An end-to-end high-dynamic target recognition method based on a self-attention mechanism
Through the end-to-end recognition method of the self-attention mechanism, the problems of degraded recognition performance and slow inference speed in high-dynamic target recognition are solved, efficient and robust recognition in complex scenarios is achieved, and recognition performance and speed are improved.
Patent Information
- Application Number
- CN202411880750.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-19
- Publication Date
- 2025-10-21
- Estimated Expiration
- 2044-12-19
AI Technical Summary
Existing technologies have problems with degraded recognition performance and slow inference speed in highly dynamic target recognition, especially in complex scenarios with multiple targets, high noise, and high nonlinearity, making it difficult to achieve robust recognition.
An end-to-end recognition method based on the self-attention mechanism is adopted to capture the data dependency in the RCS time series through the self-attention mechanism. Combined with deep feature extraction and classifier recognition, a sliding window sampling strategy for high-dynamic target RCS data is designed to automatically extract high-dimensional features and perform classification.
It achieves efficient and robust high-dynamic target recognition in multi-target, high-noise complex scenarios, improves recognition performance and inference speed, and avoids the reliance on manually designed features and the inability of convolutional networks to capture temporal relationships.
Smart Images
Figure CN119805394B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of high-dynamic target recognition, and in particular to an end-to-end recognition method based on a self-attention mechanism. Background Art
[0002] Accurately and rapidly identifying valid targets and decoys from non-cooperative parties is crucial for effectively intercepting valid targets and safeguarding national security. The attack process of a high-dynamic target can generally be divided into the boost phase, midcourse phase, and reentry phase. The midcourse phase, which can last 80% to 90% of the entire flight, is the longest and most easily identified phase. Radar cross-section (RCS) data of a target, obtained through radar, directly reflects the target's physical characteristics, including shape, size, and material. It also provides rich information about the target's attitude, including azimuth and elevation. Therefore, midcourse RCS data is often used for high-dynamic target recognition. Currently, high-dynamic target recognition falls into two main categories: feature extraction-based machine learning methods and data-driven deep learning methods. The former uses artificially designed features to extract classifiable features from the target and then uses a machine learning classifier to identify the target. The latter uses a neural network model trained on a high-dynamic target electromagnetic dataset to directly determine the target type from the target's electromagnetic data.
[0003] While these methods can achieve high-dynamic target classification to a certain extent, their recognition performance deteriorates when faced with complex scenarios involving multiple targets, high noise levels, and high nonlinearity. This is because feature extraction-based machine learning methods rely on the distribution of the original data and expert experience to design effective features for classification. When the original data is highly nonlinear or contains a lot of noise, complex feature design is required to achieve classification, which poses significant challenges for the designer. Furthermore, because feature extraction-based methods require the calculation of target features on the input data, their inference speed still needs to be improved, limiting their application in scenarios requiring real-time classification results. With the successful application of deep learning methods in many fields, the application of neural networks to high-dynamic target classification has attracted considerable attention. Convolutional neural networks (CNNs) have been used to classify high-dynamic targets based on RCS data. The CNN architecture automatically extracts high-dimensional features, eliminating the need for manual feature design based on the original data. The trained network can directly output the target category upon inputting the target RCS data. Although the introduction of CNN network structures has improved the performance of high-dynamic target recognition algorithms, convolutional networks do not consider the sequential nature of time series data. Therefore, when processing RCS time series, they cannot effectively capture the temporal relationships contained in the data, and their performance is poor when processing RCS data with strong temporal correlation. Therefore, there is an urgent need to design a high-dynamic target recognition algorithm that can effectively extract time series variation characteristics, has high inference speed, and can still achieve robust recognition performance in complex scenarios with multiple targets and high noise. Summary of the Invention
[0004] In light of this, the present invention provides an end-to-end high-dynamic target recognition method based on a self-attention mechanism. This method can effectively extract time series variation features, achieves high inference speed, and achieves robust recognition performance in complex scenarios with multiple targets and high noise.
[0005] To achieve the above object, the technical solution of the present invention includes the following steps:
[0006] Step 1: Obtain radar cross-section data of highly dynamic targets, perform preprocessing, and divide them into time series segments of different scales as multi-scale samples.
[0007] Step 2: Perform deep feature extraction on multi-scale time series data to obtain feature codes of target RCS at different scales, which are recorded as deep RCS features.
[0008] Step 3: Classify the deep RCS features to obtain target recognition results.
[0009] Furthermore, preprocessing includes removing outliers and normalizing data.
[0010] Furthermore, the time series segments are divided into different scales as multi-scale samples. The specific method is as follows:
[0011] First, determine the time windows of different scales, divide the preprocessed data by sliding windows, and set the sliding window step to half of the window length. Each time window intercepts a continuous piece of data, and multi-scale samples are obtained after sliding window sampling.
[0012] Furthermore, step 2 is specifically as follows:
[0013] For the multi-scale samples, the longest sequence length of the multi-scale samples is used as a benchmark. For sequences shorter than the benchmark, a padding value is added to the end to align them with the longest sequence length of the multi-scale samples to obtain an input sequence.
[0014] At the same time, a mask vector of the same length as the input sequence is created to record the valid data length of the input sample. The mask value 1 indicates valid input and 0 indicates the padding part.
[0015] The input sequence is input to a feature encoder; and multiple feature encoders are stacked for use.
[0016] The feature encoder includes a self-attention module, a residual connection module and a layer normalization module.
[0017] The self-attention module adopts a multi-head self-attention mechanism, which is implemented by parallel computing multiple independent self-attentions; the number of heads is set to h, and each head has an independent linear transformation matrix of query, key and value. Perform h independent self-attention calculations on the sample data to obtain h groups of attention weights head1, head2, ..., head h , Finally, all the obtained weights are concatenated into a vector and passed through a linear transformation matrix W O The final attention weight Z is obtained as follows:
[0018] Z=Concat(head1,head2,...,head h )W O
[0019] After the residual connection module obtains the attention weight Z, it performs a residual connection with the original input X, that is, O=X+Z.
[0020] The matrix O is input to the layer normalization module. The function of layer normalization is to normalize the hidden layer in the neural network to a standard normal distribution. By normalizing the output of each layer, the normalized output has zero mean and unit variance.
[0021] The final output of the feature encoder is represented as X hidden .
[0022] Furthermore, multiple feature encoders are stacked. Specifically, 6 feature encoders are stacked, and the output of the previous encoder is used as the input of the next encoder, and finally the deep RCS features of highly dynamic targets are obtained.
[0023] Furthermore, the classifier recognition module includes a fully connected layer, a linear activation layer and a classification layer.
[0024] The fully connected layer converts the deep RCS features into a fixed-length vector.
[0025] The linear activation layer performs a linear transformation on the output of the fully connected layer, adjusts the output dimension to the number of categories that high-dynamic targets can be classified into, and activates it using the RuLU function.
[0026] The classification layer uses the Softmax function to output the probability of each category.
[0027] Finally, the classification result is the index of the element position with the largest value in the probability vector output by the classification layer.
[0028] Beneficial effects:
[0029] The present invention provides an end-to-end high-dynamic target recognition method based on a self-attention mechanism, which includes three modules: input data processing, deep feature extraction, and classifier recognition. A sliding window sampling strategy for high-dynamic target RCS data is designed to obtain multi-scale time series samples from the original data, capturing the feature information of the original RCS data at different time scales. Compared with machine learning methods based on feature extraction, this method uses the performance of deep neural networks to automatically extract high-dimensional features, and directly calculates features from radar cross-section RCS data using network parameters, avoiding the dependence of manually designed features on experience and the problem of being unable to characterize target characteristics when the data has a high degree of nonlinearity, while having a faster reasoning speed. Compared with deep learning methods based on convolutional neural networks, this method uses a self-attention mechanism to capture data dependencies in RCS time series, can better focus on the intrinsic features in different high-dynamic target RCS sequences, and exhibits better high-dynamic target recognition performance in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 A framework for end-to-end high-dynamic target recognition. DETAILED DESCRIPTION
[0031] The present invention is described in detail below with reference to the accompanying drawings and embodiments.
[0032] The present invention provides a high-dynamic target recognition algorithm for complex scenes, which can still achieve robust recognition performance in complex scenes with multiple targets and high noise. The specific method framework is as follows Figure 1 shown.
[0033] A high-dynamic target is one that moves much faster than other targets in the background or the environment. The speed of a high-dynamic target varies depending on the target's range, ranging from 1.5 to 7 km / s.
[0034] The end-to-end high-dynamic target recognition algorithm of the present invention consists of three modules: an input data processing module, a deep feature extraction module, and a classifier recognition module. First, the input data is preprocessed to improve the data quality and divided into time series segments of different scales. Secondly, deep feature extraction is performed on the multi-scale time series data to obtain feature encodings of target RCS at different scales. Finally, the category of the target is output through the classifier. In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention is further described in detail below with reference to specific implementation cases.
[0035] Step 1: Obtain radar cross-section data of highly dynamic targets, and perform data preprocessing and multi-scale data segmentation. Data preprocessing includes removing outliers and normalizing the data. Multi-scale data segmentation requires first determining time windows of different scales, selecting shorter windows for fast-moving highly dynamic targets, and longer windows for rotating or oscillating targets. The original RCS data is divided by a sliding window, with the sliding window step set to half the window length. Each time window will intercept a continuous piece of data. The multi-scale samples obtained after sliding window sampling are conducive to the method capturing the characteristic information of the original RCS data at different time scales.
[0036] Step 2: Input the RCS sequence samples after sliding window sampling in step 1 into the deep feature extraction module. Since the length of the RCS sample data is inconsistent, for shorter sequences, a padding value (usually 0) is added to the end to align it with the longest sequence length. At the same time, a mask vector of the same length as the input sequence is created to record the effective data length of the input sample. The mask value is 1 for valid input and 0 for the padding part. When calculating self-attention, the mask is used to ignore the padding part, so that the data input to the network is a valid RCS sampling vector. Let the RCS sequence of the input model be X, with a shape of (l sample ,1), the deep feature dimension is l hidden . l sample is the sequence length of the longest RCS sample.
[0037] The feature encoder includes the main network layers such as the self-attention module, residual connection, and normalization. In the self-attention module, for each moment of the sequence data, by calculating its correlation with other data in the sequence (attention weight), and generating a representation that integrates global information, it can capture the dependency between different positions in the sequence and better process long time series. First, the shape of (l sample ,1) The input data becomes the shape of (l1,l2), satisfying l1*l2=l sample , and then through three independent linear transformation matrices W Q ,W K ,W V ∈(l2,l hidden ) respectively generate Q=XW Q Query, key K = XW K , value vector V = XW V ∈(l1,l hidden ). Each linear change transforms the last dimension l2 of the input data into the dimension l of the hidden layer feature hidden , the first dimension remains unchanged. Next, the dot product of the query and the key is calculated and scaled, and then the attention weight Z is obtained through the Sotfmax function, as follows:
[0038]
[0039] where d k The vector dimensions of the query and key.
[0040] In order to capture more levels of sequence information, a multi-head self-attention mechanism is adopted, which is achieved by parallel calculation of multiple independent self-attentions. The number of heads is set to h, and each head has an independent linear transformation matrix of query, key and value Perform h independent self-attention calculations on the sample data to obtain h groups of attention weights head1, head2, ..., head h Finally, all the obtained weights are concatenated into a vector Concat(head1,head2,...,head h ), and obtain the final attention weight Z through a linear transformation matrix, as follows:
[0041] MultiHead(Q,K,V)=Concat(head1,head2,...,head h )W O
[0042] in, W O It is a fully connected layer that transforms the multi-head attention concatenation result to be consistent with the input dimension.
[0043] Residual connections introduce "skip connections" to enable more direct gradient transfer, alleviating the vanishing and exploding gradient problems common in deep networks. Furthermore, residual connections allow information to be passed directly from the previous layer to the next, making it easier for the model to learn the identity mapping and helping to prevent information from being excessively distorted or lost in multi-layer nonlinear transformations. After obtaining the attention weights, they are connected to the original input X through a residual connection, i.e., O = X + Z.
[0044] The role of layer normalization is to normalize the hidden layer in the neural network to a standard normal distribution, so as to speed up the training and convergence. By normalizing the output of each layer, the normalized output has zero mean and unit variance. First, calculate the mean of each column of the matrix O and variance o ij is the element in the i-th row and j-th column of matrix O, and m is the number of rows in matrix O. The output of the normalization layer can be expressed as Where ∈ is a very small number to prevent the denominator from being zero, for example, ∈ can be set to 10 -8 or 10 -6 The normalized output is forward propagated and activated using the ReLU function. The expression of the ReLU activation function is ReLU(x)=max(0,x).
[0045] The final output of a feature encoder can be expressed as X hidden =Linear(ReLU(LayerNorm(O)));
[0046] X hidden Represents high-dimensional features learned by the network. Feature encoders are often stacked to increase network depth, enhance interactions between features, capture more complex patterns and features, and significantly enhance the model's expressive power. This method uses a stack of six feature encoders, with the output of the previous encoder serving as the input to the next encoder, ultimately obtaining deep RCS features for highly dynamic objects.
[0047] Step 3: Classify the deep RCS features. The classifier recognition module contains a fully connected layer, a linear activation layer, and a classification layer. The fully connected layer converts the deep features into a fixed-length vector. The linear activation layer performs a linear transformation on the output of the fully connected layer, adjusts the output dimension to the number of categories that high-dynamic targets can be classified into, and activates it using the RuLU function. The classification layer uses the Softmax function to output the probability of each category. The public declaration of the Softmax function is where z i is the score of category i. Ultimately, the classification result is the index of the element position with the largest value in the probability vector output by the classification layer.
[0048] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. An end-to-end high-dynamic target recognition method based on self-attention mechanism, characterized by: include: Step 1: Obtain radar cross-section data of highly dynamic targets, perform preprocessing, and divide it into time series segments of different scales as multi-scale samples. The specific method is as follows: A short window is selected for highly dynamic targets, and a long window is selected for rotating or oscillating targets. First, time windows of different scales are determined. The preprocessed data is divided using a sliding window method with a sliding window step size set to half the window length. Each time window intercepts a continuous segment of data, and multi-scale samples are obtained after sliding window sampling. Step 2: Perform deep feature extraction on multi-scale time series data to obtain feature codes of target RCS at different scales, which are recorded as deep RCS features; Step 3: Classify the deep RCS features to obtain target recognition results.
2. The end-to-end high-dynamic target recognition method based on the self-attention mechanism according to claim 1, characterized in that: The preprocessing includes deleting outliers and normalizing data.
3. The end-to-end high-dynamic target recognition method based on the self-attention mechanism according to claim 1, characterized in that: The step 2 is specifically as follows: For the multi-scale samples, taking the longest sequence length of the multi-scale samples as a benchmark, for sequences shorter than the benchmark, adding padding values at the end thereof to align them with the longest sequence length of the multi-scale samples, thereby obtaining an input sequence; At the same time, a mask vector of the same length as the input sequence is created to record the valid data length of the input sample. The mask value is 1 for valid input and 0 for padding. Inputting the input sequence into a feature encoder; Use multiple feature encoders in stacking; The feature encoder includes a self-attention module, a residual connection module and a layer normalization module; The self-attention module adopts a multi-head self-attention mechanism, which is implemented by parallel computing multiple independent self-attentions; the number of heads is set to h, and each head has an independent linear transformation matrix of query, key and value. Perform h independent self-attention calculations on the sample data to obtain h groups of attention weights head1, head2, ..., head h , Finally, all the obtained weights are concatenated into a vector and passed through a linear transformation matrix W O The final attention weight Z is obtained as follows: Z=Concat(head1,head2,...,head h )W O After the residual connection module obtains the attention weight Z, it performs a residual connection with the original input X, that is, O = X + Z; The matrix O is input to the layer normalization module. The function of layer normalization is to normalize the hidden layer in the neural network to a standard normal distribution. By normalizing the output of each layer, the normalized output has zero mean and unit variance. The final output of the feature encoder is represented as X hidden .
4. The end-to-end high-dynamic target recognition method based on the self-attention mechanism according to claim 3, characterized in that: The method uses a stack of multiple feature encoders, specifically: using a stack of 6 feature encoders, with the output of the previous encoder serving as the input of the next encoder, and ultimately obtaining deep RCS features of highly dynamic targets.
5. The end-to-end high-dynamic target recognition method based on the self-attention mechanism according to claim 4, characterized in that: The classifier recognition module contains a fully connected layer, a linear activation layer, and a classification layer; The fully connected layer converts the deep RCS features into a fixed-length vector; The linear activation layer performs a linear transformation on the output of the fully connected layer, adjusts the output dimension to the number of categories that high-dynamic targets can be classified into, and uses the RuLU function for activation; The classification layer uses the Softmax function to output the probability of each category; Finally, the classification result is the index of the element position with the largest value in the probability vector output by the classification layer.