Radar target template signal establishment method based on depth model adaptive segmentation

Abstract

The invention discloses a radar target template signal establishment method based on depth model adaptive segmentation, mainly relates to the technical field of template signals, and is used for solving the problems that a target contour is easy to fracture or false detection by clutters is easy to cause when a time-frequency map is subjected to binary segmentation by a simple threshold slice in the prior art. And the threshold parameter is very sensitive to the signal-to-clutter ratio and the environmental change. Comprising the following steps: reading a one-dimensional radar echo sequence, and mapping the one-dimensional radar echo sequence to a two-dimensional time-frequency domain to obtain a time-frequency image; multi-scale features are obtained, and a matrix is prompted; obtaining a binary mask corresponding to the mask feature through the multi-scale feature and the prompt matrix; determining the binary mask corresponding to the highest confidence score as a final mask; screening time-frequency transformation data corresponding to the radar echo sequence by using the final mask to obtain output data; and recovering the output data into a time domain signal by using inverse short-time Fourier transform, and taking the time domain signal as a target template signal.

Description

Technical Field

[0001] This application relates to the field of template signal technology, and in particular to a method for establishing radar target template signals based on depth model adaptive segmentation. Background Technology

[0002] When performing radar target detection and parameter estimation in sea clutter or ground clutter environments, constructing a pure target reference template signal is an important task. This template needs to fully preserve the Doppler variations and amplitude fluctuations of the target, while eliminating clutter, noise, and interference as much as possible, thereby improving the separability of the target from the background.

[0003] With the development of digital signal processing technology, time-frequency analysis tools such as Short Time Fourier Transform (STFT) and Continuous Wavelet Transform (CWT) have been widely used in radar preprocessing. However, relying solely on simple threshold slicing for binary segmentation of the time-frequency map can easily lead to broken target contours or false detection by clutter. Furthermore, the threshold parameter is highly sensitive to changes in the signal-to-clutter ratio (SCR) and the environment, making it difficult to guarantee the stability and reliability of the template. Summary of the Invention

[0004] This application provides a radar target template signal establishment method based on depth model adaptive segmentation to solve the problems of existing schemes that use simple threshold slicing to perform binary segmentation of time-frequency maps, which can easily lead to target contour breakage or false detection by clutter. Furthermore, the threshold parameters are very sensitive to signal-to-clutter ratio and environmental changes, making it difficult to guarantee the stability and reliability of the template.

[0005] Firstly, this application provides a method for establishing radar target template signals based on depth model adaptive segmentation, the method comprising: Read the one-dimensional radar echo sequence, map it to the two-dimensional time-frequency domain to obtain a complex matrix; after time-frequency transformation and pseudo-color mapping of the complex matrix, obtain the time-frequency image; Load the pre-trained weights of SAM2, call the image encoder to perform multi-scale processing on the input time-frequency image, and obtain multi-scale features; The cue matrix is ​​obtained by encoding the cue points injected from the time-frequency image; By using multi-scale features and cue matrices, several mask features are obtained, and then the binary mask corresponding to the mask features is obtained. Calculate the confidence scores corresponding to the mask features, and determine the binary mask corresponding to the highest confidence score as the final mask; Using the final mask, the time-frequency transform data corresponding to the radar echo sequence is filtered to obtain the filtered time spectrum. The time spectrum is restored to the original frequency arrangement order through frequency inverse rearrangement to obtain the output data. The output data is restored to the time domain signal using inverse short-time Fourier transform, and the time domain signal is used as the target template signal.

[0006] In one implementation of this application, a one-dimensional radar echo sequence is read, mapped to a two-dimensional time-frequency domain, and a complex matrix is ​​obtained; the complex matrix is ​​then subjected to time-frequency transformation and pseudo-color mapping to obtain a time-frequency image, specifically including: Reading one-dimensional radar echo sequences Wherein, the sampling rate of the radar echo sequence is equal to the pulse repetition frequency, denoted as . ; Using preset formulas: By mapping the radar echo sequence to the two-dimensional time-frequency domain, a complex matrix is ​​obtained. ; in, For length is Hamming window function, and The first The first time window center and the first Each frequency point, The total length of the signal is denoted by m; the value of m ranges from [1, M], where M represents the total number of preset frequency points. Through the formula , obtain along The frequency axis performs a centered rearrangement of the time-frequency transformation data; in, This indicates a frequency rearrangement operation; Through the formula Take the complex spectrum Logarithmic compression is performed on the amplitude to obtain compressed data; Where ε is a preset small constant to prevent underflow caused by a logarithm of zero; Will Linear normalization to the (0,1) interval; Using Jet pseudocolor function : , The radar echo sequence is mapped to a three-channel color space to obtain a time-frequency image; Where ⌊⋅⌋ represents rounding down.

[0007] In one implementation of this application, pre-trained weights of SAM2 are loaded, and an image encoder is invoked to perform multi-scale processing on the input time-frequency image to obtain multi-scale features, specifically including: Set the height and width of the time-frequency image as follows: and The image encoder outputs a new feature map after each downsampling, ultimately obtaining multi-scale features. ; in, , For scale indexing, For the first Layer feature tensor, The total number of feature layers, For the first The number of channels in the layer.

[0008] In one implementation of this application, a cue matrix is ​​obtained based on the cue point encoding injected from the time-frequency image, specifically including: Mark P groups of positive and negative cue points on the time-frequency image to form a set. ; in, , It is the first The pixel coordinates of the prompt point These are preset labels, where 1 represents the foreground point and 0 represents the background point; Through the formula: Calculate to obtain the query vector ;in, The preset mapping length; All Vertical concatenation yields the hint matrix. .

[0009] In one implementation of this application, several mask features are obtained through multi-scale features and a cue matrix, and then a binary mask corresponding to the mask features is obtained, specifically including: The mask decoder extracts image features With hint matrix Generate through cross-attention fusion Group preset low-resolution mask features ,index ; The original time-frequency image size is sampled from each set of mask features and then processed by the Sigmoid function. Activation yields a pixel-level probability mask. ; right Generate a binary mask with a threshold of 0.5 .

[0010] In one implementation of this application, the confidence score corresponding to the mask feature is calculated, and the binary mask corresponding to the highest confidence score is determined as the final mask, specifically including: Through the formula: Calculate the confidence score for each mask feature. ; in, , and It refers to the linear layer weights and biases. Indicates global average pooling; from The binary mask corresponding to the mask feature with the highest confidence score among the mask features is selected as the final mask.

[0011] In one implementation of this application, the time-frequency transform data corresponding to the radar echo sequence is filtered using a final mask to obtain the filtered time spectrum. The time spectrum is then restored to its original frequency arrangement order via frequency inverse rearrangement to obtain output data. The output data is then restored to a time-domain signal using an inverse short-time Fourier transform, and this time-domain signal is used as the target template signal. Specifically, this includes: Through the formula: Obtain the filtered time spectrum; in, This represents the time-frequency transform data after centering and rearranging along the frequency axis of the complex matrix. Through the formula: Obtain the output data; in, This indicates inverse centralization of the frequency dimension; The output data is restored to a time-domain signal using the inverse short-time Fourier transform, and the target template signal is finally obtained. .

[0012] As can be seen from the above technical solutions, this application has the following advantages: This application achieves multi-scale feature processing of time-frequency images by loading an image encoder with SAM2 pre-trained weights. Compared to the contour breakage problem caused by traditional threshold slicing methods, this application can adaptively capture multi-level information from local micro-Doppler features to global motion trajectories. The multi-scale feature fusion mechanism effectively overcomes the limitations of single-scale analysis for complex moving targets, especially for dynamic targets with harmonic characteristics such as helicopter rotors and vehicle vibrations, maintaining their physical continuity. This deep learning-based feature extraction method fundamentally avoids the sensitivity problem caused by manually setting thresholds.

[0013] The cue point encoding mechanism introduced in this application transforms the operator's prior knowledge of the target area into a mathematical cue matrix, which, in conjunction with multi-scale features, generates a binary mask. This design offers two advantages: first, it guides attention through human-computer interaction, significantly reducing the false detection rate in cluttered areas; second, the confidence scoring mechanism automatically selects the optimal mask, ensuring that the selected time-spectrum has the highest signal-to-clutter ratio. Compared to the sensitivity of traditional schemes to environmental changes, this data-driven approach combined with expert knowledge enables the template construction to maintain stable output even in strong cluttered backgrounds.

[0014] The frequency inverse rearrangement technique employed in the final output stage resolves the physical information distortion problem caused by frequency axis rearrangement during time-frequency transformation. By rigorously restoring the original Doppler frequency arrangement order, coupled with inverse short-time Fourier transform, it ensures that the generated time-domain template signal retains both the intrinsic motion characteristics of the target (such as the micro-Doppler modulation period) and eliminates non-coherent noise. This closed-loop processing flow enables the template signal to accurately reflect the kinematic characteristics of the target, providing a physically interpretable input source for subsequent identification and classification. Attached Figure Description

[0015] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a radar target template signal establishment method based on depth model adaptive segmentation provided in an embodiment of this application.

[0017] Figure 2 This is a raw signal time-frequency image provided in an embodiment of this application.

[0018] Figure 3 This is an interactive point annotation map provided in an embodiment of this application.

[0019] Figure 4 This is an automatic segmentation mask provided in an embodiment of this application.

[0020] Figure 5 This is a target energy band retention diagram provided in an embodiment of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.

[0023] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0024] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0025] The embodiment provides a method for establishing radar target template signals based on depth model adaptive segmentation, such as Figure 1 As shown in the embodiments of this application, the method mainly includes the following steps: Step 110: Read the one-dimensional radar echo sequence, map it to the two-dimensional time-frequency domain to obtain a complex matrix; after time-frequency transformation and pseudo-color mapping of the complex matrix, obtain the time-frequency image.

[0026] In some embodiments, this step may specifically be as follows: Reading one-dimensional radar echo sequences Wherein, the sampling rate of the radar echo sequence is equal to the pulse repetition frequency, denoted as . ; Using preset formulas: By mapping the radar echo sequence to the two-dimensional time-frequency domain, a complex matrix is ​​obtained. ; in, For length is Hamming window function, and The first The first time window center and the first Each frequency point, The total length of the signal is denoted by m; the value of m ranges from [1, M], where M represents the total number of preset frequency points. Through the formula , obtain along The frequency axis performs a centered rearrangement of the time-frequency transformation data; in, This indicates a frequency rearrangement operation; Through the formula Take the complex spectrum Logarithmic compression is performed on the amplitude to obtain compressed data; Where ε is a preset small constant to prevent underflow caused by a logarithm of zero; Will Linear normalization to the (0,1) interval; Using Jet pseudocolor function : , The radar echo sequence is mapped to a three-channel color space to obtain a time-frequency image (such as...). Figure 2 (as shown) Where ⌊⋅⌋ represents floor function, and (k,m) represents the image pixel coordinates, corresponding to the discrete index of the signal. .

[0027] Step 120: Load the pre-trained weights of SAM2, call the image encoder to perform multi-scale processing on the input time-frequency image, and obtain multi-scale features.

[0028] Load the pre-trained weights of SAM2, and call the image encoder to perform multi-scale processing on the input time-frequency image to obtain multi-scale features, specifically including: Set the height and width of the time-frequency image as follows: and The image encoder outputs a new feature map after each downsampling, ultimately obtaining multi-scale features. ; in, , For scale indexing, For the first Layer feature tensor, The total number of feature layers, For the first The number of channels in the layer.

[0029] It should be added that all layers together provide multi-layered information from fine textures to large structures, laying the foundation for subsequent mask generation.

[0030] Step 130: Obtain the cue matrix based on the cue point encoding injected from the time-frequency image.

[0031] Based on the cue point encoding injected from the time-frequency image (e.g.) Figure 3 As shown), the hint matrix is ​​obtained, specifically including: Mark P groups of positive and negative cue points on the time-frequency image to form a set. ; in, , It is the first The pixel coordinates of the prompt point These are preset labels, where 1 represents the foreground point and 0 represents the background point; Through the formula: Calculate to obtain the query vector ;in, The preset mapping length (here, it is normalized to align the coordinates with the network input scale) is used to ensure the coordinates are aligned with the network input scale. Then through two layers of sensors Mapped to a length of query vector ); All Vertical concatenation yields the hint matrix. .

[0032] Step 140: Obtain several mask features through multi-scale features and cue matrix, and then obtain the binary mask corresponding to the mask features.

[0033] In some embodiments, several mask features are obtained through multi-scale features and a cue matrix, and then a binary mask corresponding to the mask features is obtained, specifically including: The mask decoder extracts image features With hint matrix Generate through cross-attention fusion Group preset low-resolution mask features ,index ; The original time-frequency image size is sampled from each set of mask features and then processed by the Sigmoid function. Activation yields a pixel-level probability mask. ; in, ; right Generate a binary mask with a threshold of 0.5 ( ).

[0034] Step 150: Calculate the confidence score corresponding to the mask features, and determine the binary mask corresponding to the highest confidence score as the final mask.

[0035] Calculate the confidence scores corresponding to the mask features, and determine the binary mask corresponding to the highest confidence score as the final mask. Specifically, this includes: Through the formula: Calculate the confidence score for each mask feature. It should be noted that, The larger the value, the higher the probability that the mask matches the real foreground.

[0036] in, , and It refers to the linear layer weights and biases. Indicates global average pooling; from The binary mask corresponding to the mask feature with the highest confidence score among the mask features is selected as the final mask. The automatic segmentation mask image corresponding to the final mask (final veneer) is shown below. Figure 4 As shown.

[0037] Step 160: Using the final mask, filter the time-frequency transform data corresponding to the radar echo sequence to obtain the filtered time spectrum. The time spectrum is restored to the original frequency arrangement order through frequency inverse rearrangement to obtain the output data. The output data is restored to the time domain signal using the inverse short-time Fourier transform, and the time domain signal is used as the target template signal.

[0038] This step specifically includes: Through the formula: Obtain the filtered time spectrum; in, This represents the time-frequency transform data after centering and rearranging along the frequency axis of the complex matrix. Through the formula: Obtain the output data; in, This indicates inverse centralization of the frequency dimension; The output data is restored to a time-domain signal using the inverse short-time Fourier transform, and the target template signal is finally obtained. The target energy band retention diagram corresponding to the target template signal is shown below. Figure 5 As shown.

[0039] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for establishing radar target template signals based on depth model adaptive segmentation, characterized in that, The method includes: Read the one-dimensional radar echo sequence, map it to the two-dimensional time-frequency domain to obtain a complex matrix; after time-frequency transformation and pseudo-color mapping of the complex matrix, obtain the time-frequency image; Load the pre-trained weights of SAM2, call the image encoder to perform multi-scale processing on the input time-frequency image, and obtain multi-scale features; The cue matrix is ​​obtained by encoding the cue points injected from the time-frequency image; By using multi-scale features and cue matrices, several mask features are obtained, and then the binary mask corresponding to the mask features is obtained. Calculate the confidence scores corresponding to the mask features, and determine the binary mask corresponding to the highest confidence score as the final mask; Using the final mask, the time-frequency transform data corresponding to the radar echo sequence is filtered to obtain the filtered time spectrum. The time spectrum is restored to the original frequency arrangement order through frequency inverse rearrangement to obtain the output data. The output data is restored to the time domain signal using inverse short-time Fourier transform, and the time domain signal is used as the target template signal.

2. The radar target template signal establishment method based on depth model adaptive segmentation according to claim 1, characterized in that, Read the one-dimensional radar echo sequence and map it to the two-dimensional time-frequency domain to obtain a complex matrix; The time-frequency image is obtained by performing a time-frequency transform and pseudo-color mapping on a complex matrix, specifically including: Reading one-dimensional radar echo sequences Wherein, the sampling rate of the radar echo sequence is equal to the pulse repetition frequency, denoted as . ; Using preset formulas: By mapping the radar echo sequence to the two-dimensional time-frequency domain, a complex matrix is ​​obtained. ; in, For length is Hamming window function, and Respectively The first time window center and the first Each frequency point, The total length of the signal is denoted by m; the value of m ranges from [1, M], where M represents the total number of preset frequency points. Through the formula , obtain along The frequency axis performs a centered rearrangement of the time-frequency transformation data; in, This indicates a frequency rearrangement operation; Through the formula Take the complex spectrum Logarithmic compression is performed on the amplitude to obtain compressed data; Where ε is a preset small constant to prevent underflow caused by a logarithm of zero; Will Linear normalization to the (0,1) interval; Using Jet pseudocolor function : ， The radar echo sequence is mapped to a three-channel color space to obtain a time-frequency image; Where ⌊⋅⌋ represents rounding down.

3. The radar target template signal establishment method based on depth model adaptive segmentation according to claim 1, characterized in that, Load the pre-trained weights of SAM2, and call the image encoder to perform multi-scale processing on the input time-frequency image to obtain multi-scale features, specifically including: Set the height and width of the time-frequency image as follows: and The image encoder outputs a new feature map after each downsampling, ultimately obtaining multi-scale features. ; in, , For scale indexing, For the first Layer feature tensor, The total number of feature layers, For the first The number of channels in the layer.

4. The radar target template signal establishment method based on depth model adaptive segmentation according to claim 1, characterized in that, Based on the coding of cue points injected from the time-frequency image, a cue matrix is ​​obtained, specifically including: Mark P groups of positive and negative cue points on the time-frequency image to form a set. ; in, , It is The pixel coordinates of the prompt point These are preset labels, where 1 represents the foreground point and 0 represents the background point; Through the formula: Calculate to obtain the query vector ;in, The preset mapping length; All Vertical concatenation yields the hint matrix. .

5. The radar target template signal establishment method based on depth model adaptive segmentation according to claim 1, characterized in that, By using multi-scale features and a cue matrix, several masking features are obtained, and then the corresponding binary mask is obtained, specifically including: The mask decoder will extract image features With hint matrix Generate through cross-attention fusion Group preset low-resolution mask features ,index ; The original time-frequency image size is sampled from each set of mask features and then processed by the Sigmoid function. Activation yields a pixel-level probability mask. ; right Generate a binary mask with a threshold of 0.5 .

6. The radar target template signal establishment method based on depth model adaptive segmentation according to claim 1, characterized in that, Calculate the confidence scores corresponding to the mask features, and determine the binary mask corresponding to the highest confidence score as the final mask. This specifically includes: Through the formula: Calculate the confidence score for each mask feature. ; in, , and It refers to the linear layer weights and biases. Indicates global average pooling; from The binary mask corresponding to the mask feature with the highest confidence score among the mask features is selected as the final mask.

7. The radar target template signal establishment method based on depth model adaptive segmentation according to claim 1, characterized in that, Using the final mask, the time-frequency transformation data corresponding to the radar echo sequence is filtered to obtain the filtered time spectrum. The time spectrum is then restored to its original frequency arrangement order through frequency inverse rearrangement to obtain the output data. The output data is restored to a time-domain signal using the inverse short-time Fourier transform, and this time-domain signal is used as the target template signal. Specifically, this includes: Through the formula: Obtain the filtered time spectrum; in, This represents the time-frequency transform data after centering and rearranging along the frequency axis of the complex matrix. Through the formula: Obtain the output data; in, This indicates inverse centralization of the frequency dimension; The output data is restored to a time-domain signal using the inverse short-time Fourier transform, and the target template signal is finally obtained. .

Images