Metallic cargo hold live detection method and system

Abstract

This invention provides a method and system for liveness detection within a metal chamber, relating to the field of detection technology. The method includes: simultaneously acquiring multi-source heterogeneous sensor data streams within the chamber using a multi-band microwave radar, an infrared thermal imaging array, a broadband ultrasonic sensor, and a high-sensitivity micro-vibration sensor array deployed inside the metal chamber; and, based on the multi-source heterogeneous sensor data streams, employing an adaptive signal enhancement algorithm based on frequency domain subband decomposition to suppress reverberation noise and multipath interference caused by metal cavity resonance, outputting a coherently enhanced target signal cluster. This invention improves the accuracy of liveness detection.

Description

Technical Field

[0001] This invention relates to the field of detection technology, and in particular to a method and system for detecting live organisms inside a metal chamber. Background Technology

[0002] Traditional liveness detection methods are mostly applicable to open or non-metallic environments, and some exhibit limitations in metallic cabins. For example, detection techniques based on a single sensor, such as microwave radar detection, are susceptible to reverberation noise and multipath interference generated by the resonance of the metal cavity, causing the target signal to be submerged in noise and making it difficult to accurately extract weak vital signs such as breathing and heartbeat. Infrared thermal imaging technology may cause thermal image distortion due to the strong thermal reflection characteristics of metal surfaces, making it impossible to effectively distinguish live targets from the environmental background. Ultrasonic sensors are prone to standing wave interference caused by multiple reflections of sound waves in a metal enclosed space, leading to misjudgments of the movement status of live objects.

[0003] Although multi-sensor fusion-based liveness detection solutions have emerged in existing technologies, several challenges remain in the unique scenario of a metal cabin. Firstly, multi-source heterogeneous sensor data (such as microwave, infrared, ultrasonic, and vibration signals) exhibit varying attenuation characteristics and noise distribution patterns in a metallic environment. Traditional signal enhancement algorithms struggle to adaptively suppress environmental noise, resulting in insufficient accuracy in extracting the effective target signal. Secondly, existing feature fusion methods often employ simple weighting or splicing, failing to fully consider the differences in feature reliability among different sensors in a metallic environment. This hinders the highlighting of the contribution of key vital signs, leading to weak discriminative power in the joint feature vector. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for detecting live bodies in a metal chamber, which improves the accuracy of live body detection.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] A first aspect includes a method for detecting live organisms inside a metal chamber, the method comprising:

[0007] Step 1: Simultaneously collect multi-source heterogeneous sensor data streams inside the cabin by deploying multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensor and high-sensitivity micro-vibration sensor group inside the metal cabin.

[0008] Step 2: Based on the multi-source heterogeneous sensing data stream, an adaptive signal enhancement algorithm based on frequency domain subband decomposition is adopted to suppress reverberation noise and multipath interference caused by metal cavity resonance and output coherently enhanced target signal clusters.

[0009] Step 3: Based on the coherent enhanced target signal cluster, the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target are extracted by time-frequency analysis technology, and a joint feature vector is generated by an attention mechanism weighted feature fusion method.

[0010] Step 4: Based on the joint feature vector, analyze the energy distribution characteristics of the live target in the cabin and dynamically construct a multi-sensor coverage confidence matrix; perform singular value decomposition on the confidence matrix and combine it with an adaptive bee colony division optimization algorithm, using the sensor weight allocation parameters as the bee colony's foraging targets, with each bee representing a set of candidate weight parameters. Utilize the division of labor and cooperation of the bee colony and dynamically adjust the search strategy in conjunction with the live energy distribution characteristics in the cabin. Through iterative optimization, the weight parameters converge to the final solution, generating the sensor weight allocation parameters.

[0011] Step 5: Based on the sensor weight allocation parameters and joint feature vector, construct a metal environment-adaptive decision model, adaptively adjust the sensor sampling strategy, and output the probability of the presence of a living organism and quantitative indicators of vital signs.

[0012] Furthermore, multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensor, and high-sensitivity micro-vibration sensor array deployed inside the metal cabin are used to simultaneously acquire multi-source heterogeneous sensor data streams within the cabin, including:

[0013] Step 1.1: Generate a three-dimensional spatial coordinate mapping table based on the structural parameters of the metal cabin, and deploy the sensor network according to the layout of microwave radar covering the longitudinal section, infrared array covering the horizontal section, ultrasonic sensor covering the corner blind area, and micro-vibration sensor group attached to the cabin wall.

[0014] Step 1.2: Based on the three-dimensional spatial coordinate mapping table, start the multi-sensor hardware synchronization triggering mechanism, generate a global timestamp through a high-precision clock source, and strictly align the microwave radar pulse transmission phase, infrared array exposure period, ultrasonic wave transmission sequence and vibration sampling time in the time domain to obtain the original sensing data.

[0015] Step 1.3: Receive the raw sensing data, and based on the sensor position parameters recorded in the three-dimensional spatial coordinate mapping table, add a cabin material attenuation compensation coefficient to the microwave radar signal, add an ambient temperature drift correction parameter to the infrared thermal imaging data, add an air density compensation factor to the ultrasonic signal, and add a mechanical coupling gain coefficient to the micro-vibration signal to generate a primary data stream with physical compensation.

[0016] Step 1.4: The primary data stream is matched to the position according to the three-dimensional spatial coordinate mapping table, wherein microwave radar data is mapped to the polar coordinate system, infrared data is mapped to the Cartesian coordinate system, ultrasonic data is mapped to the spherical coordinate system, and micro-vibration data is mapped to the structural grid coordinate system, and the spatiotemporally correlated multi-source heterogeneous sensing data stream is output.

[0017] Furthermore, based on the multi-source heterogeneous sensing data stream, an adaptive signal enhancement algorithm based on frequency domain subband decomposition is employed to suppress reverberation noise and multipath interference caused by metal cavity resonance, outputting a coherently enhanced target signal cluster, including:

[0018] Step 2.1: Based on the multi-source heterogeneous sensing data stream, after millimeter-wave penetration compensation is performed on the microwave radar signal, the entire frequency band is decomposed into multiple sub-bands according to the resonant mode distribution of the metal cabin; using the decomposition result, the ultrasonic signal is divided into critical frequency bands according to the attenuation coefficient of the sound wave in the metal medium; further, based on the cabin structure transfer function, the micro-vibration signal is grouped into octave bands to match the mechanical vibration propagation characteristics.

[0019] Step 2.2: Based on the sub-band, frequency band, and grouping results, construct a coherence matrix across sensor channels within each sub-band, and separate the reverberation noise subspace through eigenvalue decomposition; dynamically generate sub-band weighting coefficients with the phase continuity of the target signal cluster between adjacent sub-bands as a constraint.

[0020] Step 2.3: Based on the sub-band signal with weighted coefficients, extract the time delay characteristic peaks formed by multiple reflections from the metal wall, calculate the anti-phase cancellation beam to cancel multipath interference; use spatiotemporal adaptive filtering to eliminate specular reflection artifacts on the metal surface; output the sub-band signals after interference suppression.

[0021] Step 2.4: Reconstruct the time-domain signal clusters from each sub-band signal through inverse transformation, and calculate the mutual information entropy value between the reconstructed signal and the original sensing data; if the entropy value is lower than the dynamic threshold, output the target signal cluster for coherent enhancement; otherwise, trigger the data re-acquisition mechanism for closed-loop correction.

[0022] Furthermore, based on the coherently enhanced target signal cluster, time-frequency analysis techniques are used to extract the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target. An attention-weighted feature fusion method is then employed to generate a joint feature vector, including:

[0023] Step 3.1: Based on the coherent enhanced target signal cluster, extract the microwave radar and micro-vibration signals; separate the respiratory harmonic components through time-frequency analysis technology, and generate the time-domain periodic feature vector and frequency-domain main harmonic energy ratio feature vector of the respiratory waveform envelope by using envelope detection and adaptive baseline drift correction.

[0024] Step 3.2: Based on the respiratory feature extraction results and corresponding signal segments, the cardiac impact component is separated from the microwave radar signal using cardiac harmonic reconstruction technology; at the same time, the micro-vibration signal is frequency-domain compensated in combination with the vibration transmission characteristics of the cabin structure, and the fundamental frequency stability feature vector and harmonic complexity feature vector of the cardiac impact spectrum are extracted.

[0025] Step 3.3: Based on the cardiac impact feature extraction results and the processed signal clusters, the infrared thermal imaging signal is dynamically segmented to eliminate metal thermal reflection interference, and the time-frequency feature vector of the live hot spot displacement trajectory is extracted; simultaneously, the output ultrasonic signal is used to analyze the short-time zero-crossing rate characteristics of its Doppler frequency shift, and a motion pattern classification feature vector is generated.

[0026] Step 3.4: Construct a feature quality assessment matrix based on the respiratory feature vector, cardiac impact feature vector, and motion feature vector; dynamically calculate the weights of respiratory features, cardiac impact features, and motion features by using the intra-class dispersion of respiratory features, the inter-class separability of cardiac impact features, and the confidence of motion features as the basis for attention weight allocation; and concatenate the weighted feature vectors into a joint feature vector and normalize it.

[0027] Furthermore, based on the joint feature vector, the energy distribution characteristics of the living target within the cabin are analyzed, and a multi-sensor coverage confidence matrix is ​​dynamically constructed, including:

[0028] Step 4.1: Receive the joint feature vector, extract the respiratory waveform envelope features, cardiac impact spectrum features and motion pattern time-frequency feature components from it, and map these feature components to the three-dimensional spatial discretized grid nodes of the metal cabin to generate feature distribution data with spatial location annotations.

[0029] Step 4.2: Based on the feature distribution data with spatial location annotations, dynamic thermodynamic data reflecting the spatial energy density of the living target is generated by calculating the temporal stability index and spatial continuity measure of the feature amplitude of each grid node.

[0030] Step 4.3: Using the distribution information of each energy peak region in the dynamic thermal data, combined with the preset sensor position parameters and their signal attenuation physical laws, calculate the coverage coefficient of each sensor for each spatial grid node.

[0031] Step 4.4: Normalize the capability coefficients and construct an initial coverage confidence matrix with spatial grid nodes as row indices and sensor channels as column indices. The matrix element values ​​represent the credible monitoring weights of the corresponding sensors at specific locations.

[0032] Step 4.5: Based on the metal obstruction compensation parameters provided by the pre-set database of the cabin structure, perform multipath reflection loss correction on the initial coverage confidence matrix and output the optimized multi-sensor coverage confidence matrix.

[0033] Furthermore, singular value decomposition is performed on the confidence matrix, and combined with an adaptive bee colony division optimization algorithm, the sensor weight allocation parameters are used as the foraging targets of the bee colony. Each bee represents a set of candidate weight parameters. Utilizing the division of labor and cooperation within the bee colony, and combining the energy distribution characteristics of living organisms within the capsule, the search strategy is dynamically adjusted. Through iterative optimization, the weight parameters converge to the final solution, generating the sensor weight allocation parameters, including:

[0034] Step 4.6: Perform singular value decomposition on the multi-sensor coverage confidence matrix to extract the principal component vectors characterizing the spatial coverage characteristics of the sensors and their corresponding singular value spectrum distributions.

[0035] Step 4.7: Based on the characteristic peak interval of the singular value spectrum distribution, determine the initial search boundary of the sensor weight parameters, and map each weight parameter combination to the position coordinates of individual bees in the bee colony optimization.

[0036] Step 4.8: Based on the cabin space coverage characteristics reflected by the principal component vectors and combined with the spatial gradient changes of the living body energy distribution, the search area is divided into a high-gradient fine search area and a low-gradient broad exploration area to obtain the partitioning results.

[0037] Step 4.9: Dynamically configure the roles of the bee colony based on the partitioning results. Deploy scout bees to high gradient regions to perform neighborhood depth optimization; deploy follower bees to low gradient regions to perform heat tracking exploration; after each iteration, redistribute the region search task based on the fitness of the weight parameters. When the change in weight parameters in consecutive iterations is lower than the preset convergence threshold, extract the weight allocation parameters corresponding to the current optimal bee position coordinates.

[0038] Furthermore, based on sensor weight allocation parameters and joint feature vectors, a metal environment-adaptive decision model is constructed to adaptively adjust the sensor sampling strategy and output quantitative indicators of the probability of presence of living organisms and vital signs, including:

[0039] Step 5.1: Receive the sensor weight allocation parameters, use them as input layer neurons to initialize weights and inject them into the decision model, and simultaneously load the joint feature vector into the model feature input channel;

[0040] Step 5.2: Based on the constructed initial decision model, call the historical working condition database of the cabin to build a training sample set, and fine-tune the hidden layer structure parameters of the model through the transfer learning mechanism to make the model adapt to the feature distribution shift in the metal cavity environment and generate a metal environment-adaptive decision model.

[0041] Step 5.3: Based on the metal environment-adaptive decision model, monitor the activation state of its hidden layer in real time, generate sensor resource allocation instructions according to the spatial distribution pattern of the activation state, and dynamically allocate microwave radar scanning frame rate, infrared array resolution, ultrasonic transmission power and vibration sensor sensitivity level according to the weight parameter ratio.

[0042] Step 5.4: Collect new sensor data streams according to the adaptive sampling strategy, input the real-time data into the metal environment-adaptive decision model, update the model parameters through the online incremental learning mechanism, and output the probability value of the existence of living organisms in the cabin and the quantitative indicators of vital signs composed of respiratory rate, heart rate coefficient of variation, and body movement intensity.

[0043] Step 5.5: When the probability value of the presence of a living organism exceeds a preset threshold, analyze the quantitative indicators of vital signs corresponding to the probability value and generate a three-dimensional parameter analysis report including respiratory rate abnormality, heart rate variability stability, and body movement intensity trend.

[0044] Secondly, a liveness detection system for a metal chamber includes:

[0045] The acquisition module is used to simultaneously acquire multi-source heterogeneous sensor data streams inside the metal cabin through a multi-band microwave radar, infrared thermal imaging array, wideband ultrasonic sensor, and high-sensitivity micro-vibration sensor group deployed inside the cabin; and to output coherently enhanced target signal clusters by using an adaptive signal enhancement algorithm based on frequency domain subband decomposition based on the multi-source heterogeneous sensor data streams.

[0046] The processing module is used to extract the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target based on the coherent enhanced target signal cluster through time-frequency analysis technology, and to generate a joint feature vector using an attention mechanism weighted feature fusion method.

[0047] The computation module is used to analyze the energy distribution characteristics of the live target in the cabin based on the joint feature vector, dynamically construct the multi-sensor coverage confidence matrix, perform singular value decomposition on the confidence matrix, and combine it with the adaptive bee colony division optimization algorithm to use the sensor weight allocation parameters as the foraging target of the bee colony. Each bee represents a set of candidate weight parameters. By utilizing the division of labor and cooperation of the bee colony and combining the live energy distribution characteristics in the cabin, the search strategy is dynamically adjusted. Through iterative optimization, the weight parameters converge to the final solution, and the sensor weight allocation parameters are generated.

[0048] The intelligent decision-making module is used to construct a metal environment-adaptive decision-making model based on sensor weight allocation parameters and joint feature vectors, adaptively adjust sensor sampling strategies, and output quantitative indicators of the probability of the presence of living organisms and vital signs.

[0049] Thirdly, a computing device includes:

[0050] One or more processors;

[0051] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0052] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0053] The above-described solution of the present invention has at least the following beneficial effects:

[0054] By coordinating and synchronously acquiring data from a multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensor, and high-sensitivity micro-vibration sensor group, combined with three-dimensional spatial coordinate mapping and physical compensation mechanisms, the limitations of single sensor coverage and insufficient data correlation within the metal cabin are effectively addressed, ensuring the spatiotemporal consistency and accuracy of multi-source heterogeneous sensor data streams. An adaptive signal enhancement algorithm based on frequency domain subband decomposition specifically suppresses reverberation noise and multipath interference caused by metal cavity resonance, significantly improving the coherence of the target signal. An attention-weighted feature fusion method accurately extracts key living features such as respiration, cardiac impact, and motion from the enhanced signal and dynamically assigns weights, highlighting the recognizability of vital signs. By constructing a multi-sensor coverage confidence matrix and iteratively optimizing sensor weight parameters using singular value decomposition and adaptive bee colony optimization algorithms, intelligent allocation and synergistic efficiency of sensor resources are achieved. A metal environment-adaptive decision model can adaptively adjust sampling strategies, adapting to metal environment feature shifts through transfer learning and online incremental learning, accurately outputting the probability of living presence. Attached Figure Description

[0055] Figure 1 This is a schematic flowchart of a live body detection method in a metal chamber provided by an embodiment of the present invention.

[0056] Figure 2 This is a schematic diagram of a live body detection system in a metal chamber provided by an embodiment of the present invention. Detailed Implementation

[0057] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0058] like Figure 1 As shown, an embodiment of the present invention proposes a method for detecting live organisms inside a metal chamber, the method comprising the following steps:

[0059] Step 1: Simultaneously collect multi-source heterogeneous sensor data streams inside the cabin by deploying multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensor and high-sensitivity micro-vibration sensor group inside the metal cabin.

[0060] Step 2: Based on the multi-source heterogeneous sensing data stream, an adaptive signal enhancement algorithm based on frequency domain subband decomposition is adopted to suppress reverberation noise and multipath interference caused by metal cavity resonance and output coherently enhanced target signal clusters.

[0061] Step 3: Based on the coherent enhanced target signal cluster, the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target are extracted by time-frequency analysis technology, and a joint feature vector is generated by an attention mechanism weighted feature fusion method.

[0062] Step 4: Based on the joint feature vector, analyze the energy distribution characteristics of the live target in the cabin and dynamically construct a multi-sensor coverage confidence matrix; perform singular value decomposition on the confidence matrix and combine it with an adaptive bee colony division optimization algorithm, using the sensor weight allocation parameters as the bee colony's foraging targets, with each bee representing a set of candidate weight parameters. Utilize the division of labor and cooperation of the bee colony and dynamically adjust the search strategy in conjunction with the live energy distribution characteristics in the cabin. Through iterative optimization, the weight parameters converge to the final solution, generating the sensor weight allocation parameters.

[0063] Step 5: Based on the sensor weight allocation parameters and joint feature vector, construct a metal environment-adaptive decision model, adaptively adjust the sensor sampling strategy, and output the probability of the presence of a living organism and quantitative indicators of vital signs.

[0064] In this embodiment of the invention, efficient and accurate liveness detection is achieved through multi-dimensional technical means. The collaborative acquisition of multi-source heterogeneous sensors ensures comprehensive coverage of the cabin space and spatiotemporal correlation of data, solving the detection limitations of a single sensor in a metallic environment. The signal enhancement algorithm based on frequency domain subband decomposition effectively suppresses reverberation noise and multipath interference caused by metal resonance, improving the clarity of the target signal. The attention mechanism-weighted feature fusion enhances the recognition of key liveness features such as breathing, cardiac impact, and movement, avoiding interference from invalid features. The dynamic optimization of sensor weights achieves efficient collaboration of multi-source data through bee colony algorithm. Finally, the intelligent decision model, combined with adaptive sampling strategy and online learning, realizes accurate detection throughout the entire process from liveness identification to state assessment in a metallic cabin.

[0065] In a preferred embodiment of the present invention, step 1 above may include:

[0066] Step 1.1: Generate a three-dimensional spatial coordinate mapping table based on the structural parameters of the metal cabin, and deploy the sensor network according to the layout of microwave radar covering the longitudinal section, infrared array covering the horizontal section, ultrasonic sensor covering the corner blind area, and micro-vibration sensor group attached to the cabin wall.

[0067] Step 1.2: Based on the three-dimensional spatial coordinate mapping table, start the multi-sensor hardware synchronization triggering mechanism, generate a global timestamp through a high-precision clock source, and strictly align the microwave radar pulse transmission phase, infrared array exposure period, ultrasonic wave transmission sequence and vibration sampling time in the time domain to obtain the original sensing data.

[0068] Step 1.3: Receive the raw sensing data, and based on the sensor position parameters recorded in the three-dimensional spatial coordinate mapping table, add a cabin material attenuation compensation coefficient to the microwave radar signal, add an ambient temperature drift correction parameter to the infrared thermal imaging data, add an air density compensation factor to the ultrasonic signal, and add a mechanical coupling gain coefficient to the micro-vibration signal to generate a primary data stream with physical compensation.

[0069] Step 1.4: The primary data stream is matched to the position according to the three-dimensional spatial coordinate mapping table, wherein microwave radar data is mapped to the polar coordinate system, infrared data is mapped to the Cartesian coordinate system, ultrasonic data is mapped to the spherical coordinate system, and micro-vibration data is mapped to the structural grid coordinate system, and the spatiotemporally correlated multi-source heterogeneous sensing data stream is output.

[0070] In this embodiment of the invention, a high-quality data foundation is laid for liveness detection within a metal chamber through scientific layout and precise processing: multiple types of sensors are deployed in zones according to the structural characteristics of the chamber, and a three-dimensional coordinate mapping table is used to achieve comprehensive coverage of the longitudinal, horizontal, corner, and chamber walls, eliminating blind spots; a high-precision clock synchronization mechanism ensures strict alignment of the time domain of multiple sensors, avoiding data time deviation; physical compensation coefficients such as material attenuation and temperature drift are added to correct signal distortion for the characteristics of the metal environment; and data from different coordinate systems are uniformly mapped to a spatiotemporal correlation framework to achieve the organic fusion of multi-source heterogeneous data. The overall process ensures both the comprehensiveness of data acquisition and spatiotemporal consistency.

[0071] In this embodiment of the invention, the specific steps include:

[0072] Step 1.1: Based on the structural parameters of the metal cabin, such as length, width, height, and cabin wall thickness, construct a three-dimensional spatial model of the cabin interior, divide it into mesh units, and generate a three-dimensional spatial coordinate mapping table that records the coordinates of each spatial point; according to the principle of "microwave radar covering the longitudinal section, infrared array covering the horizontal section, ultrasonic blind spot filling, and micro-vibration attached to the cabin wall", calculate the optimal installation coordinates of each sensor (e.g., microwave radar installed on the longitudinal axis of the middle section of the cabin, infrared array installed at the top center, etc.), and write them into the mapping table.

[0073] Step 1.2: Generate a global time base using a high-precision clock source (such as a 10MHz temperature-controlled crystal oscillator) and calculate the trigger time of each sensor:

[0074] The trigger time of the microwave radar pulse transmission phase = global timestamp + radar channel delay compensation value;

[0075] The start time of the infrared array exposure cycle = global timestamp + infrared detector response delay value;

[0076] The start time of the ultrasonic emission sequence = global timestamp + sound wave propagation warm-up time;

[0077] The sampling time of the micro-vibration sensor is equal to the global timestamp (without additional delay, as vibration needs to be captured in real time).

[0078] The aforementioned delay compensation values ​​were pre-determined through calibration experiments to ensure that the time-domain alignment error of each sensor is ≤10 microseconds.

[0079] Step 1.3, Carrier Material Attenuation Compensation Coefficient (Microwave Radar): First, multiply the four values: material attenuation constant, square root of microwave frequency, square root of electromagnetic conductivity of carrier metal, and path length of microwave signal penetrating metal, to obtain a product result; then divide the product result by 20 to obtain a new value; finally, use 10 as the base and this new value as the exponent to perform a power operation, and the result is the carrier material attenuation compensation coefficient.

[0080] Ambient temperature drift correction parameter (infrared thermal imaging): First, calculate the difference between the real-time ambient temperature and the reference temperature; multiply this difference by the temperature drift coefficient of the infrared sensor to obtain a product result; then add 1 to the product result, and the final value is the ambient temperature drift correction parameter.

[0081] Ultrasonic air density compensation factor: Calculate the real-time air density, and divide the obtained real-time air density by the air density under standard conditions (1.225 kg / m³). The quotient is the ultrasonic air density compensation factor.

[0082] Mechanical coupling gain coefficient (micro-vibration sensor): The mechanical coupling gain coefficient is the product of the coupling coefficient, the contact area between the sensor and the bulkhead, and the elastic modulus of the coupling material.

[0083] Step 1.4: Map microwave radar data to polar coordinates:

[0084] The coordinates in the polar coordinate system are represented by (r, θ), where r is the radar target distance after compensation for attenuation caused by the material of the carrier; and θ is the radar scanning angle directly obtained by the radar motor rotation sensor.

[0085] Infrared thermal imaging data is mapped to a Cartesian coordinate system: coordinates in a Cartesian coordinate system are represented by (x, y). The value of x is calculated as: (the column number of the target pixel in the infrared array minus the column number of the array center) multiplied by the size of a single pixel.

[0086] The calculation method for y is: (the row number of the target pixel in the infrared array minus the row number of the array center) multiplied by the size of a single pixel;

[0087] The above calculation results need to be corrected for coordinate offset based on the installation position of the infrared array to finally obtain the coordinates in the Cartesian coordinate system.

[0088] Ultrasonic sensor data is mapped to a spherical coordinate system: the coordinates in the spherical coordinate system are represented by (r, φ, θ). Here, r is the ultrasonic ranging value after air density compensation; φ is the horizontal rotation angle of the ultrasonic sensor; and θ is the vertical elevation angle of the ultrasonic sensor.

[0089] Micro-vibration sensor data is mapped to a structural grid coordinate system: the coordinates in the structural grid coordinate system are represented by (i, j), where i corresponds to the row index of the grid and j corresponds to the column index of the grid. The specific coordinate value is determined by the structural grid node index corresponding to the sensor installation location, and the signal value collected by the micro-vibration sensor will be directly mapped to the grid node corresponding to (i, j).

[0090] In a preferred embodiment of the present invention, step 2 above may include:

[0091] Step 2.1: Based on the multi-source heterogeneous sensing data stream, after millimeter-wave penetration compensation is performed on the microwave radar signal, the entire frequency band is decomposed into multiple sub-bands according to the resonant mode distribution of the metal cabin; using the decomposition result, the ultrasonic signal is divided into critical frequency bands according to the attenuation coefficient of the sound wave in the metal medium; further, based on the cabin structure transfer function, the micro-vibration signal is grouped into octave bands to match the mechanical vibration propagation characteristics.

[0092] Step 2.2: Based on the sub-band, frequency band, and grouping results, construct a coherence matrix across sensor channels within each sub-band, and separate the reverberation noise subspace through eigenvalue decomposition; dynamically generate sub-band weighting coefficients with the phase continuity of the target signal cluster between adjacent sub-bands as a constraint.

[0093] Step 2.3: Based on the sub-band signal with weighted coefficients, extract the time delay characteristic peaks formed by multiple reflections from the metal wall, calculate the anti-phase cancellation beam to cancel multipath interference; use spatiotemporal adaptive filtering to eliminate specular reflection artifacts on the metal surface; output the sub-band signals after interference suppression.

[0094] Step 2.4: Reconstruct the time-domain signal clusters from each sub-band signal through inverse transformation, and calculate the mutual information entropy value between the reconstructed signal and the original sensing data; if the entropy value is lower than the dynamic threshold, output the target signal cluster for coherent enhancement; otherwise, trigger the data re-acquisition mechanism for closed-loop correction.

[0095] In this embodiment of the invention, by specifically addressing the unique problems of resonance reverberation and multipath interference in metal cabins, precise frequency band division is first performed based on the propagation characteristics of different sensor signals in the metal environment (such as microwave resonant modes, ultrasonic attenuation laws, and vibration propagation characteristics) to ensure that signal processing is adapted to the metal cavity environment. Then, noise is separated through coherence analysis and sub-band signals are dynamically weighted using phase continuity to enhance the consistency of the target signal. Subsequently, anti-phase cancellation and spatiotemporal filtering are used to specifically eliminate multipath reflections and specular artifacts, suppressing interference from the metal environment. Finally, mutual information entropy verification and closed-loop correction mechanisms are used to ensure the reliability of the output signal.

[0096] In this embodiment of the invention, the specific steps include:

[0097] Step 2.1, Millimeter-wave penetration compensation: First, multiply the millimeter-wave penetration attenuation coefficient in metal, the metal bulkhead thickness, and the millimeter-wave frequency to obtain a product. Then, divide this product by 20 to obtain a new value. Finally, exponentiate the new value with 10 as the base to obtain the penetration compensation coefficient. Multiplying the original radar signal by this penetration compensation coefficient yields the compensated signal.

[0098] Microwave subband decomposition: The method for calculating the nth resonant frequency is as follows: multiply the positive integer n by the propagation speed of electromagnetic waves within the metal, and then divide this product by (twice the characteristic dimension of the cabin cavity). The quotient is the nth resonant frequency. Using adjacent resonant frequencies as boundaries, the entire microwave frequency band is divided into multiple subbands, ensuring that each subband avoids the concentrated frequency band of resonant interference from the metal cabin.

[0099] Ultrasonic critical frequency band division: Multiply the medium constant by the k-th power of the ultrasonic frequency (k is approximately equal to 2 in metals), and the resulting product is the attenuation coefficient.

[0100] The method for calculating the critical band boundary (f_c) is as follows: divide the maximum acceptable attenuation value (α_max) by the medium constant to obtain the quotient; then take the k-th power of the quotient, and the result is the critical band boundary. The frequency band below this critical band boundary is divided into the effective frequency band of ultrasound.

[0101] Micro-vibration octave band grouping: Based on the cabin structure transfer function H(f), the first peak frequency of the function is found as the fundamental frequency (f0). Starting from this fundamental frequency, the frequency range of the micro-vibration signal is divided into multiple octave band intervals. The upper limit frequency of each interval is twice the lower limit frequency, that is, the first interval is [f0, 2f0], the second interval is [2f0, 4f0], and so on, so that each frequency interval matches the propagation characteristics of mechanical vibration in the cabin structure.

[0102] Step 2.2, Coherence Matrix Construction: First, calculate the expected cross-correlation of the two sensor channel signals; then calculate the standard deviation of the two channel signals and multiply them; finally, divide the expected cross-correlation by the product of the standard deviations. The result is the (i, j)th element of the coherence matrix, which reflects the correlation between the two channel signals.

[0103] Noise subspace separation: Eigenvalue decomposition is performed on the coherence matrix to obtain a series of eigenvalues, which are then sorted from largest to smallest. A threshold is set to 1 / 10 of the largest eigenvalue, and eigenvectors corresponding to all eigenvalues ​​smaller than this threshold are selected. The space formed by these eigenvectors is the reverberant noise subspace.

[0104] Sub-band weighting coefficient generation: The method for calculating the phase difference between adjacent sub-band signals is as follows: take the phase values ​​of the center frequencies of two adjacent sub-bands, calculate their absolute difference, and obtain the phase difference.

[0105] The method for calculating the subband weighting coefficient is as follows: add 1 to the phase difference to obtain a value; then divide 1 by this value, and the result is the subband weighting coefficient. The smaller the phase difference, the larger the subband weighting coefficient, thus ensuring that subbands with continuous phase receive higher weights.

[0106] Step 2.3, Time Delay Feature Peak Extraction: First, calculate the integral product of the signal and the signal after its own delay τ; then calculate the energy of these two signals, multiply them, and take the square root; finally, divide the integral product by the square root of the energy product to obtain the cross-correlation function. The delay time τ corresponding to the peak value of the cross-correlation function is the time delay feature peak of the metal wall reflection.

[0107] Anti-phase cancellation beam calculation: First, calculate the product of the imaginary unit j with 2π, the signal frequency f, and the reflection delay τ. Then, take the negative sign of this product and exponentiate it with the natural constant e as the base. The result is the anti-phase beam weight. Superimposing this weight on the original signal will cancel out multipath reflections.

[0108] Spatiotemporal adaptive filtering: First, calculate the inverse of the noise covariance matrix; then multiply this inverse matrix by the desired signal direction vector. The result is the filter weight vector. This vector is used to filter artifact signals generated by specular reflection from metal surfaces.

[0109] Step 2.4, Time-domain signal reconstruction: ,in, It is the first The frequency domain signal of each sub-band It is a sub-band index, with values ​​ranging from 0 to N-1. It is a time variable. It is the imaginary unit. It is a natural constant. It represents the total number of sub-bands.

[0110] Mutual information entropy calculation: For all possible values ​​of the signal, calculate the product of the probability and the logarithm to the base 2; add all these products together and take the negative value, the result is the entropy of the signal.

[0111] The method for calculating mutual information entropy is as follows: add the entropy of the original signal to the entropy of the reconstructed signal, and then subtract the joint entropy of the original signal and the reconstructed signal. The result is the mutual information entropy.

[0112] Dynamic threshold comparison: First, calculate the mean of historical mutual information entropy; then calculate the standard deviation of historical mutual information entropy and multiply it by 2; finally, add the mean to twice the standard deviation, and the result is the dynamic threshold. Compare the current mutual information entropy with this dynamic threshold. If it is lower than the threshold, output a coherent enhancement signal; otherwise, trigger the data re-acquisition mechanism.

[0113] In a preferred embodiment of the present invention, step 3 above may include:

[0114] Step 3.1: Based on the coherent enhanced target signal cluster, extract the microwave radar and micro-vibration signals; separate the respiratory harmonic components through time-frequency analysis technology, and generate the time-domain periodic feature vector and frequency-domain main harmonic energy ratio feature vector of the respiratory waveform envelope by using envelope detection and adaptive baseline drift correction.

[0115] Step 3.2: Based on the respiratory feature extraction results and corresponding signal segments, the cardiac impact component is separated from the microwave radar signal using cardiac harmonic reconstruction technology; at the same time, the micro-vibration signal is frequency-domain compensated in combination with the vibration transmission characteristics of the cabin structure, and the fundamental frequency stability feature vector and harmonic complexity feature vector of the cardiac impact spectrum are extracted.

[0116] Step 3.3: Based on the cardiac impact feature extraction results and the processed signal clusters, the infrared thermal imaging signal is dynamically segmented to eliminate metal thermal reflection interference, and the time-frequency feature vector of the live hot spot displacement trajectory is extracted; simultaneously, the output ultrasonic signal is used to analyze the short-time zero-crossing rate characteristics of its Doppler frequency shift, and a motion pattern classification feature vector is generated.

[0117] Step 3.4: Construct a feature quality assessment matrix based on the respiratory feature vector, cardiac impact feature vector, and motion feature vector; dynamically calculate the weights of respiratory features, cardiac impact features, and motion features by using the intra-class dispersion of respiratory features, the inter-class separability of cardiac impact features, and the confidence of motion features as the basis for attention weight allocation; and concatenate the weighted feature vectors into a joint feature vector and normalize it.

[0118] In this embodiment of the invention, by extracting and intelligently fusing core living features in a hierarchical manner, the interference problem of feature extraction in a metal cabin environment is specifically addressed, which has multiple benefits: First, three types of core living features—respiration, cardiac impact, and motion—are accurately separated from the enhanced signal. Among them, the respiratory feature is combined with time-frequency analysis and baseline correction to ensure the stability of periodic and harmonic features. The cardiac impact feature is adapted to the cabin vibration transmission characteristics through frequency domain compensation to extract the fundamental frequency stability and complexity. The motion feature eliminates metal thermal reflection by dynamic region segmentation and captures motion patterns through Doppler zero-crossing rate, thereby achieving environmental adaptability and specificity of feature extraction. Then, attention weights are dynamically allocated based on feature quality (intra-class dispersion, inter-class separability, and confidence), so that effective features receive higher weights and the influence of noise interference features is weakened. The final generated normalized joint feature vector retains the unique information of each feature and enhances the overall recognizability through fusion.

[0119] In this embodiment of the invention, the specific steps include:

[0120] Step 3.1: Extract microwave radar signals and micro-vibration signals from the coherent enhanced target signal cluster. Separate the breathing-related harmonic components (usually concentrated in the 0.1-0.5Hz frequency band) using time-frequency analysis techniques (such as short-time Fourier transform). Perform envelope detection on the breathing harmonic components (preserving the signal amplitude variation trend), and then perform adaptive baseline drift correction (calculate the mean within the signal sliding window as the baseline, and subtract the baseline from the original signal) to obtain a stable breathing waveform envelope.

[0121] Feature vector generation:

[0122] Time-domain periodic feature vector: Calculate the time difference between adjacent peaks (or valleys) in the envelope of the respiratory waveform, take the average value of multiple cycles as the periodic feature, and then calculate the standard deviation, maximum value and minimum value of these cycles to form a time-domain periodic feature vector.

[0123] Frequency domain primary harmonic energy ratio eigenvector: Calculate the energy of each harmonic in the respiratory harmonics (signal power integral of each harmonic frequency band), calculate the integral of the instantaneous power of the signal (square of the signal value) on the time axis within the frequency band (integration interval is 10 seconds to ensure coverage of multiple respiratory cycles), obtain the energy of each harmonic, take the energy of the primary harmonic with the largest energy, and the ratio of the energy of the primary harmonic to the total energy of all harmonics as the primary harmonic energy ratio, and then combine it with the energy ratio of the secondary harmonic to the primary harmonic to form the frequency domain eigenvector.

[0124] Step 3.2: Based on the extracted respiratory features corresponding to the signal segments (excluding periods of strong respiratory interference), the cardiac impact component is separated from the microwave radar signal using heartbeat harmonic reconstruction technology (retaining the 1-3Hz frequency band, which is the main frequency band of the cardiac impact signal). Combining the vibration transmission characteristics of the cabin structure (such as the vibration attenuation coefficient at the cabin wall), the frequency domain amplitude of the micro-vibration signal is compensated (low-frequency vibrations are multiplied by a smaller coefficient, and high-frequency vibrations are multiplied by a larger coefficient to offset structural transmission losses).

[0125] Fundamental frequency stability eigenvector: Calculate the value of the fundamental frequency (main peak frequency) in the cardiac impact spectrum within a continuous time window, find the ratio of the standard deviation to the mean of these values ​​(the smaller the ratio, the more stable the frequency), and combine it with the fluctuation range of the fundamental frequency to form the fundamental frequency stability eigenvector.

[0126] Harmonic complexity feature vector: Count the number of harmonics above the fundamental frequency in the shock spectrum, calculate the ratio of each harmonic energy to the fundamental frequency energy, and take the proportion of the number of harmonics and the total energy of the ratio greater than a threshold (such as 0.1) to form the harmonic complexity feature vector.

[0127] Step 3.3: Based on the extracted cardiac impact features corresponding to the living area, perform dynamic region segmentation on the infrared thermal imaging signal (set temperature threshold: areas 5-10℃ higher than the ambient temperature are judged as living hot spots) to eliminate high-temperature artifact areas reflected from the metal surface.

[0128] Hot spot displacement time-frequency characteristics: Track the change of hot spot center coordinates over time, calculate the average velocity, acceleration and frequency components of the displacement trajectory in different time windows (obtained through time-frequency analysis), and combine them into a hot spot displacement time-frequency characteristic vector.

[0129] Ultrasonic Doppler frequency shift analysis: Short-time zero-crossing rate (the number of times the signal crosses the zero point per unit time) is calculated for the Doppler frequency shift component of the ultrasonic signal. The mean, peak and rate of change of the zero-crossing rate within different time windows are statistically analyzed to generate motion pattern classification feature vectors (e.g., stationary corresponds to low zero-crossing rate, and moving corresponds to high zero-crossing rate).

[0130] Step 3.4, Feature quality assessment matrix construction: The matrix rows correspond to the three types of features: respiration, cardiac impact, and exercise, and the columns correspond to the feature quality indicators (intra-class consistency, inter-class discrimination, and noise tolerance). The matrix elements are the quantitative scores (between 0 and 1) of each feature on the corresponding indicator.

[0131] Attention weight calculation:

[0132] Respiratory feature weight: It is inversely proportional to the intra-class dispersion of respiratory features (variance of feature vector) (the smaller the dispersion, the greater the weight). The calculation formula is "Respiratory weight = 1 ÷ (1 + intra-class dispersion of respiratory features)".

[0133] Cardiac impact feature weight: It is proportional to the inter-class separability of cardiac impact features (the difference in mean values ​​of different target features divided by the sum of variances within the class) (the greater the separability, the greater the weight). The calculation formula is "cardiac impact weight = inter-class separability of cardiac impact ÷ (intra-class dispersion of respiratory function + inter-class separability of cardiac impact + confidence of motion features)".

[0134] Motion feature weight: It is proportional to the confidence level of motion features (the success rate of feature extraction, such as the accuracy of hot spot recognition). The calculation formula is "motion weight = confidence level of motion features ÷ (the sum of the above three indicators)" (the sum of the three weights is 1, which is ensured by normalization).

[0135] Joint feature vector generation: The respiratory, cardiac impact, and motion feature vectors are multiplied by their respective weights and then concatenated to form a total feature vector. This vector is then normalized (each element is subtracted from the mean and divided by the standard deviation) to output the final joint feature vector.

[0136] In a preferred embodiment of the present invention, step 4 above may include:

[0137] Step 4.1: Receive the joint feature vector, extract the respiratory waveform envelope features, cardiac impact spectrum features and motion pattern time-frequency feature components from it, and map these feature components to the three-dimensional spatial discretized grid nodes of the metal cabin to generate feature distribution data with spatial location annotations.

[0138] Step 4.2: Based on the feature distribution data with spatial location annotations, dynamic thermodynamic data reflecting the spatial energy density of the living target is generated by calculating the temporal stability index and spatial continuity measure of the feature amplitude of each grid node.

[0139] Step 4.3: Using the distribution information of each energy peak region in the dynamic thermal data, combined with the preset sensor position parameters and their signal attenuation physical laws, calculate the coverage coefficient of each sensor for each spatial grid node.

[0140] Step 4.4: Normalize the capability coefficients and construct an initial coverage confidence matrix with spatial grid nodes as row indices and sensor channels as column indices. The matrix element values ​​represent the credible monitoring weights of the corresponding sensors at specific locations.

[0141] Step 4.5: Based on the metal obstruction compensation parameters provided by the pre-set database of the cabin structure, perform multipath reflection loss correction on the initial coverage confidence matrix and output the optimized multi-sensor coverage confidence matrix.

[0142] In this embodiment of the invention, by transforming abstract feature vectors into concrete spatial energy distributions, a precise correlation between living features and the physical space of the cabin is achieved, which has multiple benefits: First, respiratory, cardiac impact, and motion features are mapped onto a three-dimensional grid, giving the features clear spatial location attributes; then, dynamic thermal data is generated through temporal stability and spatial continuity analysis, which intuitively reflects the energy distribution and changing trends of the living target, providing a spatial basis for sensor coverage assessment; next, the coverage capability coefficient of the sensors to each grid node is calculated and a confidence matrix is ​​constructed to quantify the monitoring effectiveness of different sensors in the cabin; finally, the metal obstruction compensation parameter correction matrix is ​​combined to eliminate the interference of metal structures on sensor signals, making the matrix more consistent with the actual monitoring scenario.

[0143] In this embodiment of the invention, the specific steps include:

[0144] Step 4.1: Extract three types of feature components from the joint feature vector according to the preset dimensions: respiratory waveform envelope feature component (corresponding to the time domain period and frequency domain energy ratio feature generated in step 3.1), cardiac impact spectrum feature component (corresponding to the fundamental frequency stability and harmonic complexity feature generated in step 3.2), and motion pattern time-frequency feature component (corresponding to the hot spot displacement and time-frequency feature generated in step 3.3).

[0145] Based on the three-dimensional spatial discretization mesh of the metal cabin (e.g., a 10cm×10cm×10cm cube mesh), and combined with the sensor layout coordinates from step 1.1, the feature components are mapped to the corresponding mesh nodes:

[0146] Microwave radar-related features (respiration, cardiac impact) are mapped to the grid covered by its longitudinal section;

[0147] Infrared thermal imaging features (moving hotspots) are mapped to a grid covered by their horizontal profile;

[0148] Ultrasonic and micro-vibration characteristics are mapped to their blind spots and bulkhead mesh.

[0149] The final generated feature distribution data for each grid node includes "spatial coordinates (x, y, z) + three types of feature component values".

[0150] Step 4.2, for the feature amplitude of each grid node (the weighted sum of the three types of feature components, with the weights being the feature weights in Step 3.4), calculate the temporal stability index within a continuous time window (e.g., 5 seconds / window, sliding step size of 2 seconds): Temporal stability index = (standard deviation of feature amplitude within the window) ÷ (mean of feature amplitude within the window). (The smaller the index, the more stable the feature is in time, and the more likely it is to come from a living target.)

[0151] For each grid node, calculate the difference in characteristic amplitude between it and its 6 adjacent grids (front and back, left and right, up and down): Spatial continuity measure = 1 - [(sum of absolute differences between the characteristic amplitude of adjacent grids and the characteristic amplitude of the current node) ÷ (6 × mean of the characteristic amplitude of the current node)], (the closer the measure is to 1, the more continuous the spatial features are, and the more likely it is to be the energy distribution of the same living target).

[0152] The energy density value of each grid node is calculated as (characteristic amplitude) × (temporal stability index) × (spatial continuity measure). The energy density values ​​of all grid nodes are arranged according to spatial coordinates to form dynamic thermodynamic data that is updated over time (the higher the energy, the greater the possibility of the existence of a living organism).

[0153] Step 4.3: From the dynamic thermal data, extract regions with energy density values ​​exceeding twice the global average as energy peak regions, and record the coordinates of the grid nodes contained in these regions. For each sensor (microwave radar, infrared array, ultrasonic, micro-vibration sensor) and each grid node, combine the sensor position parameters (step 1.1) and the physical laws of signal attenuation to calculate the coverage capability coefficient: Coverage capability coefficient = (maximum monitoring distance of sensor / straight-line distance from sensor to grid node) × (1 - signal attenuation rate) × (peak region weight).

[0154] Signal attenuation rate: For microwaves or ultrasound, it is calculated as "distance × medium attenuation coefficient" (e.g., microwave attenuation rate in metal = 0.05 × distance), for infrared, it is calculated as "the reciprocal of the square of the distance", and for micro-vibration, it is calculated as "bullet transmission loss coefficient × distance".

[0155] Peak region weight: If a grid node belongs to the energy peak region, the weight is 1.2; otherwise, it is 1.0 (enhancing the coverage assessment of high-energy regions).

[0156] Step 4.4: For each grid node, normalize the coverage coefficient of all sensors: Normalization coefficient = (coverage coefficient of a certain sensor) ÷ (sum of coverage coefficients of all sensors in the grid node), (ensure that the sum of the sensor coefficients corresponding to each grid node is 1, with a value range of 0-1).

[0157] Using the 3D grid nodes as row indices (numbered in spatial coordinate order) and the sensor channels as column indices (numbered in the order of microwave, infrared, ultrasonic, and micro-vibration), the (i, j)th element in the matrix is ​​the normalized coefficient of the jth sensor corresponding to the i-th grid node, which is the initial coverage confidence matrix (the larger the element value, the higher the confidence of the sensor in monitoring the grid).

[0158] Step 4.5: Extract the metal shading compensation parameters corresponding to each grid node from the pre-set database of the cabin structure: shading coefficient (0-1, 1 means no shading, 0 means complete shading) and multipath reflection loss ratio (0-0.5, such as 0.3 for corner grid reflection loss).

[0159] Each element (i, j) of the initial coverage confidence matrix is ​​corrected: Corrected element value = (initial matrix element value) × (occlusion coefficient) × (1 - multipath reflection loss ratio). (The occlusion coefficient reduces the sensor weight of the grid nodes blocked by metal, and the reflection loss ratio corrects the signal attenuation caused by multipath interference, and finally outputs the optimized multi-sensor coverage confidence matrix.)

[0160] In a preferred embodiment of the present invention, step 4 above may include:

[0161] Step 4.6: Perform singular value decomposition on the multi-sensor coverage confidence matrix to extract the principal component vectors characterizing the spatial coverage characteristics of the sensors and their corresponding singular value spectrum distributions.

[0162] Step 4.7: Based on the characteristic peak interval of the singular value spectrum distribution, determine the initial search boundary of the sensor weight parameters, and map each weight parameter combination to the position coordinates of individual bees in the bee colony optimization.

[0163] Step 4.8: Based on the cabin space coverage characteristics reflected by the principal component vectors and combined with the spatial gradient changes of the living body energy distribution, the search area is divided into a high-gradient fine search area and a low-gradient broad exploration area to obtain the partitioning results.

[0164] Step 4.9: Dynamically configure the roles of the bee colony based on the partitioning results. Deploy scout bees to high gradient regions to perform neighborhood depth optimization; deploy follower bees to low gradient regions to perform heat tracking exploration; after each iteration, redistribute the region search task based on the fitness of the weight parameters. When the change in weight parameters in consecutive iterations is lower than the preset convergence threshold, extract the weight allocation parameters corresponding to the current optimal bee position coordinates.

[0165] In this embodiment of the invention, efficient and accurate solutions for sensor weight parameters are achieved through scientific matrix analysis and intelligent optimization strategies. Then, reasonable search boundaries are determined based on the characteristic peaks of the singular value spectrum to ensure that the search range of the weight parameters conforms to the actual coverage characteristics and avoids ineffective exploration. By combining spatial coverage characteristics and the live energy gradient to divide the search area, a fine-grained search is used in high-gradient areas (where live energy changes drastically and requires focused monitoring), while a broad-grained search is used in low-gradient areas (where energy is moderate and can be widely explored), balancing optimization accuracy and efficiency. Finally, through dynamic role allocation and iterative fitness adjustment of the bee colony, scout bees are allowed to cultivate key areas and follower bees expand potential areas, achieving adaptive optimization of the weight parameters and rapidly converging to the final solution.

[0166] In this embodiment of the invention, the specific steps include:

[0167] Step 4.6: Perform singular value decomposition on the multi-sensor coverage confidence matrix (denoted as matrix A, with dimensions of grid node number × sensor channel number). The decomposition result is expressed as follows: matrix A can be decomposed into the product of three matrices, where the first matrix (U) is the left singular vector matrix, the second matrix (Σ) is a diagonal matrix (the elements on the diagonal are singular values, and are sorted in descending order), and the third matrix (Vᵀ) is the transpose of the right singular vector matrix.

[0168] The principal component vectors are the column vectors in the left singular vector matrix (U) that correspond to singular values ​​greater than the threshold (threshold = maximum singular value × 0.1). These vectors reflect the main characteristics of the sensor's spatial coverage (such as coverage range and overlapping area). The singular value spectrum distribution is a sequence of all singular values ​​in the diagonal matrix arranged in their original order, where larger singular values ​​are principal components that contribute more to the coverage characteristics.

[0169] Step 4.7: From the singular value spectrum distribution, extract the continuous intervals where the singular values ​​are greater than "maximum singular value × 0.3" as the feature peak intervals (this interval contains the singular values ​​that play a dominant role in coverage characteristics). The initial search boundary for the sensor weight parameters (each sensor corresponds to one weight, such as microwave radar weight, infrared array weight, etc.) is determined according to the following rules:

[0170] The minimum value of each weight parameter = the minimum singular value within the feature peak interval × 0.5;

[0171] The maximum value of each weight parameter = the maximum singular value within the feature peak interval × 1.5;

[0172] Each weight parameter combination (w1, w2, ..., w n (where n is the number of sensors) is mapped to the position coordinates of individual bees in the bee colony optimization: the coordinate of a bee in the i-th dimension = (the value of the i-th weight parameter - the minimum value of that parameter) ÷ (the maximum value of that parameter - the minimum value of that parameter) × the search space dimension range (e.g., 0-100); (may map the actual value range of the weight parameters to the search coordinate space of the bee colony algorithm, which facilitates the algorithm's optimization).

[0173] Step 4.8, the spatial gradient change of the living energy distribution is calculated as follows: the gradient value of each grid node = (the sum of the absolute differences in energy density between the node and its 6 adjacent grid nodes) ÷ (6 × the average energy density of the node); (the larger the gradient value, the more drastic the change in living energy in this region, and the more accurate the sensor weights are required).

[0174] Based on the dense sensor coverage areas reflected by the principal component vectors (regions with high principal component vector magnitudes), the search area is divided into:

[0175] The high gradient fine search region is the region where the gradient value is greater than the mean gradient of all grids, and this region is the core region covered by the principal component vector;

[0176] Low gradient breadth exploration region = region where gradient value ≤ the average gradient of all grids, or edge region covered by principal component vector.

[0177] Step 4.9: Deploy according to the ratio of "60% of scout bees in high gradient regions and 60% of follower bees in low gradient regions": The initial positions of scout bees are randomly distributed within the weight parameter range of the high gradient region. In each iteration, new candidate parameters are generated in the neighborhood of the current position (e.g., within ±5% of the parameter value). The initial positions of follower bees are distributed in the low gradient region based on high fitness parameters (parameters that performed well in the previous iteration). In each iteration, they move to the parameter position with higher fitness (movement step size = distance from the current position to the high fitness position × 0.3).

[0178] Fitness function calculation: Fitness of weight parameters = 1 ÷ (1 + sensor coverage error), where: Sensor coverage error = (sum of squared differences between the actual coverage confidence matrix and the coverage matrix after weighting by weight parameters) ÷ (total number of matrix elements); (The larger the fitness, the better the weight parameters can optimize the sensor coverage effect).

[0179] After each iteration, the change range of the weight parameters is calculated: Change range = (sum of absolute differences between the weight parameter values ​​of the current iteration and the weight parameter values ​​of the previous iteration) ÷ (total number of weight parameters).

[0180] When the change magnitude of three consecutive iterations is less than the preset convergence threshold (e.g., 0.001), the iteration stops, and the weight parameter combination corresponding to the location coordinates of the bee with the highest fitness is extracted as the final sensor weight allocation parameter.

[0181] In a preferred embodiment of the present invention, step 5 above may include:

[0182] Step 5.1: Receive the sensor weight allocation parameters, use them as input layer neurons to initialize weights and inject them into the decision model, and simultaneously load the joint feature vector into the model feature input channel;

[0183] Step 5.2: Based on the constructed initial decision model, call the historical working condition database of the cabin to build a training sample set, and fine-tune the hidden layer structure parameters of the model through the transfer learning mechanism to make the model adapt to the feature distribution shift in the metal cavity environment and generate a metal environment-adaptive decision model.

[0184] Step 5.3: Based on the metal environment-adaptive decision model, monitor the activation state of its hidden layer in real time, generate sensor resource allocation instructions according to the spatial distribution pattern of the activation state, and dynamically allocate microwave radar scanning frame rate, infrared array resolution, ultrasonic transmission power and vibration sensor sensitivity level according to the weight parameter ratio.

[0185] Step 5.4: Collect new sensor data streams according to the adaptive sampling strategy, input the real-time data into the metal environment-adaptive decision model, update the model parameters through the online incremental learning mechanism, and output the probability value of the existence of living organisms in the cabin and the quantitative indicators of vital signs composed of respiratory rate, heart rate coefficient of variation, and body movement intensity.

[0186] Step 5.5: When the probability value of the presence of a living organism exceeds a preset threshold, analyze the quantitative indicators of vital signs corresponding to the probability value and generate a three-dimensional parameter analysis report including respiratory rate abnormality, heart rate variability stability, and body movement intensity trend.

[0187] In this embodiment of the invention, by constructing an intelligent decision-making system adapted to the metal environment, the entire process of liveness detection, from model initialization to dynamic output, is optimized. Sensor resources are dynamically allocated based on the activation state of the model's hidden layer (such as adjusting radar frame rate, infrared resolution, etc.), optimizing resource consumption while ensuring detection accuracy. The model parameters are updated in real time through online incremental learning, enabling the model to dynamically adapt to changes in the cabin environment and the state of the liveness, ensuring that the output probability of liveness and vital signs indicators (respiratory rate, heart rate variability, etc.) remain consistently accurate. Finally, when the probability of liveness exceeds a threshold, a three-dimensional analysis report is generated, intuitively presenting the abnormality and trend of vital signs.

[0188] In this embodiment of the invention, the specific steps include:

[0189] Step 5.1: Assign sensor weight parameters (e.g., microwave radar weight 0.4, infrared array weight 0.3, etc.) to the connection weights of neurons in the input layer of the decision model according to the sensor type (e.g., the first neuron in the input layer corresponds to microwave radar, and its initial weight is set to 0.4). Input the joint feature vector (containing respiratory, cardiac impact, and motion features) into the feature input channel of the model as the initial input data of the model.

[0190] Step 5.2: Extract historical feature data (such as liveness signals under different metal thicknesses and structural layouts) from the historical operating condition database of the cabin, and construct a training sample set according to the ratio of "70% normal operating conditions and 30% abnormal operating conditions". Freeze the input and output layer parameters of the initial decision model; use the training sample set to iteratively fine-tune the hidden layer parameters (e.g., using a mini-batch gradient descent algorithm with a learning rate of 0.001); after every 50 training batches, evaluate the model's classification accuracy of features in the metal environment on the validation set. Stop fine-tuning when the accuracy improves by <0.5% for 3 consecutive times, and generate a decision model adapted to the metal environment.

[0191] Step 5.3: Extract the activation values ​​of the hidden layers of the metal environment-adaptive decision model in real time (such as the output value of the ReLU activation function), and group and count the activation intensity according to the spatial region corresponding to the neuron (such as the front part of the cabin corresponding to the 1st to 10th neurons).

[0192] If the activation intensity of a certain area is greater than 1.5 times the global average, it is determined to be a key monitoring area, and the sensor resources of the corresponding area are increased according to the weight ratio (such as increasing the microwave radar scanning frame rate by 20% and improving the infrared resolution by 1 level).

[0193] If the activation intensity of a certain area is less than 0.5 times the global average, it is determined to be a secondary monitoring area, and resource allocation is reduced (e.g., ultrasonic transmission power is reduced by 15%).

[0194] The final resource parameters for each sensor = basic configuration parameters × corresponding sensor weight × regional adjustment coefficient.

[0195] Step 5.4: When the probability of a living organism output by the model is >0.7, the sampling frequency is increased to 10Hz (originally 5Hz).

[0196] When the fluctuation range of vital signs is greater than twice the historical average, emergency sampling is triggered (the sampling frequency is temporarily increased to 20Hz for 30 seconds).

[0197] Newly acquired sensor data is mixed with historical data at a ratio of 1:4. The model parameters are updated every 100 new samples (using an incremental learning algorithm, retaining 80% of the historical knowledge weights).

[0198] Indicator Output:

[0199] Probability of presence of live organism: The probability value of the "live organism present" category corresponding to the normalization function of the model output layer;

[0200] Respiratory rate: The reciprocal of the periodic mean extracted from the envelope features of the respiratory waveform;

[0201] Coefficient of variation of heart rate: The ratio of the standard deviation to the mean in the fundamental frequency stability characteristics of cardiac impact spectrum;

[0202] Body dynamic intensity: the average displacement velocity of hot spots in the time-frequency characteristics of motion patterns.

[0203] Step 5.5: When the probability of a live organism being present is greater than 0.8 (preset threshold), the report generation process is triggered.

[0204] Abnormality of respiratory rate = |Current respiratory rate - Mean of normal range| ÷ Standard deviation of normal range;

[0205] Heart rate variability stability = 1 - heart rate variability coefficient (the smaller the coefficient, the more stable the heart rate).

[0206] Physical activity intensity trend = current physical activity intensity - average physical activity intensity of the previous 5 minutes.

[0207] A three-dimensional coordinate system is constructed with respiratory rate abnormality as the X-axis, heart rate variability stability as the Y-axis, and body movement intensity trend as the Z-axis. The current index values ​​are mapped to the coordinate system to generate scatter data. Based on the position of the scatter points (such as falling in the upper right corner), warning information such as "rapid breathing and unstable heart rate" is automatically marked, and an analysis report containing historical trend comparisons is generated.

[0208] like Figure 2 As shown, embodiments of the present invention also provide a liveness detection system for a metal cargo compartment, comprising:

[0209] The acquisition module is used to simultaneously acquire multi-source heterogeneous sensor data streams inside the metal cabin through a multi-band microwave radar, infrared thermal imaging array, wideband ultrasonic sensor, and high-sensitivity micro-vibration sensor group deployed inside the cabin; and to output coherently enhanced target signal clusters by using an adaptive signal enhancement algorithm based on frequency domain subband decomposition based on the multi-source heterogeneous sensor data streams.

[0210] The processing module is used to extract the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target based on the coherent enhanced target signal cluster through time-frequency analysis technology, and to generate a joint feature vector using an attention mechanism weighted feature fusion method.

[0211] The computation module is used to analyze the energy distribution characteristics of the live target in the cabin based on the joint feature vector, dynamically construct the multi-sensor coverage confidence matrix, perform singular value decomposition on the confidence matrix, and combine it with the adaptive bee colony division optimization algorithm to use the sensor weight allocation parameters as the foraging target of the bee colony. Each bee represents a set of candidate weight parameters. By utilizing the division of labor and cooperation of the bee colony and combining the live energy distribution characteristics in the cabin, the search strategy is dynamically adjusted. Through iterative optimization, the weight parameters converge to the final solution, and the sensor weight allocation parameters are generated.

[0212] The intelligent decision-making module is used to construct a metal environment-adaptive decision-making model based on sensor weight allocation parameters and joint feature vectors, adaptively adjust sensor sampling strategies, and output quantitative indicators of the probability of the presence of living organisms and vital signs.

[0213] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting live organisms inside a metal chamber, characterized in that, The method includes: Step 1: Simultaneously collect multi-source heterogeneous sensor data streams inside the cabin by deploying multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensor and high-sensitivity micro-vibration sensor group inside the metal cabin. Step 2: Based on the multi-source heterogeneous sensing data stream, an adaptive signal enhancement algorithm based on frequency domain subband decomposition is used to suppress reverberation noise and multipath interference caused by metal cavity resonance, and output a coherently enhanced target signal cluster, including: Step 2.1: Based on the multi-source heterogeneous sensing data stream, after millimeter-wave penetration compensation is performed on the microwave radar signal, the entire frequency band is decomposed into multiple sub-bands according to the resonant mode distribution of the metal cabin; using the decomposition result, the ultrasonic signal is divided into critical frequency bands according to the attenuation coefficient of the sound wave in the metal medium; based on the cabin structure transfer function, the micro-vibration signal is grouped into octave bands to match the mechanical vibration propagation characteristics. Step 3: Based on the coherent enhanced target signal cluster, the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target are extracted by time-frequency analysis technology, and a joint feature vector is generated by an attention mechanism weighted feature fusion method. Step 4: Based on the joint feature vector, analyze the energy distribution characteristics of the live target in the cabin and dynamically construct a multi-sensor coverage confidence matrix; perform singular value decomposition on the confidence matrix and combine it with an adaptive bee colony division optimization algorithm, using the sensor weight allocation parameters as the bee colony's foraging targets, with each bee representing a set of candidate weight parameters. Utilize the division of labor and cooperation of the bee colony and dynamically adjust the search strategy based on the live energy distribution characteristics in the cabin. Through iterative optimization, the weight parameters converge to the final solution, generating the sensor weight allocation parameters. Step 5: Based on the sensor weight allocation parameters and joint feature vector, construct a metal environment-adaptive decision model, adaptively adjust the sensor sampling strategy, and output the probability of the presence of a living organism and quantitative indicators of vital signs.

2. The method for detecting live organisms inside a metal carrier chamber according to claim 1, characterized in that, Multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensors, and high-sensitivity micro-vibration sensor arrays deployed inside the metal cabin are used to simultaneously acquire multi-source heterogeneous sensor data streams within the cabin, including: Step 1.1: Generate a three-dimensional spatial coordinate mapping table based on the structural parameters of the metal cabin, and deploy the sensor network according to the layout of microwave radar covering the longitudinal section, infrared array covering the horizontal section, ultrasonic sensor covering the corner blind area, and micro-vibration sensor group attached to the cabin wall. Step 1.2: Based on the three-dimensional spatial coordinate mapping table, start the multi-sensor hardware synchronization triggering mechanism, generate a global timestamp through a high-precision clock source, and strictly align the microwave radar pulse transmission phase, infrared array exposure period, ultrasonic wave transmission sequence and vibration sampling time in the time domain to obtain the original sensing data. Step 1.3: Receive the raw sensing data, and based on the sensor position parameters recorded in the three-dimensional spatial coordinate mapping table, add a cabin material attenuation compensation coefficient to the microwave radar signal, add an ambient temperature drift correction parameter to the infrared thermal imaging data, add an air density compensation factor to the ultrasonic signal, and add a mechanical coupling gain coefficient to the micro-vibration signal to generate a primary data stream with physical compensation. Step 1.4: The primary data stream is matched to the position according to the three-dimensional spatial coordinate mapping table, wherein microwave radar data is mapped to the polar coordinate system, infrared data is mapped to the Cartesian coordinate system, ultrasonic data is mapped to the spherical coordinate system, and micro-vibration data is mapped to the structural grid coordinate system, and the spatiotemporally correlated multi-source heterogeneous sensing data stream is output.

3. The method for detecting live organisms inside a metal chamber according to claim 2, characterized in that, Based on the multi-source heterogeneous sensing data stream, an adaptive signal enhancement algorithm based on frequency domain subband decomposition is used to suppress reverberation noise and multipath interference caused by metal cavity resonance, and output a coherently enhanced target signal cluster, which also includes: Step 2.2: Based on the sub-band, frequency band, and grouping results, construct a coherence matrix across sensor channels within each sub-band, and separate the reverberation noise subspace through eigenvalue decomposition; dynamically generate sub-band weighting coefficients with the phase continuity of the target signal cluster between adjacent sub-bands as a constraint. Step 2.3: Based on the sub-band signal with weighted coefficients, extract the time delay characteristic peaks formed by multiple reflections from the metal wall, calculate the anti-phase cancellation beam to cancel multipath interference; use spatiotemporal adaptive filtering to eliminate specular reflection artifacts on the metal surface; output the sub-band signals after interference suppression. Step 2.4: Reconstruct the time-domain signal clusters from each sub-band signal through inverse transformation, and calculate the mutual information entropy value between the reconstructed signal and the original sensing data; if the entropy value is lower than the dynamic threshold, output the target signal cluster for coherent enhancement; otherwise, trigger the data re-acquisition mechanism for closed-loop correction.

4. The method for detecting live organisms inside a metal chamber according to claim 3, characterized in that, Based on coherently enhanced target signal clusters, respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of living targets are extracted using time-frequency analysis techniques. A joint feature vector is then generated using an attention-weighted feature fusion method, including: Step 3.1: Based on the coherent enhanced target signal cluster, extract the microwave radar and micro-vibration signals; separate the respiratory harmonic components through time-frequency analysis technology, and generate the time-domain periodic feature vector and frequency-domain main harmonic energy ratio feature vector of the respiratory waveform envelope by using envelope detection and adaptive baseline drift correction. Step 3.2: Based on the respiratory feature extraction results and corresponding signal segments, the cardiac impact component is separated from the microwave radar signal using cardiac harmonic reconstruction technology; at the same time, the micro-vibration signal is frequency-domain compensated in combination with the vibration transmission characteristics of the cabin structure, and the fundamental frequency stability feature vector and harmonic complexity feature vector of the cardiac impact spectrum are extracted. Step 3.3: Based on the cardiac impact feature extraction results and the processed signal clusters, the infrared thermal imaging signal is dynamically segmented to eliminate metal thermal reflection interference, and the time-frequency feature vector of the live hot spot displacement trajectory is extracted; simultaneously, the output ultrasonic signal is used to analyze the short-time zero-crossing rate characteristics of its Doppler frequency shift, and a motion pattern classification feature vector is generated. Step 3.4: Construct a feature quality assessment matrix based on the respiratory feature vector, cardiac impact feature vector, and motion feature vector; dynamically calculate the weights of respiratory features, cardiac impact features, and motion features by using the intra-class dispersion of respiratory features, the inter-class separability of cardiac impact features, and the confidence of motion features as the basis for attention weight allocation; and concatenate the weighted feature vectors into a joint feature vector and normalize it.

5. The method for detecting live organisms inside a metal carrier chamber according to claim 4, characterized in that, Based on the joint feature vector, the energy distribution characteristics of the living target within the cabin are analyzed, and a multi-sensor coverage confidence matrix is ​​dynamically constructed, including: Step 4.1: Receive the joint feature vector, extract the respiratory waveform envelope features, cardiac impact spectrum features and motion pattern time-frequency feature components from it, and map these feature components to the three-dimensional spatial discretized grid nodes of the metal cabin to generate feature distribution data with spatial location annotations. Step 4.2: Based on the feature distribution data with spatial location annotations, dynamic thermodynamic data reflecting the spatial energy density of the living target is generated by calculating the temporal stability index and spatial continuity measure of the feature amplitude of each grid node. Step 4.3: Using the distribution information of each energy peak region in the dynamic thermal data, combined with the preset sensor position parameters and their signal attenuation physical laws, calculate the coverage coefficient of each sensor for each spatial grid node. Step 4.4: Normalize the capability coefficients and construct an initial coverage confidence matrix with spatial grid nodes as row indices and sensor channels as column indices. The matrix element values ​​represent the credible monitoring weights of the corresponding sensors at specific locations. Step 4.5: Based on the metal obstruction compensation parameters provided by the pre-set database of the cabin structure, perform multipath reflection loss correction on the initial coverage confidence matrix and output the optimized multi-sensor coverage confidence matrix.

6. The method for detecting live organisms inside a metal carrier chamber according to claim 5, characterized in that, Singular value decomposition is performed on the confidence matrix, and combined with an adaptive bee colony division of labor optimization algorithm, the sensor weight allocation parameters are used as the foraging targets of the bee colony. Each bee represents a set of candidate weight parameters. By utilizing the division of labor and cooperation of the bee colony and combining the energy distribution characteristics of living organisms in the capsule, the search strategy is dynamically adjusted. Through iterative optimization, the weight parameters converge to the final solution, generating the sensor weight allocation parameters, including: Step 4.6: Perform singular value decomposition on the multi-sensor coverage confidence matrix to extract the principal component vectors characterizing the spatial coverage characteristics of the sensors and their corresponding singular value spectrum distributions. Step 4.7: Based on the characteristic peak interval of the singular value spectrum distribution, determine the initial search boundary of the sensor weight parameters, and map each weight parameter combination to the position coordinates of individual bees in the bee colony optimization. Step 4.8: Based on the cabin space coverage characteristics reflected by the principal component vectors and combined with the spatial gradient changes of the living body energy distribution, the search area is divided into a high-gradient fine search area and a low-gradient broad exploration area to obtain the partitioning results. Step 4.9: Dynamically configure the roles of the bee colony based on the partitioning results. Deploy scout bees to high gradient regions to perform neighborhood depth optimization; deploy follower bees to low gradient regions to perform heat tracking exploration; after each iteration, redistribute the region search task based on the fitness of the weight parameters. When the change in weight parameters in consecutive iterations is lower than the preset convergence threshold, extract the weight allocation parameters corresponding to the current optimal bee position coordinates.

7. The method for detecting live organisms inside a metal carrier chamber according to claim 6, characterized in that, Based on sensor weight allocation parameters and joint feature vectors, a metal environment-adaptive decision model is constructed to adaptively adjust the sensor sampling strategy and output quantitative indicators of the probability of presence of living organisms and vital signs, including: Step 5.1: Receive the sensor weight allocation parameters, use them as input layer neurons to initialize weights and inject them into the decision model, and simultaneously load the joint feature vector into the model feature input channel; Step 5.2: Based on the constructed initial decision model, call the historical working condition database of the cabin to build a training sample set, and fine-tune the hidden layer structure parameters of the model through the transfer learning mechanism to make the model adapt to the feature distribution shift in the metal cavity environment and generate a metal environment-adaptive decision model. Step 5.3: Based on the metal environment-adaptive decision model, monitor the activation state of its hidden layer in real time, generate sensor resource allocation instructions according to the spatial distribution pattern of the activation state, and dynamically allocate microwave radar scanning frame rate, infrared array resolution, ultrasonic transmission power and vibration sensor sensitivity level according to the weight parameter ratio. Step 5.4: Collect new sensor data streams according to the adaptive sampling strategy, input the real-time data into the metal environment-adaptive decision model, update the model parameters through the online incremental learning mechanism, and output the probability value of the existence of living organisms in the cabin and the quantitative indicators of vital signs composed of respiratory rate, heart rate coefficient of variation, and body movement intensity. Step 5.5: When the probability value of the presence of a living organism exceeds a preset threshold, analyze the quantitative indicators of vital signs corresponding to the probability value and generate a three-dimensional parameter analysis report including respiratory rate abnormality, heart rate variability stability, and body movement intensity trend.

8. A live human detection system for a metal carrier chamber, the system implementing the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to simultaneously collect multi-source heterogeneous sensor data streams inside the metal cabin by using a multi-band microwave radar, infrared thermal imaging array, broadband ultrasonic sensor and high-sensitivity micro-vibration sensor group deployed inside the metal cabin. This algorithm is used to suppress reverberation noise and multipath interference caused by metal cavity resonance and output coherently enhanced target signal clusters based on an adaptive signal enhancement algorithm based on frequency domain subband decomposition, according to multi-source heterogeneous sensing data streams. The processing module is used to extract the respiratory waveform envelope features, cardiac impact spectrum features, and motion pattern time-frequency features of the living target based on the coherent enhanced target signal cluster through time-frequency analysis technology, and to generate a joint feature vector using an attention mechanism weighted feature fusion method. The computation module is used to analyze the energy distribution characteristics of the live target in the cabin based on the joint feature vector, dynamically construct the multi-sensor coverage confidence matrix, perform singular value decomposition on the confidence matrix, and combine it with the adaptive bee colony division optimization algorithm to use the sensor weight allocation parameters as the foraging target of the bee colony. Each bee represents a set of candidate weight parameters. By utilizing the division of labor and cooperation of the bee colony and combining the live energy distribution characteristics in the cabin, the search strategy is dynamically adjusted. Through iterative optimization, the weight parameters converge to the final solution, and the sensor weight allocation parameters are generated. The intelligent decision-making module is used to construct a metal environment-adaptive decision-making model based on sensor weight allocation parameters and joint feature vectors, adaptively adjust sensor sampling strategies, and output quantitative indicators of the probability of the presence of living organisms and vital signs.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images