A miniature SAR imaging method, system and medium

Abstract

This invention discloses a miniature SAR imaging method, system, and medium, relating to the field of synthetic aperture radar imaging technology. The method includes: configuring a multi-channel receiving array on a target miniature platform and integrating miniature SAR radar sensors; receiving radar echo signals to obtain a multi-channel radar echo signal sequence array; performing intra-sequence heterogeneous separation and identification; performing homogeneous low-pass filtering and heterogeneous difference preservation processing; performing intra-sequence information fusion to obtain a multi-channel regional SAR image set; and performing multi-channel cross-enhancement fusion of image features to obtain a target region SAR image. This invention solves the technical problems of insufficient processing accuracy and low imaging resolution of multi-channel radar echo signals under miniature platforms in existing technologies, achieving accurate processing and fusion of multi-channel radar echo signals under miniature platforms and improving the SAR imaging resolution of target areas.

Description

Technical Field

[0001] This invention relates to the field of synthetic aperture radar imaging technology, specifically to a miniature SAR imaging method, system, and medium. Background Technology

[0002] Synthetic Aperture Radar (SAR), as an active microwave remote sensing device, has all-weather, 24 / 7 operational capabilities and is widely used in environmental monitoring, disaster assessment, and military reconnaissance. With the rapid development of micro-platforms such as UAVs and small satellites, micro SAR systems have become a research hotspot due to their advantages of flexible deployment and low cost. However, limited by the payload of micro-platforms, existing micro SAR systems typically face problems such as small antenna size, low transmit power, and a limited number of channels. This leads to radar echo signals being susceptible to noise interference and insufficient multi-channel signal matching accuracy, ultimately affecting imaging resolution and image quality, making it difficult to meet the requirements of high-precision observation.

[0003] Existing technologies suffer from insufficient processing accuracy and low imaging resolution of multi-channel radar echo signals on miniature platforms. Summary of the Invention

[0004] This application provides a miniature SAR imaging method, system, and medium to address the technical problems of insufficient processing accuracy and low imaging resolution of multi-channel radar echo signals on miniature platforms in the prior art.

[0005] In view of the above problems, this application provides a miniature SAR imaging method, system and medium.

[0006] A first aspect of the embodiments of this application provides a miniature SAR imaging method, the method comprising:

[0007] A multi-channel receiving array is configured on the target micro-platform, and a micro SAR radar sensor is selected for integration. The integrated micro SAR radar scans the target area according to a preset scanning path, and the radar echo signals are received in parallel through the multi-channel receiving array to obtain a multi-channel radar echo signal sequence array. Based on the multi-dimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, intra-sequence heterogeneous separation identification is performed on each multi-channel radar echo signal sequence to obtain a multi-channel identified radar echo signal sequence array. The identifiers corresponding to each multi-channel identified radar echo signal are homogeneous signal identifiers and heterogeneous signal identifiers. Based on the homogeneous signal identifiers and heterogeneous signal identifiers, the signal processing dual channels are invoked to perform homogeneous low-pass filtering and heterogeneous difference preservation processing on the multi-channel identified radar echo signal sequence array to obtain a multi-channel identified filtered radar echo signal sequence array. Through synthetic aperture radar processing methods, the multi-channel identified filtered radar echo signal sequence array is traversed to perform intra-sequence information fusion to obtain a multi-channel regional SAR image set. The multi-channel cross-enhanced fusion of image features is performed on the multi-channel regional SAR image set to obtain the target area SAR image.

[0008] A second aspect of this application provides a miniature SAR imaging system, the system comprising:

[0009] The array configuration module is used to configure a multi-channel receiving array for the target micro-platform and integrate a micro SAR radar sensor. The echo signal receiving module is used to scan the target area according to a preset scanning path using the integrated micro SAR radar and to receive radar echo signals in parallel through the multi-channel receiving array, obtaining a multi-channel radar echo signal sequence array. The heterogeneous matching separation and identification module is used to perform intra-sequence heterogeneous matching separation and identification of each multi-channel radar echo signal sequence based on the multi-dimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, obtaining a multi-channel identified radar echo signal sequence array, wherein each multi-channel identified radar echo signal corresponds to an identifier. The system includes: a signal identifier for both homo-matching and hetero-matching signals; an echo signal sequence array acquisition module, used to perform homo-matching low-pass filtering and hetero-matching difference preservation processing on the multi-channel identified radar echo signal sequence array based on homo-matching and hetero-matching signal identifiers, thereby obtaining a multi-channel identified radar echo signal sequence array; a SAR image set acquisition module, used to perform intra-sequence information fusion on the multi-channel identified radar echo signal sequence array using synthetic aperture radar processing methods, thereby obtaining a multi-channel regional SAR image set; and an image fusion module, used to perform multi-channel cross-enhanced fusion of image features on the multi-channel regional SAR image set to obtain a target region SAR image.

[0010] In a third aspect of this application, a computer-readable storage medium is provided storing a computer program for executing a miniature SAR imaging method provided in this application.

[0011] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0012] A multi-channel receiving array is configured on the target micro-platform, and a micro SAR radar sensor is selected for integration. The integrated micro SAR radar scans the target area according to a preset scanning path, and the radar echo signals are received in parallel through the multi-channel receiving array to obtain a multi-channel radar echo signal sequence array. Based on the multi-dimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, intra-sequence heterogeneous separation identification is performed on each multi-channel radar echo signal sequence to obtain a multi-channel identified radar echo signal sequence array. Based on the same-match signal identification and heterogeneous signal identification, the signal processing dual-channel is invoked to perform same-match low-pass filtering and heterogeneous difference preservation processing on the multi-channel identified radar echo signal sequence array to obtain a multi-channel identified filtered radar echo signal sequence array. Through synthetic aperture radar processing methods, the multi-channel identified filtered radar echo signal sequence array is traversed to perform intra-sequence information fusion to obtain a multi-channel regional SAR image set. The multi-channel regional SAR image set is subjected to multi-channel cross-enhancement fusion of image features to obtain a target area SAR image. This technology achieves precise processing and fusion of multi-channel radar echo signals on a micro platform, thereby improving the SAR imaging resolution of the target area. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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.

[0014] Figure 1 This is a flowchart illustrating a miniature SAR imaging method provided in an embodiment of this application.

[0015] Figure 2 This is a schematic diagram of the structure of a miniature SAR imaging system provided in an embodiment of this application.

[0016] Figure labeling: Array configuration module 10, echo signal receiving module 20, heterogeneous separation identification module 30, echo signal sequence array acquisition module 40, SAR image set acquisition module 50, image fusion module 60. Detailed Implementation

[0017] This application provides a miniature SAR imaging method, system, and medium to address the technical problems of insufficient processing accuracy and low imaging resolution of multi-channel radar echo signals on miniature platforms in the prior art.

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0019] Example 1, as Figure 1 As shown, this application provides a miniature SAR imaging method, the method comprising:

[0020] Step S100: Configure a multi-channel receiving array for the target micro platform and select a micro SAR radar sensor for integration.

[0021] Specifically, the process begins by clarifying the specific type and operating parameters of the target micro-platform. Based on the platform's weight limitations, available space, and actual imaging scenario requirements, the specifications of the multi-channel receiving array are determined, including the number of antenna elements, their arrangement, and signal reception bandwidth. This ensures the array can be installed within the platform's limited space and meets the spatial resolution requirements for subsequent imaging. Next, miniature SAR radar sensors that meet the platform's compatibility standards are selected, with a focus on the sensor's size, power consumption, weight, and signal output stability to ensure compatibility with the target micro-platform's power supply system and data transmission interface. Subsequently, the selected multi-channel receiving array and the miniature SAR radar sensors are hardware integrated, completing the circuit connections between them to achieve signal transmission and synchronous control. Simultaneously, post-integration calibration tests are conducted to ensure that each unit of the multi-channel receiving array can accurately receive the echoes from the radar sensor's emitted signals reflected by the target, forming a collaborative miniature SAR radar hardware system. This lays the hardware foundation for subsequent target area scanning and echo signal acquisition.

[0022] Step S200: The integrated miniature SAR radar scans the target area according to a preset scanning path, and receives radar echo signals in parallel through a multi-channel receiving array to obtain a multi-channel radar echo signal sequence array.

[0023] Specifically, in the miniature SAR imaging process, after hardware integration, the integrated miniature SAR radar systematically scans the target area according to a pre-planned scanning path. During scanning, the radar actively emits electromagnetic waves, which generate echoes when they come into contact with various objects in the target area. A multi-channel receiving array works synchronously, capturing these echo signals in parallel. Different channels receive the echoes separately: 1. Amplitude information, reflecting the target's reflection intensity; for example, echoes from metal surfaces are strong, while echoes from smooth or distant targets are weak; 2. Phase information, indicating the relative position of the target and the radar, used for distance calculation; 3. Doppler shift, velocity information, generated by the target's relative motion, which can determine relative velocity and is crucial for monitoring dynamic targets such as aircraft and vehicles; 4. Spatial orientation information, assisting in determining the target's azimuth and elevation angles. Multiple channels receive signals from different angles, improving the spatial resolution of the image. These echo signals are arranged in an orderly manner according to time and channel sequences, ultimately forming a multi-channel radar echo signal sequence array, providing raw data containing rich target characteristics for subsequent signal processing.

[0024] Step S300: Based on the multidimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, perform intra-sequence heterogeneous separation identification of each multi-channel radar echo signal sequence to obtain a multi-channel identified radar echo signal sequence array, wherein the identifier corresponding to each multi-channel identified radar echo signal is a homogeneous signal identifier and a heterogeneous signal identifier.

[0025] Specifically, based on the acquired multi-channel radar echo signal sequence array, multi-dimensional features are extracted from each sequence, encompassing target reflection intensity reflected by echo amplitude, target relative position reflected by phase, target relative velocity associated with Doppler frequency shift, and target azimuth and elevation angles determined by spatial orientation. Next, based on these multi-dimensional features, signals within a single sequence are finely differentiated. Signals exhibiting background region characteristics and dominated by low-frequency information are marked as compatible signals and assigned a compatible signal identifier; signals exhibiting target region characteristics and dominated by high-frequency information are marked as discompatible signals and added a discompatible signal identifier. By performing the above operations on all sequences in the multi-channel radar echo signal sequence array, a multi-channel identified radar echo signal sequence array is finally formed. This array can serve as input data for subsequent synthetic aperture radar imaging, distinguishing between background and target, and providing fundamental support for improving imaging accuracy and enhancing target detection and resolution.

[0026] Step S400: Based on the same-match signal identifier and the different-match signal identifier, call the dual-channel signal processing to perform same-match low-pass filtering and different-match difference preservation processing on the multi-channel identifier radar echo signal sequence array to obtain a multi-channel identifier filtered radar echo signal sequence array.

[0027] Specifically, the process is based on a multi-channel labeled radar echo signal sequence array. For sequences with matching signal identifiers, a low-pass filter channel is activated, employing the Butterworth low-pass filter algorithm. The cutoff frequency is set according to the signal spectral characteristics, allowing low-frequency signals reflecting stable target features, such as the reflection amplitude of static ground objects and phase information at fixed positions, to pass smoothly. Simultaneously, high-frequency noise is attenuated, making the matching signals cleaner while preserving their basic characteristics. For sequences with dissimilar signal identifiers, a difference-preserving channel is triggered. Singular value decomposition (SVD) is used to decompose the matrix constructed from the dissimilar signals, extracting eigenvectors with larger singular values. This enhances key information distinguishing them from matching signals, such as the unique Doppler frequency shift of dynamic targets and spatial orientation differences of special targets, highlighting these differences. After dual-channel processing, all sequences are integrated to obtain a multi-channel labeled radar echo signal sequence array, providing optimized and more clearly defined signal support for subsequent imaging.

[0028] Step S500: Using the synthetic aperture radar processing method, the multi-channel identifier filter radar echo signal sequence array is traversed to perform intra-sequence information fusion to obtain a multi-channel regional SAR image set.

[0029] Specifically, a synthetic aperture radar (SAR) processing method is employed to traverse the obtained multi-channel marker-filtered radar echo signal sequence array. For each multi-channel marker-filtered radar echo signal sequence in the array, range compression processing is first performed. A matched filtering algorithm is used to compress the range dimension of the echo signal, focusing the target's signal energy in the range direction and improving range resolution. Next, azimuth focusing processing is performed, using Doppler frequency shift information to perform azimuth phase correction, eliminating phase errors caused by platform motion and concentrating the target's signal energy in the azimuth direction, further optimizing imaging clarity and obtaining a multi-channel marker-filtered preprocessed radar echo signal sequence array. Subsequently, intra-sequence information splicing and fusion are performed on each multi-channel marker-filtered preprocessed radar echo signal sequence. Signal segments acquired at different times and locations in the sequence are integrated according to temporal and spatial correlations, filling information gaps in individual signal segments and forming a complete fused sequence signal. Then, the fused one-dimensional echo signal is converted into a two-dimensional spatial image by back projection. Each multi-channel identifier filtered radar echo signal sequence generates a corresponding SAR image covering a specific area. After summing up the SAR images generated by all sequences, a multi-channel regional SAR image set is obtained.

[0030] Step S600: Perform multi-channel cross-enhancement fusion of image features on the multi-channel region SAR image set to obtain the target region SAR image.

[0031] Specifically, an image feature extractor is used to extract features from each image in the multi-channel regional SAR image set, obtaining a multi-channel regional SAR image feature set containing information such as image texture details, grayscale distribution, edge contours, and scattering intensity. Next, all image features in this feature set are paired and combined to form a multi-channel regional SAR image feature combination set. For each feature combination in the set, image feature similarity analysis is first performed to calculate the similarity coefficients between features and form an image feature similarity coefficient set. Then, softmax processing is applied to the similarity coefficient set, and a cross-enhancement fusion matrix is ​​constructed based on the processing results. Based on this matrix, each image feature in the feature combination is enhanced to obtain the corresponding cross-enhancement fusion feature combination, and all combinations are summarized into a multi-channel regional SAR image cross-enhancement fusion feature combination set. Subsequently, using the original multi-channel regional SAR image feature set as an index, the mean of multiple enhanced fusion features corresponding to the same image feature in the cross-enhancement fusion feature combination set is calculated to generate a multi-channel regional SAR image cross-enhancement fusion feature mean set. Based on this mean set, image enhancement is performed on the multi-channel regional SAR image set to improve image detail clarity and contrast, resulting in an enhanced multi-channel regional SAR image set. Finally, image mean filtering is applied to the enhanced image set to eliminate local noise interference, ultimately obtaining a target area SAR image that clearly presents the ground features of the target area without excessive smoothing.

[0032] In one possible implementation, step S300 further includes:

[0033] Step S310: According to the preset signal multidimensional features, perform feature extraction on each multi-channel radar echo signal in the first multi-channel radar echo signal sequence of the multi-channel radar echo signal sequence array to obtain the first multi-channel radar echo signal multidimensional feature sequence, wherein the preset signal multidimensional features include frequency domain features, time domain features, Doppler frequency shift features and spatial direction features.

[0034] Step S320: Obtain the multi-layer sensing matching prototype library, and perform similarity identification between each multi-channel radar echo signal multi-dimensional feature in the first multi-channel radar echo signal multi-dimensional feature sequence and each multi-layer sensing matching prototype in the multi-layer sensing matching prototype library. When the similarity identification result has a similarity greater than or equal to the preset similarity identification result threshold, the corresponding multi-channel radar echo signal is identified as a matching signal to obtain the first multi-channel matching signal identified radar echo signal subsequence.

[0035] Step S330: When there is no similarity result greater than or equal to the preset similarity result threshold, the corresponding multi-channel radar echo signal is marked with a different matching signal to obtain the first multi-channel different matching signal marked radar echo signal subsequence.

[0036] Step S340: The first multi-channel co-matching signal identification radar echo signal sub-sequence and the first multi-channel hetero-matching signal identification radar echo signal sub-sequence are fused in chronological order from front to back, and the fused first multi-channel identification radar echo signal sequence is added to the multi-channel identification radar echo signal sequence array.

[0037] Specifically, for a multi-channel radar echo signal sequence array, the "first multi-channel radar echo signal sequence" here does not specify a particular order, but rather refers to any sequence within the array. During processing, based on a pre-defined multi-dimensional signal feature system encompassing frequency domain features, time domain features, Doppler frequency shift features, and spatial direction features, feature extraction is performed on each multi-channel radar echo signal in the selected sequence. Fourier transform is used to analyze the signal spectrum, converting the time-domain signal to the frequency domain to extract frequency domain features. This allows for the identification of stable, low-frequency-dominated homologous signal features, as well as significantly changing, high-frequency-prominent heterologous signal features. The signal time-domain waveform is observed to capture time-domain features such as the rate of change and amplitude fluctuations of signal amplitude over time, thus distinguishing the differences between dynamic targets and static backgrounds. Based on the principle that the relative motion of dynamic targets causes changes in signal frequency, Doppler frequency shift features are extracted to identify frequency changes caused by dynamic targets. Finally, combined with the target's spatial distribution, including its trajectory, location, and elevation angle, spatial direction features are extracted to distinguish the spatial distribution differences between dynamic and static targets. By extracting these dimensional features from each signal in the sequence, the original echo signal is transformed into a multi-dimensional feature sequence of multi-channel radar echo signals that contains rich information. This provides accurate and comprehensive feature basis for subsequent processes such as distinguishing between homologous and heterologous signals and conducting similarity recognition.

[0038] A multi-layer sensing matching prototype library is acquired. This library is a standard feature set generated through multi-layer perceptron model training, feature aggregation, and prototype refinement from a large amount of radar echo data from background areas, i.e., typical scenarios of matching signals. It covers typical feature patterns of background area echo signals in the frequency domain, time domain, Doppler frequency shift, and spatial direction. Next, for the extracted multi-dimensional feature sequence of the first multi-channel radar echo signal, each multi-dimensional feature of the multi-channel radar echo signal is extracted sequentially and compared with each multi-layer sensing matching prototype in the multi-layer sensing matching prototype library for similarity identification. The degree of matching is determined by calculating the distance between features, such as Euclidean distance or cosine distance. If the similarity identification result of a certain multi-dimensional feature with any multi-layer sensing matching prototype reaches or exceeds a pre-set similarity identification result threshold, it indicates that the echo signal possesses matching signal characteristics dominated by background area and low-frequency information. The corresponding multi-channel radar echo signal is then labeled as a matching signal. All echo signals that have been matched and identified are arranged in chronological order according to the original sequence to form the first multi-channel matched signal identification radar echo signal subsequence, which provides a clear signal classification basis for subsequent differentiation of background and target and improvement of imaging accuracy.

[0039] For each feature in the multidimensional feature sequence of the first multi-channel radar echo signal, if the similarity identification result obtained when comparing it with all multi-layer sensing matching prototypes in the multi-layer sensing matching prototype library does not reach the preset similarity identification result threshold, it means that the multi-channel radar echo signal corresponding to that feature does not possess the typical characteristics of matching signals in the background area, but rather belongs to the target area, a signal type dominated by high-frequency information. In this case, such signals are identified as mismatched signals, and all echo signals identified as mismatched signals are integrated according to their chronological order in the original sequence to obtain the first multi-channel mismatched signal identified radar echo signal subsequence, providing clearly classified signal data for subsequent differentiation of targets from the background and further imaging processing.

[0040] The two sub-sequences are spliced ​​and fused according to the chronological order of signal generation, ensuring that the original order of the homologous and heterologous signals on the time axis is preserved, forming a complete first multi-channel labeled radar echo signal sequence containing background and target differentiation markers. This fused sequence is then added to the multi-channel labeled radar echo signal sequence array, enriching the array's data content and providing signal input with clear background-target identification for subsequent applications such as synthetic aperture radar imaging based on this array, thus helping to improve the accuracy of background and target differentiation during imaging.

[0041] In one possible implementation, step S320 further includes:

[0042] Step S321: Obtain multiple sets of radar echo signals with matching signal identifiers and corresponding sets of multidimensional features of radar echo signals with matching signal identifiers. Then, perform similar aggregation on the multiple sets of radar echo signals with matching signal identifiers according to a preset aggregation scale. Based on the similar aggregation results, perform mapping aggregation on the multidimensional feature sets of radar echo signals with matching signal identifiers to obtain multiple aggregated sets of multidimensional feature sets of radar echo signals with matching signal identifiers.

[0043] Step S322: Extract the prototype within the set for each of the multiple aggregated sample radar echo signal multidimensional feature sets to obtain multiple prototypes of radar echo signals with the same matching signal.

[0044] Step S323: Summarize the multidimensional feature prototypes of the multiple matching signal identifiers radar echo signals to construct the multilayer sensing matching prototype library.

[0045] Specifically, a large amount of sample data from background areas, such as stable terrain and conventional static features, is first collected. This data includes multiple sets of sample radar echo signals identifying the same signal, and a multi-dimensional feature set of the sample radar echo signals identifying the same signal, corresponding to each signal set. The features cover the frequency domain, time domain, Doppler frequency shift, and spatial direction dimension. Subsequently, according to a preset aggregation scale, which is determined based on the background scene type and signal feature similarity threshold, such as dividing aggregation categories according to terrain type and signal frequency range, multiple sets of sample radar echo signals identifying the same signal are aggregated into the same category. Signal sets with similar signal features and corresponding background scenes are grouped into the same category, forming several groups of similar signal sets. Based on this, a mapping relationship between signal sets and feature sets is established according to the aggregation results of the same type. The multi-dimensional feature sets of radar echo signals corresponding to each signal set of the same type are mapped and aggregated. The multi-dimensional features within the same group are statistically integrated according to the feature dimensions. The mean, variance, etc. of the features of the same dimension are calculated, and the common information of the same type of features is retained. Finally, multiple aggregated sample radar echo signal multi-dimensional feature sets are obtained, which provide a regular and representative feature data foundation for the subsequent extraction of the prototype of the same type of signal features.

[0046] For each aggregated sample's matched signal identifier radar echo signal multidimensional feature set, a prototype extraction operation is performed within the set. During this operation, all multidimensional feature data within a single aggregated feature set are first traversed, and statistical analysis is performed along four dimensions: frequency domain features, time domain features, Doppler frequency shift features, and spatial direction features. The mean value of the feature data in each dimension is calculated to determine the initial feature benchmark for the aggregated set in each dimension. Next, according to a preset iteration scale, starting from the initial feature mean value, random iterative operations are performed on the feature data in each dimension within the aggregated feature set, generating multiple sets of iterated aggregated feature data. Subsequently, the iteration density of the iterated aggregated feature data is compared with the iteration density of the initial feature mean value. If the density after iteration is greater than or equal to the initial density, the iteration process is repeated according to the preset scale until the preset number of iterations is reached, ensuring that the extracted prototype can cover the common information of similar features within the aggregated set to the greatest extent possible. After all iterations are completed, the most representative feature combinations are selected from the final iterative feature data. These are the core data that centrally represent the matching signal characteristics of the background scene corresponding to the aggregated set, and are identified as the multi-dimensional feature prototypes of the matching signal identifier radar echo signals corresponding to the multi-dimensional feature set of the matching signal identifier radar echo signals of the aggregated sample. By performing the above operation on each of the multi-dimensional feature sets of the matching signal identifier radar echo signals of all aggregated samples, multiple multi-dimensional feature prototypes of matching signal identifier radar echo signals are finally obtained, laying the foundation for the subsequent construction of a multi-layer sensing matching prototype library.

[0047] Multiple multi-dimensional feature prototypes of radar echo signals with matching signal identifiers extracted from various aggregated sample multi-dimensional feature sets are centrally summarized. The feature dimension information corresponding to each prototype is sorted out to ensure that each prototype fully contains the four core feature data categories: frequency domain features, time domain features, Doppler frequency shift features, and spatial direction features, and that the data format is consistent with the requirements of subsequent similarity recognition. Subsequently, all the summarized multi-dimensional feature prototypes are classified and organized, and secondary grouped according to the type of background scene corresponding to the prototype, such as stable terrain background, conventional static building background, etc., or feature similarity, making the internal structure of the prototype library clearer and facilitating rapid retrieval and matching later. At the same time, an index tag is added to each multi-dimensional feature prototype, containing key information such as the background scene category corresponding to the prototype and the aggregation set number from which the feature extraction originated, improving the positioning efficiency during subsequent calls. Finally, all the classified, organized, and indexed multi-dimensional feature prototypes of radar echo signals with matching signal identifiers are integrated and stored according to a preset data storage format, such as matrix form or feature vector group form, to construct a multi-layer sensing matching prototype library that can be directly used for subsequent similarity recognition operations.

[0048] In one possible implementation, step S322 further includes:

[0049] Step S3221: Traverse the set of multidimensional features of radar echo signals with matching signals of multiple aggregated samples and calculate the mean value to determine the mean value of the multidimensional features of radar echo signals with matching signals of multiple aggregated samples.

[0050] Step S3222: According to a preset iteration scale, randomly iterate the mean value of the multidimensional features of the radar echo signals of the multiple aggregated sample matching signal identifiers in the set of multidimensional features of the multiple aggregated sample matching signal identifiers to obtain multiple iterated multidimensional features of the radar echo signals of the multiple aggregated sample matching signal identifiers.

[0051] Step S3223: When the iteration density of the multidimensional features of the radar echo signal of the multiple iterative aggregated sample matching signal identifier is greater than or equal to the iteration density of the average multidimensional features of the radar echo signal of the multiple aggregated sample matching signal identifier, continue to randomly iterate the multidimensional features of the radar echo signal of the multiple iterative aggregated sample matching signal identifier according to the preset iteration scale until the preset number of iterations is met, and obtain the prototype of the multidimensional features of the radar echo signal of the multiple matching signal identifier.

[0052] Specifically, the process iterates through multiple aggregated sample matching signal identifier radar echo signal multidimensional feature sets. For the multidimensional feature data within each set, covering four dimensions—frequency domain, time domain, Doppler frequency shift, and spatial direction—the mean is calculated for each feature dimension. That is, the mean of the frequency domain feature data of all signals within a single aggregated set is calculated to obtain the mean of the frequency domain feature of that set. Similarly, the mean of the time domain feature, the mean of the Doppler frequency shift feature, and the mean of the spatial direction feature are calculated sequentially. Then, the means of the four dimensions are integrated to form the mean of the multidimensional feature of the aggregated sample matching signal identifier radar echo signal corresponding to each aggregated set. Finally, the mean of the multidimensional feature of each aggregated set is determined, providing an initial benchmark for subsequent iterative calculations.

[0053] Based on the distribution characteristics of the multidimensional features of the aggregated sample matching signal identifying radar echo signals, preset iteration scales are set for the standard deviation and data range of each dimension feature. For example, the iteration adjustment range of frequency domain features is limited to within ±5% of the mean, the iteration step size of time domain features is set to 0.01 seconds, the iteration interval of Doppler frequency shift features is controlled within ±0.5Hz of the mean, and the iteration angle accuracy of spatial direction features is retained to 0.1°, ensuring that the iteration does not deviate from the common range of matching signal features, while also covering a reasonable variation range. Next, a Monte Carlo random sampling algorithm is used to randomly generate 10-20 sets of adjustment parameters within each aggregated sample matching signal identifying radar echo signal multidimensional feature set, centered on the calculated mean of the multidimensional features, according to the preset iteration scale. Each set of parameters corresponds to the adjustment amount in four dimensions: frequency domain, time domain, Doppler frequency shift, and spatial direction. Subsequently, each set of adjustment parameters is applied to the initial multidimensional feature mean. Through feature dimension mapping operations, such as frequency domain feature superposition adjustment, time domain feature translation correction, Doppler frequency shift feature compensation, and spatial direction feature angle fine-tuning, corresponding new feature data is generated. Finally, the validity of the generated new feature data is verified, outlier data that exceeds the feature distribution range of the aggregated set is removed, and valid data that conforms to the characteristic rules of the matching signal are retained. Ultimately, multidimensional features of radar echo signals identifying matching signals from multiple iterative aggregated samples are obtained, providing diverse candidate features for subsequent density comparison and prototype screening.

[0054] A kernel density estimation algorithm is used to calculate the iterative density of the multidimensional features of the radar echo signals identified by the matched signals of multiple iterative aggregated samples, as well as the iterative density of the mean of the multidimensional features of the matched signals of the radar echo signals identified by the matched signals of multiple aggregated samples. The iterative density reflects the concentration of the feature data distribution in the multidimensional space; a higher density means that the set of features is more representative of the commonalities of the matched signals in the corresponding aggregate set. Next, the two sets of density data are compared one by one. If the density of a certain iterative feature is greater than or equal to the density of its corresponding initial mean, it indicates that the iterative feature is more representative, and the random iteration operation should continue using this iterative feature as the new benchmark, according to a preset iteration scale, such as ±5% in the frequency domain and 0.01 seconds in the time domain. If the density of the iterative feature is less than the initial mean density, the iteration direction is adjusted, such as reducing the adjustment range in the frequency domain and optimizing the sampling range in the time domain before iterating again. Throughout the process, the feature data and density values ​​of each iteration are continuously recorded until the preset number of iterations, such as 10-20, is completed. Finally, from the final iterative features of each aggregation set, the set of data with the highest density that conforms to the characteristic pattern of the matching signal in that set is selected and identified as the multidimensional feature prototype of the matching signal identification radar echo signal. In the end, multiple multidimensional feature prototypes of the matching signal identification radar echo signal belonging to different aggregation sets are obtained, providing core data support for the subsequent construction of a multilayer sensing matching prototype library.

[0055] In one possible implementation, step S400 further includes:

[0056] Step S410: Call the low-pass filter channel in the dual-channel signal processing to perform low-pass filtering on the multi-channel identifier radar echo signals with matching signal identifiers in the multi-channel identifier radar echo signal sequence array, and call the IW filter channel in the dual-channel signal processing to perform difference preservation processing on the multi-channel identifier radar echo signals with dissimilar signal identifiers in the multi-channel identifier radar echo signal sequence array, to obtain the multi-channel identifier filtered radar echo signal sequence array.

[0057] Specifically, based on the identifiers (same / different) of each signal in the multi-channel radar echo signal sequence array, a signal classification index table is constructed to clarify the identifier type of the signal at each channel and time node, providing data positioning basis for dual-channel parallel processing. For echo signals with same-match signal identifiers, a Butterworth low-pass filter is used when calling the low-pass filter channel. Based on the characteristic that same-match signals are dominated by low-frequency background features, the cutoff frequency is set to 500Hz, which can be adjusted according to the actual scenario. Convolution operation allows low-frequency signals to pass smoothly while filtering out high-frequency noise. At the same time, zero-phase filtering technology is used to avoid signal phase distortion and ensure the integrity of the background features of same-match signals. For echo signals with different-match signal identifiers, when the IW filter channel is enabled, the signal is first decomposed into wavelet coefficients of different scales by wavelet transform. The high-frequency wavelet coefficients reflecting the target difference features are assigned a weight coefficient of 1.2 to enhance the difference information, while the low-frequency coefficients are assigned a weight coefficient of 0.8 to suppress background interference. Then, the signal is reconstructed through inverse wavelet transform to preserve and enhance the difference features of different-match signals. Finally, the homologous and heterologous signals processed by the dual channels are re-integrated according to the channel affiliation and time order of the original sequence to generate a multi-channel identifier-filtered radar echo signal sequence array. This array not only ensures the purity of the homologous signals but also highlights the target characteristics of the heterologous signals, laying a high-quality signal foundation for subsequent imaging.

[0058] In one possible implementation, step S500 further includes:

[0059] Step S510: Perform range compression and azimuth focusing processing on each multi-channel labeled radar echo signal sequence in the multi-channel labeled radar echo signal sequence array to obtain a multi-channel labeled radar echo signal sequence array.

[0060] Step S520: Perform intra-sequence signal splicing and fusion on the multi-channel identifier filtering preprocessed radar echo signal sequence array, and convert the fused signal into a two-dimensional spatial image by back projection method to obtain the multi-channel regional SAR image set.

[0061] Specifically, for the obtained multi-channel labeled radar echo signal sequence array, range compression and azimuth focusing processing are performed on each multi-channel labeled radar echo signal sequence. The range compression stage employs a matched filtering algorithm to compress the broadened signal in the range direction of the echo signal to the actual target range size. By matching the echo signal with the transmitted signal, the range energy is focused, significantly improving range resolution. The azimuth focusing processing is based on the Doppler frequency shift principle, utilizing Doppler information generated by platform motion to perform azimuth phase correction on the signal, eliminating the azimuth signal broadening caused by platform motion errors, and concentrating the target energy in the azimuth direction, further optimizing imaging clarity. After these two processes, each sequence is transformed into a signal sequence with better focusing effect, forming a multi-channel labeled radar echo signal sequence array.

[0062] Based on the correlation between the timestamps of signal acquisition and spatial coordinates, preprocessed echo signals acquired at different time periods and spatial locations within each sequence are matched segment by segment. First, the time deviation of each signal segment is calibrated using a time synchronization algorithm to ensure that signals in the same scene are aligned in the time dimension. Then, based on spatial location information (such as the latitude, longitude, altitude, and attitude angle data of the radar platform), spatial registration is performed on the signals to eliminate spatial misalignment of signals caused by minor fluctuations in the platform's motion trajectory. Subsequently, a weighted splicing strategy is adopted to assign weights to signals in overlapping areas according to their signal-to-noise ratio (SNR). Signal segments with higher SNR have a higher weight, achieving seamless splicing of signal segments while preserving the low-frequency stability characteristics of the same-matched signals and the high-frequency detail characteristics of the different-matched signals, forming a complete sequence fusion signal covering the corresponding area. Next, the fused signal is converted into a two-dimensional spatial image using back projection. The specific process is as follows: First, a two-dimensional spatial coordinate system is established, using the geographic coordinates of the target area, such as the UTM coordinate system, as a reference. A uniform pixel grid is then divided to determine the spatial coordinates of each pixel. Then, the motion parameters of the radar platform during signal acquisition are retrieved, including the position, velocity, attitude angle, and radar system parameters such as wavelength, pulse repetition frequency, and antenna gain at each sampling moment. The slant range, azimuth angle, and other geometric relationships between the radar and each pixel in the two-dimensional grid at each sampling moment are calculated. Based on these geometric relationships, the fused one-dimensional echo signal is decomposed according to the "range-azimuth" dimension, and the echo signal energy at each sampling moment is back-projected. At the corresponding pixel, for each pixel, the contribution values ​​of all echo signals that can illuminate that point are superimposed through integration. At the same time, the gray value of the pixel is calculated by combining the amplitude and phase information of the radar echo. The amplitude information corresponds to the gray intensity, and the phase information is used to optimize the gray uniformity. During the projection process, range migration correction is also required to eliminate the signal offset in the range direction caused by the relative motion between the target and the radar, as well as azimuth ambiguity suppression, to avoid mutual interference between signals of adjacent pixels. After completing the back projection operation of all fused signals, each multi-channel label filtering preprocessed radar echo signal sequence generates a corresponding two-dimensional spatial image. After all these images are summarized, a multi-channel regional SAR image set is obtained.

[0063] In one possible implementation, step S600 further includes:

[0064] Step S610: Use an image feature extractor to extract image features from the multi-channel region SAR image set to obtain a multi-channel region SAR image feature set.

[0065] Step S620: Perform pairwise enumeration and combination of the multi-channel region SAR image feature set to obtain a multi-channel region SAR image feature combination set.

[0066] Step S630: Extract the first multi-channel region SAR image feature combination from the multi-channel region SAR image feature combination set, perform multi-channel cross-enhancement fusion of image features to obtain the first multi-channel region SAR image cross-enhancement fusion feature combination, and add the first multi-channel region SAR image cross-enhancement fusion feature combination to the multi-channel region SAR image cross-enhancement fusion feature combination set.

[0067] Step S640: Using the multi-channel region SAR image feature set as an index, perform mean processing on multiple multi-channel region SAR image cross-enhancement fusion features corresponding to the same multi-channel region SAR image feature in the multi-channel region SAR image cross-enhancement fusion feature combination set to obtain a multi-channel region SAR image cross-enhancement fusion feature mean set.

[0068] Step S650: Based on the mean set of cross-enhancement fusion features of the multi-channel region SAR images, perform inversion image enhancement on the multi-channel region SAR image set to obtain an enhanced multi-channel region SAR image set.

[0069] Step S660: Perform image mean filtering on the enhanced multi-channel regional SAR image set to obtain the target region SAR image.

[0070] Specifically, a multi-channel regional SAR image set is used as the processing object, and feature extraction is performed using a kernel-based image feature extractor. First, based on the characteristics of the multi-channel regional SAR images, such as differences in ground object scattering and texture distribution patterns, appropriate convolution kernel parameters are configured. For example, 3×3 or 5×5 convolution kernels are selected. Gradient-type convolution kernels, such as Sobel kernels, are set for image edge features, Gaussian kernels are set for texture features, and mean kernels are set for scattering intensity features, ensuring that different convolution kernels can accurately capture feature information of different dimensions of the image. Subsequently, each image in the multi-channel regional SAR image set is sequentially input into the image feature extractor. Feature extraction is achieved through sliding convolution operations between the convolution kernel and the image: the convolution kernel slides row by row and column by column on the image pixel matrix according to a preset stride. At each position, it performs a product summation operation with the pixels of the corresponding region, generating feature values ​​reflecting the local features of that region. During this process, different types of convolution kernels are applied to the image: gradient convolution kernels highlight the grayscale changes of ground object boundaries in the image, strengthening edge features; Gaussian convolution kernels smooth image noise while preserving the detailed distribution of ground object textures; and mean convolution kernels integrate the pixel grayscale of local regions, reflecting the mean characteristics of ground object scattering intensity. After feature extraction by all convolution kernels for each image, the extracted edge, texture, and scattering intensity features are integrated into a set of feature vectors. This vector encompasses the core feature expressions of the image in different dimensions. Finally, the feature vectors corresponding to all images in the multi-channel regional SAR image set are summarized to form a multi-channel regional SAR image feature set.

[0071] For all feature vectors in the feature set of multi-channel regional SAR images, enumeration operations are performed according to the pairwise combination rule. That is, any two feature vectors in the set are selected in turn to form a feature combination, covering all possible combination forms between image features of different channels. Finally, a multi-channel regional SAR image feature combination set containing multiple feature combinations is generated, providing rich feature interaction data for subsequent cross-enhancement fusion.

[0072] From the obtained set of multi-channel regional SAR image feature combinations, the first set of feature combinations, namely the first multi-channel regional SAR image feature combination, is extracted as the processing object. Next, following the feature fusion specification, three core operations are performed: First, image feature similarity analysis is conducted on the two sets of image features from different channels in the feature combination. By calculating indicators such as cosine similarity and Pearson correlation coefficient between feature vectors, the similarity between the two sets of features in dimensions such as texture, edge, and scattering intensity is quantified, generating a set of image feature similarity coefficients containing multiple sets of similarity data. Second, the obtained set of image feature similarity coefficients is processed using softmax to convert the similarity coefficients into a probability distribution form, highlighting the feature association weights corresponding to high similarity coefficients. Then, a cross-enhancement fusion matrix is ​​constructed based on the processed probability distribution. The matrix elements intuitively reflect the fusion priority and association strength between the two sets of features in each dimension. Third, based on the cross-enhancement fusion matrix, image feature enhancement is performed on the two sets of features in the first multi-channel region SAR image feature combination. Feature dimensions with higher weights in the matrix, such as edge feature dimensions with high similarity, are strengthened, while redundant dimensions with lower weights are appropriately suppressed. Finally, a first multi-channel region SAR image cross-enhancement fusion feature combination with the advantages of both sets of features and richer details is generated. Finally, the fusion feature combination is added to the pre-created multi-channel regional SAR image cross-enhancement fusion feature combination set according to the data storage format requirements, providing high-quality fusion feature data support for subsequent mean processing and target region image generation.

[0073] Using the feature vectors in the feature set of multi-channel regional SAR images as indices, all cross-enhancement fusion feature combinations associated with each index feature vector are selected from the set of cross-enhancement fusion feature combinations of multi-channel regional SAR images. These are different fusion combinations formed by the same original image features. The mean of these combinations is calculated according to the feature dimension to eliminate random differences between different combinations. The mean of cross-enhancement fusion features corresponding to each index feature vector is obtained. All the mean values ​​are integrated to form the set of mean values ​​of cross-enhancement fusion features of multi-channel regional SAR images.

[0074] A mapping model between feature means and image pixels is established. By analyzing the correspondence between features of various dimensions in the cross-enhanced fusion feature mean set, such as texture, edge, and scattering intensity mean, and the original image pixel grayscale and spatial distribution, the influence weight of feature means on pixel optimization is determined. For example, the pixel gradient weight corresponding to areas with high edge feature means is increased to enhance the clarity of ground feature outlines. Subsequently, for each original image in the multi-channel regional SAR image set, an inversion operation is performed based on the above mapping model: using the image pixel coordinates as indexes, the cross-enhanced fusion feature means at the corresponding position is called to dynamically adjust the grayscale value of the original pixels. For areas with high scattering intensity mean, the pixel grayscale is appropriately increased to highlight strongly reflective targets; for areas with rich texture feature means, a pixel neighborhood enhancement algorithm is used to preserve detailed textures; for areas with obvious edge feature means, gradient enhancement technology is used to sharpen ground feature boundaries. Image fidelity constraints are also introduced during the inversion process to avoid geometric deformation of ground features or generation of false features due to feature enhancement. After completing the inversion enhancement of all original images, the processed images are arranged according to their channel affiliation and the original sequence order, ultimately forming an enhanced multi-channel region SAR image set with clearer details, better contrast, and higher target recognition, laying a high-quality image foundation for subsequent mean filtering to generate target region SAR images.

[0075] Based on the detail richness and noise distribution of the enhanced image, the neighborhood window parameters for mean filtering are determined. A 3×3 or 5×5 square window is typically chosen, with the window size selected to effectively filter local noise without damaging key details such as small targets and edge textures. Subsequently, each image from the enhanced multi-channel region SAR image set is sequentially input into the filtering module. The filtering operation is performed using a sliding window to calculate the mean: taking each pixel as the center, the arithmetic mean of all pixel grayscale values ​​within the neighborhood window's coverage area is calculated to obtain the filtered grayscale value of that central pixel, which then replaces the original pixel value. For edge pixels where the neighborhood window extends beyond the image boundary, boundary extension strategies, such as mirror extension or zero-padding, are used to supplement missing pixel information within the window, ensuring the integrity and consistency of the filtering effect. During the filtering process, it is crucial to control the window sliding step size, typically set to 1, to avoid loss of image detail or blocky distortion due to an excessively large step size. After mean filtering of all enhanced images, the processed images are integrated with the region association according to the original channel order, and finally a target region SAR image with low noise, clear details and stable overall image quality is generated to meet the application needs of subsequent accurate observation, identification or monitoring of the target region.

[0076] In one possible implementation, step S630 further includes:

[0077] Step S631: Perform image feature similarity analysis on the combination of SAR image features in the first multi-channel region to obtain a set of image feature similarity coefficients.

[0078] Step S632: Perform softmax processing on the set of image feature similarity coefficients, and construct a cross-enhancement fusion matrix based on the processing result.

[0079] Step S633: Based on the cross-enhancement fusion matrix, perform image feature enhancement on the first multi-channel region SAR image feature combination to obtain the first multi-channel region SAR image cross-enhancement fusion feature combination.

[0080] Specifically, for the image features of two different channels in the first multi-channel region SAR image feature combination, such as the texture features of channel A and the edge features of channel B, a feature similarity calculation algorithm is used for analysis: by calculating the cosine similarity, Euclidean distance, or Pearson correlation coefficient between the two sets of feature vectors, the similarity between them in various feature dimensions, such as gray-level distribution, texture frequency, edge gradient, etc., is quantified. A similarity coefficient close to 1 indicates that the two sets of features are highly similar in that dimension, while a similarity coefficient close to 0 indicates significant differences. The similarity calculation results of all dimensions are organized into an ordered set of data to form an image feature similarity coefficient set, providing a quantitative basis for subsequent fusion weight allocation.

[0081] For the obtained set of image feature similarity coefficients, softmax normalization is performed: the softmax function converts each similarity coefficient in the set into a probability value ranging from [0, 1], and the sum of all probability values ​​is 1. This eliminates the difference in numerical magnitude between different coefficients, highlighting the priority of feature associations corresponding to high similarity coefficients. For example, for two dimensions with original similarity coefficients of 0.8 and 0.2, after softmax processing, the probability value of the former will be significantly higher than that of the latter, more clearly reflecting the difference in feature association strength. Subsequently, an initially empty two-dimensional cross-enhancement fusion matrix is ​​created, with the row and column dimensions of the matrix consistent with the dimensions of the two sets of features in the first multi-channel region SAR image feature combination. According to the correspondence of feature dimensions, the normalized probability values ​​after softmax processing are filled into the matrix one by one: the element value in the i-th row and j-th column of the matrix is ​​the fusion weight of the i-th dimension of the first set of features and the j-th dimension of the second set of features. After filling all normalized values, a cross-enhancement fusion matrix that can accurately quantify the association strength of each dimension of the two sets of features is formed. This matrix provides a clear weight basis for subsequent feature enhancement operations based on convolutional networks.

[0082] The constructed cross-enhancement fusion matrix is ​​used as the core parameter of the convolutional network, such as the kernel weights. Two sets of features from the first multi-channel region SAR image feature combination, such as the texture features of channel A and the edge features of channel B, are input into the feature enhancement layer of the convolutional network. Through local connectivity and parameter sharing mechanisms, the convolutional network enhances the two sets of features dimensionally, guided by the weights of each element in the cross-enhancement fusion matrix: for feature dimensions with higher weights, such as the texture-edge correspondence dimension reflecting strong correlations, the network strengthens the feature response of that dimension, highlighting feature details through convolution operations; for redundant dimensions with lower weights, their feature intensity is appropriately suppressed to avoid invalid information interfering with the fusion effect. Simultaneously, the convolutional network performs non-linear mapping on the enhanced features to further optimize the feature discrimination and expressive power, ensuring that the two sets of features retain their respective advantages while achieving efficient integration of complementary information during the fusion process. Finally, after the convolutional network enhancement process, the two sets of features form a set of feature data that combines multi-dimensional advantages, rich details, and strong correlations—the cross-enhancement fusion feature combination of the first multi-channel region SAR image.

[0083] Example 2, based on the same inventive concept as the miniature SAR imaging method in the foregoing examples, such as... Figure 2 As shown, this application provides a miniature SAR imaging system. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0084] The array configuration module 10 is used to configure a multi-channel receiving array for the target micro platform and to select a micro SAR radar sensor for integration.

[0085] The echo signal receiving module 20 is used to scan the target area according to a preset scanning path using the integrated miniature SAR radar, and to receive radar echo signals in parallel through a multi-channel receiving array to obtain a multi-channel radar echo signal sequence array.

[0086] The heterogeneous separation identification module 30 is used to perform intra-sequence heterogeneous separation identification of each multi-channel radar echo signal sequence based on the multi-dimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, thereby obtaining a multi-channel identified radar echo signal sequence array, wherein the identification corresponding to each multi-channel identified radar echo signal is a homogeneous signal identification and a heterogeneous signal identification.

[0087] The echo signal sequence array acquisition module 40 is used to perform low-pass filtering on the multi-channel identifier radar echo signal sequence array based on the same-match signal identifier and the different-match signal identifier, and to perform difference preservation processing on the different-match signal identifier, so as to obtain a multi-channel identifier filtered radar echo signal sequence array.

[0088] The SAR image set acquisition module 50 is used to perform intra-sequence information fusion by traversing the multi-channel identifier filter radar echo signal sequence array through synthetic aperture radar processing method to obtain a multi-channel regional SAR image set.

[0089] The image fusion module 60 is used to perform multi-channel cross-enhancement fusion of image features on the multi-channel region SAR image set to obtain the target region SAR image.

[0090] Furthermore, the system is also used to implement the following functions:

[0091] Based on preset multidimensional signal features, feature extraction is performed on each multi-channel radar echo signal in the first multi-channel radar echo signal sequence of the multi-channel radar echo signal sequence array to obtain a first multi-channel radar echo signal multidimensional feature sequence. The preset multidimensional signal features include frequency domain features, time domain features, Doppler frequency shift features, and spatial direction features. A multi-layer sensing matching prototype library is acquired, and the multidimensional features of each multi-channel radar echo signal in the first multi-channel radar echo signal multidimensional feature sequence are compared with each multi-layer sensing matching prototype in the multi-layer sensing matching prototype library for similarity identification. When the similarity identification result has a similarity greater than or equal to a preset similarity identification result threshold, the similarity is determined. When similarity is achieved, the corresponding multi-channel radar echo signals are labeled with matching signals to obtain the first multi-channel matching signal labeled radar echo signal subsequence. When the similarity recognition result does not have a similarity greater than or equal to the preset similarity recognition result threshold, the corresponding multi-channel radar echo signals are labeled with dissimilar signals to obtain the first multi-channel dissimilar signal labeled radar echo signal subsequence. The first multi-channel matching signal labeled radar echo signal subsequence and the first multi-channel dissimilar signal labeled radar echo signal subsequence are fused in chronological order from front to back, and the fused first multi-channel labeled radar echo signal sequence is added to the multi-channel labeled radar echo signal sequence array.

[0092] Furthermore, the system is also used to implement the following functions:

[0093] Multiple sets of radar echo signals with matching signal identifiers and corresponding sets of multidimensional features of these signals are obtained. These sets are then aggregated according to a preset aggregation scale. Based on the aggregation results, the multidimensional feature sets are mapped and aggregated to obtain multiple aggregated sets of multidimensional feature sets. In-set prototype extraction is performed on each of these aggregated sets to obtain multiple multidimensional feature prototypes. Finally, these prototypes are summarized to construct the multi-layer sensing matching prototype library.

[0094] Furthermore, the system is also used to implement the following functions:

[0095] The mean value of the multidimensional features of the radar echo signals of the multiple aggregated sample matching signals is calculated by traversing the multiple aggregated sample matching signal identifier multidimensional feature sets. According to a preset iteration scale, the mean value of the multidimensional features of the multiple aggregated sample matching signal identifier radar echo signals is randomly iterated within the multiple aggregated sample matching signal identifier radar echo signal multidimensional feature sets to obtain multiple iterative aggregated sample matching signal identifier radar echo signal multidimensional features. When the iteration density of the multiple iterative aggregated sample matching signal identifier radar echo signal multidimensional features is greater than or equal to the iteration density of the mean value of the multiple aggregated sample matching signal identifier radar echo signal multidimensional features, the multiple iterative aggregated sample matching signal identifier radar echo signal multidimensional features are randomly iterated according to the preset iteration scale until the preset number of iterations is met, thus obtaining the prototype of the multiple matching signal identifier radar echo signal multidimensional features.

[0096] Furthermore, the system is also used to implement the following functions:

[0097] The low-pass filter channel in the dual-channel signal processing is invoked to perform low-pass filtering on the multi-channel identifiable radar echo signals with the same matching signal identifier in the multi-channel identifiable radar echo signal sequence array, and the IW filter channel in the dual-channel signal processing is invoked to perform difference preservation processing on the multi-channel identifiable radar echo signals with the different matching signal identifier in the multi-channel identifiable radar echo signal sequence array, thereby obtaining the multi-channel identifiable radar echo signal sequence array.

[0098] Furthermore, the system is also used to implement the following functions:

[0099] Range compression and azimuth focusing are performed on each multi-channel labeled radar echo signal sequence in the multi-channel labeled radar echo signal sequence array to obtain a multi-channel labeled radar echo signal sequence array. The multi-channel labeled radar echo signal sequence array is then spliced ​​and fused within the sequence, and the fused signal is converted into a two-dimensional spatial image using the back projection method to obtain the multi-channel regional SAR image set.

[0100] Furthermore, the system is also used to implement the following functions:

[0101] Image features are extracted from the multi-channel region SAR image set using an image feature extractor to obtain a multi-channel region SAR image feature set. The multi-channel region SAR image feature sets are then combined pairwise to obtain a multi-channel region SAR image feature combination set. A first multi-channel region SAR image feature combination is extracted from the multi-channel region SAR image feature combination set and subjected to multi-channel cross-enhancement fusion to obtain a first multi-channel region SAR image cross-enhancement fusion feature combination. This first multi-channel region SAR image cross-enhancement fusion feature combination is added to the multi-channel region SAR image cross-enhancement fusion feature combination set. Using the multi-channel region SAR image feature set as an index, the mean values ​​of multiple multi-channel region SAR image cross-enhancement fusion features corresponding to the same multi-channel region SAR image feature in the multi-channel region SAR image cross-enhancement fusion feature combination set are averaged to obtain a multi-channel region SAR image cross-enhancement fusion feature mean set. Based on the multi-channel region SAR image cross-enhancement fusion feature mean set, the multi-channel region SAR image set is inverted for image enhancement to obtain an enhanced multi-channel region SAR image set. Finally, the enhanced multi-channel region SAR image set is subjected to image mean filtering to obtain the target region SAR image.

[0102] Furthermore, the system is also used to implement the following functions:

[0103] Image feature similarity analysis is performed on the feature combination of the first multi-channel region SAR image to obtain a set of image feature similarity coefficients; softmax processing is performed on the set of image feature similarity coefficients, and a cross-enhancement fusion matrix is ​​constructed based on the processing results; image feature enhancement is performed on the feature combination of the first multi-channel region SAR image based on the cross-enhancement fusion matrix to obtain the cross-enhancement fusion feature combination of the first multi-channel region SAR image.

[0104] Example 3: Based on the same inventive concept as the micro SAR imaging method in the foregoing examples, this example provides a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the micro SAR imaging method in this application. The processor executes various functional applications and data processing of the computer device by running the software programs, instructions, and modules stored in the memory, thereby implementing the aforementioned micro SAR imaging method.

[0105] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0106] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0107] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A miniature SAR imaging method, characterized in that, The method includes: A multi-channel receiver array was configured for the target micro platform, and a micro SAR radar sensor was selected for integration. The integrated miniature SAR radar scans the target area according to a preset scanning path, and receives radar echo signals in parallel through a multi-channel receiving array to obtain a multi-channel radar echo signal sequence array. Based on the multidimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, intra-sequence heterogeneous separation identification is performed for each multi-channel radar echo signal sequence to obtain a multi-channel identified radar echo signal sequence array. The identification corresponding to each multi-channel identified radar echo signal is a homogeneous signal identification and a heterogeneous signal identification. Signals with background area characteristics and dominated by low-frequency information are marked as homogeneous signals and assigned homogeneous signal identification. Signals with target area characteristics and dominated by high-frequency information are marked as heterogeneous signals and added heterogeneous signal identification. Based on the same-match signal identifier and the different-match signal identifier, the signal processing dual channel is invoked to perform same-match low-pass filtering and different-match difference preservation processing on the multi-channel identifier radar echo signal sequence array to obtain a multi-channel identifier filtered radar echo signal sequence array. The different-match difference preservation processing is as follows: for the sequence with different-match signal identifier, the difference preservation channel is triggered, and the singular value decomposition method is used to decompose the matrix constructed by the different-match signal, extract the feature vector with the larger corresponding singular value, strengthen the key information that is different from the same-match signal, and highlight the difference. By using synthetic aperture radar processing methods, the multi-channel identifier filter radar echo signal sequence array is traversed to perform intra-sequence information fusion to obtain a multi-channel regional SAR image set. The multi-channel regional SAR image set is subjected to multi-channel cross-enhancement fusion of image features to obtain the target region SAR image.

2. The miniature SAR imaging method as described in claim 1, characterized in that, Based on the multi-dimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, intra-sequence heterogeneous separation identification is performed for each multi-channel radar echo signal sequence to obtain a multi-channel identified radar echo signal sequence array. Each multi-channel identified radar echo signal corresponds to both a homogeneous signal identifier and a heterogeneous signal identifier, including: According to the preset signal multidimensional features, feature extraction is performed on each multi-channel radar echo signal in the first multi-channel radar echo signal sequence of the multi-channel radar echo signal sequence array to obtain the first multi-channel radar echo signal multidimensional feature sequence. The preset signal multidimensional features include frequency domain features, time domain features, Doppler frequency shift features, and spatial direction features. A multi-layer sensing matching prototype library is obtained. The multi-dimensional features of each multi-channel radar echo signal in the first multi-channel radar echo signal multi-dimensional feature sequence are similar to each multi-layer sensing matching prototype in the multi-layer sensing matching prototype library. When the similarity of the similarity recognition result is greater than or equal to the preset similarity recognition result threshold, the corresponding multi-channel radar echo signal is marked as a matching signal to obtain the first multi-channel matching signal marked radar echo signal subsequence. When the similarity recognition result does not have a similarity greater than or equal to the preset similarity recognition result threshold, the corresponding multi-channel radar echo signal is marked with a different matching signal to obtain the first multi-channel different matching signal marked radar echo signal subsequence. The first multi-channel same-match signal identification radar echo signal subsequence and the first multi-channel different-match signal identification radar echo signal subsequence are fused in chronological order from front to back, and the fused first multi-channel identification radar echo signal sequence is added to the multi-channel identification radar echo signal sequence array.

3. The miniature SAR imaging method as described in claim 2, characterized in that, Obtain the multi-layer perception matching prototype library, including Multiple sets of radar echo signals with matching signal identifiers and corresponding sets of multidimensional features of radar echo signals with matching signal identifiers are obtained. The sets of radar echo signals with matching signal identifiers are aggregated according to a preset aggregation scale. Based on the aggregation results, the sets of multidimensional features of radar echo signals with matching signal identifiers are mapped and aggregated to obtain multiple aggregated sets of multidimensional features of radar echo signals with matching signal identifiers. The prototypes within the sets of the multiple aggregated sample radar echo signal multidimensional feature sets are extracted to obtain multiple prototypes of the radar echo signal multidimensional features. The multidimensional feature prototypes of the radar echo signals identified by the multiple matching signals are summarized to construct the multilayer sensing matching prototype library.

4. The miniature SAR imaging method as described in claim 3, characterized in that, For each of the multiple aggregated sample radar echo signal multidimensional feature sets with matching signal identifiers, prototype extraction within the set is performed to obtain multiple prototype multidimensional features of radar echo signals with matching signal identifiers, including: The mean value of the multidimensional features of the radar echo signals identified by the multiple aggregated sample is calculated by traversing the set of multidimensional features of the radar echo signals identified by the multiple aggregated sample. According to a preset iteration scale, the mean value of the multidimensional features of the radar echo signals identified by the multiple aggregated sample matching signals is randomly iterated in the set of multidimensional features of the radar echo signals identified by the multiple aggregated sample matching signals to obtain the multidimensional features of the radar echo signals identified by the multiple iterated aggregated sample matching signals. When the iteration density of the multidimensional features of the radar echo signal identified by the multiple iterative aggregated samples is greater than or equal to the iteration density of the average of the multidimensional features of the radar echo signal identified by the multiple aggregated samples, the multidimensional features of the radar echo signal identified by the multiple iterative aggregated samples continue to be randomly iterated according to the preset iteration scale until the preset number of iterations is met, and the prototype of the multidimensional features of the radar echo signal identified by the multiple iterative aggregated samples is obtained.

5. The miniature SAR imaging method as described in claim 1, characterized in that, The low-pass filter channel in the dual-channel signal processing is invoked to perform low-pass filtering on the multi-channel identifiable radar echo signals with the same matching signal identifier in the multi-channel identifiable radar echo signal sequence array, and the IW filter channel in the dual-channel signal processing is invoked to perform difference preservation processing on the multi-channel identifiable radar echo signals with the different matching signal identifier in the multi-channel identifiable radar echo signal sequence array, thereby obtaining the multi-channel identifiable radar echo signal sequence array.

6. The miniature SAR imaging method as described in claim 1, characterized in that, By employing synthetic aperture radar (SAR) processing methods, the multi-channel identifier-filtered radar echo signal sequence array is traversed to perform intra-sequence information fusion, resulting in a multi-channel regional SAR image set, including: Each multi-channel labeled radar echo signal sequence in the multi-channel labeled radar echo signal sequence array is subjected to range compression and azimuth focusing processing to obtain a multi-channel labeled preprocessed radar echo signal sequence array. The multi-channel identifier filtering preprocessed radar echo signal sequence array is spliced ​​and fused within the sequence, and the fused signal is converted into a two-dimensional spatial image by back projection method to obtain the multi-channel regional SAR image set.

7. A miniature SAR imaging method as described in claim 1, characterized in that, The multi-channel SAR image set of the multi-channel region is subjected to multi-channel cross-enhancement fusion of image features to obtain the SAR image of the target region, including: Image feature extractors are used to extract image features from the multi-channel region SAR image set to obtain a multi-channel region SAR image feature set. The multi-channel region SAR image feature set is enumerated and combined in pairs to obtain the multi-channel region SAR image feature combination set. Extract the first multi-channel region SAR image feature combination from the multi-channel region SAR image feature combination set, perform multi-channel cross-enhancement fusion of image features, obtain the first multi-channel region SAR image cross-enhancement fusion feature combination, and add the first multi-channel region SAR image cross-enhancement fusion feature combination to the multi-channel region SAR image cross-enhancement fusion feature combination set. Using the multi-channel region SAR image feature set as an index, the mean value processing is performed on multiple multi-channel region SAR image cross-enhancement fusion features corresponding to the same multi-channel region SAR image feature in the multi-channel region SAR image cross-enhancement fusion feature combination set to obtain the mean set of multi-channel region SAR image cross-enhancement fusion features. Based on the mean set of cross-enhancement and fusion features of the multi-channel region SAR images, the multi-channel region SAR image set is inverted to enhance the image, thereby obtaining an enhanced multi-channel region SAR image set. The enhanced multi-channel regional SAR image set is subjected to image mean filtering to obtain the target region SAR image.

8. A miniature SAR imaging method as described in claim 7, characterized in that, Extracting a first multi-channel region SAR image feature combination from the multi-channel region SAR image feature combination set and performing multi-channel cross-enhancement fusion of image features to obtain a first multi-channel region SAR image cross-enhancement fusion feature combination, including: Image feature similarity analysis is performed on the combination of SAR image features in the first multi-channel region to obtain a set of image feature similarity coefficients; The image feature similarity coefficient set is subjected to softmax processing, and a cross-enhancement fusion matrix is ​​constructed based on the processing result; Based on the cross-enhancement fusion matrix, image feature enhancement is performed on the SAR image feature combination of the first multi-channel region to obtain the cross-enhancement fusion feature combination of the SAR image of the first multi-channel region.

9. A miniature SAR imaging system, characterized in that, The system is used to implement a miniature SAR imaging method according to any one of claims 1-8, the system comprising: The array configuration module is used to configure a multi-channel receiving array for the target micro platform and to select and integrate a micro SAR radar sensor. The echo signal receiving module is used to scan the target area according to a preset scanning path using the integrated miniature SAR radar, and to receive radar echo signals in parallel through a multi-channel receiving array to obtain a multi-channel radar echo signal sequence array. The heterogeneous signal separation and identification module is used to perform intra-sequence heterogeneous signal separation and identification of each multi-channel radar echo signal sequence based on the multi-dimensional signal characteristics of each multi-channel radar echo signal sequence in the multi-channel radar echo signal sequence array, thereby obtaining a multi-channel identified radar echo signal sequence array. The identification corresponding to each multi-channel identified radar echo signal is a homogeneous signal identification and a heterogeneous signal identification. Signals with background area characteristics and dominated by low-frequency information are marked as homogeneous signals and assigned homogeneous signal identification, while signals that present target area characteristics and are dominated by high-frequency information are marked as heterogeneous signals and added heterogeneous signal identification. The echo signal sequence array acquisition module is used to perform low-pass filtering on the multi-channel identifiable radar echo signal sequence array based on the same-match signal identifier and the different-match signal identifier, and to perform difference preservation processing on the different-match signal identifier. The result is a multi-channel identifiable radar echo signal sequence array. The difference preservation processing is performed by triggering the difference preservation channel for the sequence with the different-match signal identifier, using the singular value decomposition method to decompose the matrix constructed by the different-match signal, extracting the feature vector with the larger corresponding singular value, strengthening the key information that is different from the same-match signal, and highlighting the difference. The SAR image set acquisition module is used to perform intra-sequence information fusion by traversing the multi-channel identifier-filtered radar echo signal sequence array through synthetic aperture radar processing method to obtain a multi-channel regional SAR image set. The image fusion module is used to perform multi-channel cross-enhancement fusion of image features on the multi-channel region SAR image set to obtain the target region SAR image.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a micro SAR imaging method as described in any one of claims 1-8.

Images