A method for processing security radar target data

Abstract

The present invention discloses a method for processing security radar target data, including: S1, preprocessing the radar echo signal, and converting the signal into frequency-domain data through denoising, amplitude normalization, and spectrum analysis; S2, using a sparse attention mechanism to extract significant region features, ignoring background data, and generating a sparse feature representation; S3, optimizing the feature representation through a self-supervised learning task to achieve the learning of deep features of radar signals; S4, using a two-stream architecture to dynamically weight-fuse the sparse feature stream and the global feature stream to generate a joint feature representation; S5, iteratively enhancing the significant region features through a recursive optimization method to generate optimized target features; S6, performing target classification and detection based on the optimized target features, and outputting the category and position information of the target. The present invention improves the efficiency and accuracy of radar target data processing and is applicable to the field of security monitoring technology.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method for processing security radar target data. Background Art

[0002] With the continuous development of security technologies, radars are increasingly widely used in the field of security monitoring. Radars can monitor the movement trajectories, speeds, and spatial positions of target objects in real time, providing important technical support for security systems. However, in the existing technologies, security radar systems still face many technical challenges in target data processing, and these problems are particularly prominent in terms of processing efficiency and detection accuracy.

[0003] In the existing technologies, traditional methods for processing radar target data mainly rely on rule-based signal analysis and fixed feature extraction mechanisms, and these methods are inadequate when dealing with complex security scenarios. Specifically, the existing technologies have obvious deficiencies in the following aspects:

[0004] 1. Low feature extraction efficiency: Traditional methods usually perform global processing on radar echo signals, making it difficult to effectively extract the significant features of the target area. This not only increases the computational complexity but also leads to the omission of key feature information.

[0005] 2. Insufficient ability to utilize unlabeled data: Existing methods usually rely on labeled data for training, and in security scenarios, the acquisition cost of labeled data is high and limited, resulting in limited deep feature representation ability of radar signals.

[0006] 3. Weak information fusion ability: Traditional methods are difficult to achieve efficient fusion of global and local information in target detection, and this single-feature-stream processing mode often leads to insufficient accuracy in target classification and detection.

[0007] 4. Limited feature enhancement ability: The existing technologies lack a recursive optimization mechanism and are difficult to focus on key features through multiple iterations, resulting in weak feature expression ability in significant regions and ultimately affecting the accuracy of target detection and classification.

[0008] 5. Insufficient real-time performance: In embedded security radar devices, the high computational complexity of existing algorithms leads to insufficient real-time processing ability and is difficult to meet the requirements of actual application scenarios.

[0009] Therefore, how to provide a method for processing security radar target data is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0010] An object of the present invention is to propose a method for processing target data of a security radar. The present invention adopts technologies such as sparse attention mechanism, self-supervised learning, dual-stream architecture, and recursive feature enhancement, and details the processing methods for extracting significant region features from radar signals, dynamically fusing global and local information, and recursively optimizing target features, which has the advantages of high computational efficiency, strong target detection accuracy, and efficient utilization of unlabeled data.

[0011] A method for processing target data of a security radar according to an embodiment of the present invention includes the following steps:

[0012] S1. Obtain the radar echo signal to be processed, and perform data preprocessing operations on the radar echo signal, including denoising, amplitude normalization, and spectrum analysis, to obtain the preprocessed radar frequency-domain data;

[0013] S2. Use the sparse attention mechanism to extract significant features from the preprocessed radar frequency-domain data, and generate a sparse feature representation by calculating attention weights;

[0014] S3. Construct a self-supervised learning task, and generate an optimized deep feature representation by predicting predefined signal perturbations;

[0015] S4. Use the dual-stream architecture to perform dynamic feature fusion on the optimized deep feature representation. Among them, the sparse feature stream extracts significant region features, the global feature stream extracts overall context information, and the sparse feature stream and the global feature stream are fused through a dynamic weight allocation strategy to generate a joint feature representation;

[0016] S5. Perform recursive feature enhancement processing on the joint feature representation, and gradually strengthen the feature representation of the target region by iteratively optimizing the expression ability of the significant region features multiple times to generate a further optimized target feature;

[0017] S6. Based on the further optimized target features, perform target classification and detection processing, determine the target category through the support vector machine classification algorithm, and output the spatial position information and category information of the target to complete the target detection and recognition task.

[0018] Optionally, the S2 specifically includes:

[0019] S21. Based on the preprocessed radar frequency-domain data, divide the radar frequency-domain data into multiple feature partitions, extract the frequency-domain features of each partition as input features , construct an attention weight matrix , and calculate the feature response value :

[0020] ;

[0021] Among them, represents the characteristic response value of the th partition in the radar frequency domain data, is the input feature of the th partition extracted from the preprocessed radar frequency domain data, is the attention weight matrix, used to weight the input features, is the bias term, is the non-linear activation function;

[0022] S22. According to the calculated characteristic response value , through the sparsification function to sparsify the attention weight matrix , only retain the partition features whose characteristic response value is greater than or equal to the preset sparsification threshold ;

[0023] S23. The sparsified attention weight matrix is called the significant region feature, and weighted summation is performed on the significant region feature to generate the sparse feature representation :

[0024] ;

[0025] Among them, is the input feature of the th partition, represents the total number of retained significant partitions, is the sparse feature representation, used to characterize the significant target information in the radar frequency domain data, is the sparsification function;

[0026] S24. Perform a normalization operation on the sparse feature representation .

[0027] Optionally, the S22 specifically includes:

[0028] S221. Perform statistical analysis on each characteristic response value of the attention weight matrix to calculate the mean and the standard deviation :

[0029] ;

[0030] ;

[0031] Among them, represents the total number of partitions of the radar frequency domain data;

[0032] S222. Dynamically determine the sparsification threshold based on the statistical analysis results :

[0033] ;

[0034] wherein, is a regulation factor used to control the sensitivity of the threshold;

[0035] S223. Apply the sparsification function to the characteristic response values of the attention weight matrix to calculate the sparsified attention weight matrix :

[0036] ;

[0037] wherein, represents the sparsified attention weight matrix.

[0038] Optionally, the specific steps of S3 include:

[0039] S31. Based on the sparse feature representation , construct a self-supervised learning task, design a signal perturbation prediction task, apply a predefined perturbation to the input sparse feature representation to generate a perturbed feature :

[0040] ;

[0041] wherein, is a random perturbation vector used to simulate the random noise existing in the radar signal;

[0042] S32. According to the sparse feature representation and the perturbed feature , construct a loss function based on the weighted Euclidean distance, which is used to quantify the difference between the perturbed feature and the sparse feature representation :

[0043] ;

[0044] wherein, represents the total dimension of the sparse feature representation, and respectively represent the original value and the perturbed value of the th feature, represents the Euclidean distance, is the feature weight;

[0045] By calculating the feature response values of each feature and the mean value of the deviation degree, the feature weight is determined

[0046] ;

[0047] Among them, represents the feature response value of the th feature, and are the mean value and standard deviation of the feature response value respectively, represents the absolute value operation, is the exponential function;

[0048] S33. Loss function based on weighted Euclidean distance , by dynamically adjusting the feature weight and optimizing the parameters of the self-supervised learning network, minimizing the difference between the perturbed feature and the sparse feature representation , generating an optimized deep feature representation .

[0049] Optionally, the S4 specifically includes:

[0050] S41. Based on the optimized deep feature representation , constructing a two-stream architecture, and respectively extracting a sparse feature stream and a global feature stream , where the sparse feature stream is extracted from the sparse feature representation , representing the feature representation of local significant targets, and the global feature stream extracts the overall context information from the optimized deep feature representation ;

[0051] S42. Based on the feature response value, calculate the dynamic weight allocation strategy, and respectively calculate the dynamic weights and for the sparse feature stream and ;

[0052] S43. Perform dynamic fusion processing on the sparse feature stream and the global feature stream to generate a joint feature representation :

[0053] ;

[0054] Among them, Denote the joint feature representation, which is used to simultaneously characterize the significant target features and the global context information;

[0055] S44. Regularize the joint feature representation to adjust the range of the feature distribution.

[0056] Optionally, S42 specifically includes:

[0057] S421. Calculate the feature response values of the sparse feature representation and the feature response values of the global feature stream :

[0058] ;

[0059] ;

[0060] Among them, represents the -th feature in the sparse feature representation, represents the -th feature in the global feature stream, and are the weights of the features respectively, and are the numbers of features of the sparse feature representation and the global feature stream respectively;

[0061] S422. Calculate the dynamic weights and of the sparse feature stream and the dynamic weights and of the global feature stream based on the feature response values

[0062] ;

[0063] ;

[0064] Among them, represents the exponential function;

[0065] S423. Normalize the dynamic weights of the sparse feature stream and the dynamic weights of the global feature stream to make .

[0066] Optionally, S5 specifically includes:

[0067] S51. Initialize the recursive feature representation based on the joint feature representation , and set the recursive iteration number threshold And recursively optimize the weight matrix and the bias vector ;

[0068] S52. In each iteration, use the recursive optimization function to perform significant region optimization on the recursive feature representation to generate an optimized recursive feature representation ;

[0069] S53. After each iteration, apply a non-linear activation function to the optimized recursive feature representation to generate an activated recursive feature representation :

[0070] ;

[0071] wherein, is a predefined activation function, represents the number of iterations;

[0072] S54. When the number of iterations reaches the recursive iteration number threshold , output the finally recursively optimized feature representation , as the further optimized target feature for the target classification and detection tasks.

[0073] Optionally, the S52 specifically includes:

[0074] S521. Initialize the recursive optimization function parameters using the recursive optimization weight matrix and the bias vector ;

[0075] S522. In the th iteration, apply the recursive optimization function to the recursive feature representation to generate an optimized recursive feature representation :

[0076] ;

[0077] wherein, represents the result of the previous round of recursive optimization;

[0078] S523. Calculate the significant region feature response value of the optimized recursive feature representation through feature weight assignment:

[0079] ;

[0080] wherein, represents the The weight of a feature represents the total dimension of the recursive feature;

[0081] S524. Return the significant region feature response value Return the recursive optimization function and continue the recursive iteration until the preset number of iterations is reached .

[0082] The beneficial effects of the present invention are as follows:

[0083] (1) By combining the sparse attention mechanism, self-supervised learning, the two-stream feature fusion architecture, and the recursive feature enhancement technology, the present invention realizes the in-depth understanding and dynamic optimization of radar target features, enables the system to efficiently extract significant region features, dynamically combines global and local information, and gradually strengthens the expression ability of target region features through recursive optimization, thereby effectively coping with complex security monitoring scenarios and significantly improving the accuracy of target detection and classification.

[0084] (2) By using the sparse attention mechanism, the present invention significantly reduces the computational complexity, only focuses on the key target region features, and avoids redundant processing of global calculations; at the same time, through the self-supervised learning technology, unlabeled data is efficiently utilized, the problem of insufficient labeled data is solved, and the generalization ability and robustness of the system to target features are improved; the two-stream feature fusion architecture further combines the information of the sparse feature stream and the global feature stream, enabling the system to effectively understand the relationship between targets and backgrounds in complex scenarios.

[0085] (3) Through recursive feature enhancement processing, the present invention realizes multiple iterative optimizations of significant region features, gradually enhances the expression ability of target features, and further improves the accuracy of target classification and detection. At the same time, based on the low computational complexity design, this method can meet the strict requirements of embedded security radar devices for real-time performance and computing resources, thereby improving the efficiency, accuracy, and adaptability of the radar target detection system, and is widely applicable to real-time target recognition tasks in the field of security monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0087] Figure 1 is a flowchart of a method for processing security radar target data proposed by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0088] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0089] Reference Figure 1 , a method for processing security radar target data, comprising the following steps:

[0090] S1. Obtain the radar echo signal to be processed, and perform data preprocessing operations on the radar echo signal, including denoising, amplitude normalization, and spectrum analysis, to obtain the preprocessed radar frequency domain data;

[0091] S2. Use a sparse attention mechanism to extract significant features from the preprocessed radar frequency domain data, and generate a sparse feature representation by calculating attention weights;

[0092] Optionally, the S2 specifically includes:

[0093] S21. Based on the preprocessed radar frequency domain data, divide the radar frequency domain data into multiple feature partitions, extract the frequency domain features of each partition as input features , construct an attention weight matrix , and calculate the feature response value :

[0094] ;

[0095] Wherein, represents the feature response value of the th partition in the radar frequency domain data, is the input feature of the th partition extracted from the preprocessed radar frequency domain data, is the attention weight matrix for weighting the input features, is the bias term, is the non-linear activation function;

[0096] S22. According to the calculated feature response value , use a sparsification function to sparsify the attention weight matrix , and only retain the partition features whose feature response values are greater than or equal to the preset sparsification threshold ;

[0097] S23. Call the sparsified attention weight matrix the significant region feature, and perform weighted summation processing on the significant region feature to generate a sparse feature representation :

[0098] ;

[0099] Wherein, is the input feature of the th partition, Indicates the total number of significant partitions retained is a sparse feature representation used to characterize significant target information in radar frequency-domain data is a sparsification function

[0100] S24. Perform a normalization operation on the sparse feature representation

[0101] In this embodiment, by sparsifying the attention weight matrix and combining the dynamic threshold adjustment method, only the target features in the significant regions are retained, significantly reducing the computational complexity. At the same time, the capture ability for key target regions is improved, providing a more efficient and accurate feature expression for subsequent processing and optimizing the performance of security radar target detection

[0102] Optionally, S22 specifically includes

[0103] S221. Perform statistical analysis on the feature response values of the attention weight matrix to calculate the mean and the standard deviation :

[0104]

[0105]

[0106] where represents the total number of partitions of the radar frequency-domain data

[0107] S222. Dynamically determine the sparsification threshold based on the statistical analysis results :

[0108]

[0109] where is an adjustment factor used to control the sensitivity of the threshold

[0110] S223. Apply the sparsification function to the feature response values of the attention weight matrix to calculate the sparsified attention weight matrix :

[0111]

[0112] where represents the sparsified attention weight matrix

[0113] ​​​​​​​This embodiment dynamically optimizes the attention weight matrix through sparsification processing, effectively extracts the features of the significant target area by using the adaptive calculation of the feature response value and the dynamic threshold, reduces the influence of non-critical features, and significantly improves the accuracy and computational efficiency of target data processing, providing a more efficient solution for security radar target detection.

[0114] S3. Construct a self-supervised learning task, and generate an optimized deep feature representation by predicting a predefined signal perturbation.

[0115] Optionally, the S3 specifically includes:

[0116] S31. Based on the sparse feature representation , construct a self-supervised learning task, design a signal perturbation prediction task, and apply a predefined perturbation to the input sparse feature representation to generate a perturbed feature :

[0117] ;

[0118] Among them, is a random perturbation vector, used to simulate the random noise existing in the radar signal;

[0119] S32. According to the sparse feature representation and the perturbed feature , construct a loss function based on the weighted Euclidean distance, used to quantify the difference between the perturbed feature and the sparse feature representation :

[0120] ;

[0121] Among them, represents the total dimension of the sparse feature representation, and respectively represent the original value and the perturbed value of the th feature, represents the Euclidean distance, is the feature weight;

[0122] Determine the feature weight by calculating the deviation degree of the feature response value of each feature from the mean value :

[0123] ;

[0124] Among them, represents the feature response value of the th feature, and are the mean and standard deviation of the feature response values respectively, represents the absolute value operation, is the exponential function;

[0125] S33. Loss function based on weighted Euclidean distance , by dynamically adjusting the feature weights and optimizing the self-supervised learning network parameters, minimize the difference between the perturbed features and the sparse feature representation to generate an optimized deep feature representation .

[0126] In this embodiment, by constructing a self-supervised learning task, using sparse feature representation for deep feature optimization, and combining weighted Euclidean distance and adaptive weight allocation, the deep-level significant features of radar signals can be effectively captured, improving the robustness and generalization ability of target detection, and providing more accurate and efficient technical support for the processing of security radar target data.

[0127] S4. Use a two-stream architecture to perform dynamic feature fusion on the optimized deep feature representation. Among them, the sparse feature stream extracts the features of the significant region, the global feature stream extracts the overall context information, and the sparse feature stream and the global feature stream are fused through a dynamic weight allocation strategy to generate a joint feature representation;

[0128] Optionally, the S4 specifically includes:

[0129] S41. Based on the optimized deep feature representation , construct a two-stream architecture, and extract the sparse feature stream and the global feature stream respectively, where the sparse feature stream is extracted from the sparse feature representation and represents the feature representation of local significant targets, and the global feature stream extracts the overall context information from the optimized deep feature representation ;

[0130] S42. Calculate the dynamic weight allocation strategy based on the feature response values, and calculate the dynamic weights and for the sparse feature stream and the global feature stream respectively;

[0131] S43. Perform dynamic fusion processing on the sparse feature stream and the global feature stream to generate a joint feature representation :

[0132] ;

[0133] Among them, represents a joint feature representation, which is used to simultaneously characterize the significant target features and the global context information;

[0134] S44. Regularize the joint feature representation to adjust the feature distribution range.

[0135] In this embodiment, through the dual-flow dynamic fusion of the sparse feature flow and the global feature flow, a joint feature representation is generated by using the dynamic weight allocation strategy, which not only retains the local features of the significant region target but also integrates the global context information, realizes the accurate expression and efficient cooperation of features, improves the accuracy and robustness of object detection, and provides high-quality input support for subsequent classification.

[0136] Optionally, the S42 specifically includes:

[0137] S421. Calculate the feature response value of the sparse feature representation and the feature response value of the global feature flow :

[0138] ;

[0139] ;

[0140] Among them, represents the -th feature in the sparse feature representation, represents the -th feature in the global feature flow, and are the weights of the features respectively, and are the numbers of features of the sparse feature representation and the global feature flow respectively;

[0141] S422. Calculate the dynamic weight of the sparse feature flow and the dynamic weight of the global feature flow based on the feature response values and :

[0142] ;

[0143] ;

[0144] Among them, represents the exponential function;

[0145] S423. Dynamic weights of the sparse feature stream and the dynamic weights of the global feature stream are normalized so that .

[0146] In this embodiment, through the calculation of the dynamic weight distribution of the sparse feature and the global feature, the collaborative optimization of the significant region and the global context information is achieved. This not only improves the accuracy of feature fusion but also ensures the balance and stability of feature expression, providing a better feature basis for subsequent object detection tasks.

[0147] S5. Perform recursive feature enhancement processing on the joint feature representation. By iteratively optimizing the expression ability of the significant region features multiple times, the feature representation of the target region is gradually strengthened to generate further optimized target features;

[0148] Optionally, S5 specifically includes:

[0149] S51. Based on the joint feature representation , initialize the recursive feature representation , set the recursive iteration count threshold and the recursive optimization weight matrix and the bias vector ;

[0150] S52. In each iteration, use the recursive optimization function to optimize the significant region of the recursive feature representation to generate the optimized recursive feature representation ;

[0151] S53. After each iteration, apply the non - linear activation function to the optimized recursive feature representation to generate the activated recursive feature representation :

[0152] ;

[0153] Among them, is a predefined activation function, represents the iteration count;

[0154] S54. When the iteration count reaches the recursive iteration count threshold , output the finally recursively optimized feature representation , as the further optimized target feature for object classification and detection tasks.

[0155] In this embodiment, through recursive feature enhancement processing, combined with salient region optimization and non-linear activation, the expression ability of target features is gradually strengthened, making the features of the salient target region clearer, supporting higher-precision target classification and detection tasks, and improving the detection efficiency and accuracy of the security radar system.

[0156] Optionally, the S52 specifically includes:

[0157] S521. Initialize the parameters of the recursive optimization function using the recursive optimization weight matrix and the bias vector ;

[0158] S522. In the -th iteration, apply the recursive optimization function to the recursive feature representation to generate the optimized recursive feature representation :

[0159] ;

[0160] where represents the result of the previous round of recursive optimization;

[0161] S523. Calculate the salient region feature response value of the optimized recursive feature representation through feature weight allocation:

[0162] ;

[0163] where represents the weight of the -th feature, represents the total dimension of the recursive features;

[0164] S524. Return the salient region feature response value to the recursive optimization function and continue the recursive iteration until the preset number of iterations is reached.

[0165] In this embodiment, through the dynamic adjustment of the recursive optimization weight matrix and the bias vector, combined with the adaptive weight allocation of the salient region feature response value, the gradual optimization of the recursive feature representation is realized, which can effectively enhance the expression of the salient features of the target region and improve the accuracy and robustness of target classification and detection in the security radar target data processing.

[0166] S6. Based on the further optimized target features, perform target classification and detection processing, determine the target category through the support vector machine classification algorithm, and output the spatial position information and category information of the target to complete the target detection and recognition task.

[0167] Optionally, S6 specifically includes:

[0168] S61. Normalize the target features based on the further optimized target features, and adjust the numerical range of the target features to a predefined standard range;

[0169] S62. Input the normalized target features into the support vector machine classification algorithm to construct a classification decision model, and determine the class label of the target by calculating the similarity between the support vector classification algorithm and the target features;

[0170] S63. Output the class label of the target according to the calculation result of the classification decision model to complete the target classification task;

[0171] S64. Combine the target detection method to predict the spatial position of the classified target, including determining the center coordinates and the boundary box range of the target;

[0172] S65. Integrate the class label of the target and the spatial position information into the detection result, and output the target classification and position prediction information.

[0173] Embodiment 1:

[0174] To verify the feasibility and superiority of the present invention, we apply the method for processing security radar target data of the present invention to the perimeter security system of an airport. Since the airport perimeter has a wide coverage area, complex environment, diverse and frequently changing target types, traditional radar monitoring systems often fail to meet the requirements of real-time and accuracy in this scenario. Therefore, the airport management department decides to introduce the method of the present invention to improve the radar target detection and classification capabilities.

[0175] In this scenario, the radar system needs to monitor illegal intrusion targets (such as personnel and vehicles) and low-altitude flying targets (such as drones) around the runway in real time. Due to the low data processing efficiency and weak feature extraction ability of traditional methods, the system is prone to false alarms and missed alarms, especially in rainy, foggy weather or in severe environmental interference conditions. To solve these problems, we deploy a radar target data processing system based on the present invention.

[0176] First, the system receives radar echo signals, performs denoising, amplitude normalization, and spectral analysis on them to generate radar frequency-domain data. Subsequently, through the sparse attention mechanism, only the significant region features in the signals are extracted to reduce the interference of background noise and generate sparse feature representations. Next, the system uses self-supervised learning methods to perform feature pre-training on unlabeled data, automatically learning the deep feature representations of radar signals and enhancing the system's target recognition ability in complex scenarios. The dual-stream feature fusion architecture dynamically combines the advantages of the sparse feature stream and the global feature stream, generating a more accurate joint feature representation through a dynamic weight allocation strategy. Finally, the recursive feature enhancement module iteratively optimizes the joint feature representation multiple times, gradually strengthening the target feature expression ability and ultimately completing the classification and detection of the target.

[0177] During the implementation of the system, we collected actual data in different scenarios of the airport perimeter, including monitoring data in sunny days, rainy days, nights, and high-interference environments. The experimental data was processed in real-time at the airport security control center, and the results are as follows:

[0178] Table 1 Comparison of Target Detection Performance of the Airport Perimeter Security System

[0179] As can be seen from Table 1, after deploying the system of the present invention, the number of target detections has increased. In particular, the detection accuracy for illegal targets has increased significantly, from 87.4% to 96.8%. The false negative rate has decreased from 5.7% to 1.3%, and the false positive rate has decreased from 4.5% to 1.1%. The system's ability to identify illegal targets has been significantly enhanced. At the same time, the detection response time has been greatly shortened, from 18 seconds in the traditional system to 6 seconds, effectively meeting the requirements of real-time security at the airport perimeter. In addition, in rainy, foggy weather and strong-interference environments, the system of the present invention can still maintain stable detection performance, while the traditional system shows a significant decline.

[0180] In a specific case, the system of the present invention detected an abnormal target that appeared at the edge of the runway at night. The target features were quickly extracted under the sparse attention mechanism, and the feature representation optimized by the recursive feature enhancement module was accurately classified as an illegal personnel intrusion. Finally, the system completed the detection and alarm within 6 seconds, and the relevant security personnel arrived at the scene in time, avoiding potential risks.

[0181] From the above experimental data and actual application cases, it can be seen that the method of the present invention not only solves the deficiencies of traditional radar target detection methods in terms of efficiency, accuracy, and environmental adaptability, but also significantly improves the overall performance of the security system, providing reliable technical support for high-requirement scenarios such as airports.

[0182] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A security radar target data processing method, characterized in that: The steps include: S1. Acquire a radar echo signal to be processed, and perform data preprocessing operations on the radar echo signal, including denoising, amplitude normalization, and spectrum analysis, to obtain preprocessed radar frequency domain data; S2, using the sparse attention mechanism to extract significant features from the preprocessed radar frequency domain data, and generating sparse feature representation by calculating the attention weights; S3, construct a self-supervised learning task to generate optimized deep feature representations by predicting predefined signal perturbations; S4. Use a dual-stream architecture to perform dynamic feature fusion on the optimized deep feature representation, where the sparse feature stream extracts salient regional features and the global feature stream extracts overall context information. The sparse feature stream and the global feature stream are fused through a dynamic weight allocation strategy to generate a joint feature representation. S5, recursively enhancing the joint feature representation, optimizing the expression ability of the salient region features through multiple iterations, so that the feature representation of the target region is gradually enhanced, and further optimized target features are generated; S6. Based on the further optimized target features, target classification and detection processing are performed, the target category is determined through the support vector machine classification algorithm, the spatial position information and category information of the target are output, and the target detection and recognition task is completed.

2. A security radar target data processing method according to claim 1, characterized in that: The S2 specifically includes: S21, based on the preprocessed radar frequency domain data, dividing the radar frequency domain data into a plurality of feature partitions, and extracting the frequency domain features of each partition as input features , construct the attention weight matrix , calculate the characteristic response value : ； in, Indicates the radar frequency domain data The characteristic response value of each partition, is the first The input features of the partitions, is the attention weight matrix, which is used to weight the input features. is the bias term, is a nonlinear activation function; S22, based on the calculated characteristic response value , through the sparse function For the attention weight matrix Perform sparse processing and only retain feature response values ​​greater than or equal to the preset sparse threshold The partition characteristics of S23. The sparse attention weight matrix It is called the salient region feature, and the salient region features are weighted and summed to generate a sparse feature representation : ； in, For the The input features of the partitions, Indicates the total number of significant partitions retained, It is a sparse feature representation used to characterize the significant target information in radar frequency domain data. is the sparse function; S24, representing the sparse features Perform standardized operations.

3. A security radar target data processing method according to claim 2, characterized in that: The S22 specifically includes: S221, attention weight matrix The characteristic response values ​​of Perform statistical analysis and calculate the mean and standard deviation : ； ； in, Represents the total number of partitions of radar frequency domain data; S222: Dynamically determine the sparseness threshold based on the statistical analysis results : ； in, is the adjustment factor used to control the sensitivity of the threshold; S223, attention weight matrix The characteristic response value of Apply the sparsification function , calculate the sparse attention weight matrix : ； in, Represents the sparse attention weight matrix.

4. A security radar target data processing method according to claim 3, characterized in that: The S3 specifically includes: S31, based on sparse feature representation , construct self-supervised learning tasks, design signal disturbance prediction tasks, and represent the sparse features of the input Apply predefined perturbations to generate perturbation signatures : ； in, is a random disturbance vector used to simulate the random noise in the radar signal; S32. Based on sparse feature representation and disturbance characteristics , construct a loss function based on weighted Euclidean distance , used to quantify the disturbance characteristics and sparse feature representation The difference between: ； in, represents the total dimension of the sparse feature representation, and Respectively represent The original and perturbed values ​​of the features, represents the Euclidean distance, is the feature weight; By calculating the characteristic response value of each feature With the mean The degree of deviation determines the feature weight : ； in, Indicates The characteristic response value of each feature, and are the mean and standard deviation of the characteristic response values, Represents absolute value operation, is an exponential function; S33. Loss function based on weighted Euclidean distance , by dynamically adjusting feature weights and optimize the parameters of the self-supervised learning network to minimize the perturbation characteristics and sparse feature representation The difference between the two generates an optimized deep feature representation .

5. A security radar target data processing method according to claim 4, characterized in that: The S4 specifically includes: S41, based on optimized deep feature representation , build a two-stream architecture and extract sparse feature streams respectively and global feature flow , where the sparse feature flow From sparse feature representation Extracted from, the feature representation representing the local salient target, the global feature flow From the optimized deep feature representation Extract overall context information from S42, based on the characteristic response value calculation dynamic weight allocation strategy, for sparse feature flow and global feature flow Calculate dynamic weights separately and ; S43. Sparse feature flow and global feature flow Perform dynamic fusion processing to generate joint feature representation : ； in, Represents joint feature representation, which is used to simultaneously represent salient target features and global context information; S44. Representation of joint features Regularization processing is performed to adjust the feature distribution range.

6. A security radar target data processing method according to claim 5, characterized in that: The S42 specifically includes: S421. Calculate sparse feature representation The characteristic response value of and global feature flow The characteristic response value of : ； ； in, Indicates the sparse feature representation Features, Indicates the first Features, and are the weights of the features, and are the number of features for sparse feature representation and global feature flow, respectively; S422, based on characteristic response value and Computing dynamic weights for sparse feature flows and the dynamic weights of the global feature flow : ； ； in, represents the exponential function; S423. Dynamic weights for sparse feature flows and the dynamic weights of the global feature flow Normalize it so that .

7. A security radar target data processing method according to claim 6, characterized in that: The S5 specifically includes: S51. Joint feature representation , initialize the recursive feature representation , set the recursive iteration number threshold and recursively optimize the weight matrix and the bias vector ; S52, in each iteration, using the recursive optimization function to represent the recursive feature Perform significant area optimization to generate optimized recursive feature representation ; S53. After each iteration, the optimized recursive feature is expressed as Apply non-linear activation function , generating the activated recursive feature representation : ； in, is a predefined activation function, Indicates the number of iterations; S54, when the number of iterations Reached the recursive iteration threshold When , the final recursively optimized feature representation is output , as the further optimized target feature, used for target classification and detection tasks.

8. The method for processing security radar target data according to claim 7, characterized in that: The S52 specifically includes: S521. Use recursive optimization weight matrix and the bias vector Initialize recursive optimization function parameters; S522, in In the iterations, the recursive feature representation Apply the recursive optimization function to generate an optimized recursive feature representation : ； in, Indicates the result of the previous round of recursive optimization; S523, calculating the optimized recursive feature representation by feature weight allocation The significant regional characteristic response value : ； in, Indicates The weight of the feature, Represents the total dimension of the recursive feature; S524, the significant area feature response value Return to the recursive optimization function and continue recursive iteration until the preset number of iterations is reached .

Images