed independently.

[0246] In addition, although the present invention is described in the context of functional modules, it should be understood that unless otherwise stated to the contrary, one or more of the functions and / or features described may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It can also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More precisely, considering the attributes, functions and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the routine skills of an engineer. Therefore, those skilled in the art can implement the present invention as set forth in the claims without undue experimentation using ordinary skills. It can also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0247] If the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0248] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in combination with an instruction execution system, apparatus, or device.

[0249] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection parts with one or more wirings (electronic devices), portable computer disk cartridges (magnetic devices), random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memories), optical fiber devices, and portable compact disc read-only memories (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as necessary, and then storing it in a computer memory.

[0250] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following technologies well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0251] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0252] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

[0253] The above has specifically described the preferred embodiments of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A cognitive radar waveform design method for maneuvering target path prediction, characterized in that, it includes: The radar emits a narrowband waveform to detect a moving target, and obtains an echo containing target state information; Train the echo through an LSTM network to construct a time series prediction model; Perform filtered trajectory prediction on the moving target according to the time series prediction model to obtain the target trajectory; According to the target trajectory, determine the spatial position information of the moving target relative to the radar through the detection pulses emitted by the radar; According to the result of the filtered trajectory prediction and the spatial position information, determine the elevation angle information and azimuth angle information of the moving target; Calculate the RCS information of the moving target according to the elevation angle information and the azimuth angle information; Determine the inter-pulse DFC coding information according to the RCS information; Determine the transmission frequency of each pulse in each group of pulses emitted by the radar according to the inter-pulse DFC coding information.

2. The cognitive radar waveform design method for maneuvering target path prediction according to claim 1, characterized in that, The radar emits a narrowband waveform to detect a moving target, and the obtained echo containing target state information includes: By performing pulse compression on the echo of the transmitted LFM waveform pulse group, obtain the change in the distance of the target relative to the radar during the radar detection stage; Process the radar echo information at different times to obtain the target motion trajectory data set during the radar detection stage.

3. The cognitive radar waveform design method for maneuvering target path prediction according to claim 1, characterized in that, The step of determining the spatial position information of the moving target relative to the radar according to the target trajectory through the detection pulses emitted by the radar includes: According to different detection pulses emitted by the radar, obtain different RCS frequency response functions reflected by the moving target; According to the target information echoes obtained by the radar in different postures, predict the position information of the moving target at different times.

4. The cognitive radar waveform design method for maneuvering target path prediction according to claim 1, characterized in that, The step of determining the elevation angle information and azimuth angle information of the moving target according to the result of the filtered trajectory prediction and the spatial position information includes: According to the result of the filtered trajectory prediction and the spatial position information, through the calculation of trigonometric functions, obtain the elevation angle and azimuth angle of the moving target relative to the radar.

5. The cognitive radar waveform design method for maneuvering target path prediction according to claim 1, characterized in that, The step of calculating the RCS information of the moving target according to the elevation angle information and the azimuth angle information includes: Obtain the number of discrete strong scattering centers on the moving target; Obtain the scattering coefficients of each strong scattering point and the phase of each scattering point relative to the scattering center; Decompose the moving target into multiple strong scattering points; Calculate the complex target RCS information of multiple strong scattering points according to the number of strong scattering centers, the scattering coefficients and the phases.

6. The cognitive radar waveform design method for maneuvering target path prediction according to claim 5, It is characterized in that In the process of calculating the RCS information, it further includes: According to the radar transmission signal and the scattering coefficients of each scattering point, calculate the echo information at the radar receiver end and the projection distance of the scattering point on the radar line of sight from the radar receiving end.

7. A cognitive radar waveform design method for maneuvering target path prediction according to claim 1, It is characterized in that The method further includes: According to the maximum RCS frequency response function and the corresponding frequency of the moving target azimuth at the current angle, use the radar equation to calculate the optimal transmission power used by the radar at the current moment; Perform power allocation according to the optimal transmission power.

8. A cognitive radar waveform design device for maneuvering target path prediction, It is characterized in that It includes: The first module is used to detect a moving target by transmitting a narrowband waveform by the radar to obtain an echo containing target state information; The second module is used to train the echo through an LSTM network to construct a time series prediction model; The third module is used to perform filtered trajectory prediction on the moving target according to the time series prediction model to obtain the target trajectory; The fourth module is used to determine the spatial position information of the moving target relative to the radar according to the target trajectory through the detection pulses transmitted by the radar; The fifth module is used to determine the elevation angle information and azimuth angle information of the moving target according to the result of the filtered trajectory prediction and the spatial position information; The sixth module is used to calculate the RCS information of the moving target according to the elevation angle information and the azimuth angle information; The seventh module is used to determine the inter-pulse DFC coding information according to the RCS information; The eighth module is used to determine the transmission frequency of each pulse in each group of pulses transmitted by the radar according to the inter-pulse DFC coding information.

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

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

Images