A sea surface target detection method based on time-frequency feature enhancement and false alarm rate control

Abstract

This invention relates to the field of radar signal processing and target detection technology, specifically a sea surface target detection method based on time-frequency feature enhancement and false alarm rate control. The method includes: employing a fully adaptive noise ensemble empirical mode decomposition algorithm to perform multi-scale decomposition on the radar echo signal, extracting intrinsic mode components to reconstruct the signal, thereby highlighting target features; performing short-time Fourier transform on the reconstructed signal and the original signal after moving average filtering, mapping the time-frequency spectrum to an RGB image to enhance the representational ability of time-frequency features; designing a convolutional neural network with a false alarm rate control mechanism, using the fused time-frequency image as input to achieve accurate target detection and classification; effectively separating the target signal from sea clutter interference, significantly improving the identifiability of target features, and dynamically adjusting the recognition threshold.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing and target detection technology, specifically to a method for detecting sea surface targets based on time-frequency feature enhancement and false alarm rate control. Background Technology

[0002] Radar, as a primary means of target detection, tracking, and identification, plays a crucial role in maritime safety, shipping management, and military defense. With the continuous advancement of high-resolution radar technology, the accuracy of radar signal detection in complex sea conditions has improved, driving its application in maritime defense and security. Sea clutter, as a type of interference signal, contains a large number of spectral components, and its characteristics change significantly with external conditions such as wind, waves, and sea states, exhibiting non-Gaussian, nonlinear, non-stationary, and time-varying properties. The presence of sea clutter poses a significant challenge to target detection, often obscuring weak target signals. Traditional target detection algorithms primarily build models based on the amplitude statistical characteristics of sea clutter; however, in practical applications, these models cannot perfectly fit sea clutter, resulting in limited detection performance. In recent years, deep learning has been widely applied in radar target detection, capable of automatically learning high-level features from data. However, these models are susceptible to complex backgrounds, lack robustness to clutter interference, and struggle to adapt to the detection needs under different sea conditions. Summary of the Invention

[0003] The purpose of this invention is to provide a sea surface target detection method based on time-frequency feature enhancement and false alarm rate control, so as to solve the problems mentioned in the background art.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0005] A sea surface target detection method based on time-frequency feature enhancement and false alarm rate control, the method comprising:

[0006] Step 1: Obtain a radar dataset consisting of observation data samples and divide it using a sliding window.

[0007] Step 2: Decompose and reconstruct the radar signals of each sample in the radar dataset;

[0008] Step 3: Extract and map time-frequency features based on the reconstructed signal;

[0009] Step 4: Divide the radar dataset into training and testing sets according to time sequence;

[0010] The dataset is divided into training and test sets: Since the sliding window was used in step one to divide the short-term observation data samples, in order to avoid evaluation errors due to overlapping sample features, the training and test sets are divided in 8:2 according to the time order.

[0011] Step 5: Based on the training set, design a feature extraction network and a false alarm rate control module based on a convolutional neural network, and construct a convolutional neural network model.

[0012] Preferably, step one includes:

[0013] A radar dataset consisting of observation data samples was obtained. The observation time of the samples was selected to be 1.024s. The radar dataset was divided using a sliding window, and the sliding window width was set to... sliding step size The formula for generating the data samples is as follows: .

[0014] Preferably, step two includes:

[0015] S201. Obtain each sample in the radar dataset, perform CEEMDAN decomposition on each segmented sample, extract several intrinsic mode functions (IMFs), and confirm the different frequency components of the signal.

[0016] S202. Based on the IMF components obtained from the intrinsic mode functions, according to the formula: The cross-correlation coefficient between the IMF component and the radar echo signal was calculated. This allows for the determination of frequency band boundaries;

[0017] in, Represents the i-th IMF component and the original signal covariance, and They are respectively and The variance;

[0018] S203. Remove the highest frequency component and trend term: Treat all IMF components before the boundary line as high frequency components and the remaining components as low frequency IMF components.

[0019] S204. Enhance the target signal by reconstruction and suppress sea clutter interference: Smooth the noise signal by performing a moving average filter on the original signal, and generate a combination of the enhanced signal and the original signal;

[0020] The formula for the moving average filter is as follows:

[0021] ;

[0022] Where M represents the length of the sliding window.

[0023] Preferably, step three includes:

[0024] S301. Obtain the reconstructed low-frequency and high-frequency signals and the filtered signal, and perform a short-time Fourier transform on the reconstructed low-frequency and high-frequency signals and the filtered signal to convert the time series signal into a time-frequency domain representation.

[0025] S302. After normalizing the time-frequency spectrum values ​​of the low-frequency signal, high-frequency signal, and moving average filtered signal, map them into an RGB three-channel color time-frequency image.

[0026] Preferably, in step five, constructing the convolutional neural network model includes: a feature extraction network and a false alarm rate control module;

[0027] S501, The feature extraction network is used to extract target features from time-frequency images and output classification probabilities;

[0028] S502. The false alarm rate control module includes threshold calculation logic, which is used to dynamically adjust the detection threshold according to the target false alarm rate to ensure the detection performance of the model under different background complexities.

[0029] According to the dynamic control formula for false alarm rate: , Customize the classification thresholds for target signals and sea clutter signals to achieve dynamic control of the false alarm rate;

[0030] in, This represents the output result after sorting the sea clutter samples in ascending order by their first digit. This indicates the set false alarm rate. Indicates the number of sea clutter samples;

[0031] After processing by the softmax function, the output value is a two-dimensional vector; the first element represents the probability that the sample is classified as a sea clutter signal, and the second element represents the probability that the sample is classified as a target signal.

[0032] S503, Model Training and Inference: The cross-entropy loss function is used, the model optimizer is set to Adam, and the learning rate is... Inference is performed on sea clutter samples in the training set, and a feature extraction network is used to generate classification probability vectors.

[0033] ;

[0034] in This represents the probability that the i-th sample is classified as sea clutter. This represents the probability that the i-th sample is classified as the target signal. It is the number of sea clutter samples in the training set;

[0035] Then, according to the dynamic control formula for false alarm rate, the threshold T is obtained; for the test set samples, the classification probability is calculated, if... If, then it is determined to be the target signal; if If so, it is determined to be a sea clutter signal.

[0036] Preferably, step S501 in step five includes:

[0037] The feature extraction network includes a convolutional feature extraction module, a CBAM module, and a fully connected layer;

[0038] To fully extract information from different frequency bands corresponding to different channels, depthwise separable convolution is introduced. By convolving each channel individually, the independence of low-frequency, high-frequency, and local dynamic features is preserved, avoiding information interference between channels. Subsequently, cross-frequency band feature fusion is achieved through pointwise convolution. The CBAM module can adaptively adjust the importance of each channel and spatial region to enhance frequency band and target features.

[0039] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for detecting sea surface targets based on time-frequency feature enhancement and false alarm rate control.

[0040] A computer device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in the above-described method for detecting sea surface targets based on time-frequency feature enhancement and false alarm rate control.

[0041] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0042] This application provides a sea surface target detection method based on time-frequency feature enhancement and false alarm rate control. First, the CEEMDAN algorithm is used to decompose the radar signal into multiple scales, extracting intrinsic mode components to reconstruct the signal, effectively highlighting target features and reducing the impact of sea clutter interference on detection. Second, by performing short-time Fourier transform on the reconstructed and filtered signals, the time-spectrum image is mapped to RGB three channels, achieving multi-dimensional representation of target information and significantly improving feature discrimination capability. In the feature extraction process, a convolutional neural network model is used, employing a depthwise separable convolution module to efficiently extract features from different frequency bands, and combining it with a CBAM module to further enhance target region features, improving the network's target detection accuracy and robustness in complex backgrounds. For different false alarm rate requirements, this application designs a false alarm rate control module, dynamically sorting the classification probabilities of sea clutter samples and calculating the classification threshold to adjust the classification criteria between the target signal and the sea clutter signal, meeting the preset false alarm rate requirements. This application achieves small target detection in complex sea clutter backgrounds, overcoming the problems of low detection accuracy and poor adaptability of traditional methods, and is suitable for various complex application scenarios. Attached Figure Description

[0043] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0044] Figure 1 This is a flowchart of a sea surface target detection method based on time-frequency feature enhancement and false alarm rate control according to the present invention;

[0045] Figure 2 This is a structural diagram of the network model of this invention;

[0046] Figure 3 It is the solution result of the real and imaginary parts of a sample by CEEMDAN;

[0047] Figure 4 It includes the low-frequency, high-frequency, and moving average filtered STFT diagrams of the target signal, as well as the three-channel mapping diagram.

[0048] Figure 5 These are the STFT diagrams and three-channel mapping diagrams of the low-frequency, high-frequency, and moving average filtered clutter signals.

[0049] Figure 6 This is a graph showing the prediction accuracy of the present invention under different false alarm rates on the test set. Detailed Implementation

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] Please see Figures 1-6 The present invention provides the following technical solution:

[0052] Example 1:

[0053] A sea surface target detection method based on time-frequency feature enhancement and false alarm rate control, the method comprising:

[0054] Step 1: Obtain a radar dataset consisting of observation data samples and divide it using a sliding window.

[0055] Preferably, step one includes:

[0056] A radar dataset consisting of observation data samples was obtained. The observation time of the samples was selected to be 1.024s. The radar dataset was divided using a sliding window, and the sliding window width was set to... sliding step size The formula for generating the data samples is as follows: .

[0057] Step 2: Decompose and reconstruct the radar signals of each sample in the radar dataset;

[0058] Preferably, step two includes:

[0059] S201. Obtain each sample in the radar dataset, perform CEEMDAN decomposition on each segmented sample, extract several intrinsic mode functions (IMFs), and confirm the different frequency components of the signal.

[0060] S202. Based on the IMF components obtained from the intrinsic mode functions, according to the formula: The cross-correlation coefficient between the IMF component and the radar echo signal was calculated. This allows for the determination of frequency band boundaries;

[0061] in, Represents the i-th IMF component and the original signal covariance, and They are respectively and The variance;

[0062] S203. Remove the highest frequency component and trend term: Treat all IMF components before the boundary line as high frequency components and the remaining components as low frequency IMF components.

[0063] S204. Enhance the target signal by reconstruction and suppress sea clutter interference: Smooth the noise signal by performing a moving average filter on the original signal, and generate a combination of the enhanced signal and the original signal;

[0064] The formula for the moving average filter is as follows:

[0065] ;

[0066] Where M represents the length of the sliding window.

[0067] Step 3: Extract and map time-frequency features based on the reconstructed signal;

[0068] Preferably, step three includes:

[0069] S301. Obtain the reconstructed low-frequency and high-frequency signals, as well as the filtered signal, according to the formula: Short-time Fourier transform is performed on the reconstructed low-frequency and high-frequency signals and the filtered signal to convert the time series signal into a time-frequency domain representation;

[0070] Among them, sampling frequency Window length ;

[0071] S302. After normalizing the time-frequency spectrum values ​​of the low-frequency signal, high-frequency signal, and moving average filtered signal, map them into an RGB three-channel color time-frequency image.

[0072] Step 4: Divide the radar dataset into training and testing sets according to time sequence;

[0073] The dataset is divided into training and test sets: Since the sliding window was used in step one to divide the short-term observation data samples, in order to avoid evaluation errors due to overlapping sample features, the training and test sets are divided in 8:2 according to the time order.

[0074] Step 5: Based on the training set, design a feature extraction network and a false alarm rate control module based on a convolutional neural network, and construct a convolutional neural network model;

[0075] Preferably, in step five, constructing the convolutional neural network model includes: a feature extraction network and a false alarm rate control module;

[0076] S501, The feature extraction network is used to extract target features from time-frequency images and output classification probabilities;

[0077] Preferably, step S501 in step five includes:

[0078] The feature extraction network includes a convolutional feature extraction module, a CBAM module, and a fully connected layer;

[0079] To fully extract information from different frequency bands corresponding to different channels, depthwise separable convolution is introduced. By convolving each channel individually, the independence of low-frequency, high-frequency, and local dynamic features is preserved, avoiding information interference between channels. Subsequently, cross-frequency band feature fusion is achieved through pointwise convolution. The CBAM module can adaptively adjust the importance of each channel and spatial region to enhance frequency band and target features.

[0080] S502. The false alarm rate control module includes threshold calculation logic, which is used to dynamically adjust the detection threshold according to the target false alarm rate to ensure the detection performance of the model under different background complexities.

[0081] According to the dynamic control formula for false alarm rate: , Customize the classification thresholds for target signals and sea clutter signals to achieve dynamic control of the false alarm rate;

[0082] in, This represents the output result after sorting the sea clutter samples in ascending order by their first digit. This indicates the set false alarm rate. Indicates the number of sea clutter samples;

[0083] After processing by the softmax function, the output value is a two-dimensional vector; the first element represents the probability that the sample is classified as a sea clutter signal, and the second element represents the probability that the sample is classified as a target signal.

[0084] S503, Model Training and Inference: The cross-entropy loss function is used, the model optimizer is set to Adam, and the learning rate is... Inference is performed on sea clutter samples in the training set, and a feature extraction network is used to generate classification probability vectors.

[0085] ;

[0086] in This represents the probability that the i-th sample is classified as sea clutter. This represents the probability that the i-th sample is classified as the target signal. It is the number of sea clutter samples in the training set;

[0087] Then, according to the dynamic control formula for false alarm rate, the threshold T is obtained; for the test set samples, the classification probability is calculated, if... If, then it is determined to be the target signal; if If so, it is determined to be a sea clutter signal.

[0088] Based on this, the radar dataset was divided into short-time observation data samples, with the observation time set to 1.024 seconds and the sliding step size set to 32 sampling points, generating a total of 44,715 samples. CEEMDAN decomposition was performed on the real and imaginary parts of each sample to obtain multiple intrinsic mode components (IMFs). The correlation coefficient between each IMF component and the original signal was calculated to determine the frequency band boundary; in this invention, the boundary is 6. The highest frequency component and trend term were removed, and the remaining components were reconstructed, classifying low-frequency and high-frequency components separately. A moving average filter was used to smooth the original signal, and combined with the reconstructed signal, an enhanced signal was generated. Short-time Fourier transforms were performed on the enhanced signal, with a sampling frequency of 500 and a window length of 64, yielding a time-frequency domain representation. The time-frequency plot size was [missing value]. 4. The STFT results of low-frequency, high-frequency, and smoothed signals are mapped to RGB channels to synthesize a color time-frequency image to enhance the representation ability of target features. The dataset is divided into an 8:2 ratio, with the first 80% of samples used as the training set and the last 20% as the test set. The training set contains 3252 target samples and 32520 clutter samples, while the test set contains 813 target samples and 8130 clutter samples.

[0089] Following step five, a convolutional neural network model with a controllable false alarm rate is built and trained. The network model optimizer is set to Adam, with an initial learning rate of 1e-4 and 50 training epochs. The detection threshold T is dynamically adjusted according to the target false alarm rate requirement, and the classification probability P is calculated for the test set samples. If it is, it is determined to be the target signal; otherwise, it is determined to be the sea clutter signal.

[0090] After 50 rounds of training, the accuracy on the test set was 0.9987 when the false alarm rate was 0.1; 0.9876 when the false alarm rate was 0.01; and 0.9089 when the false alarm rate was 0.001.

[0091] Example 2:

[0092] The computer-readable storage medium of this embodiment stores a computer program that, when executed by a processor, implements the steps of a sea surface target detection method based on time-frequency feature enhancement and false alarm rate control in Embodiment 1.

[0093] The computer-readable storage medium in this embodiment can be an internal storage unit of the terminal, such as the terminal's hard disk or memory; the computer-readable storage medium in this embodiment can also be an external storage device of the terminal, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc. equipped on the terminal; furthermore, the computer-readable storage medium can include both the terminal's internal storage unit and external storage devices.

[0094] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0095] Example 3:

[0096] The computer device of this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the sea surface target detection method based on time-frequency feature enhancement and false alarm rate control in Embodiment 1.

[0097] In this embodiment, the processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The memory can include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0098] Those skilled in the art will understand that the content disclosed in the embodiments can be provided as a method, system, or computer program product. Therefore, this solution can take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this solution can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage) containing computer-usable program code.

[0099] This solution is described with reference to flowchart illustrations and / or schematic diagrams of methods and computer program products according to embodiments of this solution. It should be understood that each block of the flowchart illustrations and / or schematic diagrams, and combinations of blocks of the flowchart illustrations and / or schematic diagrams, can be implemented by computer program instructions; these computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the flowchart illustrations and / or block combinations. Figure 1 One or more processes and / or methods are illustrated. Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or methods are illustrated. Figure 1 The function specified in one or more boxes.

[0101] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or methods are illustrated. Figure 1 The steps of the function specified in one or more boxes.

[0102] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0103] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A sea surface target detection method based on time-frequency feature enhancement and false alarm rate control, characterized in that: The method includes: Step 1: Obtain a radar dataset consisting of observation data samples and divide it using a sliding window. Step 2: Decompose and reconstruct the radar signals of each sample in the radar dataset; Step 3: Extract and map time-frequency features based on the reconstructed signal; Step 4: Divide the radar dataset into training and testing sets according to time sequence; Step 5: Based on the training set, design a feature extraction network and a false alarm rate control module based on a convolutional neural network, and construct a convolutional neural network model; Step two includes: S201. Obtain each sample in the radar dataset, perform CEEMDAN decomposition on each segmented sample, extract several intrinsic mode functions (IMFs), and confirm the different frequency components of the signal. S202. Based on the IMF components obtained from the intrinsic mode functions, according to the formula: The cross-correlation coefficient between the IMF component and the radar echo signal was calculated. This allows for the determination of frequency band boundaries; in, Represents the i-th IMF component and the original signal covariance, and They are respectively and The variance; S203. Remove the highest frequency component and trend term: Treat all IMF components before the boundary line as high frequency components and the remaining components as low frequency IMF components. S204. Enhance the target signal by reconstruction and suppress sea clutter interference: Smooth the noise signal by performing a moving average filter on the original signal, and generate a combination of the enhanced signal and the original signal; The formula for the moving average filter is as follows: ； Where M represents the length of the sliding window; Step three includes: S301. Obtain the reconstructed low-frequency and high-frequency signals and the filtered signal, and perform a short-time Fourier transform on the reconstructed low-frequency and high-frequency signals and the filtered signal to convert the time series signal into a time-frequency domain representation. S302. After normalizing the time-frequency spectrum values ​​of the low-frequency signal, high-frequency signal, and moving average filtered signal, map them into an RGB three-channel color time-frequency image.

2. The sea surface target detection method based on time-frequency feature enhancement and false alarm rate control as described in claim 1, characterized in that, Step one includes: A radar dataset consisting of observation data samples was obtained. The observation time of the samples was selected to be 1.024s. The radar dataset was divided using a sliding window, and the sliding window width was set to... sliding step size The formula for generating the data samples is as follows: .

3. The sea surface target detection method based on time-frequency feature enhancement and false alarm rate control as described in claim 1, characterized in that, In step five, the construction of the convolutional neural network model includes: a feature extraction network and a false alarm rate control module; S501, The feature extraction network is used to extract target features from time-frequency images and output classification probabilities; S502. The false alarm rate control module includes threshold calculation logic, which is used to dynamically adjust the detection threshold according to the target false alarm rate to ensure the detection performance of the model under different background complexities. According to the dynamic control formula for false alarm rate: , Customize the classification thresholds for target signals and sea clutter signals to achieve dynamic control of the false alarm rate; in, This represents the output result after sorting the sea clutter samples in ascending order by their first digit. This indicates the set false alarm rate. Indicates the number of sea clutter samples; After processing by the softmax function, the output value is a two-dimensional vector; the first element represents the probability that the sample is classified as a sea clutter signal, and the second element represents the probability that the sample is classified as a target signal. S503, Model Training and Inference: The cross-entropy loss function is used, the model optimizer is set to Adam, and the learning rate is... Inference is performed on sea clutter samples in the training set, and a feature extraction network is used to generate classification probability vectors. ； in This represents the probability that the i-th sample is classified as sea clutter. This represents the probability that the i-th sample is classified as the target signal. It is the number of sea clutter samples in the training set; Then, according to the dynamic control formula for false alarm rate, the threshold T is obtained; for the test set samples, the classification probability is calculated, if... If, then it is determined to be the target signal; if If so, it is determined to be a sea clutter signal.

4. The sea surface target detection method based on time-frequency feature enhancement and false alarm rate control as described in claim 3, characterized in that, Step S501 in step five includes: The feature extraction network includes a convolutional feature extraction module, a CBAM module, and a fully connected layer; To fully extract information from different frequency bands corresponding to different channels, depthwise separable convolution is introduced. By convolving each channel individually, the independence of low-frequency, high-frequency, and local dynamic features is preserved, avoiding information interference between channels. Subsequently, cross-frequency band feature fusion is achieved through pointwise convolution. The CBAM module adaptively adjusts the importance of each channel and spatial region to enhance frequency band and target features.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the steps in the sea surface target detection method based on time-frequency feature enhancement and false alarm rate control as described in any one of claims 1-4.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps in the sea surface target detection method based on time-frequency feature enhancement and false alarm rate control as described in any one of claims 1-4.

Images