Machine room wild animal intelligent defense system based on millimeter wave radar

Abstract

The invention relates to the field of intelligent monitoring, and discloses a machine room wild animal intelligent defense system based on a millimeter wave radar, which comprises a signal capture module, a signal processing module, a feature extraction module, a classification tracking module and an alarm linkage module, the signal capturing module collects target reflection signals through a millimeter wave radar, the signal processing module carries out multi-target separation and interference suppression on the signals, the feature extraction module extracts time-frequency characteristics and motion track characteristics of targets, and the classification tracking module classifies the targets and tracks the motion states of the targets in real time. And the alarm linkage module evaluates the threat level according to the classification result and triggers corresponding defense measures. The system can accurately identify and track targets, timely respond and execute defense actions, has the advantages of high precision, multi-target monitoring, real-time dynamic tracking, intelligent defense and the like, and can effectively guarantee safe operation of machine room equipment.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring technology, and in particular to a machine room wild animal intelligent defense system based on millimeter wave radar. Background Art

[0002] With the rapid development of intelligent technology, many important power facilities such as computer rooms and substations have gradually achieved unattended operation. In order to improve the safety of equipment, computer rooms usually take blocking measures to prevent small animals from entering. However, with the changes in environmental conditions and the unpredictability of small animal activities, these traditional blocking measures and physical protection methods have not completely solved the problem. Especially in important facilities such as substations, the invasion of small animals such as rats, snakes, and birds often leads to equipment damage, power accidents, and safety hazards, posing a huge challenge to the reliability of the power system.

[0003] Traditional protection methods rely on manual regular inspections and plugging operations, which is not only labor-intensive and inefficient, but also difficult to monitor in real time and accurately identify potential risks. In addition, there are large blind spots in the monitoring of hidden areas such as cable trenches, and manual inspection methods are often not comprehensive, real-time, and efficient, which allows many small animals to quietly enter the machine room and damage equipment.

[0004] Most existing animal monitoring systems are based on visual or infrared technology, and the detection accuracy of these systems is easily affected by changes in ambient light, target occlusion or weather factors. In complex machine room environments, areas such as cable trenches and equipment channels are less adaptable to infrared or optical technologies, and these systems usually require a large number of sensors or cameras to be set up near the equipment, increasing the complexity and maintenance costs of the system.

[0005] In addition, although some monitoring solutions based on wireless sensor networks have been applied in physical protection, these systems are usually unable to effectively cope with complex signal processing and real-time data analysis in multi-target environments. Many existing animal identification technologies rely on a single sensor or are limited to monitoring environmental changes. They lack comprehensive solutions and cannot meet the requirements of high-precision, multi-target, and all-weather monitoring.

[0006] Therefore, the problem facing current technology is how to conduct real-time monitoring and protection against wildlife invasions through efficient and accurate technical means in a complex and dynamic computer room environment, especially in the monitoring of multi-target and hidden areas, how to avoid the blind spots of traditional monitoring methods and improve the real-time and reliability of the protection system.

[0007] The purpose of the present invention is to address the deficiencies of the prior art and to propose an intelligent defense system based on millimeter-wave radar. Summary of the invention

[0008] In view of the shortcomings of the existing technology, the present invention provides an intelligent defense system for wild animals in the computer room based on millimeter-wave radar, which solves the problem that it is difficult to achieve accurate and real-time monitoring and protection of small animals in the computer room in the existing technology, especially the effective monitoring and defense of multiple targets and hidden areas in complex environments.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: a machine room wildlife intelligent defense system based on millimeter wave radar, including a signal capture module, a signal processing module, a feature extraction module, a classification tracking module and an alarm linkage module;

[0010] The signal capture module is used to capture the millimeter wave radar reflection signal from the target through the intermediate frequency signal model;

[0011] The signal processing module is connected to the signal acquisition module and is used to perform multi-target separation and interference suppression on the target reflection signal;

[0012] The feature extraction module is connected to the signal processing module and is used to extract the time-frequency features and motion trajectory features of the target;

[0013] The classification and tracking module is connected to the feature extraction module and is used to classify the target and track its motion state according to the extracted features;

[0014] The alarm linkage module is connected to the classification tracking module and is used to trigger an alarm signal according to the classification result and to link the execution of the defense device.

[0015] Preferably, the signal capture module adopts a linear frequency modulation continuous wave millimeter wave radar, including a transmitting unit, a receiving unit and a signal mixing unit, wherein:

[0016] The transmitting unit is used to transmit a linear frequency modulation continuous wave signal and capture the distance and speed of the target based on the target distance and speed formula;

[0017] The receiving unit is used to receive the target reflected echo signal and perform preliminary filtering processing to suppress high-frequency noise;

[0018] The signal mixing unit is used to mix the target reflection signal and the transmission signal to generate an intermediate frequency signal as an output, and the intermediate frequency signal retains the distance and Doppler information of the target.

[0019] Preferably, the intermediate frequency signal model is:

[0020]

[0021] Among them, r(t) is the intermediate frequency signal generated after the target reflection signal is received and mixed, K is the number of targets, α kis the reflection intensity of the kth target, f k is the Doppler frequency, φ k is the phase, n(t) is the noise;

[0022] The target distance speed formula for calculating the distance and speed of the target is:

[0023]

[0024] Where B is the frequency modulation bandwidth, T is the frequency modulation period, and d k is the distance to the target, v k is the speed of the target, c is the speed of light, f c is the center frequency of the millimeter wave radar, Δf k is the target frequency drift.

[0025] Preferably, the signal processing module includes a sparse modeling unit and an optimization unit;

[0026] The sparse modeling unit is used to represent multiple target signals in the intermediate frequency signal as a sparse signal matrix by establishing a joint sparse matrix;

[0027] The optimization unit is used to extract and separate the target signal based on an iterative optimization algorithm, optimize the extracted and separated target signal based on a sparse representation model, and output a sparse coefficient matrix of the target signal.

[0028] Preferably, the sparse representation model is:

[0029] Y=ΦX+N

[0030] The iterative optimization algorithm is:

[0031]

[0032] in,

[0033] represents the received intermediate frequency signal matrix, M is the number of measurement signals, T is the number of time sampling points, N is the number of target signal characteristics, and the intermediate frequency signal matrix includes a measurement matrix, a sparse signal matrix and a background noise matrix;

[0034] when When , Φ represents the measurement matrix;

[0035] when When , X represents the sparse signal matrix;

[0036] when When , N represents the background noise matrix;

[0037] To extract and separate the target signal, represents a joint sparse constraint, subject to represents Subordinate to ∈ represents the noise limit threshold, which is used to ensure that background noise does not interfere with the extraction of the target signal.

[0038] Preferably, the feature extraction module includes a time-frequency analysis unit and a trajectory calculation unit;

[0039] The time-frequency analysis unit is used to perform time-frequency characteristic analysis on the target signal through a short-time Fourier transform algorithm to generate a frequency spectrum distribution of the target signal, wherein the frequency spectrum distribution of the target signal includes a peak frequency, a bandwidth, and a frequency drift characteristic of the target;

[0040] The trajectory calculation unit is used to generate the position, velocity and acceleration information of the target through the time variation characteristics in the target signal, and provide a motion trajectory description.

[0041] Preferably, the short-time Fourier transform algorithm is:

[0042] P(t,f)=|∫x(τ)e -j2πfτ h(t-τ)dτ| 2

[0043] Where P(t, f) represents the time-frequency distribution of the target signal at time t and frequency f, x(τ) represents the amplitude of the signal at time τ, h(t-τ) represents the window function centered at time t, and e -j2πfτ is the complex exponential kernel function of the frequency domain transform, dτ represents the small time increment of the integral variable;

[0044] The trajectory calculation unit calculates the speed and distance of the target based on the point cloud data, wherein the calculation formula of the speed of the target is:

[0045]

[0046] The calculation formula of the distance of the target is:

[0047]

[0048] Where v represents the speed of the target, Δf represents the frequency drift of the target signal, and f c represents the center frequency of the millimeter-wave radar, d represents the distance of the target, c represents the speed of light, and τ represents the total propagation time delay of the signal from the time it is transmitted to the time it is reflected back to the radar by the target.

[0049] Preferably, the classification and tracking module includes a classification unit and a tracking unit;

[0050] The classification unit is used to judge the target category of the target feature vector by using a sparse Bayesian classification formula based on a sparse Bayesian classifier;

[0051] The tracking unit is used to dynamically update the target state based on the judgment result of the target category of the target feature vector by using Kalman filtering, and continuously update the position, speed and acceleration of the target by combining a dynamic estimation algorithm to generate the target's trajectory data.

[0052] Preferably, the sparse Bayesian classification formula is:

[0053]

[0054] Among them, P(c k |f) represents the probability that the target belongs to the kth class, f represents the feature vector of the target, and the feature vector of the target includes the time-frequency characteristics and motion trajectory characteristics of the target. k ) indicates that the feature vector belongs to category c k The conditional probability, P(c k ) indicates category c k The prior probability of , K represents the total number of target categories, represents the normalized probability of all categories.

[0055] Preferably, the alarm linkage module includes an alarm unit and a linkage unit;

[0056] The alarm unit receives the target classification result, and is used to obtain the threat level of the target through a threat level calculation formula, and triggers a local alarm, a remote notification, or automatically records the alarm information according to the threat level of the target. The threat level calculation formula is:

[0057] L=ω1P(c k )+ω2||v||+ω3d

[0058] Where L represents the threat level, P(c k ) indicates that the target belongs to threat category c k The probability of ||v|| represents the speed of the target, which is defined as v x and v y are the velocity components of the target in the x-axis and y-axis directions respectively, d represents the distance of the target from the radar, ω1, ω2, ω3 represent the weight coefficients in the threat level calculation;

[0059] The linkage unit is used to obtain the trajectory prediction of the target based on the trajectory prediction model, and determine the start time and position of the defense device according to the trajectory prediction result of the target, and the trajectory prediction model is:

[0060]

[0061] Among them, p(t+Δt) represents the predicted target position, p(t) represents the current target position, v(t) represents the current velocity of the target, a(t) represents the current acceleration of the target, and (Δt) represents the time interval.

[0062] The present invention provides a machine room wildlife intelligent defense system based on millimeter wave radar.

[0063] Beneficial effects:

[0064] 1. Through the high-frequency signal acquisition of millimeter-wave radar and the multi-target separation technology of the signal processing module, the present invention can achieve high-precision target monitoring in complex environments. The feature extraction module can accurately extract the time-frequency characteristics and motion trajectory of the target, and accurately classify the target through the sparse Bayesian classifier, thereby effectively distinguishing different types of animals and reducing the risk of false alarms and missed alarms.

[0065] 2. The classification tracking module of the present invention is combined with the Kalman filter algorithm to update the position, velocity and acceleration information of the target in real time and predict the future position of the target. This module enables the system to dynamically track the movement trajectory of the target, thereby providing accurate triggering timing for the defense device and ensuring the timeliness and effectiveness of the defense measures.

[0066] 3. This system uses a joint sparse representation model to process target signals. It can process multiple target signals simultaneously in the same monitoring area, separate the reflected signals of different targets, avoid signal overlap or interference, and has powerful multi-target recognition and tracking capabilities. This feature enables the system to cope with complex machine room environments and ensure the monitoring and defense needs of various small animals.

[0067] 4. The alarm linkage module automatically triggers local alarms, remote notifications or the activation of defense devices through real-time assessment of the target threat level, and can make the best response according to different situations. By calculating the threat level and predicting the target trajectory, the system can flexibly adjust the defense strategy to ensure a rapid response in high-risk situations and reduce equipment damage and safety hazards.

[0068] 5. The signal processing module of the system has a strong interference suppression capability and can effectively handle multipath interference and environmental noise. By introducing multipath modeling and noise limitation technology, it ensures that the system can still maintain efficient and stable operation in various complex environments, adapt to different types and sizes of computer room environments, and improve the overall robustness of the system.

[0069] 6. The modules of the present invention are tightly integrated to form an efficient collaborative working mechanism, avoiding redundant and complex equipment configurations and simplifying system deployment and maintenance. The system can be quickly integrated into existing monitoring and protection platforms, and has strong scalability, and can easily adapt to different types of computer room environments and defense requirements as needed. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 Schematic diagram of the system architecture of the present invention. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] Please see attached Figure 1 The embodiment of the present invention provides a computer room wildlife intelligent defense system based on millimeter wave radar, which realizes intelligent monitoring, classification, tracking and defense of wildlife in the computer room through the coordinated work of signal capture, signal processing, feature extraction, classification tracking and alarm linkage modules. The system of the present invention mainly includes a signal capture module, a signal processing module, a feature extraction module, a classification tracking module and an alarm linkage module. Each module is described in detail below.

[0073] Signal capture module

[0074] The signal capture module mainly realizes the capture and preprocessing of the target reflection signal, and provides high-quality input signals for the subsequent signal processing module. In general, the signal capture module extracts information such as the distance, speed and direction of the target through the transmission, reception and mixing operations of the millimeter-wave radar. Specifically, this module uses a linear frequency modulated continuous wave (FMCW) millimeter-wave radar, which has the advantages of high precision, real-time and anti-interference capabilities, and can complete the basic parameter measurement of the target in complex environments.

[0075] In this embodiment, the signal capture module mainly includes a transmitting unit, a receiving unit and a mixing unit.

[0076] In some embodiments, the transmitting unit is used to transmit a linear frequency modulation continuous wave signal, and the signal is expressed in the form of:

[0077]

[0078] in:

[0079] f cIndicates the center frequency of the millimeter-wave radar, usually in the 24 GHz or 77 GHz band;

[0080] B represents the frequency modulation bandwidth;

[0081] T represents the frequency modulation period;

[0082] t is the time variable.

[0083] In a possible implementation, the transmitting unit transmits the frequency modulated signal to the monitoring area in the machine room through the antenna. The signal will be reflected by the target during the propagation process, thereby carrying the characteristic information of the target.

[0084] Generally, the receiving unit is used to receive the reflected signal of the target and perform preliminary filtering on it to suppress the high-frequency noise interference in the environment. The reflected signal is expressed as:

[0085]

[0086] in:

[0087] r(t) is the intermediate frequency signal generated after the target reflected signal is received and mixed;

[0088] α represents the reflection intensity of the target, which is related to the target material and shape;

[0089] represents the propagation delay of the signal, which is proportional to the distance d of the target;

[0090] c is the speed of light.

[0091] In a specific implementation, the receiving unit processes the received reflected signal through a bandpass filter, thereby eliminating interference from non-target signals and ensuring signal integrity for subsequent processing.

[0092] As an option, the signal mixing unit is used to mix the target reflected signal with the transmitted signal to generate an intermediate frequency signal. Specifically, the intermediate frequency signal obtained by the mixing operation can be expressed as:

[0093] r IF (t) = cos(2πf IF t+φ)

[0094] in:

[0095] f IF is the intermediate frequency signal frequency, which contains the distance information of the target;

[0096] φ is the phase of the signal, which contains the speed information of the target.

[0097] The intermediate frequency signal carries the target's distance and Doppler information, and can provide data support for subsequent signal processing modules.

[0098] In some embodiments, the intermediate frequency signal generated by the signal acquisition module can be further modeled as follows:

[0099]

[0100] in:

[0101] K is the number of targets;

[0102] α k is the reflection intensity of the kth target;

[0103] f k is the Doppler frequency, which indicates the frequency shift caused by the target velocity;

[0104] φ k is the phase of the kth target;

[0105] n(t) is the system noise.

[0106] In general, the distance d of the target k and speed v k Calculated by the following formula:

[0107]

[0108] in:

[0109] B represents the frequency modulation bandwidth;

[0110] T represents the frequency modulation period;

[0111] c represents the speed of light;

[0112] f c is the center frequency of the millimeter-wave radar;

[0113] Δf k Indicates the frequency drift of the target, which is related to the relative velocity of the target.

[0114] Specifically, the signal acquisition module can generate a high-quality intermediate frequency signal through the above process, and the signal completely retains the distance, speed and direction information of the target. In a possible implementation, the signal acquisition module transmits the intermediate frequency signal to the signal processing module for multi-target separation and interference suppression processing.

[0115] In some embodiments, in order to improve the adaptability and robustness of the system, the signal acquisition module can adapt to the monitoring needs of targets of different sizes by adjusting the frequency modulation bandwidth B and frequency modulation period f of the transmitting unit. For example, when monitoring smaller targets (such as rodents), the frequency modulation bandwidth is appropriately increased to improve the distance resolution; when monitoring larger targets (such as snakes), the bandwidth is appropriately reduced to optimize the stability of signal acquisition.

[0116] Signal processing module

[0117] The main function of the signal processing module is to perform multi-target separation and interference suppression on the intermediate frequency signal generated by the signal capture module, and extract the effective signal characteristics of the target for subsequent module processing. In general, the signal processing module separates sparse target signals from a multi-target environment by combining sparse modeling and optimization solution, while suppressing environmental noise and multipath interference. Specifically, the workflow of this module includes sparse modeling, optimization solution and signal enhancement processing of target signals.

[0118] In this embodiment, the signal processing module includes a sparse modeling unit and an optimization unit, and is processed in combination with a multipath interference suppression algorithm. The specific implementation method is as follows:

[0119] In some embodiments, the sparse modeling unit mathematically models the intermediate frequency signal matrix provided by the signal acquisition module by establishing a joint sparse representation model to represent the sparse characteristics of the multi-target reflection signal. Specifically, the sparse representation model can be expressed as:

[0120] Y=ΦX+N

[0121] in:

[0122] Represents the received intermediate frequency signal matrix, including M measurement signals and T time sampling points;

[0123] represents the measurement matrix;

[0124] represents a sparse signal matrix, containing N target signal features;

[0125] represents the background noise matrix.

[0126] Generally speaking, the sparse modeling unit utilizes the characteristics of joint sparse representation to decompose the features of multi-target signals into non-zero items of sparse matrices, thereby effectively extracting target information in a mixed multi-target environment.

[0127] As an option, the optimization unit uses an iterative optimization method to solve the above sparse representation model to extract the effective signal features of the target from the background noise. Specifically, the optimization problem is expressed as:

[0128]

[0129] in:

[0130] is the joint sparsity constraint;

[0131] ∈ is the noise tolerance, which is used to control the interference range of background noise;

[0132] Represents the Frobenius norm of the matrix, which is used to measure the reconstruction error of the model;

[0133] subjectto Subordinate to

[0134] The optimization unit solves the above problem and outputs the sparse coefficient matrix X of the target signal for use by the feature extraction module.

[0135] Specifically, the optimization unit can use the alternating direction method of multipliers (ADMM) to solve the above optimization problem. As a possible implementation method, the basic steps of the ADMM algorithm include:

[0136] Initialize the initial value of the sparse coefficient matrix;

[0137] Update the sparse matrix X and Lagrange multiplier respectively;

[0138] The optimal sparse matrix solution is obtained by iterative solution until convergence.

[0139] In some embodiments, the signal processing module can effectively suppress the interference of environmental noise and significantly enhance the sparsity of the target signal by solving the optimization unit, thereby ensuring signal integrity.

[0140] In general, the signal processing module can not only separate multiple target signals, but also effectively suppress multipath interference. As a possible implementation method, the sparse modeling unit introduces multipath modeling technology to represent the multipath signal components as:

[0141]

[0142] in:

[0143] α k Indicates the direct reflected signal strength of the target;

[0144] β j Indicates the strength of multipath signal;

[0145] τ k , τ jdenote the propagation delays of direct signal and multipath signal respectively;

[0146] L represents the number of multipath signals.

[0147] The optimization unit estimates the multipath parameters through the Expectation Maximum (EM) algorithm and removes the multipath components to ensure the integrity of the direct reflection signal.

[0148] In a possible implementation, the signal processing module can also adapt to the changes in the number of different targets and signal strength by dynamically adjusting the parameters of the measurement matrix Φ. For example:

[0149] When the number of targets N is large, the sparse modeling unit can select a higher-dimensional measurement matrix to ensure the resolution of the target signal;

[0150] When the signal strength is weak, the optimization unit can appropriately increase the noise limit threshold ∈ to improve the robustness of the sparse solution.

[0151] Feature extraction module

[0152] The main function of the feature extraction module is to extract key features from the target signal output by the signal processing module, and provide data support for the subsequent classification and tracking module. In general, the feature extraction module includes a time-frequency analysis unit and a trajectory calculation unit, which are used to extract the time-frequency characteristics and motion trajectory characteristics of the target, respectively. These characteristics reflect the physical state and dynamic behavior of the target, and can provide a reliable basis for target classification and state prediction.

[0153] In this embodiment, the module includes a time-frequency analysis unit and a trajectory calculation unit.

[0154] In some embodiments, the time-frequency analysis unit is used to perform a short-time Fourier transform (STFT) on the target signal, thereby generating a joint distribution of the target signal in the time domain and the frequency domain. Specifically, the mathematical expression of the short-time Fourier transform is:

[0155] P(t,f)=|∫x(τ)e -j2πfτ h(t-τ)dτ| 2

[0156] in:

[0157] P(t, f) represents the time-frequency distribution of the target signal at time t and frequency f;

[0158] x(τ) represents the amplitude of the signal at time τ;

[0159] h(t-τ) represents the window function centered at time h(t-τ), which is used to capture the local characteristics of the signal;

[0160] e -j2πfτis the complex exponential kernel function of frequency domain transform;

[0161] dτ represents the small time increment of the integrated variable.

[0162] In general, the short-time Fourier transform can generate a time-frequency spectrum of the target signal, from which the peak frequency, bandwidth, and frequency drift of the target can be extracted. Specifically, the peak frequency corresponds to the main reflection characteristic of the target, the bandwidth reflects the complexity of the target's movement, and the frequency drift is used for subsequent speed calculation.

[0163] As an option, the time-frequency analysis unit can also perform feature optimization processing on the target signal. For example, in a noisy environment, a weighted window function (such as a Hanning window or a Hamilton window) can be selected as h(t-τ) to enhance the resolution of the target signal and reduce noise interference.

[0164] In some embodiments, by analyzing the time-frequency distribution of the target signal, the characteristic differences of multiple target signals can also be distinguished. For example, fast-moving targets (such as mice) have larger frequency drifts, while stationary targets (such as snakes) appear as low-frequency narrowband signals.

[0165] Generally, the trajectory calculation unit is used to generate the position, velocity and acceleration information of the target based on the time-varying characteristics of the target signal. As a possible implementation, the trajectory calculation unit calculates the motion parameters of the target by combining the point cloud data with a mathematical model.

[0166] The target's velocity v can be calculated using the following formula:

[0167]

[0168] in:

[0169] v represents the speed of the target;

[0170] Δf represents the frequency drift of the target signal;

[0171] f c Indicates the center frequency of the millimeter-wave radar.

[0172] The distance d to the target is calculated using the following formula:

[0173]

[0174] in:

[0175] d represents the distance of the target;

[0176] c represents the speed of light, and its specific value is 3×10 8 □m / s;

[0177] τ represents the total propagation time delay of the signal from the time it is transmitted to the time it is reflected back to the radar by the target.

[0178] Specifically, the trajectory calculation unit can further calculate the acceleration a and future position p(t+Δt) of the target based on the calculated target speed and distance. In some embodiments, the acceleration is calculated based on the continuous speed change rate, and the future position of the target can be predicted by the following formula:

[0179]

[0180] in:

[0181] p(t+Δt) represents the predicted position of the target at time t+Δt;

[0182] p(t) represents the position of the target at the current time t;

[0183] v(t) represents the velocity of the target at the current moment;

[0184] a(t) represents the acceleration of the target at the current moment;

[0185] Δt represents the time interval.

[0186] In one possible implementation, the trajectory calculation unit generates the continuous motion trajectory of the target in combination with the historical trajectory of the point cloud data. Generally, the process uses Kalman filtering to recursively update the motion parameters of the target to eliminate noise interference and enhance the stability of the motion trajectory. The prediction formula of Kalman filtering is:

[0187]

[0188] in:

[0189] represents the state prediction at the current time t;

[0190] F represents the state transfer matrix;

[0191] Represents the state estimate at the previous moment.

[0192] Classification tracking module

[0193] The main function of the classification tracking module is to classify and identify the target feature data output by the feature extraction module and dynamically track it. In general, the classification tracking module uses the classification unit to accurately determine the target type, and uses the tracking unit combined with the dynamic estimation algorithm to track the target position, speed and acceleration in real time, generating complete target trajectory data and providing accurate input for the alarm linkage module.

[0194] In this embodiment, the specific implementation of the classification tracking module includes a classification unit and a tracking unit. The functions and implementation contents of each unit are as follows:

[0195] In some embodiments, the classification unit uses a sparse Bayesian classifier (SBC) to classify the target. Specifically, the classification unit determines the category of the target by using the feature vector of the target. The target feature vector f includes time-frequency characteristics (such as peak frequency, frequency drift) and motion trajectory characteristics (such as speed, acceleration).

[0196] The classification unit uses the following sparse Bayesian classification formula to classify the target:

[0197]

[0198] in:

[0199] P(c k |f) represents the probability that the target belongs to the kth category;

[0200] f represents the feature vector of the target;

[0201] P(f|c k ) indicates that the feature vector belongs to category c k The conditional probability of

[0202] P(c k ) indicates category c k The prior probability of

[0203] K represents the total number of target categories;

[0204] represents the normalized probability of all categories.

[0205] Specifically, the sparse Bayesian classifier determines the category to which the target belongs by calculating the degree of match between the target features and the distribution of features of each category. As an option, the classification unit can dynamically adjust P(c k ) to adapt to the distribution characteristics of different targets. For example, if a certain type of target (such as rodents) has a higher invasion risk, a higher weight can be assigned to its prior probability.

[0206] In one possible implementation, the tracking unit uses Kalman filtering to dynamically update the position, velocity, and acceleration of the target based on the classification result of the classification unit to generate the complete motion trajectory data of the target. In general, the basic process of Kalman filtering includes two steps: prediction and update. The specific implementation is as follows:

[0207] 1. State prediction

[0208] The tracking unit estimates the next state of the target based on the current state of the target. The prediction formula is:

[0209]

[0210] P t|t-1 =FP t-1|t-1 F T +Q

[0211] in:

[0212] represents the state prediction at the current time t;

[0213] Represents the state estimate at the previous moment;

[0214] F is the state transfer matrix;

[0215] P t|t-1 is the covariance matrix of the predicted state;

[0216] P t-1|t-1 is the covariance matrix of the previous moment;

[0217] Q is the process noise covariance matrix.

[0218] 2. Status Update

[0219] The tracking unit combines the observed value of the target to correct the prediction result. The update formula is:

[0220]

[0221] P t|t =(IK t H)P t|t-1

[0222] in:

[0223] Indicates the status update result at the current moment;

[0224] K t is the Kalman gain matrix, defined as:

[0225] K t =P t|t-1 H T (HP t|t-1 H T +R) -1

[0226] z t Represents the observation value at the current moment;

[0227] H is the observation matrix;

[0228] R is the observation noise covariance matrix;

[0229] I is the identity matrix.

[0230] In general, the classification tracking module can combine the classification results and the dynamic tracking algorithm to generate the complete motion trajectory data of the target, providing high-quality input data for the alarm linkage module. In some embodiments, the tracking unit can also separately track targets in a multi-target environment to ensure that the trajectory information of different targets does not overlap. For example, by assigning a unique target identifier (ID), the tracking unit can calculate the status update of multiple targets separately, thereby achieving dynamic tracking of multiple targets.

[0231] Alarm linkage module

[0232] The main function of the alarm linkage module is to dynamically evaluate the threat level of the target and trigger corresponding alarms or defense measures based on the output results of the classification and tracking module. Generally, the alarm linkage module realizes the full process control from threat identification to defense execution through the collaborative work of the alarm unit and the linkage unit. Specifically, the alarm unit evaluates the threat level of the target and determines the alarm method, and the linkage unit accurately controls the action of the defense device based on the target trajectory prediction results.

[0233] In this embodiment, the alarm linkage module includes an alarm unit and a linkage unit.

[0234] In some embodiments, the alarm unit receives the target classification result and motion state parameters output by the classification tracking module, and triggers different alarm strategies according to the threat level of the target. Specifically, the threat level L is calculated by the following formula:

[0235] L=ω1P(c k )+ω2||v||+ω3d

[0236] in:

[0237] L represents the threat level score;

[0238] P(c k ) indicates that the target belongs to threat category c k The probability of , output by the sparse Bayes classifier;

[0239] ||v|| represents the speed of the target and is defined as:

[0240]

[0241] Among them, v x and v y are the velocity components of the target in the x and y directions respectively;

[0242] d represents the distance of the target from the radar;

[0243] ω1, ω2, and ω3 represent weight coefficients in the threat level calculation, which can be adjusted according to the actual threat level of different targets.

[0244] Generally, when the threat level L exceeds the preset threshold, the alarm unit will trigger a local sound and light alarm and send a notification signal to the remote monitoring platform, while recording the alarm information for subsequent analysis.

[0245] As an option, the alarm unit can also adopt a multi-level alarm strategy based on different threat level scores. For example:

[0246] When L is low, only alarm information is recorded;

[0247] When L is medium, a local alarm is triggered;

[0248] When L is higher, remote notification is triggered and defense devices are linked.

[0249] In some embodiments, the alarm unit may also combine the threat level calculation results of multiple targets to generate a comprehensive threat assessment report for determining the overall security situation of the computer room.

[0250] In general, the linkage unit controls the action of the defense device (such as a trap or a repelling device) based on the instructions of the alarm unit and the target trajectory prediction result. The trajectory prediction result is calculated by the following formula:

[0251]

[0252] in:

[0253] p(t+Δt) represents the predicted target position;

[0254] p(t) represents the current target position, which is updated in real time by the tracking unit;

[0255] v(t) represents the current speed of the target;

[0256] a(t) represents the current acceleration of the target;

[0257] Δt represents the time interval.

[0258] Specifically, the linkage unit adjusts the activation time and capture position of the defense device according to the predicted target position. For example, when the target is close to the trapping device, the linkage unit will calculate the trigger time required for the device to be activated in advance to ensure that the action accurately matches the target position, thereby increasing the capture success rate of the defense device.

[0259] In some embodiments, the linkage unit can also update the trajectory prediction parameters of the target in real time to cope with changes in the target trajectory. For example, when the target accelerates, the linkage unit increases the sampling frequency of the time interval Δt to improve the accuracy of trajectory prediction, thereby optimizing the response time of the defense device.

[0260] As a possible implementation method, the linkage unit can also interact with the remote monitoring platform and upload the execution results of each action to the platform for archiving and analysis. For example:

[0261] If the defense device successfully captures the target, the linkage unit will record the target's classification information and capture time;

[0262] If the defense device fails to capture the target, the linkage unit will adjust the defense strategy to deal with similar situations.

[0263] Specifically, the alarm linkage module realizes closed-loop control from threat assessment to defense execution through the synergy of the alarm unit and the linkage unit. Generally speaking, this module can not only meet the defense needs of a single target, but also work efficiently in multi-target scenarios. For example, when multiple targets appear at the same time, the alarm linkage module will respond to high-threat targets first according to the threat level of each target, and dynamically adjust the action order of the defense device.

[0264] The invention discloses a computer room wildlife intelligent defense system based on millimeter wave radar, which realizes real-time monitoring, classification and identification, dynamic tracking and intelligent defense of wild animals in the computer room through the coordinated work of a signal capture module, a signal processing module, a feature extraction module, a classification and tracking module and an alarm linkage module.

[0265] The following is the detailed workflow of the system:

[0266] 1. Signal capture

[0267] After the system is started, the signal capture module uses the millimeter wave radar to transmit a linear frequency modulated continuous wave (FMCW) signal to the monitoring area of ​​the computer room. After the signal is reflected by the target during the propagation process, the reflected signal is collected by the receiving unit.

[0268] The reflected signal is processed by the signal mixing unit to generate an intermediate frequency signal containing target distance, speed and direction information, and the intermediate frequency signal is transmitted to the signal processing module.

[0269] 2. Signal processing

[0270] After receiving the intermediate frequency signal, the signal processing module first constructs a sparse representation model of the target signal through the sparse modeling unit to separate the reflected signals of different targets.

[0271] Next, the optimization unit uses an iterative optimization algorithm to extract the sparse signal matrix and suppress multipath interference and environmental noise at the same time. The processed signal contains the basic characteristics of the target and provides high-quality input for the feature extraction module.

[0272] 3. Feature extraction

[0273] The feature extraction module analyzes the target signal output by the signal processing module. First, the time-frequency analysis unit generates the time-frequency distribution diagram of the target signal using short-time Fourier transform and extracts characteristics such as peak frequency, bandwidth and frequency drift.

[0274] Subsequently, the trajectory calculation unit calculates the target's motion parameters, including position, velocity, and acceleration, based on the target's time-varying characteristics, and generates the target's continuous motion trajectory data. The output of this module is the target's feature vector and trajectory information, which is used by the classification and tracking module.

[0275] 4. Classification tracking

[0276] After the classification tracking module receives the output of the feature extraction module, the classification unit first classifies the target. The classification unit uses a sparse Bayesian classifier to determine the type of the target (such as rodents, snakes, etc.) using the target's feature vector and outputs the probability value of the target's category.

[0277] Subsequently, the tracking unit tracks the target in real time based on the classification results and target trajectory information, combined with the dynamic estimation algorithm, to generate complete motion trajectory data. The output of this module is the classification results and motion state information of the target, which is transmitted to the alarm linkage module.

[0278] 5. Alarm linkage

[0279] The alarm linkage module evaluates the threat level of the target based on the output data of the classification tracking module. The alarm unit calculates the threat level score based on the classification result, speed and distance of the target, and triggers the corresponding alarm strategy based on the score, such as local sound and light alarm, remote notification or recording information.

[0280] At the same time, the linkage unit controls the action of the defense device according to the target's trajectory prediction results, determines the device's start time and trigger position, and achieves the capture or expulsion of the target. The execution results are recorded in real time and uploaded to the remote monitoring platform for subsequent analysis and strategy adjustment.

[0281] In general, the signal capture module of the present invention uses millimeter wave radar to generate and receive target reflection signals, the signal processing module extracts target characteristics through sparse modeling and optimization, the feature extraction module analyzes the time-frequency characteristics and motion trajectory of the target, the classification and tracking module identifies and dynamically tracks the target, and the alarm linkage module triggers the alarm and links the defense device according to the threat level. The system has high-precision, multi-target monitoring capabilities, can classify, track and execute intelligent defense in real time, and ensure the safe operation of the computer room.

[0282] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The intelligent defense system for wild animals in the computer room based on millimeter wave radar is characterized by: It includes signal capture module, signal processing module, feature extraction module, classification tracking module and alarm linkage module; The signal capture module is used to capture the millimeter wave radar reflection signal from the target through the intermediate frequency signal model; The signal processing module is connected to the signal acquisition module and is used to perform multi-target separation and interference suppression on the target reflection signal; The feature extraction module is connected to the signal processing module and is used to extract the time-frequency features and motion trajectory features of the target; The classification and tracking module is connected to the feature extraction module and is used to classify the target and track its motion state according to the extracted features; The alarm linkage module is connected to the classification tracking module and is used to trigger an alarm signal according to the classification result and to link the execution of the defense device.

2. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 1 is characterized in that: The signal capture module adopts a linear frequency modulation continuous wave millimeter wave radar, including a transmitting unit, a receiving unit and a signal mixing unit, wherein: The transmitting unit is used to transmit a linear frequency modulation continuous wave signal and capture the distance and speed of the target based on the target distance and speed formula; The receiving unit is used to receive the target reflected echo signal and perform preliminary filtering processing to suppress high-frequency noise; The signal mixing unit is used to mix the target reflection signal and the transmission signal to generate an intermediate frequency signal as an output, and the intermediate frequency signal retains the distance and Doppler information of the target.

3. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 2 is characterized in that: The intermediate frequency signal model is: Among them, r(t) is the intermediate frequency signal generated after the target reflection signal is received and mixed, K is the number of targets, α k is the reflection intensity of the kth target, f k is the Doppler frequency, φ k is the phase, n(t) is the noise; The target distance speed formula for calculating the distance and speed of the target is: Where B is the FM bandwidth, T is the FM period, and d k is the distance to the target, v k is the speed of the target, c is the speed of light, f c is the center frequency of the millimeter wave radar, Δf k is the target frequency drift.

4. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 1 is characterized in that: The signal processing module includes a sparse modeling unit and an optimization unit; The sparse modeling unit is used to represent multiple target signals in the intermediate frequency signal as a sparse signal matrix by establishing a joint sparse matrix; The optimization unit is used to extract and separate the target signal based on an iterative optimization algorithm, optimize the extracted and separated target signal based on a sparse representation model, and output a sparse coefficient matrix of the target signal.

5. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 4 is characterized in that: The sparse representation model is: Y=ΦX+N The iterative optimization algorithm is: in, represents the received intermediate frequency signal matrix, M is the number of measurement signals, T is the number of time sampling points, N is the number of target signal characteristics, and the intermediate frequency signal matrix includes a measurement matrix, a sparse signal matrix and a background noise matrix; when When , Φ represents the measurement matrix; when When , X represents the sparse signal matrix; when When , N represents the background noise matrix; To extract and separate the target signal, represents a joint sparse constraint, subject to represents Subordinate to ∈ represents the noise limit threshold, which is used to ensure that the background noise does not interfere with the extraction of the target signal.

6. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 1 is characterized in that: The feature extraction module includes a time-frequency analysis unit and a trajectory calculation unit; The time-frequency analysis unit is used to perform time-frequency characteristic analysis on the target signal by a short-time Fourier transform algorithm to generate a frequency spectrum distribution of the target signal, wherein the frequency spectrum distribution of the target signal includes a peak frequency, a bandwidth and a frequency drift characteristic of the target; The trajectory calculation unit is used to generate the position, velocity and acceleration information of the target through the time variation characteristics in the target signal, and provide a motion trajectory description.

7. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 6 is characterized in that: The short-time Fourier transform algorithm is: P(t,f)=|∫x(τ)e -j2πfτ h(t-τ)dτ| 2 Where P(t,f) represents the time-frequency distribution of the target signal at time t and frequency f, x(τ) represents the amplitude of the signal at time τ, h(t-τ) represents the window function centered at time t, and e -j2πfτ is the complex exponential kernel function of the frequency domain transform, dτ represents the small time increment of the integral variable; The trajectory calculation unit calculates the speed and distance of the target based on the point cloud data, wherein the calculation formula of the speed of the target is: The calculation formula of the distance of the target is: Where v represents the speed of the target, Δf represents the frequency drift of the target signal, and f c represents the center frequency of the millimeter-wave radar, d represents the distance of the target, c represents the speed of light, and τ represents the total propagation time delay of the signal from the time it is transmitted to the time it is reflected back to the radar by the target.

8. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 1 is characterized in that: The classification and tracking module includes a classification unit and a tracking unit; The classification unit is used to judge the target category of the target feature vector by using a sparse Bayesian classification formula based on a sparse Bayesian classifier; The tracking unit is used to dynamically update the target state based on the judgment result of the target category of the target feature vector by using Kalman filtering, and continuously update the position, speed and acceleration of the target by combining a dynamic estimation algorithm to generate the target's trajectory data.

9. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 8 is characterized in that: The sparse Bayesian classification formula is: Among them, P(c k |f) represents the probability that the target belongs to the kth class, f represents the feature vector of the target, and the feature vector of the target includes the time-frequency characteristics and motion trajectory characteristics of the target. k ) indicates that the feature vector belongs to category c k The conditional probability, P(c k ) indicates category c k The prior probability of , K represents the total number of target categories, represents the normalized probability of all categories.

10. The intelligent defense system for wild animals in a computer room based on millimeter wave radar according to claim 1 is characterized in that: The alarm linkage module includes an alarm unit and a linkage unit; The alarm unit receives the target classification result, and is used to obtain the threat level of the target through a threat level calculation formula, and triggers a local alarm, a remote notification, or automatically records the alarm information according to the threat level of the target. The threat level calculation formula is: L=ω1P(c k )+ω2∥v∥+ω3d Where L represents the threat level, P(c k ) indicates that the target belongs to threat category c k The probability of , ∥v∥ represents the speed of the target, defined as v x and v y are the velocity components of the target in the x- and y-axis directions, d represents the distance of the target from the radar, ω1, ω2, ω3 represent the weight coefficients in the threat level calculation; The linkage unit is used to obtain the trajectory prediction of the target based on the trajectory prediction model, and determine the start time and position of the defense device according to the trajectory prediction result of the target, and the trajectory prediction model is: Among them, p(t+Δt) represents the predicted target position, p(t) represents the current target position, v(t) represents the current velocity of the target, a(t) represents the current acceleration of the target, and (Δt) represents the time interval.

Images