A multi-band radar fusion processing method and system
Through the multi-band radar fusion processing method, deep neural network and gamma correction module are used to optimize the radar signal, and information fusion is combined with autocorrelation attention network, solving the detection accuracy and interference problems of traditional single-band radar, achieving more accurate goals Identify and classify.
Patent Information
- Application Number
- CN202410418160.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-08
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-04-08
AI Technical Summary
Traditional single-band radars are difficult to accurately detect target types and are susceptible to interference, so they cannot completely extract the shape, size, material and motion characteristics of the target.
The multi-band radar fusion processing method is adopted to extract the semantic features of the radar reflected signal through the deep neural network model, and the signal is optimized using the gamma correction module, and the autocorrelation fusion of multi-source information is combined with the autocorrelation attention network to determine the target type label.
It improves the accuracy and robustness of target recognition, and can more comprehensively capture the shape, size, material and motion characteristics of the target, reduce noise interference, and enhance detection performance.
Smart Images

Figure CN118311563B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of radar, and in particular to a multi-band radar fusion processing method and system. Background Art
[0002] Radar (Radio Detection and Ranging) is a system that uses radio waves to detect objects and determine their range, speed, and direction. A radar system transmits a pulse of radio waves and then receives the signal reflected from the target object. By measuring the time delay, frequency shift, and amplitude of the reflected signal, radar can determine the range, speed, and radar cross section (RCS) of the target.
[0003] However, traditional solutions for detecting target categories through radar usually use single-band radar, but single-band radar can only extract the scattering characteristics of the target at that specific frequency, which may lead to incomplete extraction of target features such as shape, size, material and movement. In addition, single-band radar is susceptible to interference from other radar systems or environmental noise, which may reduce the accuracy of target detection.
[0004] Therefore, a multi-band radar fusion processing solution is desired. Summary of the invention
[0005] The present disclosure is made in view of the above problems. One object of the present disclosure is to provide a multi-band radar fusion processing method and system.
[0006] An embodiment of the present disclosure provides a multi-band radar fusion processing method, which includes: acquiring multiple radar reflection signals returned from a target object, wherein the multiple radar reflection signals have different bands; performing feature extraction on the multiple radar reflection signals respectively through a signal feature extractor based on a deep neural network model to obtain multiple radar reflection signal semantic feature vectors; passing the multiple radar reflection signal semantic feature vectors respectively through a nonlinear response optimizer based on a gamma correction module to obtain multiple optimized radar reflection signal semantic feature vectors; performing multi-source information autocorrelation fusion on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features; and determining the type label of the target object based on the multi-source radar signal semantic fusion features.
[0007] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, the deep neural network model is a convolutional neural network model.
[0008] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, the multiple radar reflection signal semantic feature vectors are respectively passed through a nonlinear response optimizer based on a gamma correction module to obtain multiple optimized radar reflection signal semantic feature vectors, including:
[0009] The plurality of radar reflection signal semantic feature vectors are processed respectively by the nonlinear response optimizer based on the gamma correction module according to the following gamma correction formula to obtain the plurality of optimized radar reflection signal semantic feature vectors;
[0010] Wherein, the gamma correction formula is:
[0011]
[0012] in, is the feature value of the ith position of the sth radar reflection signal semantic feature vector in the multiple radar reflection signal semantic feature vectors, A, B, C and D are adjustment hyperparameters, n i It is the feature value of the i-th position of the s-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors.
[0013] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, wherein the multi-source information autocorrelation fusion is performed on the multiple optimized radar reflection signal semantic feature vectors to obtain the multi-source radar signal semantic fusion feature, including:
[0014] The multiple optimized radar reflection signal semantic feature vectors are passed through a multi-source information autocorrelation fuser based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector as the multi-source radar signal semantic fusion feature.
[0015] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, the plurality of optimized radar reflection signal semantic feature vectors are passed through a multi-source information autocorrelation fuser based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector as the multi-source radar signal semantic fusion feature, including:
[0016] The multiple optimized radar reflection signal semantic feature vectors are processed by the multi-source information autocorrelation fusion device based on the autocorrelation attention network using the following multi-source information fusion formula to obtain the multi-source radar signal semantic fusion feature vector; wherein the multi-source information fusion formula is:
[0017]
[0018] Among them, v i is the i-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors, and W i Represent the weight coefficient vector and weight coefficient matrix respectively, B i is the offset vector, tanh(·) represents the tanh function, ei is the attention score of the i-th optimized radar reflection signal semantic feature vector, exp(·) represents exponential operation, e k is the attention score of the kth optimized radar reflection signal semantic feature vector, t is the number of vectors of the multiple optimized radar reflection signal semantic feature vectors, a i is the weight value of the i-th optimized radar reflection signal semantic feature vector, and v1 is the multi-source radar signal semantic fusion feature vector.
[0019] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, determining the type label of the target object based on the semantic fusion features of the multi-source radar signal includes:
[0020] The semantic fusion feature vector of the multi-source radar signal is passed through a target identifier based on a classifier to obtain a recognition result, and the recognition result is used to represent a type label of the target object.
[0021] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, it also includes a training step: for training the signal feature extractor based on the convolutional neural network model, the nonlinear response optimizer based on the gamma correction module, the multi-source information autocorrelation fuser based on the autocorrelation attention network, and the target identifier based on the classifier.
[0022] For example, according to the multi-band radar fusion processing method of an embodiment of the present disclosure, the training step includes: acquiring training data, the training data including a plurality of training radar reflection signals returned from a target object, wherein the plurality of training radar reflection signals have different bands; performing feature extraction on the plurality of training radar reflection signals respectively through the signal feature extractor based on the convolutional neural network model to obtain a plurality of training radar reflection signal semantic feature vectors; passing the plurality of training radar reflection signal semantic feature vectors respectively through the nonlinear response optimizer based on the gamma correction module to obtain a plurality of optimized training radar reflection signal semantic feature vectors; and The signal semantic feature vector is passed through the multi-source information autocorrelation fusion device based on the autocorrelation attention network to obtain a training multi-source radar signal semantic fusion feature vector; the training multi-source radar signal semantic fusion feature vector is passed through the target identifier based on the classifier to obtain a classification loss function value; based on the classification loss function value, the signal feature extractor based on the convolutional neural network model, the nonlinear response optimizer based on the gamma correction module, the multi-source information autocorrelation fusion device based on the autocorrelation attention network and the target identifier based on the classifier are trained, wherein in each round of iteration of the training, the training multi-source radar signal semantic fusion feature vector is iteratively optimized.
[0023] An embodiment of the present disclosure also provides a multi-band radar fusion processing system, which includes: a signal acquisition module, used to acquire multiple radar reflection signals returned from a target object, wherein the multiple radar reflection signals have different bands; a semantic feature extraction module, used to perform feature extraction on the multiple radar reflection signals respectively through a signal feature extractor based on a deep neural network model to obtain multiple radar reflection signal semantic feature vectors; a nonlinear response optimization module, used to pass the multiple radar reflection signal semantic feature vectors respectively through a nonlinear response optimizer based on a gamma correction module to obtain multiple optimized radar reflection signal semantic feature vectors; a multi-source information autocorrelation fusion module, used to perform multi-source information autocorrelation fusion on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features; a type analysis module, used to determine the type label of the target object based on the multi-source radar signal semantic fusion features.
[0024] For example, according to the multi-band radar fusion processing system of an embodiment of the present disclosure, the deep neural network model is a convolutional neural network model.
[0025] According to the multi-band radar fusion processing method of the embodiment of the present disclosure, it collects radar reflection signals of multiple different bands returned from the target object, and introduces signal processing and analysis algorithms at the back end to perform collaborative correlation analysis of these radar reflection signals, so as to utilize the scattering characteristics of radar signals of different bands to capture the shape, size, material, movement and other characteristic information of the target object, and based on this, identify and detect the type of the target object. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings of the embodiments of the present disclosure. Obviously, the drawings described below only relate to some embodiments of the present disclosure, and are not intended to limit the present disclosure.
[0027] Figure 1 A schematic diagram of the application architecture of the multi-band radar fusion processing method in an embodiment of the present disclosure is shown;
[0028] Figure 2 A flow chart of a multi-band radar fusion processing method in an embodiment of the present disclosure is shown;
[0029] Figure 3 A schematic diagram of the structure of a multi-band radar fusion processing system in an embodiment of the present disclosure is shown;
[0030] Figure 4 An application scenario diagram of the multi-band radar fusion processing method in an embodiment of the present disclosure is shown;
[0031] Figure 5 A schematic diagram of a storage medium according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION
[0032] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present disclosure, rather than all embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present disclosure.
[0033] The terms used in this specification are those general terms currently widely used in the art in consideration of the functions of the present disclosure, but these terms may vary according to the intention of a person of ordinary skill in the art, precedents, or new technologies in the art. In addition, specific terms may be selected by the applicant, and in this case, their detailed meanings will be described in the detailed description of the present disclosure. Therefore, the terms used in the specification should not be understood as simple names, but rather as an overall description based on the meaning of the terms and the present disclosure.
[0034] Although the present disclosure makes various references to certain modules in the system according to embodiments of the present disclosure, any number of different modules can be used and run on a user terminal and / or server. The modules are illustrative only, and different aspects of the systems and methods can use different modules.
[0035] Flowcharts are used in this disclosure to illustrate the operations performed by the system according to the embodiments of the present disclosure. It should be understood that the preceding or following operations are not necessarily performed precisely in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.
[0036] Figure 1 A schematic diagram of the application architecture of the multi-band radar fusion processing method in an embodiment of the present disclosure is shown, including a server 100 and a terminal device 200.
[0037] The terminal device 200 and the server 100 can be connected via the Internet to achieve mutual communication. Optionally, the above-mentioned Internet uses standard communication technology and / or protocol. The Internet is usually the Internet, but it can also be any network, including but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a dedicated network or a virtual private network. In some embodiments, the data exchanged through the network is represented by technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec) can also be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.
[0038] The server 100 can provide various network services for the terminal device 200, wherein the server 100 can be a single server, a server cluster consisting of several servers, or a cloud computing center. Specifically, the server 100 may include a processor 110 (Center Processing Unit, CPU), a memory 120, an input device 130, and an output device 140, etc. The input device 130 may include a keyboard, a mouse, a touch screen, etc., and the output device 140 may include a display device, such as a liquid crystal display (Liquid Crystal Display, LCD), a cathode ray tube (Cathode Ray Tube, CRT), etc.
[0039] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides the processor 110 with program instructions and data stored in the memory 120. In the embodiment of the present disclosure, the memory 120 may be used to store the program of the multi-band radar fusion processing method in the embodiment of the present disclosure.
[0040] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any multi-band radar fusion processing method in the embodiments of the present disclosure according to the obtained program instructions.
[0041] In addition, the application architecture diagram in the embodiment of the present disclosure is intended to more clearly illustrate the technical solution in the embodiment of the present disclosure, and does not constitute a limitation on the technical solution provided in the embodiment of the present disclosure. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present disclosure is also applicable to similar problems.
[0042] The following is a non-limiting description of the multi-band radar fusion processing method provided according to at least one embodiment of the present disclosure through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present disclosure.
[0043] At present, radar signals of a single band often have difficulty in accurately distinguishing the types of targets, because different types of targets may have similar radar scattering cross sections or similar motion states. Radar signals of different bands have different scattering characteristics and can reflect different aspects of the target. For example, high-frequency radar signals are more sensitive to small targets and surface details, while low-frequency radar signals can penetrate the target and provide internal structure information. This enables them to reveal different aspects of the target, such as shape, size, material, and motion. By fusing multi-band radar signals, the comprehensive features of the target can be extracted, thereby improving the accuracy and robustness of target recognition.
[0044] Based on this, in the technical solution of the present application, a multi-band radar fusion processing method is proposed, which can utilize the complementary characteristics of radar signals of different bands to enhance the performance of target detection and recognition. Specifically, the technical concept itself is to collect radar reflection signals of multiple different bands returned from the target object, and introduce signal processing and analysis algorithms at the back end to perform collaborative correlation analysis of these radar reflection signals, so as to utilize the scattering characteristics of radar signals of different bands to capture the shape, size, material, movement and other characteristic information of the target object, and based on this, identify and detect the type of the target object. In this way, the fusion information contained in the multi-band radar signal can be used to comprehensively characterize the characteristics of the target object, thereby improving the accuracy of target recognition and classification.
[0045] Figure 2 1 shows a flowchart of a multi-band radar fusion processing method in an embodiment of the present disclosure. For example, the multi-band radar fusion processing method can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 2 As shown, according to the multi-band radar fusion processing method of the embodiment of the present disclosure, the method includes the following steps: S510, acquiring multiple radar reflection signals returned from a target object, wherein the multiple radar reflection signals have different bands; S520, performing feature extraction on the multiple radar reflection signals respectively through a signal feature extractor based on a deep neural network model to obtain multiple radar reflection signal semantic feature vectors; S530, passing the multiple radar reflection signal semantic feature vectors respectively through a nonlinear response optimizer based on a gamma correction module to obtain multiple optimized radar reflection signal semantic feature vectors; S540, performing multi-source information autocorrelation fusion on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features; and, S550, determining the type label of the target object based on the multi-source radar signal semantic fusion features.
[0046] Specifically, in the technical solution of the present application, first, a plurality of radar reflection signals returned from the target object are obtained, wherein the plurality of radar reflection signals have different bands. Next, considering that radar reflection signals of different bands contain semantic features about different aspects of the target object, for example, high-frequency radar signals are more sensitive to small targets and surface details, and they can capture feature information about the shape, texture, color, edge, etc. of the target object, while low-frequency radar signals can penetrate the target and provide internal structure information, and they can capture semantic features about the material of the target object. Based on this, in the technical solution of the present application, the plurality of radar reflection signals need to be respectively input into a signal feature extractor based on a convolutional neural network model for feature mining, so as to respectively extract the signal semantic feature information in the radar reflection signals of different bands, thereby obtaining a plurality of radar reflection signal semantic feature vectors.
[0047] Correspondingly, in step S520, the deep neural network model is a convolutional neural network model, that is, the signal feature extractor based on the deep neural network model is a signal feature extractor based on the convolutional neural network model.
[0048] It is worth mentioning that the Convolutional Neural Network (CNN) is a deep learning model that is specifically designed to process data with a grid structure. The main features of a convolutional neural network include: 1. Convolutional Layer: The convolution operation is the core part of the convolutional neural network. The convolution operation can extract features from the image. The convolution layer uses a convolution kernel (filter) to perform a convolution operation on the input data to obtain a feature map. 2. Pooling Layer: The pooling layer is used for downsampling, reducing the dimension of the feature map while retaining the most important features. Common pooling operations include maximum pooling and average pooling. 3. Activation Function: The activation function introduces nonlinear factors, allowing the neural network to learn nonlinear relationships. Common activation functions include ReLU (Rectified Linear Unit) and Sigmoid. 4. Fully Connected Layer: The fully connected layer is used to perform tasks such as classification or regression on the features extracted by the convolution layer. The fully connected layer connects all neurons in the previous layer to all neurons in the current layer. The applications of convolutional neural networks include image classification, target detection, face recognition, text classification in natural language processing, etc. In the multi-band radar fusion processing method, the convolutional neural network model is used as a signal feature extractor. By learning the features of multiple radar reflection signals, the semantic feature vector is extracted to achieve the recognition and classification of the target object. Convolutional neural networks can automatically extract features from data, reducing the complexity of manual feature engineering.
[0049] It should be understood that, considering that different radar signal receivers may have differences in sensitivity, response speed, etc., the collected radar reflection signals are biased. In addition, radar signals in different bands may have different dynamic ranges and response characteristics. Therefore, in order to eliminate or reduce the measurement errors between different signal receivers during data processing to ensure the accuracy and reliability of radar reflection signals, in the technical solution of the present application, the multiple radar reflection signal semantic feature vectors are further passed through a nonlinear response optimizer based on a gamma correction module to obtain multiple optimized radar reflection signal semantic feature vectors. It should be understood that by applying the nonlinear response optimizer based on gamma correction, radar signals in different bands can be made more consistent and comparable in the feature extraction stage. In addition, gamma correction can help eliminate the influence of nonlinear response in the data, making the data more accurate and reliable. This can enhance the discriminative power of features and make them more suitable for target category detection.
[0050] Accordingly, in step S530, the plurality of radar reflection signal semantic feature vectors are respectively processed through a nonlinear response optimizer based on a gamma correction module to obtain a plurality of optimized radar reflection signal semantic feature vectors, including: the plurality of radar reflection signal semantic feature vectors are respectively processed through the nonlinear response optimizer based on the gamma correction module using the following gamma correction formula to obtain the plurality of optimized radar reflection signal semantic feature vectors; wherein the gamma correction formula is:
[0051]
[0052] in, is the feature value of the ith position of the sth radar reflection signal semantic feature vector in the multiple radar reflection signal semantic feature vectors, A, B, C and D are adjustment hyperparameters, n i It is the feature value of the i-th position of the s-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors.
[0053] Further, considering that the multiple optimized radar reflection signal semantic feature vectors respectively represent the semantic features of different aspects of the target object contained in the radar reflection signals of different bands, in order to be able to effectively integrate these target object semantic features, so as to obtain a more comprehensive and integrated target feature representation, in the technical solution of the present application, the multiple optimized radar reflection signal semantic feature vectors are further passed through a multi-source information autocorrelation fusion device based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector. It should be understood that the autocorrelation attention network can help the model learn the correlation and importance between the semantic features of radar reflection signals of different bands. By introducing the autocorrelation attention mechanism, the interaction and dependency between the semantic features of different aspects of the target object contained in the signals of different bands can be better captured, thereby improving the effect of feature fusion. In addition, the multi-source information autocorrelation fusion device can help eliminate the noise and redundant information between the signal features of different bands, and improve the robustness and stability of feature representation. This helps to reduce the sensitivity to noise and interference and improve the accuracy of target recognition.
[0054] Accordingly, in step S540, multi-source information autocorrelation fusion is performed on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features, including: passing the multiple optimized radar reflection signal semantic feature vectors through a multi-source information autocorrelation fuser based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector as the multi-source radar signal semantic fusion feature.
[0055] It is worth mentioning that the Self-Attention Network is a variant of the attention mechanism, which is often used to process sequence data. The self-attention mechanism allows the model to take into account the correlation between different positions when processing sequence data, so as to better capture long-distance dependencies. The basic idea of the self-attention mechanism is to calculate the attention weight of each position in the sequence, and then weight these weights with the representations of other positions in the sequence to obtain the contextual representation of each position. The self-attention mechanism not only considers the relationship between different positions in the sequence, but also automatically learns the importance between different positions based on the input data. In the multi-source radar signal fusion processing, the self-attention network is used as a multi-source information autocorrelation fuser to integrate the semantic feature vectors of multiple radar signals to obtain the semantic fusion feature vector of the multi-source radar signal. Through the self-attention network, the model can learn the correlation and importance between different radar signals, so as to better fuse multiple signals and improve the recognition and classification performance of the target object. In general, the role of the self-attention network in the multi-source information fusion task includes: 1. Capturing the correlation between different positions in the sequence: The self-attention mechanism can effectively capture long-distance dependencies and help the model understand the correlation between different positions. 2. Learn the importance of different positions: By learning the attention weight of each position, the model can automatically determine the importance of different positions in the fusion process, thereby better integrating multi-source information. 3. Improve the effect of information fusion: The autocorrelation attention network can effectively integrate the semantic features of multiple signals, improve the effect of multi-source information fusion, and thus improve the accuracy of target object classification.
[0056] Specifically, the multiple optimized radar reflection signal semantic feature vectors are passed through a multi-source information autocorrelation fusion device based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector as the multi-source radar signal semantic fusion feature, including: the multiple optimized radar reflection signal semantic feature vectors are passed through the multi-source information autocorrelation fusion device based on an autocorrelation attention network to process the multi-source information autocorrelation fusion device according to the following multi-source information fusion formula to obtain the multi-source radar signal semantic fusion feature vector; wherein the multi-source information fusion formula is:
[0057]
[0058] Among them, v i is the i-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors, and W i Represent the weight coefficient vector and weight coefficient matrix respectively, B i is the offset vector, tanh(·) represents the tanh function, e iis the attention score of the i-th optimized radar reflection signal semantic feature vector, exp(·) represents exponential operation, e k is the attention score of the kth optimized radar reflection signal semantic feature vector, t is the number of vectors of the multiple optimized radar reflection signal semantic feature vectors, a i is the weight value of the i-th optimized radar reflection signal semantic feature vector, and v1 is the multi-source radar signal semantic fusion feature vector.
[0059] Then, the semantic fusion feature vector of the multi-source radar signal is passed through a target identifier based on a classifier to obtain a recognition result, and the recognition result is used to represent the type label of the target object. Specifically, the label of the classifier is the type label of the target object. Therefore, after obtaining the classification result, the type of the target object can be identified and detected based on the classification result. In this way, the fusion information contained in the multi-band radar signal can be used to comprehensively characterize the characteristics of the target object, thereby improving the accuracy of target recognition and classification.
[0060] Accordingly, in step S550, based on the multi-source radar signal semantic fusion features, determining the type label of the target object includes: passing the multi-source radar signal semantic fusion feature vector through a classifier-based target identifier to obtain a recognition result, and the recognition result is used to represent the type label of the target object.
[0061] Specifically, the multi-source radar signal semantic fusion feature vector is passed through a classifier-based target identifier to obtain a recognition result, and the recognition result is used to represent the type label of the target object, including: using the fully connected layer of the classifier-based target identifier to fully connect the multi-source radar signal semantic fusion feature vector to obtain an encoded classification feature vector; and inputting the encoded classification feature vector into the Softmax classification function of the classifier-based target identifier to obtain the recognition result.
[0062] It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression and SVM are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification. However, this is prone to errors and is not efficient. Commonly used multi-classification methods include the Softmax classification function.
[0063] Furthermore, in the technical solution of the present application, the multi-band radar fusion processing method further includes a training step: for training the signal feature extractor based on the convolutional neural network model, the nonlinear response optimizer based on the gamma correction module, the multi-source information autocorrelation fusion device based on the autocorrelation attention network, and the target identifier based on the classifier. It should be understood that the training step plays a vital role in the multi-band radar fusion processing method, and its main purposes include: 1. Model parameter learning: through the training step, the model can learn how to correctly extract signal features, optimize signal features, fuse multi-source information, and identify targets. During the training process, the model continuously adjusts parameters through the back propagation algorithm, so that the model can better extract features and make accurate classification and recognition when processing multi-band radar data. 2. Improved generalization ability: through training, the model can learn the patterns and laws in the data, thereby improving its generalization ability for new samples, which means that the model can also make accurate predictions and recognition on data other than the data it has seen. 3. Optimize model performance: The training step helps optimize the performance of the model, making it more accurate, robust, and efficient when processing multi-band radar data. Through training, the model can continuously improve its own representation capabilities to better adapt to the requirements of the task. 4. Parameter adjustment: In the training step, the parameters of each component of the model can be adjusted and optimized according to the actual data and task requirements to ensure that the various parts of the entire multi-band radar fusion processing method can work in coordination to achieve the best performance. In general, the training step is a key link in the entire multi-band radar fusion processing method. Through training, the model can be continuously optimized and learned, so as to achieve more accurate and efficient processing of multi-band radar data, and ultimately achieve the recognition and classification of target objects.
[0064] In one example, the training step includes: acquiring training data, the training data including a plurality of training radar reflection signals returned from a target object, wherein the plurality of training radar reflection signals have different bands; performing feature extraction on the plurality of training radar reflection signals respectively through the signal feature extractor based on the convolutional neural network model to obtain a plurality of training radar reflection signal semantic feature vectors; passing the plurality of training radar reflection signal semantic feature vectors respectively through the nonlinear response optimizer based on the gamma correction module to obtain a plurality of optimized training radar reflection signal semantic feature vectors; passing the plurality of optimized training radar reflection signal semantic feature vectors through the A multi-source information autocorrelation fuser based on an autocorrelation attention network is used to obtain a semantic fusion feature vector of a training multi-source radar signal; the semantic fusion feature vector of the training multi-source radar signal is passed through the target identifier based on the classifier to obtain a classification loss function value; and based on the classification loss function value, the signal feature extractor based on the convolutional neural network model, the nonlinear response optimizer based on the gamma correction module, the multi-source information autocorrelation fuser based on the autocorrelation attention network and the target identifier based on the classifier are trained, wherein in each round of iteration of the training, the semantic fusion feature vector of the training multi-source radar signal is iteratively optimized.
[0065] In the above technical solution, the multiple optimized training radar reflection signal semantic feature vectors respectively express the nonlinear response optimization presentation of the signal waveform image semantic feature distribution of the multiple training radar reflection signals. When the multiple optimized training radar reflection signal semantic feature vectors are passed through a multi-source information autocorrelation fusion device based on an autocorrelation attention network, the multi-source information autocorrelation fusion device based on an autocorrelation attention network can perform autocorrelation attention enhancement on each of the multiple optimized training radar reflection signal semantic feature vectors based on the correlation information between the waveform semantic features of the reflection signals of different bands, so as to emphasize the significance and importance of the fluctuation characteristics of the training radar reflection signals of different bands in the global feature distribution.
[0066] However, considering that the multiple training radar reflection signals have different bands, the multiple training radar reflection signals have significant signal differences in the signal source domain. After convolution coding and gamma correction, the feature distribution inconsistency between the multiple optimized training radar reflection signal semantic feature vectors will become significant. In this way, when the multiple optimized training radar reflection signal semantic feature vectors pass through the multi-source information autocorrelation fusion device based on the autocorrelation attention network, the inconsistency of the local feature distribution of the obtained training multi-source radar signal semantic fusion feature vector based on the feature value distribution affects the convergence speed of the training multi-source radar signal semantic fusion feature vector through the class regression iteration of the target identifier based on the classifier.
[0067] Therefore, the applicant of the present application iteratively optimizes the semantic fusion feature vector of the training multi-source radar signal each time the semantic fusion feature vector of the training multi-source radar signal is iterated through classification regression by a target identifier based on a classifier.
[0068] Accordingly, in one example, in each iteration of the training, the training multi-source radar signal semantic fusion feature vector is iteratively optimized using the following iterative optimization formula to obtain an optimized training multi-source radar signal semantic fusion feature vector; wherein the iterative optimization formula is:
[0069]
[0070] in, is the i-th eigenvalue of the semantic fusion feature vector of the training multi-source radar signal, Represents the probabilistic function of the eigenvalue, that is, the eigenvalue is a probabilistic function mapped to the interval [0,1], p is a class probability value obtained by a target identifier based on a classifier, L is the length of the semantic fusion feature vector of the training multi-source radar signal, and α is a weight hyperparameter, exp[·] represents an exponential operation of a value, and the exponential operation of the value represents calculating a natural exponential function value with the value as a power, is the i-th eigenvalue of the semantic fusion feature vector of the optimized training multi-source radar signal.
[0071] Here, the eigenvalues of the multi-source radar signal semantic fusion feature vector are calculated by taking the Bernoulli probability likelihood distribution as the authoritative meta-entity Based on the high-level class decision of Bernoulli probability distribution (p, 1-p), and introducing the probability density space mapping allocation relative to the whole feature set, the eigenvalue of the semantic fusion feature vector of the multi-source radar signal is determined. The key insights relative to the class probability mapping are used to realize the eigenvalues of the semantic fusion feature vector of the multi-source radar signal Based on the correctness of the matching function of the class probability distribution (p, 1-p) and the class convergence, the convergence performance is superior to promote the convergence of sample iteration and improve the classification training speed. In this way, the scattering characteristics of radar signals in different bands can be used to capture the shape, size, material, movement and other characteristic information of the target object, and based on this, the target object type can be identified and detected, thereby improving the accuracy of target recognition and classification.
[0072] Based on the above embodiments, see Figure 3 As shown, it is a structural schematic diagram of a multi-band radar fusion processing system 800 in an embodiment of the present disclosure. The multi-band radar fusion processing system 800 includes: a signal acquisition module 810, which is used to acquire multiple radar reflection signals returned from a target object, wherein the multiple radar reflection signals have different bands; a semantic feature extraction module 820, which is used to extract features from the multiple radar reflection signals respectively through a signal feature extractor based on a deep neural network model to obtain multiple radar reflection signal semantic feature vectors; a nonlinear response optimization module 830, which is used to pass the multiple radar reflection signal semantic feature vectors through a nonlinear response optimizer based on a gamma correction module to obtain multiple optimized radar reflection signal semantic feature vectors; a multi-source information autocorrelation fusion module 840, which is used to perform multi-source information autocorrelation fusion on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features; and a type analysis module 850, which is used to determine the type label of the target object based on the multi-source radar signal semantic fusion features.
[0073] In one example, in the multi-band radar fusion processing system 800, the deep neural network model is a convolutional neural network model.
[0074] Here, those skilled in the art will appreciate that the specific functions and operations of each module in the multi-band radar fusion processing system 800 have been described in detail above. Figure 2 It has been described in detail in the description of the multi-band radar fusion processing method, and therefore, its repeated description will be omitted.
[0075] Figure 4 FIG. 1 is an application scenario diagram of the multi-band radar fusion processing method according to an embodiment of the present disclosure. Figure 4 As shown, in this application scenario, first, multiple radar reflection signals (for example, Figure 4 D), wherein the multiple radar reflection signals have different bands, and then the multiple radar reflection signals are input to a server (for example, Figure 4In S) as shown in , the server is capable of using the multi-band radar fusion processing algorithm to process the multiple radar reflection signals to obtain an identification result for representing a type label of a target object.
[0076] Based on the above embodiments, another exemplary embodiment of an electronic device is also provided in the embodiments of the present disclosure. In some possible implementations, the electronic device in the embodiments of the present disclosure may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the steps of the multi-band radar fusion processing method in the above embodiments when executing the program.
[0077] For example, the electronic device is disclosed in the present invention Figure 1 Taking the server 100 in the example as an example, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.
[0078] An embodiment of the present disclosure also provides a computer-readable storage medium. Figure 5 FIG. 1 is a schematic diagram of a computer-readable storage medium 1000 according to an embodiment of the present disclosure. Figure 5 As shown, the computer-readable storage medium 1000 stores computer-executable instructions 1001. When the computer-executable instructions 1001 are executed by the processor, the multi-band radar fusion processing method according to the embodiment of the present disclosure described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc.
[0079] The embodiments of the present disclosure also provide a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the multi-band radar fusion processing method according to the embodiments of the present disclosure.
[0080] Those skilled in the art will appreciate that the contents disclosed in this disclosure may be subject to various modifications and improvements. For example, the various devices or components described above may be implemented by hardware, or by software, firmware, or a combination of some or all of the three.
[0081] In addition, although the present disclosure makes various references to certain units in the system according to embodiments of the present disclosure, any number of different units may be used and run on the client and / or server. The units are illustrative only, and different aspects of the systems and methods may use different units.
[0082] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of a software functional module. The present disclosure is not limited to any particular form of combination of hardware and software.
[0083] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meanings as commonly understood by those of ordinary skill in the art to which the present disclosure belongs. It should also be understood that terms such as those defined in a common dictionary should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formalized sense, unless explicitly defined as such herein. The above is an explanation of the present disclosure and should not be considered as a limitation thereof. Although several exemplary embodiments of the present disclosure have been described, it will be readily understood by those skilled in the art that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the present disclosure. Therefore, all these modifications are intended to be included within the scope of the present disclosure as defined in the claims. It should be understood that the above is an explanation of the present disclosure and should not be considered to be limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present disclosure is defined by the claims and their equivalents.
Claims
1. A multi-band radar fusion processing method, characterized in that: include: Acquire a plurality of radar reflection signals returned from the target object, wherein the plurality of radar reflection signals have different bands; Performing feature extraction on the multiple radar reflection signals respectively by a signal feature extractor based on a deep neural network model to obtain multiple radar reflection signal semantic feature vectors; Passing the plurality of radar reflection signal semantic feature vectors through a nonlinear response optimizer based on a gamma correction module to obtain a plurality of optimized radar reflection signal semantic feature vectors; Performing multi-source information autocorrelation fusion on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features; and Determining a type label of a target object based on the semantic fusion features of the multi-source radar signals; The deep neural network model is a convolutional neural network model; The plurality of radar reflection signal semantic feature vectors are respectively passed through a nonlinear response optimizer based on a gamma correction module to obtain a plurality of optimized radar reflection signal semantic feature vectors, including: The plurality of radar reflection signal semantic feature vectors are processed respectively by the nonlinear response optimizer based on the gamma correction module according to the following gamma correction formula to obtain the plurality of optimized radar reflection signal semantic feature vectors; Wherein, the gamma correction formula is: in, is the feature value of the ith position of the sth radar reflection signal semantic feature vector in the multiple radar reflection signal semantic feature vectors, A, B, C and D are adjustment hyperparameters, n i It is the feature value of the i-th position of the s-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors.
2. The multi-band radar fusion processing method according to claim 1, characterized in that: Performing multi-source information autocorrelation fusion on the multiple optimized radar reflection signal semantic feature vectors to obtain multi-source radar signal semantic fusion features, including: The multiple optimized radar reflection signal semantic feature vectors are passed through a multi-source information autocorrelation fuser based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector as the multi-source radar signal semantic fusion feature.
3. The multi-band radar fusion processing method according to claim 2, characterized in that: The multiple optimized radar reflection signal semantic feature vectors are passed through a multi-source information autocorrelation fuser based on an autocorrelation attention network to obtain a multi-source radar signal semantic fusion feature vector as the multi-source radar signal semantic fusion feature, including: The multiple optimized radar reflection signal semantic feature vectors are processed by the multi-source information autocorrelation fusion device based on the autocorrelation attention network using the following multi-source information fusion formula to obtain the multi-source radar signal semantic fusion feature vector; wherein the multi-source information fusion formula is: Among them, v i is the i-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors, and W i Represent the weight coefficient vector and weight coefficient matrix respectively, B i is the offset vector, tanh(·) represents the tanh function, e i is the attention score of the i-th optimized radar reflection signal semantic feature vector, exp(·) represents exponential operation, e k is the attention score of the kth optimized radar reflection signal semantic feature vector, t is the number of vectors of the multiple optimized radar reflection signal semantic feature vectors, a i is the weight value of the i-th optimized radar reflection signal semantic feature vector, and v1 is the multi-source radar signal semantic fusion feature vector.
4. The multi-band radar fusion processing method according to claim 3, characterized in that: Determining the type label of the target object based on the semantic fusion features of the multi-source radar signals includes: The semantic fusion feature vector of the multi-source radar signal is passed through a target identifier based on a classifier to obtain a recognition result, and the recognition result is used to represent a type label of the target object.
5. The multi-band radar fusion processing method according to claim 4, characterized in that: It also includes a training step: for training the signal feature extractor based on the convolutional neural network model, the nonlinear response optimizer based on the gamma correction module, the multi-source information autocorrelation fusion device based on the autocorrelation attention network, and the target identifier based on the classifier.
6. The multi-band radar fusion processing method according to claim 5, characterized in that: The training step comprises: Acquire training data, wherein the training data includes a plurality of training radar reflection signals returned from a target object, wherein the plurality of training radar reflection signals have different bands; Performing feature extraction on the multiple training radar reflection signals respectively by the signal feature extractor based on the convolutional neural network model to obtain multiple training radar reflection signal semantic feature vectors; Passing the plurality of training radar reflection signal semantic feature vectors through the nonlinear response optimizer based on the gamma correction module respectively to obtain a plurality of optimized training radar reflection signal semantic feature vectors; Passing the plurality of optimized training radar reflection signal semantic feature vectors through the multi-source information autocorrelation fusion device based on the autocorrelation attention network to obtain a training multi-source radar signal semantic fusion feature vector; Passing the training multi-source radar signal semantic fusion feature vector through the classifier-based target identifier to obtain a classification loss function value; and The signal feature extractor based on the convolutional neural network model, the nonlinear response optimizer based on the gamma correction module, the multi-source information autocorrelation fusion device based on the autocorrelation attention network, and the target identifier based on the classifier are trained based on the classification loss function value, wherein in each round of iteration of the training, the semantic fusion feature vector of the training multi-source radar signal is iteratively optimized.
7. A multi-band radar fusion processing system, characterized in that: include: A signal acquisition module, used to acquire a plurality of radar reflection signals returned from a target object, wherein the plurality of radar reflection signals have different bands; A semantic feature extraction module, used for performing feature extraction on the multiple radar reflection signals respectively through a signal feature extractor based on a deep neural network model to obtain multiple radar reflection signal semantic feature vectors; A nonlinear response optimization module, used for respectively passing the plurality of radar reflection signal semantic feature vectors through a nonlinear response optimizer based on a gamma correction module to obtain a plurality of optimized radar reflection signal semantic feature vectors; a multi-source information autocorrelation fusion module, configured to perform multi-source information autocorrelation fusion on the plurality of optimized radar reflection signal semantic feature vectors to obtain a multi-source radar signal semantic fusion feature; and A type analysis module, used to determine the type label of the target object based on the semantic fusion features of the multi-source radar signal; The deep neural network model is a convolutional neural network model; The nonlinear response optimization module is used to: The plurality of radar reflection signal semantic feature vectors are processed respectively by the nonlinear response optimizer based on the gamma correction module according to the following gamma correction formula to obtain the plurality of optimized radar reflection signal semantic feature vectors; Wherein, the gamma correction formula is: in, is the feature value of the ith position of the sth radar reflection signal semantic feature vector in the multiple radar reflection signal semantic feature vectors, A, B, C and D are adjustment hyperparameters, n i It is the feature value of the i-th position of the s-th optimized radar reflection signal semantic feature vector among the multiple optimized radar reflection signal semantic feature vectors.
