Adaptive anti-interference composite induction alarm method

Abstract

The invention provides a self-adaptive anti-interference composite induction alarm method, and relates to the technical field of self-adaptive anti-interference, and the method comprises the steps: carrying out the modal decomposition of a vibration signal through a vibration decomposition model, calculating a vibration signal decomposition optimization index, carrying out the optimization of a parameter of a recognition model according to the vibration signal decomposition optimization index, and carrying out the recognition of the vibration signal. The method comprises the following steps: acquiring target characteristics of a modal component, calculating dynamic unfamiliar attributes in a time period formed by multiple window times through a sliding window method, detecting a to-be-early-warned region by setting microwave emission initial parameters, performing time-frequency domain analysis on an echo signal, and calculating the detection quality of the echo signal, thereby realizing early warning of the to-be-early-warned region. The method comprises the following steps: optimizing microwave emission parameters through a microwave optimization model, obtaining an optimal echo signal, establishing a feature recognition model, respectively outputting feature recognition data, carrying out composite early warning, and automatically adjusting diffraction waves for detection when an obstacle shielding a microwave signal appears in microwave detection.

Description

Technical Field

[0001] The present invention relates to the field of adaptive anti-interference technology, in particular to an adaptive anti-interference composite induction alarm method. Background Art

[0002] With the increasing demand for security, especially in critical areas such as bases, prisons, factories, and commercial buildings, the performance and reliability of intrusion detection systems are becoming increasingly important. Traditional perimeter security monitoring typically relies on a single sensing method, such as vibration sensors or microwave detectors, but these methods face significant limitations. For example, vibration sensors are prone to false alarms or missed alarms when faced with complex external interference. Vibration detection is easily affected by environmental noise (such as wind, rain, vehicle traffic, and small animal activity), resulting in a high false alarm rate. Furthermore, sensitivity decreases in complex terrain or when the surface medium changes, posing a risk of missed alarms. Although microwave detection can operate 24 / 7, its signal is easily affected by inclement weather (rain, fog, snow), vegetation, and obstructions, resulting in unstable detection quality. Microwave detectors cannot accurately detect targets behind obstructions, resulting in blind spots. At the same time, traditional system parameters are fixed, lack the ability to adaptively adjust to environmental changes and interference, and lack multi-source information fusion. It is impossible to effectively distinguish between real intrusions and complex environmental interference, and it is difficult to meet the real-time monitoring requirements of low false alarms, low missed alarms, and high reliability in high-security scenarios. These technical problems have brought great challenges to practical applications. Therefore, there is an urgent need for a multimodal intrusion detection system that can simultaneously combine vibration and microwaves and can adaptively optimize performance to reduce the impact of external interference and improve detection accuracy.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0004] The purpose of the present invention is to provide an adaptive anti-interference composite induction alarm method to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: An adaptive anti-interference composite induction alarm method, the specific steps include: Step 1: Collect vibration signals from the ground in the area to be warned, perform modal decomposition on the vibration signals using a vibration decomposition model, calculate vibration signal decomposition optimization indicators based on the modal clarity, quantity, and power spectrum entropy of the decomposed targets, and optimize the parameters of the vibration decomposition model based on the vibration signal decomposition optimization indicators. Step 2: Obtain the target features of each modal component by performing deep decomposition on them. Using the sliding window method, assign window unfamiliarity attributes to the target features based on the duration and number of occurrences of each target feature within each window time, and calculate the dynamic unfamiliarity attributes within the time period formed by multiple window times. Step 3: Obtain environmental data of the area to be warned, set initial microwave emission parameters to detect the area to be warned, perform time-frequency domain analysis on the echo signal, and calculate the detection quality of the echo signal; Step 4: Based on the detection quality of the echo signal, real-time environmental data, and the parameter range of the microwave detection equipment, the microwave transmission parameters are optimized using a microwave optimization model to obtain the best echo signal; Step 5: Establish a feature recognition model, output feature recognition data separately, and issue a composite warning based on the feature recognition data. When an obstacle that blocks the microwave signal appears during microwave detection, the diffraction wave is automatically adjusted to detect what is behind the obstruction and issue a warning.

[0006] Furthermore, the vibration decomposition model is based on the variational modal decomposition method, which decomposes the signal into multiple modal functions with specific center frequencies and bandwidths; Furthermore, the method for optimizing the parameters of the vibration decomposition model according to the vibration signal decomposition optimization index is: The key parameters required for variational mode decomposition are optimized through the Grey Wolf Optimization Algorithm, and the number of decomposition layers is dynamically adjusted. and penalty factor Parameters, the specific calculation steps include constructing the wolf pack position matrix, calculating the distance vector, updating the position, and after completing the number of iterations, selecting the optimization parameters with the largest optimization index based on the vibration signal decomposition.

[0007] Furthermore, the calculation method for calculating the vibration signal decomposition optimization index based on the modal clarity, quantity and power spectrum entropy value of the decomposition target is as follows: The calculation method of power spectrum entropy is: ; ; in, is the power spectrum entropy, For frequency The power at is the modal component Fourier transform operation; The formula for calculating modal clarity is:

[0008] in, For the effective modal components, for clarity, for The Hilbert transform of The calculation method of the vibration signal decomposition optimization index is: ; in, Optimize the index for vibration signal decomposition, is the number of effective modal components.

[0009] Furthermore, the calculation method of the window unfamiliarity attribute is: ; The calculation method of dynamic stranger attribute is: ; in, is the unfamiliar attribute within the window time, is the duration sensitivity coefficient, is the target appearance duration, is the number of occurrences within the window time, is the number of windows in the time period, For the The strange properties of a window time, is the forgetting adjustment factor, The current time window sequence number.

[0010] Furthermore, the microwave transmission parameters include transmission power, center frequency, beam direction angle, and pulse width; Environmental parameters include temperature, humidity, air pressure, and particulate matter concentration.

[0011] Furthermore, the microwave optimization model is based on a sequential quadratic programming algorithm. The specific calculation steps are to construct the environment state vector and the parameter vector of microwave detection respectively, and to establish the transmission parameter optimization model and constraint conditions according to the detection quality of the echo signal. Suboptimal parameter vector , for the objective function It is approximately an optimizable form with quadratic terms.

[0012] Furthermore, the feature recognition model includes a vibration recognition model and a microwave recognition model. The feature recognition model is based on a convolutional neural network, specifically including a convolution layer, an activation layer, a pooling layer and an output layer. Feature recognition data includes; Vibration target type and location; Type, location, and target size data of microwave detection target identification; The vibration recognition model uses the modal function of the vibration signal decomposition as the vibration recognition model input, and the vibration signal type and position as the vibration recognition model output to train the model. The modal function of the vibration signal decomposition detected in real time is used as the input of the vibration recognition model to identify the type and position of the vibration target and output the confidence level of the recognition result. The microwave recognition model is trained by taking the time-frequency domain features of the echo signal as the input of the microwave recognition model and the type, position, and target size data of the microwave detection target recognition as the output of the microwave recognition model. The type, position, and target size data of the microwave detection target recognition are recognized by taking the time-frequency domain features of the real-time detected echo signal as the input of the microwave recognition model, and the confidence of the recognition result is output.

[0013] Furthermore, the method for performing compound early warning based on feature recognition data is as follows: ; in, is the confidence after fusion, 、 are the vibration signal confidence and microwave signal confidence output by the recognition model respectively. are the weights of vibration signal confidence and microwave signal confidence, respectively. They are the coordinate positions of the vibration intrusion target and microwave identification intrusion target of the same target type detected. is the position tolerance parameter, , are the correction factors for the vibration signal confidence and microwave signal confidence respectively; when 、 、 Immediately issue an early warning; in, is the fusion warning threshold, is the vibration detection warning threshold, is the microwave detection warning threshold.

[0014] Furthermore, when an obstacle that blocks the microwave signal appears during microwave detection, the diffraction wave is automatically adjusted to detect what is behind the obstacle and to issue an early warning. Based on the target identification type, location, target size data of microwave detection and the location of the detection equipment, the area behind the obstruction is detected by modulating the diffraction wave parameters. The modulation steps of the diffraction wave parameters are as follows: The calculation formula of wavelength is: ; in, is the wavelength of the diffracted wave, is the horizontal dimension of the obstruction, is the straight-line distance between the radar and the obstacle, is the angle between the microwave beam and the obstacle surface, is the diffraction optimization factor; The diffraction deflection angle of the main lobe of the beam is: ; in, is the diffraction deflection angle, is the azimuth of the obstacle edge.

[0015] Compared with the prior art, the present invention has the following beneficial effects: the present invention performs modal decomposition on the vibration signal through a vibration decomposition model, calculates the vibration signal decomposition optimization index, optimizes the parameters of the recognition model according to the vibration signal decomposition optimization index, obtains the target features of the modal components, calculates the dynamic strangeness attributes within the time period formed by multiple window times through a sliding window method, detects the warning area by setting the initial parameters of microwave emission, performs time-frequency domain analysis on the echo signal, calculates the detection quality of the echo signal, optimizes the microwave emission parameters through a microwave optimization model, obtains the best echo signal, establishes a feature recognition model, outputs feature recognition data respectively and performs a composite warning, and automatically adjusts the diffraction wave for detection when an obstacle that blocks the microwave signal appears during microwave detection; The present invention integrates vibration and microwave sensing and introduces a composite sensing alarm mechanism with adaptive anti-interference, which significantly improves the reliability and intelligence level of perimeter protection. First, by utilizing optimized vibration signal decomposition and dynamic unfamiliar attribute analysis, the interference signal caused by environmental noise is effectively eliminated, and the false alarm rate of vibration detection is greatly reduced. Secondly, by real-time analysis of environmental data and echo quality, and dynamic optimization of microwave emission parameters based on sequential quadratic programming, the stability and high quality of detection signals in severe weather are ensured, and the risk of missed reports is reduced. Then, the feature recognition model performs a composite analysis of the dual-modal data to improve the accuracy of intrusion discrimination. Finally, when the microwave is blocked, the system can automatically adjust the diffraction wave detection blind area to the rear, eliminating the blind spots of traditional microwave detection. Ultimately, this solution realizes all-weather, high-precision, and highly robust real-time monitoring of perimeter intrusions in complex and changing environments, significantly improving the effectiveness of security protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the overall method of the present invention. DETAILED DESCRIPTION

[0017] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0018] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0019] Example: See also Figure 1 , the present invention provides a technical solution: An adaptive anti-interference composite induction alarm method, the specific steps include: Step 1: Collect the vibration signal of the ground in the warning area, perform modal decomposition of the vibration signal through the vibration decomposition model, and calculate the vibration signal decomposition optimization index based on the modal clarity, quantity and power spectrum entropy value of the decomposed target. Optimize the parameters of the recognition model based on the vibration signal decomposition optimization index.

[0020] Variational mode decomposition (VMD) is a signal processing method used to decompose a complex signal into several sub-signals (called mode functions) with specific frequency characteristics. Compared with traditional empirical mode decomposition (EMD), VMD offers greater mathematical rigor and robustness, making it well-suited for the analysis and early warning systems of non-stationary vibration signals.

[0021] The primary benefit of using variational mode decomposition (VMD) to decompose vibration signals is its ability to effectively decompose complex vibration signals into multiple modal functions with specific center frequencies and bandwidths, facilitating the analysis of distinct frequency components. VMD is an adaptive decomposition technique that automatically determines the optimal number of modes and frequency range based on signal characteristics, making it particularly suitable for processing nonlinear and nonstationary vibration signals. This decomposition allows for precise extraction of target features from vibration signals with minimal interference between modes, thus reducing false positives caused by signal aliasing. Furthermore, VMD's adaptability effectively adapts to environmental variations, such as those caused by temperature and humidity, enabling the early warning system to operate stably in a variety of complex environments. Overall, using VMD to decompose vibration signals can improve the accuracy of subsequent feature extraction and target recognition, resulting in a more robust and robust interference rejection solution in complex environments, ultimately enabling more accurate and reliable intrusion detection and early warning.

[0022] In this embodiment, the vibration decomposition model is based on the variational modal decomposition method, which decomposes the signal into multiple modal functions with specific center frequencies and bandwidths. The specific calculation formula is: Variational Mode Decomposition: ; Constraint function: ; in, For the A modal function, For the The center frequency of the mode function, is the time partial derivative operator, is the time partial derivative operator, is the imaginary unit, is the complex exponential modulation term, is the collected vibration signal, is the number of levels of modal decomposition, is the time variable; The accuracy and convergence of the decomposition are guaranteed by introducing penalty terms and Lagrange multipliers: ; in, is the penalty factor, is the Lagrange multiplier, is the Lagrangian function.

[0023] The Gray Wolf Optimization (GWA) algorithm is a swarm intelligence optimization algorithm that simulates the hunting behavior of gray wolves. It is capable of finding the global optimal solution within a complex search space. In variational mode decomposition (VMD), key parameters such as the number of decomposition levels and the penalty factor directly influence the decomposition quality and clarity of the modal decomposition. Optimizing these key parameters using the GWA algorithm can make the VMD results more consistent with the characteristics of the actual signal and improve the accuracy of the modal decomposition. First, by simulating the hunting strategies of gray wolves in nature, the GWA algorithm conducts a global search within the search space, avoiding the pitfalls of local optimal solutions and ensuring that the VMD algorithm can achieve the optimal decomposition results under various noise and interference conditions. Second, the GWA algorithm possesses strong global search capabilities and can consider multiple signal characteristics during parameter adjustment, improving the model's adaptability and robustness to complex signals. This is crucial for the nonlinear and nonstationary characteristics of vibration signals. By dynamically adjusting key VMD parameters such as the number of decomposition levels and the penalty factor, not only can the decomposition quality of the vibration signal be improved, but the recognition model's immunity to interference and noise can also be enhanced, thereby improving the accuracy and reliability of the system, playing a vital role in vibration early warning systems.

[0024] In practical applications, vibration signals are often affected by noise, interference, and nonlinear factors. Optimizing these key parameters allows the system to better decompose meaningful modal components, effectively suppressing the effects of noise and interference, and improving the clarity and discernibility of signal decomposition. This approach not only allows the system to more accurately extract the target modal but also enhances its robustness to various interferences, improving the performance and reliability of the entire recognition model.

[0025] In this embodiment, the method for optimizing the parameters of the recognition model according to the vibration signal decomposition optimization index is: The key parameters required for variational mode decomposition are optimized through the Grey Wolf Optimization Algorithm, and the number of decomposition layers is dynamically adjusted. and penalty factor Parameters, the specific calculation steps are: Constructing a wolf pack position matrix : ; in, For the The number of wolf decomposition levels. For the The wolf's punishment factor, is the size of the wolf pack; Calculate the distance vector: ; Location Updates: ; in, is the distance between the gray wolf and its prey, For prey The position at the moment is the position of the optimal parameters. For The location of the wolf pack at any moment, is the random perturbation factor, is the coefficient vector; After completing the number of iterations, the optimization parameter with the largest optimization index is selected based on the vibration signal decomposition.

[0026] By calculating vibration signal decomposition optimization metrics based on the modal clarity, number, and power spectrum entropy of the decomposed targets, the quality and accuracy of vibration signal decomposition can be effectively improved, thereby enhancing the system's adaptive anti-interference capabilities. Modal clarity reflects the independence and distinguishability of different modal components. Higher clarity indicates more accurate signal decomposition, reducing the impact of noise and interference on the signal. The number of modes helps assess the adequacy of signal decomposition, ensuring that important frequency components are not missed during the decomposition process. Power spectrum entropy characterizes signal complexity. Signals with lower entropy values ​​generally indicate simpler structures, which helps improve the stability of the recognition model in complex environments. Combining these three metrics not only optimizes the signal decomposition process but also enables adaptive parameter adjustment to adapt to varying interference and noise conditions, improving the model's robustness. In dynamic environments, this vibration signal-based optimization metric calculation enables the system to more accurately adapt to changes, improves interference resistance, and ensures efficient and stable target recognition in a variety of complex scenarios.

[0027] In this embodiment, the calculation method for calculating the vibration signal decomposition optimization index based on the modal clarity, quantity, and power spectrum entropy value of the decomposed target is: The calculation method of power spectrum entropy is: ; ; in, is the power spectrum entropy, For frequency The power at is the modal component Fourier transform operation; The formula for calculating modal clarity is: ; in, For the effective modal components, for clarity, for The Hilbert transform of The calculation method of the vibration signal decomposition optimization index is: ; in, Optimize the index for vibration signal decomposition, is the number of effective modal components.

[0028] Step 2: Obtain the target features of the modal components by performing deep decomposition on each modal component. Using the sliding window method, assign window unfamiliarity attributes to the target features based on the duration and number of occurrences of each target feature within each window time, and calculate the dynamic unfamiliarity attributes within the time period formed by multiple window times.

[0029] By using a sliding window approach, targets are assigned window attributes based on the duration and number of occurrences of each target feature within each window time. This helps dynamically capture the temporal characteristics and changing patterns of target behavior, thereby enhancing the system's ability to identify abnormal signals and interference and improving adaptive anti-interference effects. Duration reflects the temporal stability of a feature, while the number of occurrences reflects its repeatability and regularity. The combination of these two can effectively distinguish real targets from sporadic interference signals. The sliding window mechanism enables dynamic updates of local features on the timeline, allowing the system to maintain high-precision recognition in the face of non-stationary interference or sudden disturbances, thereby achieving rapid response to environmental changes and adaptive parameter adjustment, effectively improving the robustness and reliability of the system and maintaining stable tracking and accurate perception of targets in complex environments.

[0030] By assigning a forgetting coefficient to each window and calculating the dynamic unfamiliarity attributes over multiple window periods, the system can effectively mitigate the impact of factors such as rain and wind on vibration detection equipment. Specifically, for interfering signals, the system can gradually reduce their impact on system judgment over time, thereby achieving adaptive anti-interference. For example, vibration signals generated by continuous rainfall may initially affect the system's judgment. The system analyzes these signals based on characteristic frequency, amplitude, and other information within the window. If these signals persist for a long time and recur repeatedly, the system gradually assigns a higher forgetting coefficient to these signals, gradually reducing the weight of historical interfering signals. Ultimately, by reassessing the target features, these vibration signals caused by non-hazardous factors such as rain or wind are ignored, thus eliminating false alarms. Furthermore, in the case of wind blowing on a railing, the dynamic unfamiliarity attributes enable the system to identify and gradually eliminate long-lasting, non-threatening vibration signatures, ensuring that no false alarms are issued due to environmental interference. Therefore, dynamic adjustment of the forgetting coefficient enables the system to flexibly adapt to environmental changes and effectively filter out non-hazardous interfering signals, enhancing the stability and accuracy of vibration detection equipment in complex environments.

[0031] In this embodiment, the method for calculating the window unfamiliarity attribute is: ; The calculation method of dynamic stranger attribute is: ; in, is the unfamiliar attribute within the window time, is the duration sensitivity coefficient, is the target appearance duration, is the number of occurrences within the window time, is the number of windows in the time period, For the The strange properties of a window time, is the forgetting adjustment factor, The current time window sequence number.

[0032] Step 3: Obtain environmental data of the area to be warned, set the initial parameters of microwave emission to detect the area to be warned, perform time-frequency domain analysis on the echo signal, and calculate the detection quality of the echo signal.

[0033] By performing time-frequency domain analysis on echo signals using wavelet transforms and calculating their detection quality, the system can effectively extract both time-domain and frequency-domain features, thereby improving the system's adaptive anti-interference capabilities. Wavelet transforms possess excellent time-frequency localization properties, balancing high temporal resolution with high frequency resolution. This is particularly important for detecting interfering signals with rapidly varying frequencies or short durations. For example, in practical applications, changes in ambient temperature, humidity, or air pressure can alter the propagation characteristics of microwave signals, affecting the quality of the echo signal. In such cases, wavelet transforms can effectively decompose the interfering and target signals. During time-frequency domain analysis, the system can identify abnormal changes in the echo signal in real time, such as short-lived high-frequency noise or low-frequency interference caused by weather changes. This time-frequency decomposition allows for accurate assessment of echo signal quality and identifies interfering and target components within the echo signal, thus preventing environmental interference from being misidentified as target signals. In addition, combining microwave emission parameters (such as power, frequency, beam angle, etc.) with environmental state parameters (such as temperature, humidity, etc.) through time-frequency analysis of wavelet transform can help further improve the robustness of the system, and can adjust the detection strategy according to real-time environmental changes, ensuring the stability and accuracy of the system in complex and dynamic environments.

[0034] The method for time-frequency domain analysis of echo signals is: ; in, is the time-frequency domain characteristics of the echo signal, is the scale factor, is the translation factor, is the wavelet basis function, is the complex conjugate.

[0035] In this embodiment, the microwave transmission parameters include transmission power, center frequency, beam direction angle, and pulse width; Environmental state parameters include temperature, humidity, air pressure, and particulate matter concentration; The detection quality of the echo signal is calculated as follows: ; ; in, is the detection quality of the echo signal, Frequency band The corresponding scale, is the time window, Frequency band The number of sampling points, Frequency band energy, is the target feature frequency band set, Frequency band The weight of is the noise suppression coefficient, is the background noise level, is the noise power.

[0036] Step 4: Based on the detection quality of the echo signal, real-time environmental data, and the parameter range of the microwave detection equipment, the microwave transmission parameters are optimized through the microwave optimization model to obtain the best echo signal.

[0037] Sequential quadratic programming is an iterative algorithm for solving nonlinear constrained optimization problems. Its basic idea is to approximate the original nonlinear problem as a quadratic programming subproblem at each iteration. By continuously solving this subproblem and updating the variables, the optimal solution is gradually approached. Sequential quadratic programming offers high efficiency and precision, making it particularly suitable for complex optimization problems involving nonlinearities in both the objective function and the constraints. Microwave emission parameter optimization involves multiple coupled variables (such as frequency, power, and pulse width), which are subject to various constraints, including hardware limitations, signal propagation patterns, and background interference. The optimization objective is to optimize the detection quality of the echo signal. This type of problem is inherently nonlinear and constrained. Therefore, the use of a sequential quadratic programming algorithm can effectively handle these complex coupling relationships and physical constraints, enabling precise control of microwave parameters to achieve the optimal echo effect and improve the performance and anti-interference capability of the detection system.

[0038] The sequential quadratic programming algorithm is used to construct a microwave optimization model and obtain the optimal echo signal. This approach aims to balance multiple constraints and objective functions, and is particularly well-suited for complex optimization problems with constraints. In microwave detection systems, echo signal quality is affected by microwave transmission parameters (such as power, frequency, and pulse width) as well as environmental factors (such as temperature, humidity, and air pressure). The sequential quadratic programming algorithm can precisely adjust microwave transmission parameters based on real-time echo signal quality and environmental data to optimize detection performance. For example, when strong interference signals (such as meteorological factors or stray noise) are present in the environment, the sequential quadratic programming algorithm can dynamically adjust microwave transmission parameters such as power and frequency, while taking into account system constraints, to optimize echo signal quality, enhance detection quality, and mitigate the effects of interference. Specifically, in actual operation, if temperature fluctuations cause changes in microwave propagation characteristics, the sequential quadratic programming algorithm can automatically calculate the optimal frequency and power combination to achieve the best echo signal quality, thereby ensuring the system maintains stable detection capabilities in harsh environments. Through this method, not only the adaptability of the detection system in complex environments is improved, but also its anti-interference ability is enhanced, ensuring that the target signal and the interference signal can be effectively distinguished, thereby improving the accuracy and reliability of detection.

[0039] In this embodiment, the microwave optimization model is based on the sequential quadratic programming algorithm. The specific calculation steps are to construct the environment state vector and the parameter vector of microwave detection respectively, and to establish the transmission parameter optimization model and constraint conditions according to the detection quality of the echo signal. Suboptimal parameter vector , for the objective function It is approximately an optimizable form with a quadratic term. The specific calculation steps are: Build the environment state vector: ; in, is the environmental state vector, is the temperature, For humidity, is the air pressure, is the concentration of particulate matter; Construct the parameter vector for microwave detection: ; in, is the parameter vector, is the transmit power, is the center frequency, is the beam direction angle, is the pulse width; Construct a launch parameter optimization model: ; Constraints: ; in, are the lower and upper limits of the transmit power respectively, Expressed as the range of center frequencies, is the range of beam direction angles, is the range of pulse width; Construct the quadratic programming subproblem: Put the Suboptimal parameter vector , for the objective function Approximately an optimizable form with quadratic terms: ; in, The increment of the parameter vector, For the The curvature of the objective function of the iteration, For the The objective function gradient of the iteration, is the transpose operation; Constrained linearization: ; in, For the The parameter vector for the iteration, is the constraint function vector, which represents the parameter vector All constraint values ​​at is the constrained Jacobian matrix, The sequence number of the current optimization step.

[0040] Step 5: Establish a feature recognition model, output feature recognition data separately, and issue a composite warning based on the feature recognition data. When an obstacle that blocks the microwave signal appears during microwave detection, the diffraction wave is automatically adjusted to detect what is behind the obstruction and issue a warning.

[0041] Convolutional neural networks (CNNs) are widely used in feature recognition, particularly for processing spatiotemporal features in signals such as vibration and microwave signals. In vibration recognition models, CNNs can automatically learn discriminative features from the signal, particularly through decomposed mode functions. These modal functions effectively extract the different frequency components of the vibration signal. The time-frequency domain features obtained through wavelet transforms further improve the signal's resolution and robustness. In microwave recognition models, CNNs process the time-frequency domain features of the echo signal obtained through wavelet transforms to accurately distinguish microwave detection targets from other background noise or interfering signals. By establishing these two models separately, the system can process and recognize different types of signals specifically, thereby improving the accuracy and reliability of target recognition, especially in complex environments or when encountering signal interference. For example, in practical applications, when the detection system is obstructed by an obstruction, the vibration recognition model can identify the physical changes caused by the vibration, while the microwave recognition model can accurately detect the target signal behind the obstruction. This adaptively adjusts the diffraction waves for more precise target detection and recognition, and issues warnings based on the recognition results. In this way, through multimodal data fusion, the system's anti-interference ability can be effectively improved, ensuring continuous and stable operation under complex conditions.

[0042] In this embodiment, the feature recognition model includes a vibration recognition model and a microwave recognition model. The feature recognition model is based on a convolutional neural network, specifically including a convolution layer, an activation layer, a pooling layer, and an output layer. Feature recognition data includes; Vibration target type and location; Type, location, and target size data of microwave detection target identification; The vibration recognition model uses the modal function of the vibration signal decomposition as the vibration recognition model input, and the vibration signal type and position as the vibration recognition model output to train the model. The modal function of the vibration signal decomposition detected in real time is used as the input of the vibration recognition model to identify the type and position of the vibration target and output the confidence level of the recognition result. The microwave recognition model is trained by taking the time-frequency domain features of the echo signal as the input of the microwave recognition model and the type, position, and target size data of the microwave detection target recognition as the output of the microwave recognition model. The type, position, and target size data of the microwave detection target recognition are recognized by taking the time-frequency domain features of the real-time detected echo signal as the input of the microwave recognition model, and the confidence of the recognition result is output.

[0043] The convolutional neural network specifically includes a convolution layer, an activation layer, a pooling layer and an output layer; Convolutional layers; ; in, The modal function of the input vibration signal decomposition and the time-frequency domain characteristics of the echo signal, is the convolution kernel, is the coordinate of the output data matrix, The convolution kernel is Row and The value of the column; The calculation formula of the activation layer is: ; in, The coordinates of the output matrix of the convolutional layer output; The calculation formula of the pooling layer is: ; in, is the maximum pooling output; The calculation formula of the output layer is: ; in, is the output of the pooling layer, is the result of feature recognition data recognition, is the weight, is the bias parameter.

[0044] Providing a composite warning based on the location and confidence of both vibration-detected intrusion targets and microwave-detected targets, respectively, can achieve complementary enhancement of multi-source information, thereby improving the accuracy and robustness of target recognition and intrusion detection in complex environments. Vibration detection is primarily sensitive to disturbances on the ground or structures, and can accurately identify the location of physical contact behaviors such as touching and climbing, but is susceptible to interference from environmental noise such as wind and mechanical vibration. Microwave detection, on the other hand, is suitable for identifying the movement trajectory of targets in the air or at long distances, and is particularly capable of penetrating a certain degree of obstruction, but is susceptible to factors such as rain, fog, and electromagnetic interference. By extracting the location and confidence information output by the two models separately, they can verify and complement each other in practical applications, improving the credibility of the overall recognition results. In particular, when the signal is incomplete or a single sensor is interfered with, the system can use information provided by another signal source to make fault-tolerant judgments and warnings, thereby enhancing adaptive anti-interference capabilities.

[0045] For example, in a certain border protection scenario, the microwave signal is severely attenuated due to an obstruction on one side. Even after the system automatically adjusts the diffraction wave, it is still unable to fully obtain a clear image. At this time, the vibration model detects obvious vibration characteristics on the ground in this area. Combined with its output position coordinates and high recognition confidence, the system confirms the existence of an intrusion risk through data fusion and promptly triggers a composite warning. Conversely, when strong winds cause the vibration sensor to falsely trigger, if the microwave detection model does not recognize a moving target at the same location or the recognition confidence is low, the system can automatically reduce the signal weight to avoid false alarms. This position and confidence fusion mechanism based on multi-source recognition results not only enhances anti-interference capabilities, but also makes the warning results more stable and reliable.

[0046] In this embodiment, the method for performing composite warning based on feature recognition data is: ; in, is the confidence after fusion, 、 are the vibration signal confidence and microwave signal confidence output by the recognition model respectively. 、 are the weights of vibration signal confidence and microwave signal confidence, respectively. They are the coordinate positions of the vibration intrusion target and microwave identification intrusion target of the same target type detected. is the position tolerance parameter, , are the correction factors for the vibration signal confidence and microwave signal confidence respectively; when 、 、 Immediately issue an early warning; in, is the fusion warning threshold, is the vibration detection warning threshold, is the microwave detection warning threshold.

[0047] In this embodiment, when an obstacle that blocks the microwave signal appears during microwave detection, the method of automatically adjusting the diffracted wave, detecting what is behind the obstruction, and issuing an early warning is as follows: Based on the target identification type, location, target size data of microwave detection and the location of the detection equipment, the area behind the obstruction is detected by modulating the diffraction wave parameters. The modulation steps of the diffraction wave parameters are as follows: The calculation formula of wavelength is: ; Center Frequency: ; Bandwidth is calculated as follows: ; in, is the wavelength of the diffracted wave, is the horizontal dimension of the obstruction, is the straight-line distance between the radar and the obstacle, is the angle between the microwave beam and the obstacle surface, is the diffraction optimization factor, is the center frequency, bandwidth, is the speed of light; The diffraction deflection angle of the main lobe of the beam is: ; in, is the diffraction deflection angle, is the azimuth of the obstacle edge.

[0048] When microwave detection encounters an obstruction, it automatically adjusts the diffraction wave for detection. This is to overcome the signal attenuation or complete loss caused by the obstacle, thereby ensuring that targets in the detection area can still be detected. Microwave signals are typically reflected, refracted, or diffracted when encountering an obstacle. Diffraction waves refer to the way the signal propagates around the obstacle. In many application scenarios, obstacles may completely block the microwave signal, making the target impossible to detect via the conventional straight-line propagation path. By adjusting the diffraction wave, the microwave detector can adjust the signal propagation path so that even if there is an obstruction, the signal can still bypass the obstacle and accurately detect the target behind it. This method effectively solves the blind spot problem that traditional microwave detectors may have when encountering obstacles, improves the system's continuous monitoring capabilities, and can maintain high detection accuracy, especially in complex environments.

[0049] For example, in an urban security system, if a microwave sensor is blocked by a building or other tall object, the system automatically adjusts the diffraction angle and propagation mode of the microwave signal to bypass the obstruction and detect the target behind it. If a person or object attempts to enter the protected area, the diffracted waves will enable the microwave signal to detect the target's presence and trigger an alarm. This adaptive adjustment enables the system to continuously monitor and reduce detection blind spots caused by obstructions, effectively improving its prevention capabilities and ensuring the stable operation of the security system in complex environments.

[0050] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0051] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0052] The units described as decomposed components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment based on actual needs.

[0053] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. An adaptive anti-interference composite induction alarm method, characterized in that: The specific steps include: Step 1: Collect vibration signals from the ground in the warning area, perform modal decomposition on the vibration signals using a vibration decomposition model, and calculate the vibration signal decomposition optimization index based on the modal clarity, number, and power spectrum entropy of the decomposed targets. Optimize the parameters of the vibration decomposition model based on the vibration signal decomposition optimization index. Step 2: Obtain the target features of each modal component by performing deep decomposition on them. Using the sliding window method, assign window unfamiliarity attributes to the target features based on the duration and number of occurrences of each target feature within each window time, and calculate the dynamic unfamiliarity attributes within the time period formed by multiple window times. Step 3: Obtain environmental data of the area to be warned, set initial microwave emission parameters to detect the area to be warned, perform time-frequency domain analysis on the echo signal, and calculate the detection quality of the echo signal; Step 4: Based on the detection quality of the echo signal, real-time environmental data, and the parameter range of the microwave detection equipment, the microwave transmission parameters are optimized using a microwave optimization model to obtain the best echo signal; Step 5: Establish a feature recognition model, output feature recognition data separately, and issue a composite warning based on the feature recognition data. When an obstacle that blocks the microwave signal appears during microwave detection, the diffraction wave is automatically adjusted to detect what is behind the obstruction and issue a warning.

2. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The vibration decomposition model is based on a variational modal decomposition method, which decomposes the signal into multiple modal functions with specific center frequencies and bandwidths.

3. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The method for optimizing the parameters of the vibration decomposition model according to the vibration signal decomposition optimization index is as follows: The key parameters required for variational mode decomposition are optimized through the Grey Wolf Optimization Algorithm, and the number of decomposition layers is dynamically adjusted. and penalty factor Parameters, the specific calculation steps include constructing the wolf pack position matrix, calculating the distance vector, updating the position, and after completing the number of iterations, selecting the optimization parameters with the largest optimization index based on the vibration signal decomposition.

4. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The calculation method for calculating the vibration signal decomposition optimization index based on the modal clarity, quantity and power spectrum entropy value of the decomposition target is as follows: The calculation method of power spectrum entropy is: ； ； in, is the power spectrum entropy, For frequency The power at is the modal component Fourier transform operation; The formula for calculating modal clarity is: ； in, For the effective modal components, for clarity, for The Hilbert transform of The calculation method of the vibration signal decomposition optimization index is: ； in, Optimize the index for vibration signal decomposition, is the number of effective modal components.

5. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The calculation method of the window strangeness attribute is: ； The calculation method of dynamic stranger attribute is: ； in, is the unfamiliar attribute within the window time, is the duration sensitivity coefficient, is the target appearance duration, is the number of occurrences within the window time, is the number of windows in the time period, For the The strange properties of a window time, is the forgetting adjustment factor, The current time window sequence number.

6. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The microwave transmission parameters include transmission power, center frequency, beam direction angle, and pulse width; Environmental parameters include temperature, humidity, air pressure, and particulate matter concentration.

7. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The microwave optimization model is based on the sequential quadratic programming algorithm. The specific calculation steps are to construct the environment state vector and the parameter vector of microwave detection respectively, and to establish the transmission parameter optimization model and constraint conditions according to the detection quality of the echo signal. Suboptimal parameter vector , for the objective function It is approximately an optimizable form with quadratic terms.

8. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The feature recognition model includes a vibration recognition model and a microwave recognition model. The feature recognition model is based on a convolutional neural network and specifically includes a convolution layer, an activation layer, a pooling layer, and an output layer. Feature recognition data includes; Vibration target type and location; Type, location, and target size data of microwave detection target identification; The vibration recognition model uses the modal function of the vibration signal decomposition as the vibration recognition model input, and the vibration signal type and position as the vibration recognition model output to train the model. The modal function of the vibration signal decomposition detected in real time is used as the input of the vibration recognition model to identify the type and position of the vibration target and output the confidence level of the recognition result. The microwave recognition model is trained by taking the time-frequency domain features of the echo signal as the input of the microwave recognition model and the type, position, and target size data of the microwave detection target recognition as the output of the microwave recognition model. The type, position, and target size data of the microwave detection target recognition are recognized by taking the time-frequency domain features of the real-time detected echo signal as the input of the microwave recognition model, and the confidence of the recognition result is output.

9. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: The method for compound early warning based on feature recognition data is: ； in, is the confidence after fusion, 、 are the vibration signal confidence and microwave signal confidence output by the recognition model respectively. 、 are the weights of vibration signal confidence and microwave signal confidence, respectively. They are the coordinate positions of the vibration intrusion target and microwave identification intrusion target of the same target type detected. is the position tolerance parameter, , are the correction factors for the vibration signal confidence and microwave signal confidence respectively; when 、 、 Immediately issue an early warning; in, is the fusion warning threshold, is the vibration detection warning threshold, is the microwave detection warning threshold.

10. The adaptive anti-interference composite induction alarm method according to claim 1, characterized in that: When an obstacle that blocks the microwave signal appears during microwave detection, the diffraction wave is automatically adjusted to detect what is behind the obstacle and to issue an early warning. Based on the target identification type, location, target size data of microwave detection and the location of the detection equipment, the area behind the obstruction is detected by modulating the diffraction wave parameters. The modulation steps of the diffraction wave parameters are as follows: The calculation formula of wavelength is: ； in, is the wavelength of the diffracted wave, is the horizontal dimension of the obstruction, is the straight-line distance between the radar and the obstacle, is the angle between the microwave beam and the obstacle surface, is the diffraction optimization factor; The diffraction deflection angle of the main lobe of the beam is: ； in, is the diffraction deflection angle, is the azimuth of the obstacle edge.

Images