Radar multi-angle multi-feature fusion target classification method

Abstract

The invention discloses a radar multi-angle multi-feature fusion target classification method, and relates to the technical field of radar signal processing, and the method comprises the steps: obtaining a multi-angle multi-feature fusion network target classification model; the multi-angle multi-feature fusion network target classification model comprises a time sequence feature extraction module, an attention fusion module and a feature fusion module; acquiring radar echo data of a target in real time; processing the radar echo data of the target to obtain processed radar echo data; the processed radar echo data comprises a time domain signal component graph of the target and a time frequency graph of the target; and inputting the processed radar echo data into a multi-angle multi-feature fusion network target classification model to obtain a target classification result. According to the invention, the accuracy of target classification can be improved.

Description

Technical Field

[0001] The present application relates to the field of radar signal processing technology, and in particular to a radar multi-angle and multi-feature fusion target classification method. Background Art

[0002] In recent years, the use of small civilian unmanned aerial vehicles has increased dramatically due to their low price and strong practicality. However, the widespread increase in drones has brought many potential threats. Many airports have experienced drone and illegal intrusion incidents, which seriously threaten public safety. Since different types of drones and flying birds pose different threats, it is imperative to identify drones and flying birds. Radar has become an effective means of target monitoring because the radar system is not easily affected by scene changes and weather changes, but there is still a lack of effective methods to identify drones and flying birds through radar.

[0003] Micro-Doppler is a Doppler effect that reflects the micro-motion characteristics of the target. It is a tiny additional modulation in the spectrum change caused by the target motion. It can capture the weak frequency offset caused by the non-rigid motion of the target (such as vibration, swing or rotation), and is an important feature in modern radar signal processing and target recognition. Under the assumption of the ideal scattering center model, the time-frequency distribution is introduced to extract the micro-motion characteristics of the rotorcraft UAV. Since the signal returned by the radar is composed of complex components, including the volume Doppler signal caused by translation, vibration interference and environmental clutter, the time-frequency representation of the micro-Doppler signal is always mixed with other signals and cannot be directly extracted. Empirical mode decomposition has advantages in solving the problem of mode mixing. Empirical mode decomposition (EMD) is a completely data-driven method that can be used to analyze nonlinear and non-stationary signals. It can adaptively decompose any signal into a complete and limited set of local amplitude frequencies. However, the EMD method is prone to modal aliasing in the process of decomposing complex signals, making it difficult to distinguish different types of UAVs. Short-time Fourier transform can extract micro-motion features of rotorcraft drones and flying birds. Since the micro-Doppler of multi-rotor drones has aliasing as the number of rotors increases, it is difficult to describe and identify and classify them through mathematical models and parameters in complex motion and environment. In recent years, with the development of deep learning and machine learning technology, feature fusion has become an important trend in the field of micro-motion feature extraction of drones. The combination of radar target classification and recognition tasks with deep learning has greatly improved the target classification and recognition capabilities. Deep learning methods extract high-dimensional features and avoid the limitations of manual feature extraction. Convolutional Neural Networks (CNN), as a classic method in deep learning, has been widely used in the field of image detection and classification. For example, in the current technology, empirical mode decomposition (EMD) and variational mode decomposition (VMD) are combined to improve the classification accuracy of the target by splicing the decomposed features and sending them to the deep learning network for classification. The classification accuracy of targets is also improved by combining EMD and LSTM (Long Short-Term Memory) to extract time domain signal sequence features. In addition, FMCW radar is used to collect drone and bird data, merge ranging spectrograms and micro-Doppler features, and use multi-channel DCNN for target classification, which improves the accuracy of drone and bird classification. It can be seen that radar and deep learning are of great significance for the classification of low, slow and small targets.In addition, by fusing micro-motion features with traditional radar features (such as distance, Doppler speed, etc.), constructing a multimodal data set, and using a deep learning model to jointly learn the features, the accuracy of drone recognition can still be significantly improved. Multi-sensor fusion technology is also widely used in the extraction of drone micro-motion features. Although it is theoretically possible to effectively reduce the defects of a single feature by fusing multi-feature radar data, thereby enhancing the stability and accuracy of micro-motion features, drone micro-motion feature extraction and fusion technology has only achieved certain results in an experimental environment. In actual applications, the classification task of low, slow and small targets still faces some challenges, and cannot truly reduce the defects of a single feature and enhance the stability and accuracy of micro-motion features. Summary of the invention

[0004] The purpose of this application is to provide a radar multi-angle and multi-feature fusion target classification method, which can enhance the stability and precision of micro-motion features, thereby improving the accuracy of target classification.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a radar multi-angle and multi-feature fusion target classification method, comprising: Acquire a multi-angle multi-feature fusion network target classification model; the multi-angle multi-feature fusion network target classification model includes: a temporal feature extraction module, an attention fusion module and a feature fusion module; Acquire the target's radar echo data in real time; Processing the radar echo data of the target to obtain processed radar echo data; the processed radar echo data includes: a time domain signal component diagram of the target and a time-frequency diagram of the target; The processed radar echo data is input into the multi-angle multi-feature fusion network target classification model to obtain a target classification result.

[0006] Optionally, the process of constructing the multi-angle and multi-feature fusion network target classification model includes: Acquire radar echo data of sample targets; Processing the radar echo data of the sample target to obtain processed radar echo data of the sample target; the processed radar echo data of the sample target includes: a time domain signal component diagram of the sample target and a time-frequency diagram of the sample target; Constructing a training data set based on the processed radar echo data of the sample target; Construct an initial multi-angle and multi-feature fusion network target classification model; The initial multi-angle multi-feature fusion network target classification model is trained based on the training data set until the sample target classification result output by the initial multi-angle multi-feature fusion network target classification model reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model; the trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

[0007] Optionally, the process of acquiring the radar echo data of the sample target includes: Use radars of different bands to collect multi-angle radar echo data for each target in the set area, and obtain multi-angle radar echo data of all targets in the set area; The radar echo data of the sample target is obtained based on the multi-angle radar echo data of all targets in the set area.

[0008] Optionally, processing the radar echo data of the sample target to obtain processed radar echo data of the sample target includes: Using moving target indication to remove negative frequencies and filter out clutter on the radar echo data of the sample target to obtain pre-processed radar echo data of the sample target; Obtaining an extraction signal based on the preprocessed radar echo data of the sample target; Performing empirical mode decomposition on the extracted signal to obtain a time domain signal component graph of the sample target; Performing short-time Fourier transform on the extracted signal to obtain a time-frequency diagram of the sample target.

[0009] Optionally, constructing a training data set based on the processed radar echo data of the sample target includes: Constructing a first data set based on the time domain signal component map of the sample target; constructing a second data set based on the time-frequency diagram of the sample target; A training data set is constructed based on the first data set and the second data set.

[0010] Optionally, the initial multi-angle multi-feature fusion network target classification model is trained based on the training data set until the sample target classification result output by the initial multi-angle multi-feature fusion network target classification model reaches a set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model, including: Using the initial multi-angle multi-feature fusion network target classification model, a sample target classification result is obtained based on the training data set; The AdamW algorithm is used to train the initial multi-angle multi-feature fusion network target classification model based on the sample target classification result until the sample target classification result reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model; The trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

[0011] Optionally, the attention fusion module includes: a CBAM attention feature extraction unit and a visual channel attention feature extraction unit; during the training process, the process of using the initial multi-angle multi-feature fusion network target classification model to obtain the sample target classification result includes: Using the temporal feature extraction module and the CBAM attention feature extraction unit, obtaining a first feature map based on the first data set; Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit, obtaining a second feature map based on the second data set; The feature fusion module is used to obtain the sample target classification result based on the first feature map and the second feature map.

[0012] Optionally, it is characterized in that the formula Performing short-time Fourier transform on the extracted signal to obtain a time-frequency diagram of the sample target; In the formula, Represents a sliding window, using a Gaussian window function; represents the integral variable, represents the time dimension, represents the frequency dimension, Represents the result after short-time Fourier transform; Represents the extraction signal; is the kernel of Fourier transform, which represents the basis functions of different frequency components.

[0013] Optionally, it is characterized in that the formula Performing empirical mode decomposition on the extracted signal to obtain a time domain signal component graph of the sample target; In the formula, represents the input signal, Indicates The eigenmode functions, represents the number of decomposed intrinsic mode functions, Represents the process of extracting the intrinsic mode function from the residual.

[0014] Optionally, it is characterized in that the attention fusion module includes: a CBAM attention feature extraction unit and a visual channel attention feature extraction unit; inputting the processed radar echo data into the multi-angle multi-feature fusion network target classification model to obtain a target classification result, including: Using the temporal feature extraction module and the CBAM attention feature extraction unit, a third feature map is obtained based on the processed radar echo data; Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit, a fourth feature map is obtained based on the processed radar echo data; The feature fusion module is used to obtain the target classification result based on the third feature map and the fourth feature map.

[0015] According to the specific embodiments provided in this application, this application has the following technical effects: The present application provides a radar multi-angle and multi-feature fusion target classification method, which obtains target classification results based on processed radar echo data by adopting a multi-angle and multi-feature fusion network target classification model, can overcome the differences in feature extraction of a single neural network, enhance the stability and accuracy of micro-motion features, and thus improve the accuracy of target classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is a flow chart of a radar multi-angle multi-feature fusion target classification method in one embodiment of the present application; Figure 2 A schematic diagram of a target classification process of a radar multi-angle multi-feature fusion target classification method provided in an embodiment of the present application; Figure 3 A schematic diagram of a multi-angle and multi-feature fusion network target classification model provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of a temporal feature extraction module and a CBAM attention feature extraction unit provided in an embodiment of the present application; Figure 5 A schematic diagram of the structure of a CBAM attention feature extraction unit and a visual channel attention feature extraction unit provided in an embodiment of the present application; Figure 6 A schematic diagram of the structure of a feature fusion module provided in one embodiment of the present application; Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0020] In an exemplary embodiment, Figure 1 As shown, a radar multi-angle multi-feature fusion target classification method is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is used as an example to illustrate target classification, including: Step 100, obtaining a multi-angle multi-feature fusion network target classification model. The multi-angle multi-feature fusion network target classification model includes: a temporal feature extraction module, an attention fusion module and a feature fusion module.

[0021] Step 200, acquiring radar echo data of the target in real time.

[0022] Step 300, the radar echo data of the target is processed to obtain the processed radar echo data. The processed radar echo data includes: a time domain signal component diagram of the target and a time-frequency diagram of the target.

[0023] Step 400, input the processed radar echo data into a multi-angle multi-feature fusion network target classification model to obtain a target classification result.

[0024] As an optional implementation, in order to improve the accuracy of the radar multi-angle multi-feature fusion target classification method, the construction process of the multi-angle multi-feature fusion network target classification model in step 100 includes: 101, obtain radar echo data of sample targets. The process of obtaining radar echo data of sample targets includes: using radars of different bands to collect multi-angle radar echo data of each target in a set area, and obtaining multi-angle radar echo data of all targets in the set area. The radar echo data of the sample target is obtained based on the multi-angle radar echo data of all targets in the set area.

[0025] 102, process the radar echo data of the sample target to obtain the processed radar echo data of the sample target. The processed radar echo data of the sample target includes: a time domain signal component diagram of the sample target and a time-frequency diagram of the sample target. Among them, the process of processing the radar echo data of the sample target to obtain the processed radar echo data of the sample target includes: using a moving target indication to remove negative frequencies and filter out clutter on the radar echo data of the sample target to obtain the preprocessed radar echo data of the sample target. Based on the preprocessed radar echo data of the sample target, an extracted signal is obtained. The extracted signal is subjected to empirical mode decomposition to obtain the time domain signal component diagram of the sample target. The extracted signal is subjected to short-time Fourier transform to obtain the time-frequency diagram of the sample target.

[0026] For example, using the formula Perform short-time Fourier transform on the extracted signal to obtain the time-frequency diagram of the sample target. Represents a sliding window, using a Gaussian window function; represents the integral variable, represents the time dimension, represents the frequency dimension, Represents the result after short-time Fourier transform; Represents the extraction signal, is the kernel of Fourier transform, which represents the basis functions of different frequency components.

[0027] Using formula Perform empirical mode decomposition on the extracted signal to obtain the time domain signal component diagram of the sample target. In the formula, represents the input signal, Indicates The eigenmode functions, represents the number of decomposed intrinsic mode functions, Represents the process of extracting the intrinsic mode function from the residual.

[0028] 103, construct a training data set based on the processed radar echo data of the sample target. For example, construct a first data set based on the time domain signal component map of the sample target. Construct a second data set based on the time-frequency map of the sample target. Construct a training data set based on the first data set and the second data set.

[0029] 104, construct an initial multi-angle and multi-feature fusion network target classification model.

[0030] 105, training the initial multi-angle multi-feature fusion network target classification model based on the training data set until the sample target classification result output by the initial multi-angle multi-feature fusion network target classification model reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model. The trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

[0031] The process of training the initial multi-angle multi-feature fusion network target classification model includes: using the initial multi-angle multi-feature fusion network target classification model to obtain the sample target classification result based on the training data set. The AdamW algorithm is used to train the initial multi-angle multi-feature fusion network target classification model based on the sample target classification result until the sample target classification result reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model. The trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

[0032] The attention fusion module includes: a CBAM attention feature extraction unit and a visual channel attention feature extraction unit. The process of using the initial multi-angle multi-feature fusion network target classification model to obtain the sample target classification result during the training process includes: using the temporal feature extraction module and the CBAM attention feature extraction unit to obtain a first feature map based on the first data set. Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit to obtain a second feature map based on the second data set. Using the feature fusion module to obtain the sample target classification result based on the first feature map and the second feature map.

[0033] The process of inputting the processed radar echo data into the multi-angle multi-feature fusion network target classification model in step 400 to obtain the target classification result includes: using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit to obtain a third feature map based on the processed radar echo data. Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit to obtain a fourth feature map based on the processed radar echo data. Using the feature fusion module to obtain the target classification result based on the third feature map and the fourth feature map.

[0034] The output vector of the multi-angle and multi-feature fusion network target classification model is ,in The model is The score of the class is calculated using the Softmax function. Probability of class , expressed as: .in, It is The score of the class, It is The predicted probability of the class.

[0035] In an exemplary embodiment, in combination with steps 101 to 105 of the above embodiment, the process of obtaining a multi-angle multi-feature fusion network target classification model, that is, training an initial multi-angle multi-feature fusion network target classification model is described as an example. Figure 2 shown.

[0036] S1, obtaining radar echo data of a sample target, including K-band radar echo data and L-band radar echo data, and processing the radar echo data of the sample target to obtain processed radar echo data of the sample target.

[0037] First, in order to eliminate the negative frequency signal in the radar echo data, the data is preliminarily processed, and a high-pass filter is used to effectively remove the DC component in the echo signal (i.e., radar echo data). This step is intended to enhance the dynamic range of the signal and ensure the accurate extraction of high-frequency components. The moving target indication (MTI) technology is used to remove negative frequencies and filter out clutter on the radar echo data, and to filter out stationary clutter on the ground and in the air, thereby improving the detection capability of the target. MTI can effectively suppress stationary or slow-moving background noise by comparing the phase changes between adjacent pulses, making the target signal more prominent, including negative frequency modulation removal, demodulation, and pulse compression processing. The radar echo data of the preprocessed sample target is obtained through the MTI technology.

[0038] After the above preprocessing steps, the radar echo data of the preprocessed sample target is obtained. Based on the preprocessed radar echo data of the sample target, the target range period diagram is drawn, and the echo signal of the distance unit where the target is located is extracted to obtain the extracted signal (including the K-band extracted signal and the L-band extracted signal). The L-band extracted signal is subjected to empirical mode decomposition (EMD) to obtain the time domain signal component diagram of IMF1 (i.e., the time domain signal component diagram of the sample target): (1) In the formula, represents the input signal, Indicates The eigenmode functions, Represents the process of extracting IMF from residuals, Represents the number of decomposed eigenmode functions.

[0039] In order to further extract the micro-motion characteristics of the target, the K-band extracted signal is subjected to short-time Fourier transform (STFT) to obtain the time-frequency diagram of the sample target. Assume that the K-band echo signal of the target is expressed as , the echo signal in the time window is subjected to fast Fourier transform (FFT) by means of a sliding window to obtain the time-frequency diagram of the sample target: (2) in, represents a sliding window, and in this embodiment, a Gaussian window function is used; represents the integral variable, represents the time dimension, represents the frequency dimension, It represents the result after short-time Fourier transform. By observing the range Doppler of the UAV, the distance unit where the target is located is selected, and the signal of the distance unit where the target is located is extracted for STFT to obtain the time-frequency diagram of the target signal at multiple angles.

[0040] S2, constructing a training data set based on the processed radar echo data of the sample target, the training data set including a first data set and a second data set. After processing the echo data of the sample target in step S1, a time domain signal component diagram of the sample target and a time-frequency diagram of the sample target are obtained. The first data set is constructed based on the time domain signal component diagram of the sample target. The second data set is constructed based on the time-frequency diagram of the sample target.

[0041] S3, construct the initial multi-angle multi-feature fusion network target classification model. The constructed target classification model includes the temporal feature extraction module (ConvLSTM), the attention fusion module and the feature fusion module (AFF). The attention fusion module includes the CBAM attention feature extraction unit and the visual channel attention feature extraction unit (SEAttention). Figure 3 As shown, a temporal feature extraction module and a CBAM attention feature extraction unit are used to obtain a first feature map based on a first data set. A CBAM attention feature extraction unit is used to obtain CBAM extracted features based on a second data set; a SEAttention unit is used to obtain SEAttention extracted features based on the second data set; and the CBAM extracted features and the SEAttention extracted features are weightedly fused to obtain a second feature map.

[0042] 1) Temporal feature extraction module (ConvLSTM).

[0043] Different types of rotorcraft UAVs exhibit different dynamic characteristics in time domain signals, so obtaining and analyzing the time series information of signals is crucial for target recognition. In order to effectively extract the features of continuous time domain signals, a time domain feature extraction method based on ConvLSTM is adopted. ConvLSTM is a deep model that combines convolutional neural network (CNN) and long short-term memory network (LSTM), which is specially used to process data with spatiotemporal dependencies. LSTM network has significant advantages in processing sequence data, especially in capturing long-distance temporal dependencies. Its special gating mechanism enables it to effectively retain important temporal information. CNN is good at extracting local spatial features from data and obtaining high-dimensional spatial dependency information through convolution operations. The ConvLSTM model integrates convolution operations into the gating unit of the LSTM model, so that it can simultaneously learn the time series features and spatial features of the data, thereby adapting to the processing requirements of spatiotemporal data. The ConvLSTM model can not only effectively capture the dynamic changes of rotorcraft UAV signals in the time dimension, but also extract the time series features of time domain signals in the spatial dimension. In all embodiments provided in this application, the ConvLSTM model is used to process the continuous time domain data of the radar echo signal, extract the dependency between signals at different times and the time series features that change over time, so that the target classification model can better distinguish different types of rotary-wing UAVs. Figure 4 As shown in Figure 2, the internal structure of the ConvLSTM model is shown, including the input gate , Forget Gate , output gate The convolution operation process can retain spatial features while strengthening the capture of temporal dependencies. The implementation of time domain feature extraction based on ConvLSTM is as follows: (3) (4) (5) (6) (7) (8) (9) in, represents the convolution operation, , , , are all convolution kernels, represents bitwise multiplication, Represents the input of the current time step, Conv() represents the convolution operation on the input, and CBAM() represents the attention feature extraction operation. represents the hidden state at the previous time step, represents the input gate, represents the cell state at the previous time step, represents the convolution operation, Represents the Sigmoid activation function, which limits the output to between [0,1]. represents the forget gate (controls which information is retained in the cell state), represents the hyperbolic tangent function, , , , Respectively represent the bias items of the forget gate, input gate, cell state, and output gate, represents the candidate cell state, Represents the cell state of the current cell step, represents the bias of the input gate, represents the output gate (controls the final hidden state The input of the ConvLSTM model is a five-dimensional tensor. By taking five consecutive frames as the input of the ConvLSTM model, the temporal features between each frame are extracted, and the spatial features of the feature map are extracted during the convolution process.

[0044] 2) Attention fusion module.

[0045] The echo data of different rotorcraft UAVs (i.e., the radar echo data of sample targets) are collected by radar, and the obtained data are subjected to STFT. Then, features are extracted through convolution operation. The obtained multi-angle target echo data are subjected to STFT to obtain the micro-motion features of the rotorcraft UAV. The micro-motion features obtained by two radars (K-band and L-band) are weightedly fused after convolution operation to obtain features with higher dimensionality.

[0046] CBAM is a lightweight and efficient attention module that can adaptively assign different importance weights to the channel space dimensions of feature maps, thereby improving network performance. Figure 5 As shown, CBAM uses the intermediate feature map (i.e., the time-frequency diagram in the second data set) is used as input, and the one-dimensional channel attention map is derived in turn according to the channel attention module and the spatial attention module. and a 2D spatial attention map ,in represents the channel dimension, represents the height of the input feature, Represents the width of the input feature. The entire attention feature extraction process can be summarized as: (10) (11) in, represents element-wise multiplication, represents channel attention feature extraction, Represents spatial attention feature extraction. represents the feature matrix after channel attention feature extraction, Represents the feature matrix after spatial attention feature extraction. The channel attention module aims to assign different weights to each channel, use global pooling to aggregate global information, and then generate the importance weight of each channel through a multi-layer perceptron (MLP). The channel attention mechanism can be expressed as: (12) in, is the Sigmoid activation function, MLP is a two-layer perceptron with shared weights, AvgPool() represents the average pooling layer operation, and MaxPool() represents the maximum pooling layer operation. The spatial attention module aims to assign different weights to each spatial position to capture the local areas that are more important to the target. The spatial attention mechanism can be expressed as: (13) in, Represents the weight matrix generated after inputting features, Indicates splicing by channel, Represents a convolution operation.

[0047] The SEAttention unit is used to extract features from the multi-angle time-frequency diagram obtained by radar, such as Figure 5 As shown, assuming that the extracted features are expressed as , its feature extraction can be expressed as: (14) in, represents a standard convolution operator, Represents the extracted feature map: (15) in, Is a c-dimensional vector representing the global features of each channel; represents the height of the input feature map, represents the width of the input feature map, represents the vertical index in the feature map, represents the horizontal index in the feature map, Represents the channel index of the feature map; Represents global pooling, extracting channel global features through the global average pooling layer operation, compressing each channel of the feature map into a scalar: (16) (17) (18) in, Represents the final output feature, Represents the global description of each channel, Indicates the output result, represents the ReLU function, Indicates that MLP calculates channel weights, and is a learnable weight matrix, Indicates that the channel weight Acting on the original feature On, complete the channel recalibration. Represents the final output feature, Indicates the indicator quantity, Represents a feature map.

[0048] In order to better extract the time-frequency spectrum features obtained from the two bands, two attention modules are combined to extract the features from the two spectrograms and then perform weighted fusion on them to obtain two feature maps containing the two features (i.e., the second feature map).

[0049] 3) Feature fusion module (AFF).

[0050] The feature fusion module can adapt to different levels of local features and global features according to the input features. AFF is a network framework for multi-scale and multi-modal data fusion, which mainly focuses on the attention problem of fusion of different scale features in different network structures. The AFF feature fusion network realizes flexible and efficient multi-scale and multi-modal feature fusion through an adaptive weighting mechanism. Through the lightweight design of global average pooling and weight learning, it ensures the accuracy and robustness of feature fusion while ensuring efficient calculation. Therefore, this module is used to combine the features obtained by the ConvLSTM model. Figure 1 (i.e., the first feature map) and the features obtained from the target's time-frequency map Figure 2 The MS-CAM (Multi-Scale Channel Attention Module) used in the AFF module follows the idea of ​​combining local and global features in CNN by ParseNet (Semantic Segmentation Network) and the idea of ​​spatial attention that aggregates multi-scale features in the attention module.

[0051] (19) in, To fusion features, represents the initial feature integral, X and Y represent the two input features. The overall framework of the AFF fusion module is as follows: Figure 6 shown.

[0052] MS-CAM combines multi-scale features and channel attention mechanism to capture the details and global information of objects by processing different scale features of the image. The channel attention mechanism calculates the weights of different channels to enable the model to automatically focus on which features are more important. The calculation formula of channel attention can be expressed as : (20) Among them, B() indicates the execution of BatchNorm operation. represents the ReLU activation function, express Point convolution, the calculated weight value is used to input features After doing the attention operation, we get the output , the local channel information keeps the same size as the input feature.

[0053] (twenty one) S4, using the initial multi-angle multi-feature fusion network target classification model, obtain the sample target classification result based on the training data set. The AdamW algorithm is used to train the initial multi-angle multi-feature fusion network target classification model based on the sample target classification result until the sample target classification result reaches the set result, and the trained initial multi-angle multi-feature fusion network target classification model is obtained. The trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

[0054] The process of using the initial multi-angle multi-feature fusion network target classification model to obtain the sample target classification result during the training process includes: using the temporal feature extraction module and the CBAM attention feature extraction unit to obtain the first feature map based on the first data set. Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit to obtain the second feature map based on the second data set. Using the feature fusion module to obtain the sample target classification result based on the first feature map and the second feature map.

[0055] When performing iterative optimization training, use the AdamW gradient descent method and give the initial learning rate parameter (i.e. scaling factor), momentum factor , , optimization parameters , the first instantaneous vector (i.e. step), the second instantaneous vector (i.e. step), progress multiplier (i.e. step learning rate).

[0056] (twenty two) (twenty three) (twenty four) (25) (26) (27) in, Indicates the number of current iterations. represents the learning rate, represents the first-order moment estimate of the gradient, represents the gradient of the current time step, represents the exponential decay rate of the first-order momentum, Indicates The second-order momentum of the step, represents the exponential decay rate of the second-order momentum. A hyperparameter that represents the strength of regularization and controls the regularization term The weight of Represents the objective function Parameters at the last moment The gradient at Indicates The model parameters of the step, Indicated in The second-order momentum of the step, Indicated in The first-order momentum estimate after step correction, Indicates that Second-order momentum estimation after step correction; express of power, with As it increases, its value approaches 0; express of Power. Represents a decimal number that prevents the denominator from being zero.

[0057] When the sample target classification result obtained by the initial multi-angle multi-feature fusion network target classification model reaches the set result, the trained initial multi-angle multi-feature fusion network target classification model is obtained. For example, when the sample target classification result is close to the type of the set sample target, the training of the initial multi-angle multi-feature fusion network target classification model is completed, and its network parameters are optimized to obtain the trained initial multi-angle multi-feature fusion network target classification model.

[0058] According to all the above embodiments, combined with the solution provided by this application, this application has the following advantages: (1) Breaking through the limitations of traditional classification methods, it can realize intelligent classification of radar targets in complex environments.

[0059] (2) It overcomes the shortcomings of single neural network feature extraction of single radar data, and simultaneously obtains the time-frequency information and time domain signal information in the echo signals collected by radars of different bands at multiple angles for feature extraction and fusion processing, thereby improving the target classification performance.

[0060] (3) The feature extraction capability of ConvLSTM is enhanced by utilizing the attention mechanism. The time-frequency feature information is extracted and fused through the attention mechanism. Finally, the spatiotemporal features and time-frequency features of the time domain information are fused through the AFF fusion module, thereby improving the accuracy of target classification and meeting the needs of actual applications.

[0061] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to the radar multi-angle multi-feature fusion target classification method. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a radar multi-angle multi-feature fusion target classification method is implemented.

[0062] Those skilled in the art will understand that Figure 7The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0063] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0064] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0065] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0066] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0067] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0068] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0069] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A radar multi-angle multi-feature fusion target classification method, characterized in that: The radar multi-angle multi-feature fusion target classification method includes: Acquire a multi-angle multi-feature fusion network target classification model; the multi-angle multi-feature fusion network target classification model includes: a temporal feature extraction module, an attention fusion module and a feature fusion module; Acquire the target's radar echo data in real time; Processing the radar echo data of the target to obtain processed radar echo data; the processed radar echo data includes: a time domain signal component diagram of the target and a time-frequency diagram of the target; The processed radar echo data is input into the multi-angle multi-feature fusion network target classification model to obtain a target classification result.

2. The radar multi-angle multi-feature fusion target classification method according to claim 1 is characterized in that: The construction process of the multi-angle and multi-feature fusion network target classification model includes: Acquire radar echo data of sample targets; Processing the radar echo data of the sample target to obtain processed radar echo data of the sample target; the processed radar echo data of the sample target includes: a time domain signal component diagram of the sample target and a time-frequency diagram of the sample target; Constructing a training data set based on the processed radar echo data of the sample target; Construct an initial multi-angle and multi-feature fusion network target classification model; The initial multi-angle multi-feature fusion network target classification model is trained based on the training data set until the sample target classification result output by the initial multi-angle multi-feature fusion network target classification model reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model; the trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

3. The radar multi-angle multi-feature fusion target classification method according to claim 2 is characterized in that: The acquisition process of the radar echo data of the sample target includes: Use radars of different bands to collect multi-angle radar echo data for each target in the set area, and obtain multi-angle radar echo data of all targets in the set area; The radar echo data of the sample target is obtained based on the multi-angle radar echo data of all targets in the set area.

4. The radar multi-angle multi-feature fusion target classification method according to claim 2 is characterized in that: Processing the radar echo data of the sample target to obtain processed radar echo data of the sample target includes: Using moving target indication to remove negative frequencies and filter out clutter on the radar echo data of the sample target to obtain pre-processed radar echo data of the sample target; Obtaining an extraction signal based on the preprocessed radar echo data of the sample target; Performing empirical mode decomposition on the extracted signal to obtain a time domain signal component graph of the sample target; Performing short-time Fourier transform on the extracted signal to obtain a time-frequency diagram of the sample target.

5. The radar multi-angle multi-feature fusion target classification method according to claim 4 is characterized in that: Constructing a training data set based on the processed radar echo data of the sample target, including: Constructing a first data set based on the time domain signal component map of the sample target; constructing a second data set based on the time-frequency diagram of the sample target; A training data set is constructed based on the first data set and the second data set.

6. The radar multi-angle multi-feature fusion target classification method according to claim 2 is characterized in that: The initial multi-angle multi-feature fusion network target classification model is trained based on the training data set until the sample target classification result output by the initial multi-angle multi-feature fusion network target classification model reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model, including: Using the initial multi-angle multi-feature fusion network target classification model, a sample target classification result is obtained based on the training data set; The AdamW algorithm is used to train the initial multi-angle multi-feature fusion network target classification model based on the sample target classification result until the sample target classification result reaches the set result, thereby obtaining the trained initial multi-angle multi-feature fusion network target classification model; The trained initial multi-angle multi-feature fusion network target classification model is used as the multi-angle multi-feature fusion network target classification model.

7. The radar multi-angle multi-feature fusion target classification method according to claim 5 is characterized in that: The attention fusion module includes: a CBAM attention feature extraction unit and a visual channel attention feature extraction unit; during the training process, the process of using the initial multi-angle multi-feature fusion network target classification model to obtain the sample target classification result includes: Using the temporal feature extraction module and the CBAM attention feature extraction unit, obtaining a first feature map based on the first data set; Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit, obtaining a second feature map based on the second data set; The feature fusion module is used to obtain the sample target classification result based on the first feature map and the second feature map.

8. The radar multi-angle multi-feature fusion target classification method according to claim 4 is characterized in that: Using formula Performing short-time Fourier transform on the extracted signal to obtain a time-frequency diagram of the sample target; In the formula, Represents a sliding window, using a Gaussian window function; represents the integral variable, represents the time dimension, represents the frequency dimension, Represents the result after short-time Fourier transform; Represents the extraction signal; is the kernel of Fourier transform, which represents the basis functions of different frequency components.

9. The radar multi-angle multi-feature fusion target classification method according to claim 4 is characterized in that: Using formula Performing empirical mode decomposition on the extracted signal to obtain a time domain signal component graph of the sample target; In the formula, represents the input signal, Indicates The eigenmode functions, represents the number of decomposed intrinsic mode functions, Represents the process of extracting the intrinsic mode function from the residual.

10. The radar multi-angle multi-feature fusion target classification method according to claim 1, characterized in that: The attention fusion module includes: a CBAM attention feature extraction unit and a visual channel attention feature extraction unit; the processed radar echo data is input into the multi-angle multi-feature fusion network target classification model to obtain a target classification result, including: Using the temporal feature extraction module and the CBAM attention feature extraction unit, a third feature map is obtained based on the processed radar echo data; Using the CBAM attention feature extraction unit and the visual channel attention feature extraction unit, a fourth feature map is obtained based on the processed radar echo data; The feature fusion module is used to obtain the target classification result based on the third feature map and the fourth feature map.

Images