A method and system for processing electromagnetic micro-acoustic echo signals against metal clutter

Abstract

The application provides an anti-metal clutter electromagnetic micro-acoustic echo signal processing method and system. The method comprises: in response to receiving a query signal, generating electromagnetic micro-acoustic echo signals of the same target in continuous multiple transmission periods; constructing a complex observation matrix according to the electromagnetic micro-acoustic echo signals, constructing a real matrix based on the complex observation matrix, and performing low-rank-sparse decomposition on the real matrix to obtain a decomposition result; configuring a decomposition parameter, and generating a sparse amplitude map according to the decomposition parameter and the decomposition result; taking a unit to be detected in the sparse amplitude map as the center, setting a protection unit and a training unit on the left and right sides thereof respectively, and adopting a bilateral CA-CFAR mode for adaptive detection and local consistency secondary decision to determine the output target peak. The application solves the technical problem that in a strong metal reflection environment, it is difficult to effectively suppress multipath clutter, residual false peaks and non-target disturbances, resulting in unstable target multi-peak extraction and poor recognition and positioning reliability.

Description

Technical Field

[0001] This application belongs to the field of signal processing technology, and in particular relates to the field of electromagnetic micro-acoustic echo signal processing and passive identification and positioning technology. Background Technology

[0002] In current industrial settings, the identification and localization of ladles, converters, and other targets often employ visual, laser, or conventional wireless methods. However, these methods are prone to failure in environments with high temperatures, dust, smoke, severe obstruction, and strong metal reflection. In contrast, passive identification methods based on surface acoustic wave (SAW) tags or micro-acoustic identification chips offer advantages such as being passive, heat-resistant, coded, and suitable for harsh environments, making them ideal for target identification and localization in complex industrial environments.

[0003] Existing systems of this type typically include a query signal transmitting device, transceiver antennas, a receiving sampling module, a signal processing module, and a passive SAW tag placed on the target. The signal transmission relationship is generally as follows: the reader transmits a query signal, the tag receives it and generates a multi-peak echo with a specific time delay structure, the echo returns to the reader via the antenna, and the reader then performs matched filtering, pulse compression, peak detection, and identification / location processing on the echo.

[0004] The closest existing technology to this patent application involves directly performing threshold detection or constant false alarm rate (CFAR) detection after the echo has undergone matched filtering and pulse compression. This type of method extracts the target peak by setting training and guard units on both sides of the target cell and constructing a detection threshold using local background statistics. This method has a relatively simple structure and is effective when the background is relatively uniform.

[0005] However, in environments with strong metallic reflection, such as steel mills and smelters, walls, floors, supports, pipelines, and large equipment generate strong and multiple reflections, creating significant multipath clutter. In this situation, the effective multi-peak echo of the target tag will be superimposed with numerous clutter peaks, sidelobe tails, and non-uniform background undulations. This makes the training unit susceptible to abnormal peaks and strong clutter contamination, raising or distorting the detection threshold, leading to missed and false detections, and affecting subsequent template matching and positioning accuracy.

[0006] Another similar existing technique utilizes low-rank sparse decomposition methods for background suppression of multipulse echo data. This type of method organizes the multipulse echoes into a matrix, taking advantage of the strong cross-pulse correlation of background clutter to separate it into low-rank components, while retaining the target peak as a sparse component. This method can reduce structured background clutter to a certain extent.

[0007] However, even after using only low-rank sparse decomposition, residual spurious peaks, local anomalous peaks, and non-target perturbations may still exist, making it difficult to directly obtain stable and reliable target peak extraction results. Furthermore, some existing methods rely on complex scene modeling or propagation of prior parameters, while industrial environments are complex and frequently changing, making practical applications quite challenging.

[0008] Therefore, there is an urgent need to provide a new electromagnetic microacoustic echo signal processing method that can suppress structured background clutter in multi-pulse data and robustly detect and screen target multi-peaks without relying on precise modeling of complex environments, thereby improving the accuracy and reliability of target identification and positioning in complex industrial environments. Summary of the Invention

[0009] This application provides an anti-metal clutter electromagnetic micro-acoustic echo signal processing method and system to solve the technical problem that existing echo processing methods based on threshold detection or low-rank sparse decomposition are difficult to effectively suppress multipath clutter, residual spurious peaks and non-target disturbances in strong metal reflection environments, resulting in unstable target multi-peak extraction and poor reliability of identification and positioning.

[0010] In a first aspect, embodiments of this application provide a method for processing electromagnetic microacoustic echo signals against metallic clutter, comprising: in response to receiving a query signal, generating electromagnetic microacoustic echo signals of the same target in multiple consecutive transmission cycles; constructing a complex observation matrix based on the electromagnetic microacoustic echo signals, constructing a real matrix based on the complex observation matrix, and performing low-rank sparse decomposition on the real matrix to obtain a decomposition result; configuring decomposition parameters, and generating a sparse amplitude map based on the decomposition parameters and the decomposition result; taking the unit to be detected in the sparse amplitude map as the center, setting a protection unit and a training unit on its left and right sides respectively, and using a bilateral CA-CFAR method for adaptive detection and local consistency secondary decision-making to determine the output target peak.

[0011] In one implementation of the first aspect, constructing a complex observation matrix based on the electromagnetic microacoustic echo signal includes: performing matched filtering and pulse compression on the electromagnetic microacoustic echo signal for each transmission cycle to obtain a corresponding time delay vector; stacking multiple time delay vectors to form a multi-pulse complex observation matrix, wherein the row index of the complex observation matrix corresponds to a distance or time delay unit, and the column index corresponds to a pulse number.

[0012] In one implementation of the first aspect, constructing a real matrix based on the complex observation matrix includes: dividing and concatenating the real and imaginary parts of the complex observation matrix into blocks to construct a real matrix equivalent to the complex observation matrix.

[0013] In one implementation of the first aspect, the low-rank sparse decomposition of the real matrix to obtain the decomposition result includes: performing GoDec decomposition on the real matrix to decompose the real matrix into low-rank terms, sparse terms, and residual terms; the low-rank terms are used to characterize quasi-static background clutter, multipath trails, and structured interference formed by fixed reflectors with high correlation across pulses; the sparse terms are used to preserve significant peak structures in the target echo; and the residual terms are used to characterize noise and unmodeled random disturbances.

[0014] In one implementation of the first aspect, the configuration decomposition parameters include: configuring the rank parameter and sparsity parameter of the GoDec decomposition; wherein, the rank parameter of the GoDec decomposition is configured according to the principal component energy ratio of the multi-pulse background, and the sparsity parameter is configured according to the target peak number, the effective width of the single-peak main lobe, and the number of pulses.

[0015] In one implementation of the first aspect, generating a sparse amplitude map based on the decomposition parameters and the decomposition result includes: restoring the sparse terms output by GoDec to a complex sparse matrix based on the decomposition parameters and the decomposition result, and calculating its amplitude to obtain a sparse amplitude map after background separation.

[0016] In one implementation of the first aspect, the adaptive detection using the bilateral CA-CFAR method includes: estimating the local background mean using the bilateral CA-CFAR method, and calculating the adaptive detection threshold of the current unit under test based on a preset false alarm rate; when the amplitude of the unit under test is higher than the detection threshold, it is determined to have passed the primary detection and is used as a candidate target peak for output; when the amplitude of the unit under test is lower than the detection threshold, it is determined to be background clutter or noise.

[0017] In one implementation of the first aspect, the local consistency secondary decision includes: determining the output target peak by combining at least one of the following information with the candidate target peak: local mean in the neighborhood of the candidate peak, local fluctuation degree, peak significance, cross-pulse consistency, and degree of matching with the preset target multi-peak template.

[0018] In one implementation of the first aspect, the output target peak includes the output target peak position, peak amplitude, and arrival time information.

[0019] Secondly, embodiments of this application provide an anti-metallic clutter electromagnetic micro-acoustic echo signal processing system, comprising: an echo signal module, a query signal transmitting and receiving module, and a signal processing module; the echo signal module, in response to receiving a query signal, acquires electromagnetic micro-acoustic echo signals of the same target in multiple consecutive transmission cycles; the query signal transmitting and receiving module transmits a query signal to the echo signal module and receives the electromagnetic micro-acoustic echo signals from the echo signal module; the signal processing module constructs a complex observation matrix based on the electromagnetic micro-acoustic echo signals, constructs a real matrix based on the complex observation matrix, and performs low-rank sparse decomposition on the real matrix to obtain the decomposition result; configures decomposition parameters, and generates a sparse amplitude map based on the decomposition parameters and the decomposition result; with the unit to be detected in the sparse amplitude map as the center, a protection unit and a training unit are respectively set on its left and right sides, and a bilateral CA-CFAR method is used for adaptive detection and local consistency secondary decision to determine the output target peak.

[0020] The anti-metal clutter electromagnetic micro-acoustic echo signal processing method and system provided in this application have the following beneficial effects:

[0021] This application effectively separates structured clutter from target echoes in strong metallic environments through low-rank sparse decomposition, significantly improving the contrast between the target and the background without requiring complex environment modeling. Furthermore, it incorporates adaptive constant false alarm rate (CFAR) detection, avoiding clutter contamination threshold estimation and reducing the probability of missed and false detections in non-uniform backgrounds. Further, through local consistency-based secondary decision-making, isolated spurious peaks and noise disturbances are eliminated, making the output target peak more stable and clear. This provides a high-quality data foundation for subsequent template matching and localization calculations, effectively solving the technical problem that existing echo processing methods based on threshold detection or low-rank sparse decomposition struggle to effectively suppress multipath clutter, residual spurious peaks, and non-target disturbances in strong metallic reflection environments, leading to unstable target multi-peak extraction and poor reliability in identification and localization. Attached Figure Description

[0022] Figure 1 The diagram shown is an overall flowchart of an embodiment of the electromagnetic microacoustic echo signal processing method for resisting metal clutter in this application.

[0023] Figure 2 The diagram shown illustrates the principle of constructing a complex observation matrix in an anti-metallic clutter electromagnetic microacoustic echo signal processing method according to an embodiment of this application.

[0024] Figure 3 The diagram shown illustrates the principle of adaptive detection using a bilateral CA-CFAR method in an embodiment of this application for processing electromagnetic micro-acoustic echo signals against metal clutter.

[0025] Figure 4The diagram shows an implementation process of an anti-metallic noise electromagnetic micro-acoustic echo signal processing method according to an embodiment of this application.

[0026] Figure 5 The diagram shown is a schematic diagram of the principle structure of an anti-metallic clutter electromagnetic micro-acoustic echo signal processing system according to an embodiment of this application.

[0027] Component designation explanation

[0028] 100 Anti-metallic noise electromagnetic micro-acoustic echo signal processing system 110 echo signal module 120 Query signal transmission and reception module 130 Signal processing module S100~S400 step Detailed Implementation

[0029] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0030] Before providing a detailed description of this embodiment, the nouns and terms used in this embodiment are explained, and the nouns and terms used in this embodiment are subject to the following interpretations:

[0031] SAW stands for Surface Acoustic Wave.

[0032] CFAR stands for Constant False Alarm Rate.

[0033] CA-CFAR stands for Cell Averaging Constant False Alarm Rate.

[0034] GoDec is a low-rank sparse matrix factorization algorithm.

[0035] Existing methods for identifying and locating industrial targets such as steel ladles and converters often employ vision, laser, or conventional wireless methods, but these are prone to failure in high-temperature, dusty, and strong metal reflection environments. While passive identification methods based on SAW tags have advantages, their echo processing often relies on threshold or constant false alarm rate (CFAR) detection, which can easily lead to threshold distortion due to contamination of training units under strong multipath clutter. Alternatively, low-rank sparse decomposition may be used, but residual spurious peaks remain, making it difficult to stably extract target peaks. This embodiment provides a fast query method based on a combination of linear frequency modulation and time inversion to address the technical problem that existing echo processing methods based on threshold detection or low-rank sparse decomposition struggle to effectively suppress multipath clutter, residual spurious peaks, and non-target disturbances in strong metal reflection environments, resulting in unstable target multi-peak extraction and poor reliability in identification and location.

[0036] Specifically, the anti-metal clutter electromagnetic micro-acoustic echo signal processing method and system provided in this embodiment is an anti-metal clutter signal processing method and system for electromagnetic micro-acoustic echoes in industrial environments. Its principle is as follows: First, for the echo data after multi-pulse compression, a complex observation matrix is ​​constructed, arranging each pulse echo in columns to retain amplitude and phase information. Next, the matrix is ​​decomposed into low-rank sparse components, utilizing the strong cross-pulse correlation of background clutter to separate it into low-rank components, while retaining the multi-peak echoes of the target label as sparse components. Then, a sparse amplitude map is generated based on the decomposed sparse components, further increasing the contrast between the target peak and the residual background. On this basis, bilateral CA-CFAR adaptive detection is adopted, setting protection and training units on both sides of the unit to be detected, generating a detection threshold based on local background statistical results, and effectively extracting candidate target peaks. Finally, through local consistency secondary decision-making, combining information such as neighborhood mean, local fluctuation degree, peak significance, and cross-pulse consistency, isolated pseudo-peaks and local abnormal disturbances are eliminated, thereby outputting a stable and reliable target multi-peak position. The entire process does not rely on precise modeling of complex industrial environments, and can significantly suppress strong metal reflections and multipath clutter interference, thereby improving the accuracy of target identification and positioning.

[0037] The following will refer to the appendices in the embodiments of this application. Figure 1 To be continued Figure 5 This application provides a detailed description of the technical solutions for the anti-metal clutter electromagnetic micro-acoustic echo signal processing method and system in the embodiments of this application. This allows those skilled in the art to understand and implement the anti-metal clutter electromagnetic micro-acoustic echo signal processing method and system of this embodiment without inventive effort.

[0038] This embodiment provides a method for processing electromagnetic micro-acoustic echo signals against metal clutter. It is a method based on low-rank sparse decomposition and constant false alarm rate detection, which is applicable to steel mills, smelters and other industrial scenarios with dense metal structures. It is applied to wireless passive SAW identification chips or SAW tags, and is used for clutter suppression, target peak extraction, identification and location calculation of echo signals from wireless passive surface acoustic wave tags or micro-acoustic identification chips. Figure 1 The flowchart shown is a process for processing electromagnetic microacoustic echo signals against metal clutter, as described in an embodiment of this application. Figure 1 As shown, the anti-metallic noise electromagnetic micro-acoustic echo signal processing method provided in this application includes the following steps S100 to S400.

[0039] Step S100: In response to receiving the query signal, generate electromagnetic microacoustic echo signals of the same target in multiple consecutive transmission cycles.

[0040] Step S200: Construct a complex observation matrix based on the electromagnetic micro-acoustic echo signal, construct a real matrix based on the complex observation matrix, and perform low-rank sparse decomposition on the real matrix to obtain the decomposition result;

[0041] Step S300: Configure decomposition parameters and generate a sparse amplitude map based on the decomposition parameters and the decomposition results;

[0042] Step S400: Taking the unit to be inspected in the sparse amplitude map as the center, a protection unit and a training unit are set on its left and right sides respectively. Adaptive detection and local consistency secondary decision are performed using the bilateral CA-CFAR method to determine the target peak of the output.

[0043] The method in this embodiment can balance query speed, estimation accuracy, and stability even when the resonant frequency has an unknown shift. First, a linear frequency-modulated query signal covering a preset query frequency band is generated. A single broadband excitation is sufficient to adapt to scenarios where the resonant frequency of the surface acoustic wave sensor is unknown or drifting, eliminating the need for point-by-point frequency sweeping. After transmitting the query signal to the sensor and receiving the first echo, time-reversal processing is performed on the first echo to construct an adaptive secondary transmission signal. This coherently enhances the resonant-related frequency components, effectively improving the echo signal-to-noise ratio and frequency component contrast. The secondary transmission signal is then transmitted to the sensor to acquire the second echo. Finally, frequency domain transformation and peak search are performed on the second echo, allowing the sensor resonant frequency estimate to be obtained with only one frequency domain processing step. The entire process eliminates the need for multiple transmissions and repetitive frequency domain calculations, achieving fast broadband querying while significantly improving frequency estimation accuracy and stability, shortening the measurement cycle, and making it more suitable for dynamic measurement scenarios.

[0044] The following combination Figures 2 to 4The steps S100 to S400 of the anti-metallic clutter electromagnetic microacoustic echo signal processing method in this embodiment will be described in detail.

[0045] Step S100: In response to receiving the query signal, generate electromagnetic microacoustic echo signals of the same target in multiple consecutive transmission cycles.

[0046] In this embodiment, a software-defined radio device, such as the USRP-B200, can be used to generate the radio frequency query signal.

[0047] The software-defined radio device transmits a preset radio frequency query signal to the target area. Upon receiving the query signal, a passive SAW tag or micro-acoustic identification chip installed on the target generates an electromagnetic micro-acoustic echo containing multiple reflection peaks. During its propagation in the industrial field, this echo is superimposed with multipath clutter and non-stationary background noise from walls, the ground, metal supports, equipment surfaces, and other reflectors.

[0048] The purpose of this step is to acquire raw observation data, including the target's valid echo, strong environmental clutter, and random noise, to provide input for subsequent signal processing.

[0049] Step S200: Construct a complex observation matrix based on the electromagnetic micro-acoustic echo signal, construct a real matrix based on the complex observation matrix, and perform low-rank sparse decomposition on the real matrix to obtain the decomposition result.

[0050] like Figure 2 As shown, in one implementation of this embodiment, constructing a complex observation matrix based on the electromagnetic micro-acoustic echo signal includes: performing matched filtering and pulse compression on the electromagnetic micro-acoustic echo signal for each transmission cycle to obtain a corresponding time delay vector; stacking multiple time delay vectors to form a multi-pulse complex observation matrix, wherein the row index of the complex observation matrix corresponds to a distance or time delay unit, and the column index corresponds to a pulse number.

[0051] Existing anti-clutter methods often rely on scene geometry modeling, propagation path prediction, or prior reflection parameters. However, industrial environments are typically characterized by dense equipment, complex structures, frequent state changes, and difficulty in accurately obtaining parameters, making it challenging to deploy such methods in practice.

[0052] This embodiment does not rely on precise industrial environment modeling. Instead, it directly processes the multi-pulse echo data acquired at the front end and after matching filtering / pulse compression, achieving background separation and target peak extraction through a data-driven approach.

[0053] The direct technical effect of matching filtering and pulse compression of the electromagnetic microacoustic echo signal in each transmission cycle is that the method has good adaptability to scene changes. Even when the field equipment moves, the reflection conditions change, or the environmental parameters are difficult to measure accurately, it can still maintain good clutter suppression and target detection capabilities. Therefore, it is more suitable for promotion and application in steel plants, smelters and similar complex metal environments.

[0054] Specifically, in one implementation embodiment, the process of constructing a complex observation matrix based on the electromagnetic microacoustic echo signal is as follows:

[0055] First, the electromagnetic microacoustic echo signals of the same target are acquired within multiple consecutive transmission cycles. Specifically, using a detection system such as radar or sonar, the corresponding electromagnetic microacoustic echo signals are received within multiple consecutive transmission cycles for the same target. The "multiple consecutive transmission cycles" refer to several transmit-receive cycles arranged sequentially according to the repeat interval of the transmit pulses. The number of these cycles can be set according to actual detection requirements, such as 16, 32, or 64 cycles.

[0056] Secondly, in this embodiment, to improve processing robustness, multiple sets of echo data from the same target across several consecutive transmission cycles are typically acquired and subjected to matched filtering and pulse compression respectively. This processing enhances the time delay resolution of the echoes, making the target's multi-peak structure and clutter peak structure clearer on the time delay axis, laying the foundation for subsequent matrix construction and background separation. Specifically, matched filtering and pulse compression are performed on the electromagnetic microacoustic echo signals received in each transmission cycle. Matched filtering uses a filter that matches the transmitted signal waveform to maximize the signal-to-noise ratio of the received signal; pulse compression converts wide pulse signals into narrow pulse signals through time-domain or frequency-domain compression, thereby achieving high resolution along the range dimension (or time delay dimension). After matched filtering and pulse compression, a time delay vector is obtained for each transmission cycle. This time delay vector reflects the energy distribution of the target echo and clutter along the range or time delay dimension. Each element in the time delay vector represents the complex amplitude of the echo at different time delays (i.e., different range units), and its length is determined by the number of range gates.

[0057] Then, following the two-dimensional organization of range cell-pulse number, the time delay vectors corresponding to multiple transmission cycles are stacked to form a complex observation matrix X. Specifically, the row index of each time delay vector corresponds to a range cell or time delay cell, and the column index corresponds to the pulse number of the transmission cycle (i.e., the slow time dimension). Thus, the element in the i-th row and j-th column of matrix X represents the complex echo value received in the i-th range cell within the j-th pulse cycle. This matrix is ​​a complex matrix and contains both amplitude and phase information of the echo.

[0058] Since fixed reflective backgrounds in industrial scenarios typically exhibit high correlation across multi-pulse dimensions, background clutter manifests as structured and repetitive components in matrix X. In contrast, target multi-peak echoes occupy only a limited number of peak positions in the range dimension, thus exhibiting relatively sparse structural characteristics. Constructing a multi-pulse complex observation matrix transforms the one-dimensional processing problem of single-pulse data into a two-dimensional structured processing problem of multi-pulse data, enabling the background and target to be separable in terms of matrix structure.

[0059] In other embodiments, the echo observation matrix may also be composed of multiple frames of echoes that have been preprocessed, downsampled, or windowed.

[0060] In this embodiment, by constructing the complex observation matrix X, the statistical structure of the target echo and background clutter in the multi-pulse dimension (slow time dimension) can be completely preserved, including the correlation, non-stationarity, and phase change patterns between range units and between pulses. This statistical structure is crucial for subsequent matrix decomposition processing (e.g., robust principal component analysis, nonnegative matrix decomposition, or singular value decomposition): on the one hand, the target signal typically exhibits low-rank or sparse characteristics in the multi-pulse dimension; on the other hand, background clutter and noise exhibit different low-rank or random distribution characteristics. The observation matrix X provides a data foundation for subsequent algorithms that combines range resolution and inter-pulse continuity, thereby facilitating the effective separation of the target's micro-motion features from the static background and slowly varying clutter, ultimately achieving accurate extraction and identification of the target's electromagnetic and acoustic features.

[0061] This embodiment significantly improves the distinguishability of the target and background in multiple dimensions through multi-cycle pulse accumulation and matrix structured organization, laying a reliable data foundation for improving the detection performance and parameter estimation accuracy of the detection system.

[0062] To facilitate matrix factorization in the real domain, in one implementation of this embodiment, constructing a real matrix based on the complex observation matrix includes: concatenating the real and imaginary parts of the complex observation matrix in blocks to construct a real matrix equivalent to the complex observation matrix. This step preserves the complex information of the echoes while meeting the implementation requirements of the low-rank sparse decomposition algorithm.

[0063] This transformation allows for the subsequent low-rank sparse solution to be completed without losing the amplitude and phase information of the original complex echo. This real-valued transformation also transforms the subsequent low-rank sparse decomposition problem into a real-domain solution problem without disrupting the original complex signal structure.

[0064] In one implementation of this embodiment, the low-rank sparse decomposition of the real matrix to obtain the decomposition result includes: performing GoDec decomposition on the real matrix to decompose the real matrix into low-rank terms, sparse terms, and residual terms; the low-rank terms are used to characterize quasi-static background clutter, multipath trails, and structured interference formed by fixed reflectors with high correlation across pulses; the sparse terms are used to preserve significant peak structures in the target echo; and the residual terms are used to characterize noise and unmodeled random disturbances.

[0065] In other embodiments, the low-rank sparse decomposition may employ robust principal component analysis, low-rank matrix recovery, or other low-rank sparse decomposition methods with similar background separation capabilities, in addition to GoDec decomposition.

[0066] This embodiment addresses the issues of strong reflections, multiple reflections, and non-uniform backgrounds caused by metal structures such as walls, floors, supports, pipelines, and large equipment in industrial settings. First, the multi-pulse compression results are processed by low-rank sparse decomposition. Quasi-static background clutter, multipath trails, and structured interference formed by fixed reflectors with strong correlation across pulses are separated into low-rank terms, while the multi-peak echoes corresponding to the target label are retained in the sparse terms.

[0067] This enables the structured separation of background clutter and the effective peak of the target directly from measured echo data without relying on precise modeling of complex industrial environments, solving the problems of strong background clutter, easy obscuring of the target peak, and difficulty in distinguishing clutter from the target in existing technologies.

[0068] The direct technical effect of low-rank sparse decomposition of real matrices is that the background mean in the processed detection input is significantly reduced, large-scale continuous clutter and multipath tails are suppressed, and the contrast between the target peak and the background is significantly improved, thus providing a cleaner data foundation for subsequent stable detection and localization.

[0069] Step S300: Configure decomposition parameters and generate a sparse amplitude map based on the decomposition parameters and the decomposition results.

[0070] The purpose of this step is to separate the structured background clutter from the effective target peak in the original complex observation, thereby reducing the interference of the background on subsequent threshold estimation and target detection.

[0071] In one implementation of this embodiment, the configuration decomposition parameters include: configuring the rank parameter and sparsity parameter of the GoDec decomposition; wherein, the rank parameter of the GoDec decomposition is configured according to the principal component energy ratio of the multi-pulse background, and the sparsity parameter is configured according to the number of target peaks, the effective width of the single-peak main lobe, and the number of pulses, so that the background principal energy preferentially enters the low-rank term, while the target peak principal energy is preferentially retained in the sparse term, thereby realizing the separable representation of the background clutter and the target peak structure.

[0072] In one implementation of this embodiment, generating a sparse amplitude map based on the decomposition parameters and the decomposition result includes: restoring the sparse terms output by GoDec into a complex sparse matrix based on the decomposition parameters and the decomposition result, and calculating its amplitude to obtain a sparse amplitude map after background separation. This step aims to generate input data suitable for constant false alarm rate (CFAR) detection, making the local background statistics on which the detector is based closer to a stationary state, thereby improving the accuracy of the detection threshold estimation.

[0073] Compared to detection directly on the original pulse compression results, this sparse amplitude map has a lower background mean, more prominent target peaks, and suppresses large-scale fluctuations and strong correlation tails. It can reduce the contamination of subsequent threshold estimation by strong clutter, while the multi-peak structure of the target is relatively more prominent.

[0074] Step S400: Taking the unit to be inspected in the sparse amplitude map as the center, a protection unit and a training unit are set on its left and right sides respectively. Adaptive detection and local consistency secondary decision are performed using the bilateral CA-CFAR method to determine the target peak of the output.

[0075] The purpose of this step is to achieve adaptive detection of target peaks under different local background conditions, avoiding the problems of missed detection and false detection caused by using a fixed threshold.

[0076] like Figure 3 As shown, in one implementation of this embodiment, constant false alarm rate (CFAR) primary detection is performed on the sparse amplitude map after low-rank sparse decomposition. The CFAR primary detection adopts bilateral CA-CFAR adaptive detection. The adaptive detection using bilateral CA-CFAR includes: estimating the local background mean using bilateral CA-CFAR, and calculating the adaptive detection threshold of the current test unit based on a preset false alarm rate; when the amplitude of the test unit is higher than the detection threshold, it is determined to pass the primary detection and is used as a candidate target peak for output; when the amplitude of the test unit is lower than the detection threshold, it is determined to be background clutter or noise.

[0077] The purpose of this step is to further reduce false alarms and improve the stability and reliability of target peak extraction.

[0078] The constant false alarm rate (CFAR) detection method is not limited to bilateral CA-CFAR, but can also employ other adaptive CFAR detection methods.

[0079] This embodiment sets up protection units and training units on both sides of the unit to be inspected, performs statistical estimation of the local background, and adaptively generates a detection threshold based on a preset false alarm rate. Compared with existing methods of fixed threshold detection or direct threshold detection on the original pulse compression result, this embodiment first weakens the structured background clutter and then performs constant false alarm rate detection. Therefore, the training unit is closer to the local stable background, which can reduce the lifting effect of abnormal peaks, background fluctuations, and multipath remnants on the threshold estimation. In complex industrial environments, it can effectively reduce the probability of missed detection and false detection, and enable the target multi-peaks to be stably detected under different background conditions, thereby improving the accuracy of echo peak extraction and the reliability of system operation.

[0080] In this embodiment, the number of training units, the number of protection units, the target false alarm rate, and the local consistency decision threshold of the bilateral CA-CFAR can be adaptively configured according to the target distance resolution, the on-site noise level, and the degree of background undulation.

[0081] This embodiment addresses the potential for local fluctuations, outliers, and non-uniform backgrounds in industrial settings. It further combines the neighborhood mean, local fluctuation level, and consistency between adjacent pulses to make a secondary judgment on candidate peaks that pass the initial detection. Peaks that are significantly more prominent than the local background and exhibit stable cross-pulse performance are retained, while isolated spurious peaks and clutter peaks caused by local abnormal fluctuations are eliminated.

[0082] In one implementation of this embodiment, the local consistency secondary decision includes: determining the output target peak by combining at least one of the following information with the candidate target peak: local mean in the neighborhood of the candidate peak, local fluctuation degree, peak significance, cross-pulse consistency, and degree of matching with the preset target multi-peak template.

[0083] Candidate peaks that stand out significantly from the local background and are stable between adjacent pulses are retained; spurious peaks that appear only occasionally in a single pulse or are caused by local abnormal fluctuations are eliminated.

[0084] In this embodiment, the local consistency secondary decision can be implemented by one or more combinations of local mean comparison, local variance comparison, cross-pulse consistency test or template matching score.

[0085] This implementation, based on the constant false alarm rate (CFAR) primary detection, also incorporates a secondary decision step based on local consistency. For candidate peaks that pass the primary detection, a second screening process is performed, incorporating information such as the neighborhood mean, local fluctuation level, peak significance, and cross-pulse consistency. Target peaks that are significantly prominent compared to the local background and exhibit stability across multiple pulse dimensions are retained, while spurious peaks caused solely by local spikes, sporadic noise, or abrupt background changes are eliminated. This addresses the problem in existing technologies where, even after a threshold detection, false alarms are still easily generated due to strong local fluctuations, outliers, and non-uniform background interference.

[0086] The direct technical effect of the local consistency secondary decision is that the final output target peak position is more stable, the peak shape is clearer, and the cross-pulse consistency is better, which is beneficial to subsequent template matching, identity recognition, arrival time estimation and positioning calculation, and improves the engineering application value of the overall system.

[0087] This embodiment employs a combined processing approach of low-rank background separation, adaptive constant false alarm rate (CFAR) detection, and local consistency decision to stably preserve and prominently output the multi-peak echo structure of the target label. The processed peak positions, peak amplitudes, and arrival times are clearer and more stable, which is more beneficial for subsequent multi-peak template matching, identity recognition, and localization calculations.

[0088] Therefore, this embodiment not only improves the clutter suppression effect in the front-end signal processing stage, but also improves the accuracy and stability of subsequent identification and positioning results at the system level, reducing misidentification and positioning errors caused by clutter peak confusion, peak position shift and weak peak submersion.

[0089] In this embodiment, the output target peak can be used for SAW tag localization, as well as for identity recognition, state discrimination, alignment control, or other industrial sensing tasks based on multi-peak echo structures.

[0090] Specifically, in one implementation of this embodiment, the output target peak includes the output target peak position, peak amplitude, and arrival time information. That is, the target peak position, peak amplitude, and arrival time information retained after the above processing are output and used for subsequent multi-peak template matching, identification, target positioning, or alignment control. The purpose of this process is to transform the aforementioned signal processing results into effective output information that can directly serve industrial applications, both for offline data processing and for deployment in online industrial inspection or positioning systems.

[0091] like Figure 4 As shown in the figure, the implementation principle of the anti-metallic clutter electromagnetic micro-acoustic echo signal processing method described in this embodiment is as follows:

[0092] 1) Transmit the query signal and receive the raw echo.

[0093] A preset query signal is transmitted to the target detection area. This query signal can be a linear frequency modulated signal, a phase-coded signal, or other pulse signal with a large time-bandwidth product. Simultaneously, the raw echo signal formed by the reflection of this signal from the target and the background environment is received. This operation acquires raw data including the target's electromagnetic micro-acoustic characteristics, background clutter, and noise.

[0094] 2) Perform matched filtering and pulse compression on the raw echo.

[0095] For the raw echo received in each transmission cycle, matched filtering is performed. The transfer function of the matched filter is matched with the complex conjugate of the transmitted signal to maximize the output signal-to-noise ratio. Subsequently, pulse compression is performed, using time-domain convolution or frequency-domain multiplication-inverse transform to compress the wide pulse echo into a narrow pulse, thereby achieving high resolution in the time delay dimension (i.e., the range dimension). After the above processing, each transmission cycle yields a time delay vector, where the elements correspond to the complex echo amplitude at different range cells.

[0096] 3) Organize the processing results of multiple transmission cycles into a multi-pulse complex observation matrix.

[0097] The time delay vectors of multiple consecutive transmission cycles (e.g., N cycles, N≥2) are accumulated and stacked in a two-dimensional structure of "range cell - pulse number" to form a complex observation matrix X∈ℂ^{M×N}, where M is the total number of range cells and N is the total number of pulses. The row index of the matrix corresponds to the range cell (or time delay cell), and the column index corresponds to the pulse number (i.e., the slow time dimension). This complex observation matrix simultaneously preserves the amplitude and phase information of the echo, reflecting the statistical correlation between the target and the background in multiple pulse dimensions.

[0098] 4) Convert the complex observation matrix to real numbers.

[0099] To accommodate subsequent low-rank sparse decomposition algorithms in the real domain (such as GoDec), the complex observation matrix X is converted into a real matrix X_real. A preferred method for realization is to stack the real and imaginary parts of the complex number separately, or to construct a real matrix with double the number of rows. Specifically, let X = A + jB, where A and B are the real and imaginary part matrices, respectively. Then the realized matrix can be represented as X_real = [A; B] ∈ ℝ^{2M×N}. This process can completely preserve the amplitude and phase relationships of the complex data while satisfying the real input requirements of the decomposition algorithm.

[0100] 5) Perform GoDec low-rank sparse decomposition on the real matrix to obtain the low-rank background term, sparse objective term, and residual term.

[0101] The realized observation matrix X_real is fed into the GoDec (Go Decomposition) decomposer. The GoDec model decomposes the input matrix into three parts: X_real = L + S + R, where L is a low-rank matrix representing background clutter components (which are strongly correlated in the multi-pulse dimension, change slowly over time, and have low rank); S is a sparse matrix representing target echo components (the target exhibits sparse multi-peak or isolated micro-motion features in the range-pulse plane); and R is a residual matrix representing random noise and unmodeled components. During the decomposition process, the low-rank approximation and the hard-threshold sparse approximation are optimized by alternating projections, iterating until convergence. The core function of this operation is to utilize the different statistical structures of the background and the target in the multi-pulse dimension to achieve effective separation between them, eliminating the interference of strong clutter on target detection.

[0102] 6) Restore the sparse terms and generate a sparse amplitude map.

[0103] The sparse matrix S (corresponding to the real space) obtained from the decomposition is restored to a complex form sparse echo matrix S_complex using the inverse process of 4). Then, the modulus (or amplitude) of each element of S_complex is taken to obtain a non-negative two-dimensional amplitude matrix, denoted as A_sparse ∈ ℝ^{M×N}. This matrix is ​​the sparse amplitude map, where the horizontal axis represents the pulse number (slow time), the vertical axis represents the distance cell, and the pixel value represents the effective echo amplitude of the target retained after decomposition. Compared with the original echo amplitude map, the background clutter in the sparse amplitude map has been significantly suppressed, and the target peak is more prominent.

[0104] 7) Perform bilateral CA-CFAR adaptive detection on sparse amplitude maps

[0105] On the sparse amplitude map A_sparse, cell-averaged constant false alarm rate (CA-CFAR) detection is performed along both the distance dimension (row direction) and the slow time dimension (column direction). Specifically, for each cell to be detected, the local noise power is estimated within reference windows on both sides of its distance dimension and slow time dimension, and the estimates from both sides are combined to form an adaptive detection threshold. This two-sided CA-CFAR can adapt to different interference statistical characteristics in the distance and slow time dimensions, effectively extracting candidate target peaks. Local maxima that satisfy the threshold conditions are marked as candidate target peaks.

[0106] 8) Perform a second-order decision on local consistency for candidate peaks.

[0107] For the candidate target peak set output in step 7), a local consistency secondary decision is performed to eliminate isolated false peaks and locally anomalous peaks. Specifically, for each candidate peak, it is checked whether there are at least K other candidate peaks (K≥1) within its preset neighborhood (e.g., within a distance dimension ±1 cell or a slow time dimension ±2 pulses), or the amplitude correlation and phase consistency indicators between the peak and other peaks in the neighborhood are calculated. If the consistency condition is met, the peak is retained; otherwise, it is eliminated as a false detection result. This operation can effectively suppress isolated false alarms caused by sudden noise or decomposition residues, enhancing the stability and reliability of the output target peaks.

[0108] 9) Output the final target peak information and use it for subsequent identification and localization.

[0109] The final target peak information, confirmed in step 8), is output in a structured manner. Each target peak includes at least a range cell index, a pulse index (or the corresponding slow-time position), a normalized amplitude, and optional phase information. The output data is sent to subsequent advanced processing modules, including but not limited to: target micro-motion feature recognition (e.g., micro-Doppler extraction through phase differences between multiple peaks), target spatial localization (combined with range-angle measurements), multi-target tracking, or motion parameter estimation. This completes the entire process of high clutter suppression, adaptive detection, and target information extraction for electromagnetic microacoustic echo signals.

[0110] The dynamic relationships between the various operations can be further summarized as follows:

[0111] Together with 1) and 2) they complete the acquisition of the original echo and high-resolution mapping in the time delay domain, providing the original data basis for the coexistence of the target and background clutter and subsequent separation.

[0112] 3) and 4) transform the time-delay domain data into a matrix form suitable for structured decomposition by using multi-pulse organization and realization processing, while preserving the statistical dependencies in the slow time dimension.

[0113] 5) Utilizing the strong correlation (low rank) of background clutter in the multi-pulse dimension and the sparsity of the target echo, the two are separated into low-rank and sparse terms through GoDec decomposition; this operation is the core of the whole method and directly determines the background suppression capability.

[0114] 6) Restore the sparse terms obtained from the decomposition to the amplitude map, complete the transformation from the complex domain to the detection domain, and make the target peak stand out against the low background.

[0115] 7) Based on the bilateral local statistical characteristics, the detection threshold is adaptively formed to realize the initial extraction of candidate target peaks, taking into account the interference changes in the distance dimension and the slow time dimension.

[0116] 8) Perform a second local consistency judgment on the primary test results to further eliminate isolated false peaks and abnormal peaks, thereby improving the stability and reliability of the test results.

[0117] 9) The enhanced and filtered target peak information is then output to the identification, positioning, or control module, forming a closed loop from data acquisition to target information application.

[0118] Therefore, this embodiment is not simply a combination of several conventional algorithms, but rather, based on the signal characteristics of "strong background correlation, sparse target peaks, and local background non-uniformity" in industrial strong metal clutter environments, it constructs a joint processing flow with clear connections and functional divisions. Through this dynamic processing relationship, the synergistic effect of background clutter suppression, target peak enhancement, and stable detection can be achieved.

[0119] In one non-limiting embodiment, a low-sound identification chip is used as the target tag, and the chip contains five reflective gratings. The query system transmits a query signal with a pulse width of 1 microsecond, a bandwidth of 10 MHz, and a sampling rate of 96 MHz, and the system antenna gain is 6 dBi. After performing matched filtering and pulse compression on the received echo, the front-end system obtains time delay vectors corresponding to multiple transmission cycles, and stacks the processing results of multiple consecutive pulses to form a range-pulse complex observation matrix X.

[0120] The real and imaginary parts of the complex observation matrix X are concatenated to form a real matrix Xr, and GoDec decomposition is performed on Xr to obtain a low-rank term L, a sparse term S, and a residual term G. L primarily characterizes the background clutter and multipath tails that are stable across pulse morphologies, while S mainly retains the target's multi-peak echoes and a small number of significant effective components. A sparsity parameter k is set based on the number of target peaks, the effective width of a single peak, and the number of pulses, and a rank parameter r is set based on the energy proportion of the background principal components, ensuring that the target peaks remain clear in the sparse term after decomposition.

[0121] After restoring the sparse terms to complex matrices, an amplitude map is calculated and used as the detection input. Since the strongly correlated background structure has been separated into low-rank terms, the background noise and large-scale fluctuations in the detection input are significantly reduced, thus providing a more stable local statistical background for subsequent adaptive threshold detection.

[0122] Subsequently, a symmetrical training window and guard window are constructed centered on each cell to be inspected. The local background mean is estimated using bilateral CA-CFAR to form an adaptive threshold. When the amplitude of the cell to be inspected is higher than the threshold, it is recorded as a candidate peak. Then, a secondary decision is made by combining the local mean, fluctuation degree and cross-pulse consistency in the neighborhood of the candidate peak to eliminate isolated spurious peaks and local abnormal peaks.

[0123] After the above processing, the target multi-peak positions can be stably extracted under conditions of strong multipath interference, strong clutter, and non-uniform background. According to experimental results, after processing real industrial field data, this embodiment improves the sidelobe suppression ratio by 8.07 dB, the integrated sidelobe ratio by 5.27 dB, and the clutter suppression ratio by 153.82 dB. This indicates that this embodiment can significantly suppress strong background clutter, multipath tails, and abnormal interference peaks in industrial scenarios, and effectively enhance the discernibility of the target peak structure. This demonstrates that this embodiment is not only feasible in principle but has also exhibited good clutter suppression capability and detection stability in real-world scenarios, possessing strong engineering implementation value and promising industrial application prospects.

[0124] The scope of protection of the anti-metal clutter electromagnetic micro-acoustic echo signal processing method described in this application is not limited to the execution order of the steps listed in this embodiment. Any scheme implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the scope of protection of this application. Any technical solution that adopts the overall technical concept of low-rank background separation, adaptive constant false alarm detection, and candidate peak secondary robust decision to suppress clutter and extract target peaks in electromagnetic micro-acoustic echoes should be considered to fall within the scope of protection of this embodiment.

[0125] like Figure 5 As shown in the figure, this application provides an anti-metal clutter electromagnetic micro-acoustic echo signal processing system 100, which includes: an echo signal module 110, a query signal transmitting and receiving module 120, and a signal processing module 130.

[0126] The echo signal module 110, in response to receiving a query signal, acquires electromagnetic micro-acoustic echo signals of the same target within multiple consecutive transmission cycles; the query signal transmitting and receiving module 120 transmits a query signal to the echo signal module 110 and receives the electromagnetic micro-acoustic echo signals from the echo signal module 110; the signal processing module 130 constructs a complex observation matrix based on the electromagnetic micro-acoustic echo signals, constructs a real matrix based on the complex observation matrix, and performs low-rank sparse decomposition on the real matrix to obtain the decomposition result; configures decomposition parameters, and generates a sparse amplitude map based on the decomposition parameters and the decomposition result; with the unit to be detected in the sparse amplitude map as the center, protection units and training units are set on its left and right sides respectively, and adaptive detection and local consistency secondary decision are performed using a bilateral CA-CFAR method to determine the output target peak.

[0127] In this embodiment, the query signal transmitting and receiving module 120 is used to realize the radiation of the query signal to the target area and the reception of the target echo; the echo signal module 110 includes a wireless passive SAW tag or a low-sound identification chip, which is set on the identified or located target, and generates an echo signal with multi-peak structure characteristics after receiving the query signal.

[0128] In one specific implementation, the query signal transmitting and receiving module 120 may be implemented using a software-defined radio device, such as the USRP-B200.

[0129] In this embodiment, the query signal transmitting and receiving module 120 includes a transceiver antenna for radiating the query signal to the target area and receiving the target echo. The transceiver antenna can be a single-antenna time-division multiplexing structure or a dual-antenna structure with separate transmitting and receiving antennas. A signal transmission link is established between the transceiver antenna and the wireless passive SAW identification chip via radio electromagnetic coupling. After receiving the query signal, the SAW identification chip generates an echo with a multi-peak structure, which is then radiated back to the transceiver antenna via the antenna. In this embodiment, the signal processing module 130 includes a data acquisition unit and a signal processing unit. The data acquisition unit is connected to the query signal transmitting and receiving module 120 and is used to sample, digitize, and buffer the received echo signal. The signal processing unit is connected to the data acquisition unit and is used to sequentially perform matched filtering, pulse compression, low-rank sparse decomposition, constant false alarm rate detection, and local consistency decision on the digitized multi-pulse echo data, and output the target peak position, peak amplitude, and corresponding arrival time information. The signal processing unit can be implemented by a computer, industrial control computer, DSP, FPGA, or system-on-a-chip. The software-defined radio transceiver is connected to the transceiver antenna and is used to up-convert the query signal and send it to the antenna, and down-convert the target echo received by the antenna and output it to the computer for processing.

[0130] This application effectively separates structured clutter from target echoes in strong metallic environments through low-rank sparse decomposition, significantly improving the contrast between the target and the background without requiring complex environment modeling. Furthermore, it incorporates adaptive constant false alarm rate (CFAR) detection, avoiding clutter contamination threshold estimation and reducing the probability of missed and false detections in non-uniform backgrounds. Further, through local consistency-based secondary decision-making, isolated spurious peaks and noise perturbations are eliminated, making the output target peak more stable and clear. This provides a high-quality data foundation for subsequent template matching and localization calculations, effectively solving the technical problem that existing echo processing methods based on threshold detection or low-rank sparse decomposition struggle to effectively suppress multipath clutter, residual spurious peaks, and non-target perturbations in strong metallic reflection environments, leading to unstable target multi-peak extraction and poor reliability in identification and localization. Therefore, this application effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0131] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for processing electromagnetic micro-acoustic echo signals against metallic clutter, characterized in that, include: In response to receiving a query signal, electromagnetic microacoustic echo signals of the same target are generated in multiple consecutive transmission cycles. A complex observation matrix is ​​constructed based on the electromagnetic micro-acoustic echo signal. A real matrix is ​​then constructed based on the complex observation matrix, and a low-rank sparse decomposition is performed on the real matrix to obtain the decomposition result. Configure decomposition parameters and generate a sparse amplitude map based on the decomposition parameters and the decomposition results; Centered on the cell to be inspected in the sparse amplitude map, protection cells and training cells are set on its left and right sides respectively. Adaptive detection and local consistency secondary decision are performed using the bilateral CA-CFAR method to determine the target peak of the output.

2. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 1, characterized in that, The construction of the complex observation matrix based on the electromagnetic micro-acoustic echo signal includes: The electromagnetic microacoustic echo signal for each transmission cycle is subjected to matched filtering and pulse compression to obtain the corresponding time delay vector; Multiple delay vectors are stacked to form a complex observation matrix with multiple pulses, wherein the row index of the complex observation matrix corresponds to the distance or delay unit, and the column index corresponds to the pulse number.

3. The anti-metallic clutter electromagnetic micro-acoustic echo signal processing method according to claim 1, characterized in that, The construction of a real matrix based on the complex observation matrix includes: The real and imaginary parts of the complex observation matrix are divided into blocks and concatenated to construct a real matrix equivalent to the complex observation matrix.

4. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 1, characterized in that, The low-rank sparse decomposition of the real matrix yields the following results: The GoDec decomposition of the real matrix decomposes the real matrix into low-rank terms, sparse terms, and residual terms. The low-rank terms are used to characterize quasi-static background clutter, multipath tails, and structured interference formed by fixed reflectors with high correlation across pulses. The sparse terms are used to preserve significant peak structures in the target echo. The residual terms are used to characterize noise and unmodeled random disturbances.

5. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 1, characterized in that, The configuration decomposition parameters include: configuring the rank parameter and sparsity parameter of the GoDec decomposition; wherein, the rank parameter of the GoDec decomposition is configured according to the principal component energy ratio of the multi-pulse background, and the sparsity parameter is configured according to the target peak number, the effective width of the single-peak main lobe, and the number of pulses.

6. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 5, characterized in that, The step of generating a sparse magnitude map based on the decomposition parameters and the decomposition result includes: Based on the decomposition parameters and the decomposition results, the sparse terms output by GoDec are restored to complex sparse matrices, and their magnitudes are calculated to obtain a sparse magnitude map after background separation.

7. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 1, characterized in that, The adaptive detection using the bilateral CA-CFAR method includes: The local background mean is estimated using a two-sided CA-CFAR method, and the adaptive detection threshold of the current unit under inspection is calculated based on the preset false alarm rate. When the amplitude of the unit under test is higher than the detection threshold, it is determined that it has passed the primary detection and is used as a candidate target peak for output. When the amplitude of the unit under test is lower than the detection threshold, it is determined to be background clutter or noise.

8. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 7, characterized in that, The local consistency secondary decision includes: The candidate target peaks in the output are combined with at least one of the following pieces of information to determine the target peak of the output: The local mean, local fluctuation, peak significance, cross-pulse consistency, and matching degree with the preset target multi-peak template within the candidate peak neighborhood.

9. The method for processing electromagnetic micro-acoustic echo signals against metallic clutter according to claim 8, characterized in that, The output target peak includes the target peak position, peak amplitude, and arrival time information.

10. A system for processing electromagnetic micro-acoustic echo signals against metallic clutter, characterized in that, include: Echo signal module, query signal transmission and reception module, and signal processing module; The echo signal module responds to receiving a query signal by acquiring electromagnetic micro-acoustic echo signals of the same target in multiple consecutive transmission cycles. The query signal transmitting and receiving module transmits a query signal to the echo signal module and receives the electromagnetic micro-acoustic echo signal from the echo signal module. The signal processing module constructs a complex observation matrix based on the electromagnetic micro-acoustic echo signal, constructs a real matrix based on the complex observation matrix, and performs low-rank sparse decomposition on the real matrix to obtain the decomposition result. Configure decomposition parameters and generate a sparse amplitude map based on the decomposition parameters and the decomposition results; Centered on the cell to be inspected in the sparse amplitude map, protection cells and training cells are set on its left and right sides respectively. Adaptive detection and local consistency secondary decision are performed using the bilateral CA-CFAR method to determine the target peak of the output.

Images