Mine rescue multivariate information life detection system and method

Abstract

The invention relates to the technical field of mine rescue detection, and discloses a mine rescue multivariate information life detection system and method.The mine rescue multivariate information life detection method comprises the steps that a heterogeneous sensor network is deployed, a belief propagation algorithm is applied to construct a self-organizing network, and dynamic topological optimization is achieved; an environment adaptability calibration system is constructed, and dynamic calibration of the sensor in complex environment changes is achieved; the sensor performance is quantified through a functional space model, degradation evaluation is carried out, and intelligent redistribution of sensing tasks is achieved; constructing a cooperative detection framework by using a belief propagation algorithm, and fusing multi-sensor data through a distributed decision algorithm; performing multi-dimensional feature extraction and analysis on the life signal, performing reliability evaluation on a result, and outputting a high-reliability detection result; according to the invention, the technical problems of sensor performance drift, function degradation and insufficient topological structure adaptability of a traditional mine rescue detection system in a complex environment are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of mine rescue detection, and more specifically, it relates to a mine rescue multi - information life detection system and method. Background Art

[0002] During the mine accident rescue process, it is necessary to accurately detect the trapped personnel in the complex and changeable environment to provide a scientific decision - making basis. The existing mine rescue detection methods mainly rely on sensor networks deployed fixedly. These networks often face three core problems in the post - disaster environment: the performance of sensors drifts due to environmental factors (such as temperature, humidity, air pressure, etc.); the functions of sensors are partially degraded after long - term use or under extreme conditions; the network topology is difficult to dynamically adapt to the post - disaster terrain changes. These problems severely limit the accuracy, reliability, and sustainability of life detection, affecting the rescue efficiency.

[0003] In the prior art, a single sensor array is often used for life signal detection. For example, seismic wave sensors are used to detect knocking signals, infrared sensors are used to detect heat sources, and gas sensors are used to monitor the change of carbon dioxide concentration, etc. However, these systems usually rely on preset fixed algorithms and threshold judgments, and cannot adapt to the complex and changeable environment after mine disasters, nor do they have the ability to adaptively adjust to the degradation of sensor performance. For example, in a high - temperature and high - humidity environment, the accuracy of sensors will significantly decrease. Even when most sensors fail, the system often cannot continue to provide effective detection support. In addition, the sensor networks in the prior art usually adopt a fixed topology structure, which is difficult to cope with the terrain changes and node failures after mine disasters. Once the sensors in some areas are damaged or the communication is interrupted, the detection ability of the entire network will be greatly reduced, even leading to detection blind spots, making it impossible to timely detect the trapped personnel.

[0004] Therefore, there is a need for a multi - information life detection system and method with environmental adaptability, function self - reconstruction, and topology self - optimization capabilities in the complex and changeable mine rescue environment, so as to improve the accuracy, reliability, and sustainability of life detection and provide more reliable technical support for rescue decision - making. Summary of the Invention

[0005] The present invention provides a mine rescue multi - information life detection system and method, which solves the technical problems of sensor performance drift, function degradation, and insufficient adaptability of the topology structure in the related art.

[0006] The present invention provides a mine rescue multi - information life detection method, including the following steps: Deploy a heterogeneous sensor network and apply the belief propagation algorithm to construct a self - organizing network to achieve dynamic topology optimization; Based on the optimized network topology, an environmental adaptability calibration system is constructed to achieve dynamic calibration of sensors in complex environmental changes; For the calibrated sensors, the performance of the sensors is quantified through a functional space model and degradation assessment is carried out to achieve intelligent reallocation of sensing tasks; Based on the reallocated sensing tasks, a collaborative detection framework is constructed using the belief propagation algorithm, and multi-sensor data is fused through a distributed decision algorithm; For the fused data, multi-dimensional feature extraction and analysis of vital signs are carried out, and the reliability of the results is evaluated, and high-confidence detection results are output.

[0007] In a preferred embodiment, in the step of deploying a heterogeneous sensor network and applying the belief propagation algorithm to construct a self-organizing network, the implementation method of the belief propagation algorithm is as follows: Each node sends information to its potential neighbor nodes, indicating the confidence in the state of the neighbor nodes; The node updates its own belief according to all the neighbor information received; By iteratively transmitting belief information, the network gradually converges to a stable state, forming an optimal communication topology structure.

[0008] In a preferred embodiment, the steps of constructing the environmental adaptability calibration system include: Deploy an environmental parameter monitoring subsystem to collect data on key environmental factors affecting sensor performance; Construct an environmental parameter and sensor performance impact model to establish a mapping relationship between environmental factors and sensor performance drift; Implement a distributed calibration algorithm to perform mutual calibration between sensors using spatial correlation; Dynamically select a calibration reference node group based on environmental similarity to adapt to local environmental changes.

[0009] In a preferred embodiment, in the step of quantifying the performance of the sensors through the functional space model, the functional space model is defined as: A quantitative evaluation model for four core performance dimensions of measurement range, sensitivity, accuracy, and precision; Each dimension obtains a performance score between 0 and 1 through a standardized test; Calculate the comprehensive health status index of the sensor through a weighting function.

[0010] In a preferred embodiment, the steps of implementing intelligent reallocation of sensing tasks include: Establish a sensing task model and define the demand vectors for each functional dimension; Calculate the matching degree between the remaining functions of the sensor and the task requirements; Evaluate the suitability of tasks and the remaining functions of sensors using a matching degree function; Construct a bipartite graph matching model and use the Hungarian algorithm to solve the optimal task allocation scheme.

[0011] In a preferred embodiment, the steps of constructing a collaborative detection framework using the belief propagation algorithm include: Divide the detection area into multiple sub-regions, and each sub-region is jointly monitored by multiple sensor nodes; Define a node state vector to represent the detection states of nodes for different types of information; Construct a node compatibility function to quantify the consistency of the detection states of two nodes; Optimize the detection state of each node by iteratively updating information to improve the overall detection accuracy.

[0012] In a preferred embodiment, in the step of fusing multi-sensor data through a distributed decision algorithm, weighted Dempster-Shafer evidence theory is used to fuse multi-modal features, including: Extract features from each sensor data and convert them into the degree of support for each hypothesis set; Define a trust function for each data source to measure the total support for the hypothesis; Introduce a weight vector and apply a weighted combination rule for feature fusion; Generate the fused detection result and its confidence score.

[0013] In a preferred embodiment, the steps of evaluating the reliability of the results include: Define a reliability evaluation index system, including signal strength, signal consistency, feature stability, and environmental interference degree; Establish a reliability evaluation model based on evidence theory to calculate the overall reliability of the detection results; Implement a context-aware data credibility evaluation model to calculate the credibility of data sources in the current environment; Generate the trapped person status information and the reliability evaluation report.

[0014] In a preferred embodiment, the method for detecting multi-source information for mine rescue further includes an adaptive working mode switching step based on the energy state: Establish a node energy state model to monitor battery power, energy consumption rate, and expected life; Define multiple working modes, including high-performance mode, standard mode, energy-saving mode, and sleep mode; Design an energy efficiency optimization function to balance performance and energy consumption; Based on the current energy state and detection requirements, dynamically adjust the working modes of each node.

[0015] In a preferred embodiment, a multi - information life detection system for mine rescue, which is used to execute the above - mentioned multi - information life detection method for mine rescue, includes: A heterogeneous sensor network self - organizing unit, which is used to realize dynamic topology optimization and self - perception of node positions; An environmental adaptability calibration unit, which is used to realize dynamic parameter optimization of sensors in complex environmental changes; A function degradation perception and reconstruction unit, which is used to quantify sensor performance and realize intelligent task re - allocation; A multi - point collaborative detection unit, which is used to realize collaborative detection of multi - sensor nodes and distributed decision - making; A life signal analysis and reliability evaluation unit, which is used to generate highly reliable life detection results and reliability reports.

[0016] The beneficial effects of the present invention are as follows: Improve detection accuracy and reliability: Through the multi - point collaborative detection framework and multi - modal signal fusion technology, combined with the environmental adaptability calibration system, the life detection accuracy is improved, and the false alarm rate is significantly reduced, providing more reliable data support for rescue decision - making.

[0017] Enhance environmental adaptability: By adopting the self - organizing network construction method and topology optimization technology, the system can adapt to the complex and changeable environment after a mine accident, the detection coverage rate is improved, and an effective life detection function can still be maintained even when some nodes fail.

[0018] Prolong the effective working time: Through the adaptive working mode switching technology based on the energy state and the intelligent task re - allocation after function degradation, the energy efficiency of the system is improved, the continuous working time is extended, and the needs of long - term rescue are met.

[0019] Optimize the resource utilization efficiency: Based on the sensor performance quantification technology based on the functional space model and the task adaptability evaluation model, the system can accurately perceive the function degradation of sensors and make full use of the remaining functions, improving the resource utilization efficiency of the system.

[0020] Provide decision reliability evaluation: Through the reliability evaluation model based on the evidence theory and the contextual data credibility evaluation model, the system quantitatively evaluates the reliability of the detection results, generates a detailed reliability evaluation report, helps rescue decision - makers understand the credibility of the data, and reduces decision - making risks. Brief Description of the Drawings

[0021] Figure 1 is a flowchart of a multi - information life detection method for mine rescue of the present invention; Detailed Embodiments

[0022] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and that changes can be made to the functions and arrangements of the elements discussed without departing from the scope of protection of the content of this specification. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0023] In at least one embodiment of the present invention, a multi - information life detection method for mine rescue is disclosed, as Figure 1 shown, and it includes the following steps: Step 1, deploy a heterogeneous sensor network and apply the belief propagation algorithm to construct a self - organizing network to achieve dynamic topology optimization; Specifically, it includes the following steps: Step 1.1, deploy a heterogeneous sensing network including acoustic sensors, thermal imaging sensors, gas sensors, and electromagnetic sensors for collecting multi - dimensional life signal data.

[0024] Among them, the acoustic sensor is used to collect acoustic wave signals and detect weak human activity sounds; The thermal imaging sensor collects thermal radiation data to locate temperature anomaly regions; The gas sensor collects gas concentration data such as carbon dioxide and oxygen to detect breathing activities; The electromagnetic sensor collects electromagnetic field change data to detect human movement and heartbeat.

[0025] Step 1.2, adopt a node communication system based on the belief propagation algorithm to enable each sensor node to construct a self - organizing network by exchanging information; The belief propagation algorithm is specifically implemented as follows: Each node sends information to its potential neighbor nodes indicating the confidence of node in the state of node ; Node updates its own belief according to all the neighbor information received : ; Among them, represents the belief value of node in its own state , that is, the probability estimate that node thinks it is in state , is the node The local evidence function, representing the initial assessment of the node's state based on its own sensing data; ; For node The neighbor set, containing all nodes that communicate directly with node ; Is the normalization constant to ensure that the sum of belief probabilities is 1; Represents the product operation on the information of all neighbor nodes .

[0026] By iteratively transmitting confidence information, the network gradually converges to a stable state, forming an optimal communication topology.

[0027] Step 1.3, realizing self-sensing of node positions and self-establishment of neighbor relationships; The specific method is as follows: Calculate the relative positions of nodes according to the signal strength, propagation delay, and multi-point triangulation algorithm between nodes; Construct an initial neighbor relationship graph based on the relative position information; Dynamically adjust the neighbor relationship according to the communication quality and energy consumption.

[0028] Step 1.4, automatically adjust the network structure using the topology optimization function, and the optimization function is expressed as: ; Wherein, Represents finding the topology structure that maximizes the objective function ; , Is the optimal topology structure, representing the network connection method finally selected by the system; Is the set of all possible topology structures; Is the coverage quality function of topology , evaluating the coverage effect of the network on the target area; Is the energy consumption function of topology ; Represents the importance of balancing the coverage quality and energy consumption. The coverage quality function Is comprehensively evaluated by calculating the sensor coverage density and sensing data quality in the key area, and the energy consumption function Considers the node communication frequency, transmission distance, and data volume.

[0029] Through the above steps, the system forms a heterogeneous sensor network with an adaptive topology structure, realizing multi-dimensional information collection and efficient transmission in complex environments, laying a foundation for subsequent sensor calibration and vital sign detection.

[0030] Step 2, based on the optimized network topology, construct an environmental adaptability calibration system to achieve dynamic calibration of sensors in complex environmental changes; Specifically, it includes the following steps: Step 2.1, deploy the environmental parameter monitoring subsystem; Collect data on key environmental factors affecting sensor performance: The temperature sensor collects environmental temperature data; The humidity sensor collects environmental humidity data; The barometric pressure sensor collects environmental barometric pressure data; The gas concentration sensor collects gas concentration data such as methane and carbon monoxide in the environment.

[0031] Step 2.2, construct an environmental parameter and sensor performance impact model, and establish a mapping relationship between environmental factors and sensor performance drift; The environmental parameter and sensor performance impact model adopts a hierarchical Bayesian network structure, expressed as: ; Among them, represents the posterior probability distribution of the sensor performance parameter under the condition of known environmental parameters ; represents the likelihood probability of observing the environmental parameter under the condition of known sensor performance parameter ; represents the prior probability distribution of the sensor performance parameter , that is, the initial estimate of the sensor performance before observing the environmental parameters; represents the marginal probability distribution of the environmental parameter , which is used as a normalization factor to ensure that the sum of posterior probabilities is 1.

[0032] The specific implementation steps are as follows: Collect a large amount of standard performance data of sensors under different environmental conditions to construct a training dataset; Use the maximum likelihood estimation method to learn the conditional probability distribution , that is, the probability of observing specific environmental parameters under the condition of known sensor performance parameters; Combine prior knowledge to determine the prior distribution of the sensor performance parameter , reflecting the prediction of the initial performance state of the sensor; Calculate the posterior distribution of the sensor performance parameter under the given environmental conditions through Bayesian inference, that is , which serves as the theoretical basis for sensor calibration.

[0033] The detailed implementation method of the hierarchical Bayesian network structure is as follows: The first layer (environmental parameter layer): contains environmental variable nodes such as temperature, humidity, air pressure, etc., and each node represents an environmental parameter; The second layer (intermediate correlation layer): contains environmental factor combined effect nodes to capture the interactive effects of multiple environmental factors; The third layer (sensor parameter layer): contains sensor performance parameter nodes such as sensitivity, accuracy, precision, etc.

[0034] The conditional probability tables in the network are constructed as follows: Use expert knowledge to initialize the dependencies between nodes and determine the network topology; Map the collected training data into the network structure and calculate the conditional probabilities between nodes; Apply the Expectation-Maximization (EM) algorithm to optimize the conditional probability parameters; Evaluate the network performance through cross-validation and iteratively optimize the network structure.

[0035] In the mine rescue scenario, an application example of this hierarchical Bayesian network is as follows: When a sudden change in temperature (rising from 25°C to 40°C) and an increase in humidity (rising from 50% to 85%) are detected in a specific area, the system automatically infers that the sensitivity of the acoustic sensor may decrease by 18% and the transmission delay may increase by 250 ms; Based on the inference results, the system pre-adjusts the amplification factor and signal processing parameters of the acoustic sensor to maintain the quality of data acquisition; At the same time, the system increases the sampling frequency of the thermal imaging sensor to supplement the possibly affected acoustic data and ensure the detection performance.

[0036] Step 2.3, implement a distributed calibration algorithm to perform mutual calibration between sensors using spatial correlation; Divide the sensor network into multiple calibration groups, and each group contains multiple spatially adjacent sensors; Within each calibration group, select the sensor with the most stable environmental conditions as the reference node; Other nodes calculate the calibration parameters by comparing their data with that of the reference node; The calibration parameters are optimized through Bayesian estimation: ; Among them, represents the sensor calibration parameter vector, which contains parameters such as gain and offset that need to be adjusted; represents the set of observed data containing the reference node and the nodes to be calibrated, is the likelihood function, which represents the probability of observing the data under the condition of the given calibration parameter ; is the prior distribution of the parameters, representing the initial belief distribution of the calibration parameters before observing the data; is the posterior probability, representing the updated belief of the calibration parameters after observing the data ; represents a direct proportional relationship.

[0037] Step 2.4: Dynamically select the calibration reference node group based on the environmental similarity to adapt to local environmental changes; Calculate the similarity between the current environmental parameters and the historical environmental parameters: ; where represents the similarity between the current environmental parameters and the historical environmental parameters, is the current environmental parameter vector, is the historical environmental parameter vector, represents the square of the Euclidean distance between the two environmental parameter vectors, that is, the sum of the squares of the differences of each parameter, is the scaling parameter for environmental similarity calculation, controlling the sensitivity of the similarity function. A smaller value makes the similarity more sensitive to environmental changes, is the exponential function; Select the node group that is most suitable as the calibration reference according to the environmental similarity and the sensor status score; When the environmental change exceeds the threshold, the process of reselecting the reference node group is automatically triggered.

[0038] Through the above steps, the system can monitor the changes of environmental parameters in real time, accurately establish the relationship model between environmental factors and sensor performance drift, and dynamically optimize the sensor parameters through the distributed calibration algorithm, effectively overcoming the problem of sensor performance drift caused by environmental changes and improving the accuracy of data acquisition.

[0039] Step 3: For the calibrated sensor, quantify the sensor performance through the functional space model and conduct degradation assessment to achieve intelligent reallocation of sensing tasks; Specifically, it includes the following steps: Step 3.1: Construct a sensor functional space model to comprehensively quantify the sensor performance; Define the functional space model: ; where represents the functional space model, represents the measurement range, indicating the range of physical quantities that the sensor can measure; represents the sensitivity, indicating the response degree of the sensor to input changes; ​Accuracy represents the closeness of the measured value to the true value; Precision represents the repeatability of the measurement result.

[0040] For different types of sensors, corresponding evaluation methods for functional space parameters are established.

[0041] The detailed implementation method of the functional space model is as follows: Construct a multi-dimensional performance feature vector: For each type of sensor, according to its working principle and characteristics, a feature vector containing 4 - 8 core performance dimensions is established. For example, for acoustic sensors, the frequency response range and signal-to-noise ratio dimensions are added; Establish performance quantization criteria: Standardize the performance parameters of each dimension to values between 0 and 1, where 1 represents the ideal state and 0 represents complete failure; Implement correlation analysis between dimensions: Calculate the mutual influence relationship between each performance dimension through the covariance matrix to identify key performance bottlenecks; Construct a comprehensive performance scoring function: ; Among them, is the weight of each dimension, representing the importance of each performance dimension in the overall score. The value range is 0 - 1, and the sum of all weights is 1, which is dynamically adjusted according to the task type; is for the th standardized score of the performance dimension, representing the performance level of this dimension. The value range is 0 - 1, 1 represents the best performance, and 0 represents complete failure; is the comprehensive performance score of the sensor. The value range is 0 - 1, and the closer it is to 1, the better the overall performance of the sensor.

[0042] Application example of the functional space model in the mine rescue environment: For gas sensors, the system automatically identifies that its sensitivity drops to 0.72 (standard value 0.95) in a high-humidity environment, while the measurement range and precision basically remain normal (0.92 and 0.88 respectively); The system determines that this sensor is still suitable for detecting the trend of gas concentration changes, but not suitable for accurately measuring the absolute concentration value; Accordingly, the system adjusts the task of this sensor to monitor the breathing pattern instead of measuring the absolute concentration of carbon dioxide, and at the same time increases its sampling frequency to compensate for the decrease in sensitivity; The system also configures adjacent thermal imaging sensors as an auxiliary verification means to form a complementary detection strategy.

[0043] Step 3.2, realize real-time monitoring and degradation assessment of sensor performance; Perform self - test regularly and collect performance index data of the sensor; Compare with the reference performance and calculate the performance degradation vector: ; wherein, represents the current performance vector of the sensor, including current performance parameters such as measurement range, sensitivity, accuracy and precision; represents the reference performance vector of the sensor, including standard performance parameters after factory or calibration, represents the performance degradation vector, indicating the degradation degree of each performance dimension relative to the reference state; Adopt the Principal Component Analysis (PCA) method to map the multi - dimensional degradation vector to the significant feature space and extract key degradation features. This method identifies the main performance dimensions that contribute the most to the degradation state by calculating the eigenvalues and eigenvectors of the covariance matrix of the degradation data; Calculate the health state index of the sensor according to the degradation features: ; where represents the Euclidean norm of the degradation vector, that is, the square root of the sum of the squares of each component of the degradation vector, used to quantify the overall degradation degree, represents the Euclidean norm of the reference performance vector, that is, the square root of the sum of the squares of each component of the reference performance vector, used to standardize the degradation degree, represents the health state index of the sensor, with a value range of 0 - 1. The closer the value is to 1, the healthier the sensor is, and the closer the value is to 0, the more serious the performance degradation of the sensor is.

[0044] Step 3.3, accurately locate the function degradation dimension and degree; Normalize each component of the degradation vector to obtain the normalized degradation vector: ; wherein, represents the normalized degradation vector, , , , respectively represent the normalized degradation values of the measurement range dimension, sensitivity dimension, accuracy dimension and precision dimension, with a numerical range of 0 - 1. The larger the value, the more serious the dimension degradation; Set the degradation threshold vector for each dimension: ; wherein, represents the degradation threshold vector for each dimension, , , , respectively represent the maximum degradation thresholds acceptable for the measurement range dimension, sensitivity dimension, accuracy dimension, and precision dimension. Exceeding this value will affect the normal function of the sensor; Compare with to identify the dimension exceeding the threshold as the main degradation dimension; Calculate the percentage of degradation for each dimension , which represents the percentage of degradation of each performance dimension relative to its threshold and is used to visually evaluate the severity of degradation. Among them, represents the percentage of degradation for each dimension, represents the normalized degradation vector, represents the degradation threshold vector for each dimension.

[0045] Step 3.4, adopt a task adaptability evaluation model to dynamically reallocate sensing tasks; Establish a sensing task model , where , , respectively represent the , , th sensing tasks, represents the number of sensing tasks; Each task includes a demand vector for each functional dimension , where , , respectively represent the minimum required values of this task for the measurement range dimension, sensitivity dimension, accuracy dimension, and precision dimension of the sensor; Calculate the remaining functions of the sensor , where represents the reference performance vector of the sensor, which contains the standard performance parameters after factory or calibration; represents the performance degradation vector, indicating the degree of degradation of each performance dimension relative to the reference state; represents the current remaining functional performance vector of the sensor; Use a matching function to evaluate the adaptability of the task to the remaining functions of the sensor: ; Among them, represents the matching degree between task and sensor , and the value range is 0 - 1. The closer the value is to 1, the higher the matching degree; is the weight coefficient for each functional dimension, indicating the importance of different performance dimensions for task completion, and satisfying ; represents the task 's demand value for the functional dimension ; represents the remaining performance value of the sensor on the functional dimension ; represents taking the smaller value between the task demand and the remaining performance of the sensor to ensure that it does not exceed the task demand; represents the proportion of the remaining performance of the sensor that meets the task demand; Construct a bipartite graph matching model and use the Hungarian algorithm to solve the optimal task allocation scheme to maximize the overall matching degree , where is a 0-1 variable. When the value is 1, it means that the task is allocated to the sensor , and when the value is 0, it means no allocation; represents the overall matching degree of the entire system, which is the optimization objective function.

[0046] Through the above steps, the system can real-time sense the functional degradation of sensors, accurately locate the degradation dimension and degree, and based on the task adaptability evaluation model, achieve intelligent reallocation of sensing tasks, make full use of the remaining functions of sensors, and ensure the stability and reliability of the system during long-term operation.

[0047] Step 4: Based on the reallocated sensing tasks, use the belief propagation algorithm to construct a collaborative detection framework and fuse multi-sensor data through a distributed decision-making algorithm; Specifically, it includes the following steps: Step 4.1: Construct a multi-node collaborative detection framework and use the belief propagation algorithm to achieve information sharing and collaboration; Divide the detection area into multiple sub-areas, and each sub-area is jointly monitored by multiple sensor nodes; Define the node state vector , where represents the state vector of the sensor node , including the detection states of this node for various types of information; , , respectively represent the detection states of the node for the , , types of information; represents the total number of information categories that the system can detect; Construct a collaboration model between nodes based on the belief propagation algorithm: ; Among them, represents the message passed from node to node , including the inference information of node about the state of node ; represents the summation operation on all possible states of node ; represents the compatibility function between node and node , quantifying the consistency degree of the detection states of the two nodes; represents the set of neighbor nodes of node ; represents the set after excluding node from the set of neighbor nodes of node ; represents the product of the messages received by node from all neighbor nodes except , integrating the inferences of other nodes about the state of node ; By iteratively updating the information, the detection state of each node is optimized, and the overall detection accuracy is improved.

[0048] The detailed implementation method of the multi-node collaborative detection framework is as follows: Construct a factor graph representation: Model the sensor network as a bipartite graph structure, where variable nodes represent sensor states and factor nodes represent the interaction relationships between sensors; Define a message passing protocol: Design two types of message formats, variable-to-factor messages and factor-to-variable messages, including state estimation and confidence information; Implement an adaptive message scheduling: Dynamically adjust the message passing order and frequency according to node importance, message novelty, and energy status; Design a convergence judgment criterion: Define a confidence change threshold, and when the confidence change is less than the threshold in consecutive iterations, the algorithm is considered to converge.

[0049] Application example of the collaborative detection framework based on the belief propagation algorithm: In the mine collapse area, a group of acoustic sensors detected faint knocking sounds (confidence 0.65), but this information alone cannot determine whether it is made by trapped personnel; The system starts the collaborative detection process, and adjacent thermal imaging sensors report detecting possible human heat sources (confidence 0.58), but there is also uncertainty; Gas sensors detected periodic changes in the carbon dioxide concentration in this area (confidence 0.72), which conforms to the human breathing pattern; Through the belief propagation algorithm for state fusion, the data of the three sensors are mutually verified, and the final comprehensive confidence level is increased to 0.91, and the system confirms that there are trapped persons at this location; At the same time, the system identifies that the acoustic signal is synchronous with the breathing pattern, infers that the trapped person is in a waking state, and provides the rescue personnel with accurate position coordinates.

[0050] Step 4.2, implement the extraction of life signal features and multimodal fusion; Apply time-frequency analysis and pattern recognition algorithms to the acoustic signal to extract the sound features of human activities; Apply thermal anomaly detection and morphological analysis to the thermal imaging data to identify the human thermal radiation characteristics; Apply fluctuation analysis and trend recognition to the gas concentration data to extract the breathing activity characteristics; Apply spectrum analysis and pattern matching to the electromagnetic field data to extract the heartbeat and motion characteristics; Adopt the weighted Dempster-Shafer evidence theory to fuse multimodal features: ; Among them, represents the basic probability assignment to the hypothesis set after fusion; represents the basic probability assignment function of the th type of sensing data to the hypothesis set ; represents that the intersection of all hypothesis sets is equal to the hypothesis set ; represents that the intersection of all hypothesis sets is an empty set; represents the product of the basic probability assignments of all sensors; represents the Dempster combination rule operator.

[0051] The detailed implementation method of the weighted Dempster-Shafer evidence theory multimodal fusion is as follows: Recognition and judgment framework construction: Define a complete set of mutually exclusive hypotheses , for example , and the corresponding power set , among which, represents a complete set of mutually exclusive hypotheses, , , respectively represent the , , th different hypotheses; represents the total number of hypotheses; Basic probability assignment calculation: For each type of sensing data, design a specific feature probability mapping function to convert the extracted features into the degree of support for each hypothesis set; Belief function construction: For each data source , define the belief function to measure the total degree of support for the hypothesis , where represents the belief function value of the data source for the hypothesis set ; represents the basic probability assignment value of the data source for the hypothesis subset ; represents the summation operation for all subsets of the hypothesis set ; Weighted fusion implementation: Introduce the weight vector , where , , respectively represent the weights of the , -th sensor data sources, represents the number of sensor data sources; Apply the weighted combination rule: ; where represents the basic probability assignment for the hypothesis set after weighted fusion; represents the weight of the -th sensor data source; represents the weight exponentiation operation on the basic probability assignment value of the -th sensor; represents that the intersection of all hypothesis sets is equal to the hypothesis set ; represents that the intersection of all hypothesis sets is an empty set; represents the product of the basic probability assignments of all weighted sensors, reflecting the degree to which multiple sensors jointly support a certain hypothesis.

[0052] Example of multimodal fusion application in mine rescue environment: At the scene of a mine accident, the system simultaneously collects data from four sensors for life signal detection: The acoustic sensor detects a rhythmic knocking sound, and its basic probability assignment is: , , ; The thermal imaging sensor detects a weak heat source, and its basic probability assignment is as follows: , , ; The gas sensor detects periodic carbon dioxide fluctuations, and its basic probability assignment is as follows: , , ; The electromagnetic sensor does not detect an obvious signal, and its basic probability assignment is as follows: , , .

[0053] The system assigns a weight vector according to the current environmental conditions and the status of each sensor ; Apply the weighted Dempster-Shafer fusion algorithm to obtain the fused basic probability assignment: , , ; Based on this, the system confirms that the possibility of a trapped person existing at this location is extremely high and initiates a directional rescue procedure.

[0054] Step 4.3: Implement a distributed decision-making algorithm to adaptively adjust the detection strategy; Establish a hierarchical decision-making structure, including a node layer, a region layer, and a global layer; The decision-making at the node layer makes a preliminary judgment based on local sensing data and confidence; The decision-making at the region layer fuses the judgment results of multiple nodes within the same region; The decision-making at the global layer integrates the decision results of multiple regions to form a final conclusion on life signal detection; Adopt an adaptive strategy optimization model based on the Markov decision process (MDP): ; Among them, represents the optimal decision-making strategy in state ; represents the current state of the system; represents the possible action choices; represents the next state that the system may transfer to after executing action ; represents the probability of transferring from state to state after taking action ; represents the immediate reward obtained by transferring from state to state by taking action ; is the discount factor (ranging from 0 to 1), which is used to balance the importance of immediate rewards and long-term benefits; represents the state of the optimal value function, which represents the maximum cumulative reward that can be obtained starting from state ; represents selecting the action that maximizes the subsequent expression .

[0055] Step 4.4, implement adaptive working mode switching based on the energy state; Establish a node energy state model to monitor battery power, energy consumption rate, and expected life; Define multiple working modes, including high-performance mode, standard mode, energy-saving mode, and sleep mode; Design an energy efficiency optimization function: ; where is the optimal working mode, that is, the operating mode with the highest energy efficiency that the system should select; is the set of all possible working modes; represents a specific working mode in the set ; is the performance function of mode , which quantifies the detection performance and response ability of the system in this mode; is the trade-off coefficient, which is used to balance the weight relationship between performance and energy consumption, and reflects the emphasis on performance and endurance in the current task; represents selecting the working mode that maximizes the subsequent expression .

[0056] Based on the current energy state and detection requirements, dynamically adjust the working modes of each node to maximize energy efficiency.

[0057] Through the above steps, the system realizes the collaborative detection and distributed decision-making of multi-sensor nodes, can effectively fuse various life signal characteristics, adaptively adjust the detection strategy, and optimize the working mode according to the energy state, while improving the detection accuracy, extending the system working time, and enhancing the adaptability of the system in complex environments.

[0058] Step 5, perform multi-dimensional feature extraction and analysis of the fused data, and conduct reliability assessment on the results, and output highly reliable detection results; Specifically, it includes the following steps: Step 5.1, Implement multi-dimensional feature extraction and analysis of life signals: Apply multi-scale time-frequency analysis to the original sensing data to extract time-domain, frequency-domain, and time-frequency domain features; For acoustic features, use principal component analysis and independent component analysis to separate possible acoustic features of human activities; For thermal imaging features, use a thermal target tracking algorithm to identify thermal radiation sources with human characteristics; For gas concentration features, apply periodic analysis and fluctuation pattern recognition to extract respiratory features; For electromagnetic field features, use spectrum analysis and wavelet transform to extract heartbeat signal features; Construct a feature vector , where represents the comprehensive feature vector of life signals; , , respectively represent the , , th life signal features; represents the total number of features; Each feature has been standardized for subsequent pattern recognition and classification analysis.

[0059] Step 5.2, Apply a reliability evaluation model to quantify the confidence level of the detection results; Define a reliability evaluation index system, including dimensions such as signal strength, signal consistency, feature stability, and environmental interference degree; Establish a reliability evaluation model based on evidence theory: ; where represents the overall reliability score of the detection result ; represents the life signal detection result to be evaluated; represents the total number of evaluation indicators; represents the th reliability score of the detection result under the th evaluation indicator, with a value range of 0 to 1; represents the weight coefficient of the th evaluation indicator; Use historical data and expert knowledge to determine the scoring criteria and weight coefficients of each indicator; Classify the detection results into three levels: high confidence, medium confidence, and low confidence according to the reliability score.

[0060] The detailed implementation method of the reliability evaluation model based on evidence theory is as follows: Construct a multi - layer evaluation index system: The first layer: Overall reliability; The second layer: Four dimensions of data quality, model applicability, environmental impact, and result consistency; The third layer: Specific evaluation indicators under each dimension, a total of 12 - 15 refined indicators; Design a scoring function mapping: Design a specific scoring function for each refined indicator to convert the observed value into a standardized score within the range of 0 - 1; Implement AHP (Analytic Hierarchy Process) weight determination: Construct a judgment matrix through expert evaluation and pairwise comparison, and calculate the weights of each level of indicators; Establish an evidence accumulation process: Introduce the DS theory framework, use the scores of each indicator as supporting evidence, and accumulate through the combination rule to form an overall reliability evaluation.

[0061] Application example of the reliability evaluation model based on evidence theory: After detecting a suspected life signal at a coal mine accident site, the system starts the reliability evaluation process: Evaluation of the data quality dimension (weight 0.30): Signal strength indicator: The multi - point average signal - to - noise ratio is 8.3 dB, corresponding to a score of 0.76; Data integrity indicator: The sensor coverage rate is 92%, corresponding to a score of 0.85; Sampling sufficiency indicator: The number of samples is sufficient, corresponding to a score of 0.90; Weighted score of the data quality dimension: 0.83; Evaluation of the model applicability dimension (weight 0.25): Model - environment matching degree: The current dust concentration affects the model performance, corresponding to a score of 0.72; Parameter fitness: Automatically adjust parameters to reach the optimum, corresponding to a score of 0.88; Weighted score of the model applicability dimension: 0.80; Evaluation of the environmental impact dimension (weight 0.25): Interference source intensity: There is equipment vibration interference, corresponding to a score of 0.65; Environmental stability: The temperature fluctuates greatly, corresponding to a score of 0.70; Weighted score of the environmental impact dimension: 0.68; Evaluation of the result consistency dimension (weight 0.20): Multi - source consistency: The data of four types of sensors support each other, corresponding to a score of 0.92; Temporal consistency: The signal is continuously and stably detected for 20 minutes, corresponding to a score of 0.88; Weighted score of the result consistency dimension: 0.90; Overall reliability calculation: ; The system determines this detection result as the "highly credible" level (threshold is 0.75) and preferentially allocates rescue resources.

[0062] Step 5.3: Implement a contextual data credibility assessment model to calculate the credibility of the data source in the current environment; Establish an environmental context model , including environmental parameters such as temperature, humidity, air pressure, and noise level; Collect the credibility annotation data of the data source in different environmental contexts to construct a training set ; Use the contextual similarity weighted algorithm to calculate the credibility of the data source in the current environment: ; where represents the credibility of the data source in the environment ; represents the th data source; represents the current environmental context model; represents the th environmental context in the historical record; represents the total number of historical environmental context samples; represents the environment = and the similarity of the historical environment , with a value range from 0 to 1, and the larger the value, the more similar the two environments are; represents the known credibility of the data source in the historical environment ; is the normalized weight coefficient used to balance the importance of different historical environment samples; Comprehensively evaluate the credibility of multiple data sources to form a final detection result credibility report.

[0063] Step 5.4: Generate the trapped person status information and reliability assessment report; Based on multi-dimensional life signal characteristics, determine the location coordinates, quantity, and physiological status of the trapped persons; The location information is calculated using the triangulation and signal strength weighted average algorithm, and the positioning error estimate is given; The number of people is estimated through cluster analysis and multi-source information verification, and the estimated confidence interval is provided; The physiological status analysis includes the assessment of activity status (static / moving), consciousness status (awake / comatose), and vital signs (normal / abnormal); Generate a structured reliability assessment report, including the following content: The overall reliability score and confidence level of the detection results; The credibility of each data source and its applicability assessment in the current environment; Analysis of uncertain factors that may affect the detection results; Suggestions for result improvement, such as additional sensing data to be supplemented or detection parameters to be adjusted.

[0064] Through the above steps, the system can comprehensively analyze the collected multi-dimensional life signal data, accurately locate the trapped personnel, evaluate their quantity and physiological status, and conduct a reliability assessment of the detection results, providing highly credible life detection results and a detailed reliability report to provide a scientific basis for rescue decision-making.

[0065] Application example of this embodiment: This embodiment has been applied in the rescue process of a coal mine collapse accident in Shanxi. The depth of the mine is about 580 meters, the area of the collapsed area is about 1200 square meters, the geological structure is complex, and multiple enclosed spaces are formed after the collapse. When the accident occurred, 23 miners were trapped. The collapse caused the communication system to be interrupted, part of the ventilation system to be damaged, and some areas to be flooded. The rescue environment was extremely complex. The main technical challenges faced in the rescue included: Harsh environmental conditions: The on-site temperature fluctuates (20 - 42 °C), the humidity is extremely high (85 - 98%), there are a large amount of dust and harmful gases, and the performance of traditional sensing devices is severely limited; Topographical structure changes: The channels formed after the collapse are narrow and unstable, and some areas are only 40 - 60 cm wide, making it difficult to lay conventional fixed networks; Long-term rescue requirements: It is expected that the rescue time will take 72 - 96 hours, and the sensing devices need to work stably in a harsh environment for a long time.

[0066] In this rescue mission, a multi-information life detection system of this embodiment was deployed, including 85 heterogeneous sensing nodes, covering the collapsed area and the surrounding areas where there may be living spaces, in order to accurately locate the trapped personnel and evaluate their physiological status, providing a decision-making basis for scientific rescue. Example of the implementation process:

[0067] Implementation process of multi-information acquisition and network self-organization: Within 2 hours after the start of the rescue operation, the rescue team deployed the heterogeneous sensor network shown in Table 1. The sensors were placed at different positions in the collapsed area through rescue channels, drill holes, and robots.

[0068] Table 1: Composition of heterogeneous sensor nodes deployed at the rescue site;

[0069] After the deployment is completed, the system automatically starts the network self-organization process. In the initial stage, due to terrain limitations and signal attenuation, only 63% of the nodes successfully establish communication connections. The system performs topology optimization through the belief propagation algorithm. After 27 minutes and 139 rounds of iteration, the network topology performance shown in Table 2 is finally formed.

[0070] Table 2: Comparison of performance before and after network topology optimization;

[0071] During the topology optimization process, the system also automatically completes node position perception, with an average error of 3.8 meters and a maximum error not exceeding 7.2 meters, meeting the requirements of life detection and positioning.

[0072] In practical applications, the self-organizing network demonstrates excellent adaptability. At the 28th hour of the rescue process, due to a small-scale secondary collapse, 9 sensor nodes failed and 17 node positions changed. The system automatically completed network reconstruction within 8 minutes, re-established a stable communication topology, ensured the continuity of data collection, and avoided the large-scale paralysis that might occur in traditional fixed networks in this situation.

[0073] Implementation process of environmental parameter monitoring and sensor dynamic calibration: The environmental parameters at the rescue site fluctuate significantly, especially temperature, humidity, and gas concentration. Taking Area 12 (one of the most important potential survival spaces) as an example, the changes in environmental parameters within 48 hours are shown in Table 3.

[0074] Table 3: Changes in environmental parameters in Area 12 within 48 hours;

[0075] These environmental changes cause obvious drifts in sensor performance. For example, the sensitivity of acoustic sensors decreases by an average of 23% in high-temperature and high-humidity environments (temperature > 38°C, humidity > 92%), and the penetration ability of electromagnetic sensors decreases by about 17% in high methane concentration environments (> 1.8%).

[0076] The system effectively addresses these challenges through the dynamic calibration process. The hierarchical Bayesian network calculates the impact of environmental factors on sensor performance in real-time and optimizes the parameters. When the high-temperature and high-humidity area is first detected, the system automatically adjusts the parameters of the acoustic sensors, including increasing the gain and adjusting the filter parameters. The performance comparison of the acoustic sensors in the same area before and after calibration is shown in Table 4.

[0077] Table 4: Comparison of performance of acoustic sensors before and after calibration under environmental changes;

[0078] The system dynamically selects calibration reference nodes according to environmental similarity. During the rescue process, the calibration reference node group in Area 12 changed dynamically 8 times, ensuring calibration accuracy. In this way, the system maintained more than 85% of its sensing performance under extreme environmental conditions, far exceeding the approximately 60% level of traditional systems.

[0079] Implementation process of sensor function degradation perception and capability reconstruction: During the long-term rescue process, multiple sensor nodes showed varying degrees of function degradation. The system evaluated the health status of sensors through a functional space model. Table 5 shows the statistics of the function degradation of various sensors during the rescue process.

[0080] Table 5: Statistics of sensor function degradation during 72-hour rescue;

[0081] The system accurately identified the function degradation dimensions and degrees of each sensor using the functional space model. Taking the No. 5 thermal imaging sensor as an example, the analysis of its function degradation dimensions is shown in Table 6.

[0082] Table 6: Analysis of function degradation dimensions of the No. 5 thermal imaging sensor (after 72 hours);

[0083] Based on the function degradation assessment, the system adjusted the sensing task allocation strategy in real time. For the No. 5 thermal imaging sensor, since its sensitivity decreased significantly but the measurement range remained basically the same, the system adjusted its task to large-scale thermal anomaly screening instead of precise temperature measurement. At the same time, the system increased the sampling frequency of other sensors with better performance in this area to ensure data quality.

[0084] Through this intelligent task allocation strategy, the system maintained 87% of its overall detection ability even when more than 25% of the nodes showed severe degradation or failure, ensuring the continuous progress of the rescue operation.

[0085] Implementation process of multi-point collaborative detection and distributed decision-making: In the rescue operation, the multi-point collaborative detection and distributed decision-making framework played a key role in accurately identifying life signals. At the 18th hour of the rescue, the acoustic sensor in Area C3 captured a weak knocking signal, but the signal-to-noise ratio was only 4.2 dB. It was difficult to determine whether it was a human activity based on this signal alone. The system initiated a collaborative detection process based on the belief propagation algorithm, integrating information from multiple surrounding sensors. The specific process is shown in Table 7.

[0086] Table 7: Collaborative detection process and confidence change in Area C3;

[0087] Based on the Dempster-Shafer evidence theory for multimodal fusion, the system successfully fused multiple single-point detections with low confidence (average confidence about 0.56) into a comprehensive judgment with high confidence (0.94), and accurately located the position of the trapped persons with an error of less than 4.8 meters.

[0088] In terms of energy management, the system dynamically adjusted the working mode according to the node power status and task priority. Table 8 shows the energy consumption and working duration of the system during the entire rescue process.

[0089] Table 8: Energy Consumption and Working Duration under Different Working Modes;

[0090] Through adaptive working mode switching, the overall energy efficiency of the system increased by approximately 280%, enabling most sensors to continue working until the rescue was completed, while a system with a traditional fixed working mode was expected to work only for 25 - 30 hours.

[0091] Implementation process of comprehensive analysis and reliability assessment of life signals: During the entire rescue process, the system detected a total of 18 possible life signals, among which 14 were confirmed as the positions of the trapped persons, 3 were false positive detections (caused by equipment vibration or animal activities), and 1 could not be confirmed. The system conducted a reliability assessment on each detection result, and Table 9 shows the reliability scores and final results of each detection position.

[0092] Table 9: Detection Positions of Life Signals and Reliability Assessment (Partial);

[0093] Through the reliability assessment model based on evidence theory, the system successfully classified all false positive detections into the "low credibility" level, while all real life signals were classified into the "medium credibility" or "high credibility" level. This enabled the rescue team to reasonably allocate resources and prioritize the high-confidence areas.

[0094] The system also used a contextual data credibility assessment model to dynamically evaluate the credibility of each data source according to the current environmental conditions. For example, in high-temperature and high-humidity areas, the system automatically reduced the credibility weight of the thermal imaging sensor and increased the weights of the acoustic and gas sensors, making the detection results more accurate.

[0095] Finally, the system output a comprehensive assessment report including location, number of people, and physiological status, providing precise guidance for the rescue operation. During this rescue operation, all 23 trapped persons were successfully rescued, among which the positions of 21 people were consistent with the system prediction (error < 5 meters), and the accuracy rate of physiological status assessment reached 92%. Verification of technical effects:

[0096] In this mine rescue case, the two most important technical effects of this implementation method have been fully verified: improved detection accuracy and reliability, and enhanced environmental adaptability.

[0097] Effect of improved detection accuracy and reliability: Through environmental adaptability calibration and multi-modal information fusion, this implementation method significantly improves the accuracy and reliability of life detection. Table 10 shows the performance comparison between this system and traditional life detection systems under the same rescue conditions.

[0098] Table 10: Comparison of detection accuracy and reliability between this implementation method and traditional systems;

[0099] As can be seen from Table 10, this implementation method significantly outperforms traditional systems in all key indicators. Especially in terms of the false alarm rate, it has decreased from 21.4% to 4.3%, a decrease of 79.9%; the position location accuracy has increased by 66.1%, which is crucial for accurately guiding rescue personnel. The overall reliability score reaches 0.89, providing a highly credible basis for rescue decision-making.

[0100] Effect of enhanced environmental adaptability: Through self-organizing network topology and function self-reconfiguration technology, this implementation method greatly enhances the system's adaptability in complex and changing environments. Table 11 shows the comparison of the system's performance retention rate under different environmental conditions.

[0101] Table 11: Comparison of the system's performance retention rate under different environmental conditions;

[0102] The data in Table 11 shows that this implementation method maintains a high system performance under various harsh environmental conditions, with an average performance retention rate of 85.0%, which is 61.3% higher than that of traditional systems. Especially in the face of extreme situations such as secondary collapses and significant losses of network nodes, the advantages of this implementation method are more prominent, with the performance retention rate increasing by 119.3% and 102.6% respectively.

[0103] During the 72-hour continuous rescue process, this system coped with 8 environmental structure changes through the dynamic adjustment of the self-organizing network topology and successfully processed 73 sensor nodes with varying degrees of degradation through function self-reconfiguration technology, ensuring the continuity and stability of the system's functions and the smooth completion of the rescue operation.

[0104] In summary, this embodiment has achieved excellent application results in actual mine rescue scenarios, breaking through multiple technical limitations of traditional systems in complex environments, and providing more reliable and efficient life detection technical support for mine rescue.

[0105] The embodiments of the present invention have been described above. However, these embodiments are not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the protection scope of this embodiment.

Claims

1. A multi - information life detection method for mine rescue, characterized in that, It includes the following steps: Deploy a heterogeneous sensor network and apply the belief propagation algorithm to construct a self-organizing network to achieve dynamic topology optimization; Based on the optimized network topology, construct an environment adaptability calibration system to achieve dynamic calibration of sensors in complex environmental changes; For the calibrated sensors, quantify the sensor performance through a functional space model and conduct degradation assessment to achieve intelligent reallocation of sensing tasks; Based on the reallocated sensing tasks, use the belief propagation algorithm to construct a collaborative detection framework and fuse multi-sensor data through a distributed decision-making algorithm; For the fused data, extract and analyze multi-dimensional features of vital signs, and conduct reliability assessment on the results to output highly reliable detection results.

2. The method for detecting multi - element information of life in mine rescue according to claim 1, wherein, In the step of deploying a heterogeneous sensor network and applying the belief propagation algorithm to construct a self-organizing network, the implementation method of the belief propagation algorithm is as follows: Each node sends information to its potential neighbor nodes, indicating the confidence in the states of the neighbor nodes; The node updates its own belief according to all the neighbor information received; Through iterative transmission of confidence information, the network gradually converges to a stable state, forming an optimal communication topology structure.

3. A method for detecting the life of multiple information in mine rescue according to claim 1, characterized in that, The steps of constructing the environment adaptability calibration system include: Deploy an environmental parameter monitoring subsystem to collect data on key environmental factors affecting sensor performance; Construct an environmental parameter and sensor performance impact model to establish a mapping relationship between environmental factors and sensor performance drift; Implement a distributed calibration algorithm to perform mutual calibration between sensors using spatial correlation; Dynamically select a calibration reference node group based on environmental similarity to adapt to local environmental changes.

4. A method for detecting the life of multiple information in mine rescue according to claim 1, characterized in that, In the step of quantifying sensor performance through a functional space model, the functional space model is defined as: A quantitative evaluation model for four core performance dimensions of measurement range, sensitivity, accuracy, and precision; Each dimension obtains a performance score between 0 and 1 through a standardized test; Calculate the comprehensive health status index of the sensor through a weighting function.

5. A multi - information life detection method for mine rescue according to claim 1, characterized in that, The steps of achieving intelligent reallocation of sensing tasks include: Establish a sensing task model and define the demand vectors for each functional dimension; Calculate the matching degree between the remaining functions of the sensor and the task requirements; Use a matching degree function to evaluate the suitability of the task and the remaining functions of the sensor; Construct a bipartite graph matching model and use the Hungarian algorithm to solve the optimal task allocation scheme.

6. The method for detecting multiple information of life for mine rescue according to claim 1, characterized in that, The steps of constructing a collaborative detection framework using the belief propagation algorithm include: Divide the detection area into multiple sub-areas, and each sub-area is jointly monitored by multiple sensor nodes; Define a node state vector to represent the detection states of the node for different types of information; Construct a node compatibility function to quantify the consistency of the detection states of two nodes; Optimize the detection state of each node through iterative update of information to improve the overall detection accuracy.

7. A method for detecting the life of multiple information in mine rescue according to claim 1, characterized in that, In the step of fusing multi-sensor data through a distributed decision-making algorithm, the weighted Dempster-Shafer evidence theory is used to fuse multi-modal features, including: Extract features from the data of each sensor and convert them into the degree of support for each hypothesis set; Define a trust function for each data source to measure the total support for the hypothesis; Introduce a weight vector and apply a weighted combination rule for feature fusion; Generate the fused detection results and their confidence scores.

8. A method for detecting multi - element information of life in mine rescue according to claim 1, characterized in that, The steps for reliability assessment of the results include: Define a reliability assessment index system, including signal strength, signal consistency, feature stability, and environmental interference level; Establish a reliability assessment model based on evidence theory to calculate the overall reliability of the detection results; Implement a contextual data credibility assessment model to calculate the credibility of data sources in the current environment; Generate the status information of trapped persons and a reliability assessment report.

9. A method for detecting multi - source information of life in mine rescue according to claim 1, characterized in that It also includes an adaptive working mode switching step based on the energy state: Establish a node energy state model to monitor battery power, energy consumption rate, and expected lifespan; Define multiple working modes, including high-performance mode, standard mode, energy-saving mode, and sleep mode; Design an energy efficiency optimization function to balance performance and energy consumption; Based on the current energy state and detection requirements, dynamically adjust the working modes of each node.

10. A multi - information life detection system for mine rescue, which is used to execute a mine rescue multi - information life detection method according to any one of claims 1 - 9, and is characterized in that, It includes: A heterogeneous sensor network self-organization unit for realizing dynamic topology optimization and self-awareness of node positions; An environmental adaptability calibration unit for realizing dynamic parameter optimization of sensors in complex environmental changes; A function degradation perception and reconstruction unit for quantifying sensor performance and realizing intelligent task reallocation; A multi-point collaborative detection unit for realizing collaborative detection and distributed decision-making of multiple sensor nodes; A life signal analysis and reliability assessment unit for generating highly credible life detection results and reliability reports.

Images