Self-starting forest fire monitoring device and monitoring method

Abstract

The application discloses a self-starting forest fire monitoring device and a monitoring method, and relates to the technical field of forest fire monitoring.The device comprises a multi-physical field gradient sensing module, a volatile organic compound concentration gradient detection module, an ion wind sensing module, a fire electromagnetic characteristic identification module, a multi-source abnormal signal fusion processing module, a cooperative verification communication module and a hierarchical self-starting control module.The multi-physical field gradient sensing module monitors the gradient change rate of temperature, air pressure and humidity, generates a hierarchical trigger signal when the gradient change rate exceeds a dynamic threshold value, and automatically starts the subsequent detection module; the volatile organic compound concentration gradient detection module determines a fire source through divergence analysis; the ion wind sensing module analyzes ion field disturbance characteristics; the multi-source abnormal signal fusion processing module calculates a fire probability by using Bayesian inference; and the cooperative verification communication module confirms a fire through distributed voting.The application realizes early self-starting monitoring of forest fires through hierarchical triggering and multi-dimensional verification, and improves detection accuracy.

Description

Technical Field

[0001] This invention belongs to the field of forest fire monitoring technology, and particularly relates to a self-starting forest fire monitoring device and monitoring method. Background Technology

[0002] Forest fires are a major disaster threatening the ecological environment and human life and property. Existing forest fire monitoring technologies mainly include satellite remote sensing, video surveillance, and wireless sensor networks. However, these technologies have the following shortcomings: 1. The existing monitoring system adopts a continuous monitoring mode, in which all sensors and processing units are always in operation. This not only consumes a lot of energy, but also generates a large amount of redundant data, increasing the system burden and false alarm rate.

[0003] 2. Traditional systems use a simple judgment mechanism with fixed thresholds, which cannot adapt to the dynamic changes in the forest environment. They frequently give false alarms during seasonal changes and weather changes, and lack intelligent adaptive capabilities.

[0004] 3. Most monitoring devices use a single sensor or a single physical quantity for detection, making it difficult to distinguish between a real fire and environmental interference, and thus unable to accurately identify fire characteristics in the early stages of a fire.

[0005] 4. Existing technologies lack an effective graded response mechanism, and use the same processing procedures for all environmental changes. They cannot dynamically adjust monitoring strategies according to threat levels, resulting in low system resource utilization efficiency.

[0006] Therefore, there is an urgent need for an intelligent forest fire monitoring device and method that can automatically activate corresponding functions based on environmental changes and has graded response capabilities. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a self-starting forest fire monitoring device and method, which combines multi-physics gradient monitoring with chemical-electromagnetic composite detection to achieve intelligent, graded, and self-starting monitoring of forest fires.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A self-starting forest fire monitoring device includes a multi-physics gradient sensing module, a volatile organic compound concentration gradient detection module, an ion wind sensing module, a fire electromagnetic feature identification module, a multi-source abnormal signal fusion processing module, a collaborative verification communication module, and a hierarchical self-starting control module.

[0009] The multiphysics gradient sensing module, as the primary triggering unit of the device, uses four integrated sensing nodes arranged in a tetrahedral vertex pattern within the monitoring area. Each node integrates a temperature sensor, a pressure sensor, and a humidity sensor. It calculates the spatial gradient vectors of the temperature, pressure, and humidity fields using the central difference method, monitors the rate of change of the gradient over time, and generates a graded trigger signal when the rate of change of the gradient exceeds a dynamic threshold determined based on the statistical characteristics of historical data. The trigger signal output of the multiphysics gradient sensing module is connected to the trigger input of the volatile organic compound concentration gradient detection module, and the gradient anomaly signal output is connected to the first input of the multi-source anomaly signal fusion processing module.

[0010] The volatile organic compound (VOC) concentration gradient detection module is activated upon receiving a trigger signal. It uses a metal oxide semiconductor gas sensor array to detect terpenes, aldehydes, and ketones, which are plant pyrolysis products. A three-dimensional VOC concentration field is constructed using the Kriging interpolation algorithm. The divergence of the concentration gradient vector field is calculated. When the divergence is negative and its absolute value is greater than a set threshold, an ignition source is determined to exist and its direction is estimated. The VOC abnormal signal output terminal of the volatile organic compound concentration gradient detection module is connected to the second input terminal of the multi-source abnormal signal fusion processing module, and the abnormal confirmation output terminal is connected to the start input terminal of the ion wind sensing module.

[0011] The ion wind sensing module is activated after VOCs detection shows an abnormality. It uses a parallel plate capacitive ion detector to measure the concentration of positive and negative ions respectively, calculates the ion concentration difference, analyzes the time evolution characteristics of ion concentration through an exponential growth model, and extracts the time constant and the maximum concentration difference as fire characteristic parameters. The ion abnormality signal output terminal of the ion wind sensing module is connected to the third input terminal of the multi-source abnormal signal fusion processing module.

[0012] The fire electromagnetic feature identification module detects environmental electromagnetic field disturbances through a triaxially orthogonally configured electric field sensor and magnetic field sensor, performs a fast Fourier transform on the acquired signal to obtain a frequency domain representation, calculates the power spectral density, and analyzes the matching degree between the measured spectrum and the fire electromagnetic feature template through correlation coefficient analysis. The electromagnetic anomaly signal output terminal of the fire electromagnetic feature identification module is connected to the fourth input terminal of the multi-source anomaly signal fusion processing module, and the start input terminal is connected to the third start control terminal of the graded self-start control module.

[0013] The multi-source anomaly signal fusion processing module adopts a Bayesian inference framework, using gradient anomaly signals, VOCs anomaly signals, ion anomaly signals, and electromagnetic anomaly signals as observational evidence. It calculates the posterior probability of a fire through conditional probability decomposition and the total probability formula, and generates three-level judgment results: normal, suspected, and confirmed, based on the posterior probability values. The suspected fire output terminal of the multi-source anomaly signal fusion processing module is connected to the verification request input terminal of the collaborative verification communication module, the judgment result output terminal is connected to the judgment input terminal of the graded self-starting control module, and the verification input terminal is connected to the verification result output terminal of the collaborative verification communication module.

[0014] The collaborative verification communication module is activated when a suspected fire is detected locally. It broadcasts verification requests to neighboring nodes using LoRa modulation technology and forms a distributed consensus through a weighted voting mechanism. The voting weights are determined based on node distance and historical accuracy. At the same time, it uses multi-node gradient vector information to locate the fire source using the least squares method. The verification result output of the collaborative verification communication module is connected to the verification input of the multi-source abnormal signal fusion processing module, and the consensus result output is connected to the confirmation input of the hierarchical self-starting control module.

[0015] The hierarchical self-starting control module maintains five working states: sleep state, monitoring state, detection state, confirmation state, and alarm state. State transitions are controlled based on trigger conditions, and a dynamic priority scheduling algorithm is used for resource allocation. The priority increment is calculated based on the real-time requirements of the task. The state control output of the hierarchical self-starting control module is connected to the control input of the multi-physics gradient sensing module. The first start control terminal is connected to the start input of the volatile organic compound concentration gradient detection module, the second start control terminal is connected to the start input of the ion wind sensing module, and the third start control terminal is connected to the start input of the fire electromagnetic feature identification module.

[0016] Furthermore, the present invention also provides a self-starting forest fire monitoring method, comprising the following steps: S1. Environmental baseline learning: Collect temperature, air pressure, humidity and VOCs concentration data of the monitoring area at different time periods and under different weather conditions, use a Gaussian mixture model to describe the statistical distribution of environmental parameters, estimate model parameters through the expectation-maximization algorithm, and establish an environmental baseline model.

[0017] S2. Multiphysics gradient sensing and triggering: Continuously monitor environmental parameters, calculate the spatial gradient vector and time rate of change of each physical field, assess the degree of deviation of data points from the baseline model through Mahalanobis distance, and generate first-level, second-level or third-level trigger signals according to the magnitude of Mahalanobis distance.

[0018] S3. Volatile Organic Compound Concentration Gradient Detection: Upon receiving a trigger signal, the VOCs sensor array is activated. A three-dimensional concentration field is constructed through Kriging interpolation. The divergence of the concentration gradient vector field is calculated and analyzed to identify gradient convergence characteristics and estimate the location of the fire source.

[0019] S4. Ion wind sensing confirmation: Measure the concentration of positive and negative ions and calculate the concentration difference. Fit the ion concentration time evolution curve through an exponential growth model and extract the time constant to determine whether there is an ion field disturbance with fire characteristics.

[0020] S5. Fire Electromagnetic Feature Identification: Collect environmental electromagnetic signals and perform frequency domain transformation, calculate power spectral density, match with fire electromagnetic feature templates through correlation analysis, and output electromagnetic anomaly signals.

[0021] S6. Multi-source abnormal signal fusion processing: Time alignment of each detection signal, calculation of the posterior probability of fire using Bayes' theorem, and generation of graded judgment results based on the probability values.

[0022] S7. Collaborative Verification Communication: Broadcasts verification requests to neighboring nodes, receives independent judgment results from neighboring nodes, forms distributed consensus through weighted voting, and confirms the fire event.

[0023] S8. Hierarchical self-starting control optimization: Records the triggering time, response delay, detection results and resource consumption of the detection process, analyzes the effects of different triggering strategies, dynamically adjusts the triggering threshold, and manages the device state transition.

[0024] Furthermore, the method also includes precise fire source location, using multi-node gradient vector information to estimate the fire source location through the least squares method, calculating the distance from the point to the gradient line, and solving for the spatial location that minimizes the sum of squared distances.

[0025] Furthermore, the method also includes data compression transmission, employing differential coding combined with Huffman coding to calculate the difference between adjacent sampling points, and constructing a coding tree based on the probability distribution of the difference values ​​to achieve variable-length coding compression.

[0026] Beneficial effects:

[0027] 1. Intelligent hierarchical response: Through multi-level triggering mechanism and status management, the device automatically adjusts the monitoring intensity according to the severity of environmental anomalies to avoid continuous high-intensity operation.

[0028] 2. Early and accurate detection: Gradient triggering combined with VOCs detection and ion sensing can detect anomalies in the early stages of a fire, and multi-level verification improves detection accuracy.

[0029] 3. Adaptive Environment Learning: Based on Gaussian mixture model-based environmental baseline learning and Mahalanobis distance anomaly assessment, the device automatically adapts to different environmental characteristics.

[0030] 4. Reliable collaborative verification: The distributed consensus mechanism and weighted voting algorithm provide independent verification, reducing the impact of single-point misjudgment.

[0031] 5. Precise fire source location: Multi-node gradient vector intersection and least squares optimization enable accurate estimation of fire source location.

[0032] 6. Efficient data management: Differential Huffman coding compression and hierarchical storage strategies optimize data transmission and storage efficiency. Attached Figure Description

[0033] Figure 1 A schematic diagram of the structure of the device described in this invention is shown; Figure 2 A flowchart illustrating the steps of the method described in this invention is shown. Detailed Implementation

[0034] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0035] Combination Figure 1 The self-starting forest fire monitoring device provided by this invention achieves intelligent hierarchical self-starting monitoring of forest fires through the coordinated operation of a multi-physics field gradient sensing module, a volatile organic compound concentration gradient detection module, an ion wind sensing module, a fire electromagnetic feature identification module, a multi-source abnormal signal fusion processing module, a collaborative verification communication module, and a hierarchical self-starting control module.

[0036] Regarding the device connections, the trigger signal output of the multiphysics gradient sensing module is connected to the trigger input of the volatile organic compound (VOC) concentration gradient detection module, and its gradient anomaly signal output is connected to the first input of the multi-source anomaly signal fusion processing module. The VOCs anomaly signal output of the VOC concentration gradient detection module is connected to the second input of the multi-source anomaly signal fusion processing module, and its anomaly confirmation output is connected to the start input of the ion wind sensing module. The ion anomaly signal output of the ion wind sensing module is connected to the third input of the multi-source anomaly signal fusion processing module. The electromagnetic anomaly signal output of the fire electromagnetic feature identification module is connected to the fourth input of the multi-source anomaly signal fusion processing module.

[0037] The suspected fire output of the multi-source abnormal signal fusion processing module is connected to the verification request input of the collaborative verification communication module, and its judgment result output is connected to the judgment input of the graded self-starting control module. The verification result output of the collaborative verification communication module is connected to the verification input of the multi-source abnormal signal fusion processing module, and its consensus result output is connected to the confirmation input of the graded self-starting control module.

[0038] The status control output terminal of the graded self-starting control module is connected to the control input terminal of the multi-physics gradient sensing module. Its first start control terminal is connected to the start input terminal of the volatile organic compound concentration gradient detection module, its second start control terminal is connected to the start input terminal of the ion wind sensing module, and its third start control terminal is connected to the start input terminal of the fire electromagnetic feature identification module.

[0039] The modules transmit data via digital signal interfaces. Trigger and control signals use level-triggered methods, while detection data is transmitted using a serial communication protocol. Through these connections, the modules form a complete fire monitoring device with hierarchical triggering, step-by-step verification, and collaborative judgment.

[0040] In this embodiment, the multiphysics gradient sensing module serves as the primary triggering unit of the device, undertaking the task of initial environmental anomaly detection. This module arranges four integrated sensing nodes within the monitoring area in a tetrahedral configuration, with each node integrating a temperature sensor, a barometric pressure sensor, and a humidity sensor, forming a three-dimensional coverage of the monitoring space. The sensors continuously collect environmental data at a set sampling frequency; the temperature sensor uses a platinum resistance temperature detector, the barometric pressure sensor uses a MEMS piezoresistive sensor, and the humidity sensor uses a capacitive sensor.

[0041] To accurately calculate the spatial gradient of each physical field, this invention employs the central difference method for gradient calculation. For the temperature field T(x,y,z,t), the formula for calculating its gradient vector is: in: This is the gradient vector of the temperature field; Let be the partial derivative of temperature in the x-direction; Let be the partial derivative of temperature in the y-direction; Let be the partial derivative of temperature in the z-direction; It is the unit vector in the x-direction; It is the unit vector in the y-direction; is the unit vector in the z-direction; T is the temperature value; x is the spatial x-coordinate; y is the spatial y-coordinate; z is the spatial z-coordinate; t is time.

[0042] In practical calculations, the partial derivatives are obtained through the central difference approximation: in: The spatial sampling interval in the x-direction; For position The temperature value at time t; For position The temperature value at time t.

[0043] Similarly, the gradient vectors of the pressure field P(x,y,z,t) and the humidity field H(x,y,z,t) and The same method is used for calculation. To identify abnormal gradient changes caused by a fire, the device simultaneously monitors the magnitude of the gradient and its rate of change over time. The formula for calculating the rate of change of the gradient is: in: The rate of change of the temperature gradient over time; The magnitude of the temperature gradient at time t; For a moment Temperature gradient magnitude; The time sampling interval; This represents the derivative of the temperature gradient magnitude with respect to time.

[0044] When the rate of change of the gradient of any physical field exceeds the corresponding dynamic threshold, the multiphysics gradient sensing module generates a trigger signal. The dynamic threshold is determined using an adaptive algorithm based on the statistical characteristics of historical data. in: This is the dynamic threshold for the rate of change of the temperature gradient; This represents the average rate of change of the historical temperature gradient. denoted as , where is the standard deviation of the historical temperature gradient change rate; k is the sensitivity adjustment coefficient.

[0045] The generation of trigger signals follows a hierarchical strategy: a level-one trigger signal is generated when a single physical field malfunctions; a level-two trigger signal is generated when two physical fields malfunction simultaneously; and a level-three trigger signal is generated when all three physical fields malfunction. This hierarchical triggering mechanism enables the device to respond accordingly based on the severity of the environmental anomaly.

[0046] The volatile organic compound (VOC) concentration gradient detection module activates upon receiving a trigger signal, entering VOC detection mode. This module employs a metal oxide semiconductor gas sensor array to detect characteristic compounds produced by plants during high-temperature pyrolysis. The sensor array includes three sensors made of different sensitive materials, each selectively responding to terpenes, aldehydes, and ketones, respectively.

[0047] To construct a three-dimensional concentration field of VOCs, the device simultaneously collects concentration data at multiple locations within the monitoring area. Let sensor i be at location... The VOCs concentration measured at the location was A continuous concentration field is then constructed using the Kriging interpolation algorithm. : in: For any point in space Estimated VOCs concentrations; The weight coefficient for the i-th sampling point; Let be the measured concentration value of the i-th sampling point; n is the total number of sampling points; The summation symbol is used.

[0048] Weighting coefficient The determination of requires satisfying the conditions of unbiased estimation and minimum variance, which can be obtained by solving the Kriging equations: in: The semivariogram value between sampling point i and sampling point j; The semivariogram value between sampling point i and the point to be estimated; It is a Lagrange multiplier; Let be the weighting coefficients to be determined; This forms a semi-variogram matrix.

[0049] The semivariogram describes the change of spatial correlation with distance, and uses a spherical model: in: is the semivariogram value at a distance h; h is the Euclidean distance between the two points; denoted as nugget value, representing measurement error and micro-variation; c is the sill value, representing the magnitude of spatial variation; and a is the range, representing the scope of spatial correlation.

[0050] After constructing the concentration field, the device calculates the VOCs concentration gradient vector. The spatial distribution characteristics of VOCs were analyzed. VOCs release at the fire source formed the source point of the concentration field, and the gradient vector exhibited a centripetal convergence characteristic. This convergence pattern was identified by calculating the divergence of the gradient vector field. in: Let be the divergence of the concentration gradient vector field; Let C be the second partial derivative of the concentration field C in the x-direction; Let C be the second partial derivative of the concentration field C in the y-direction; Let be the second partial derivative of the concentration field C in the z-direction; div denotes the divergence operator.

[0051] When the divergence value is negative and its absolute value is large, it indicates that there is a convergence of gradient vectors in the region, which may indicate the presence of a fire source. The device further estimates the location of the fire source by tracing the convergence point of the gradient lines.

[0052] The ionization wind sensing module automatically activates after an abnormality is detected in the VOCs detection, providing an independent fire verification mechanism. The high temperatures of a fire ionize air molecules, generating a large number of positive and negative ions, forming a characteristic ion field distribution. This module uses a parallel-plate capacitive ion detector, which collects positive and negative ions separately by applying a bias voltage.

[0053] The relationship between ion current and ion concentration is as follows: Where: I is the ion current; q is the charge of a single ion; n is the ion concentration; v is the ion migration velocity; and A is the effective area of ​​the collecting electrode.

[0054] Ion migration rate is related to electric field strength and ion mobility: Where: v is the ion migration rate; denoted as ion mobility; E is the electric field strength.

[0055] In a normal environment, the concentrations of positive and negative ions are basically balanced, with a small concentration difference. The high temperatures generated by a fire cause asymmetric ionization of air molecules, resulting in a significantly higher concentration of positive ions than negative ions. The device identifies this anomaly by calculating the ion concentration difference. in: It represents the ion concentration difference; This represents the concentration of positive ions. This refers to the concentration of negative ions. It is a positive ion current; It is a negative ion current; The positive ion mobility; denoted as negative ion mobility; q as the charge of a single ion; E as the electric field strength; and A as the effective area of ​​the collecting electrode.

[0056] In addition to the static concentration difference, the device also analyzes the dynamic changes in ion concentration. The changes in ion concentration caused by a fire exhibit a specific time evolution pattern, which can be described by an exponential growth model: in: Let be the ion concentration difference at time t; This represents the initial ion concentration difference. This represents the maximum ion concentration difference. is the time constant, reflecting the rate of increase in ion concentration; e is the base of the natural logarithm; t is time.

[0057] By performing curve fitting on the measured data, the time constant was extracted. and maximum concentration difference As a characteristic parameter, the ion field generated by a fire has the characteristic of rising rapidly, while the ion concentration changes caused by other interference sources are relatively slow.

[0058] The fire electromagnetic feature recognition module works synchronously with the ionization wind sensing module, providing auxiliary evidence for fire identification from an electromagnetic field perspective. During combustion, plasma generates characteristic electromagnetic radiation, and the movement of charged particles in the flame creates a changing electromagnetic field. This module detects disturbances in the ambient electromagnetic field using electric and magnetic field sensors, respectively.

[0059] The electric field sensor employs a triaxial orthogonal configuration to measure the spatial electric field vector. in: It is the electric field vector; Let x be the electric field component in the x-direction; The electric field component is in the y-direction; Let z be the electric field component in the z-direction; It is the unit vector in the x-direction; It is the unit vector in the y-direction; It is the unit vector in the z-direction.

[0060] Perform a Fast Fourier Transform on the acquired electric field signal to obtain its frequency domain representation: in: This represents the electric field in the frequency domain. This is the time-domain signal of the electric field; ω is the angular frequency; j is the imaginary unit; t is time; dt is the time infinitesimal element; Indicates the integral symbol; Represents negative infinity; It represents positive infinity.

[0061] The electromagnetic signals generated by a fire have specific spectral characteristics, and the device calculates the power spectral density: in: Let be the power spectral density of the electric field; for The modulus; This represents the square of the modulus.

[0062] By analyzing the distribution characteristics of the power spectral density, the device extracts the electromagnetic fingerprint of a fire. The electromagnetic spectrum of a fire exhibits energy peaks in specific frequency bands, and the peak frequencies are correlated with the fire intensity. The device establishes a fire electromagnetic feature template library, and pattern matching is performed by calculating the correlation coefficient between the measured spectrum and the templates. in: The correlation coefficient; This represents the value of the i-th frequency point in the measured spectrum; This represents the value of the i-th frequency point in the template spectrum. This is the mean of the measured spectrum; is the mean of the template spectrum; m is the number of spectrum sampling points; This represents the summation from i=1 to i=m; It represents the square root.

[0063] When the correlation coefficient exceeds a set threshold, the presence of electromagnetic fire characteristics is determined. The output of this module, along with the results from other detection modules, is input into the multi-source abnormal signal fusion processing module.

[0064] The multi-source anomaly signal fusion processing module is the decision-making center of the entire device, responsible for integrating the outputs of various detection modules to generate the final fire determination result. This module adopts a Bayesian inference framework, using each detection result as observational evidence to update the posterior probability of the fire event.

[0065] Let the fire event be F, and the outputs of each detection module be: gradient anomaly signal. VOCs abnormal signals Ion abnormal signals Electromagnetic abnormal signals According to Bayes' theorem, the posterior probability of a fire is: in: Given all observed evidence, this represents the posterior probability of a fire. This represents the likelihood probability of observing all evidence under fire conditions. Let be the prior probability of a fire. Marginal probability of observed evidence; This represents the "given" in conditional probability.

[0066] Considering that each detection module is relatively independent, the likelihood probability can be decomposed as follows: in: This represents the probability of a gradient anomaly occurring under fire conditions. The probability of abnormal VOCs occurring under fire conditions; The probability of ion anomalies occurring under fire conditions; This represents the probability of electromagnetic anomalies occurring under fire conditions.

[0067] These conditional probabilities were obtained through historical data statistics and experimental testing. Marginal probabilities were calculated using the law of total probability. in: This indicates that there was no fire incident; The prior probability of no fire is calculated using the following formula: ; This represents the probability of observing all evidence under conditions of no fire.

[0068] The device generates hierarchical decisions based on posterior probability values: when When, it is judged as a normal state; when At that time, it was determined to be a suspected fire; when At that time, it was determined to be a confirmed fire.

[0069] To further improve the reliability of the judgment, when a local detection indicates a suspected fire, the collaborative verification communication module automatically activates and requests collaborative verification from adjacent monitoring nodes. This module uses LoRa modulation technology to achieve long-distance, low-power communication, and the communication protocol uses a custom lightweight format that includes node identifiers, timestamps, detection data summaries, and judgment results.

[0070] Collaborative verification employs a distributed consensus algorithm, where each participating node independently determines the fire status and broadcasts the result to other nodes. Let the set of nodes participating in collaborative verification be denoted as . ,node The judgment result is The value can be 0 (normal) or 1 (fire). The device reaches a final consensus through a weighted voting mechanism: in: This is the final consensus result; For nodes Voting weight; For nodes The judgment result; The threshold for voting is k; k is the total number of participating nodes. This represents the summation from i=1 to i=k.

[0071] The voting weight of a node is determined based on its distance from the suspected fire source and its historical accuracy in identifying fire sources. in: For nodes Voting weight; For nodes Distance from suspected fire source; The distance attenuation constant; For nodes Historical accuracy of judgments; This refers to the distance weighting coefficient. This is the accuracy weighting coefficient; ; e is the base of the natural logarithm.

[0072] Through collaborative verification, the device can effectively reduce the impact of single-point misjudgments and improve the overall reliability of fire detection. Simultaneously, the gradient information from multiple nodes can also be used for precise fire source location.

[0073] The fire source location employs the gradient vector intersection method. The temperature gradient vectors measured at each node point towards the fire source, and theoretically, all gradient lines should intersect at the fire source location. However, due to measurement errors and environmental interference, the gradient lines do not perfectly intersect at a single point. The device uses the least squares method to estimate the optimal fire source location. in: The estimated fire source location vector; It is a spatial position vector; For position The distance to the i-th gradient line; n is the number of nodes involved in the localization. This represents the parameter that minimizes the objective function.

[0074] The parametric equation of the gradient line is: in: Let be a point on the i-th gradient line; Let i be the position vector of node i; Let be the gradient vector measured at node i; t is a parameter.

[0075] point The distance to the gradient line is: in: For point Distance to the i-th gradient line; Represents the vector cross product; Represents the magnitude of a vector; From node i to point The vector.

[0076] The optimal estimate of the fire source location is obtained by solving a least-squares problem. The device also calculates the positioning uncertainty as an evaluation index of positioning accuracy.

[0077] The hierarchical self-starting control module is responsible for coordinating the operation of the entire device, enabling intelligent resource management and function scheduling. This module maintains the device's state machine, defining five operating states: hibernation, monitoring, detection, confirmation, and alarm. State transitions are driven by triggering conditions and follow predefined transition rules.

[0078] In sleep mode, only the multiphysics gradient sensing module operates at minimum power consumption, while all other modules are shut down. Upon detecting a Level 1 trigger signal, the device enters monitoring mode, increasing the sampling frequency of the gradient sensing module. Upon detecting a Level 2 or 3 trigger signal, the device enters detection mode, activating the VOCs detection module. If VOCs detection confirms an anomaly, the device enters confirmation mode, simultaneously activating the ion sensing module and the electromagnetic feature recognition module. Once a fire is confirmed, the device enters alarm mode, activating all functional modules and continuously tracking the fire's development.

[0079] The device's resource scheduling strategy is dynamically adjusted based on the current state and task priority. In detection mode, the VOCs detection module receives the highest CPU time slice allocation, ensuring rapid completion of concentration field construction and gradient analysis. In confirmation mode, the multi-source signal fusion processing module's priority is increased to guarantee the real-time performance of Bayesian inference. Resource allocation employs a dynamic priority scheduling algorithm. in: Let i be the dynamic priority of module i at time t; The basic priority of module i; This is the priority increment.

[0080] The calculation of priority increments takes into account the real-time requirements of the task: in: This is the priority increment of module i; t represents the arrival time of task i; t represents the current time. Let be the deadline for task i; Let be the urgency coefficient of task i; For time weighting parameters; This is the urgency weighting parameter.

[0081] Combination Figure 2 The self-starting forest fire monitoring method of the present invention includes the following steps: S1. Environmental Baseline Learning: After deployment, the device first performs environmental baseline learning, collecting environmental data of the monitoring area at different time periods, including temperature, air pressure, humidity, and VOCs concentration. The data collection process continues for at least one full seasonal cycle to ensure coverage of various weather conditions. The device uses a Gaussian mixture model to describe the statistical distribution of environmental parameters. in: Environmental parameter vector The probability density; J is the number of Gaussian components; The mixing weights for the j-th component satisfy the following condition: ; The mean is The covariance matrix is The multidimensional Gaussian distribution.

[0082] The environmental baseline model is established by estimating model parameters using the expectation-maximization algorithm and is used for subsequent anomaly detection, providing a basis for the dynamic setting of gradient triggering threshold.

[0083] S2. Multiphysics Gradient Sensing and Triggering: The multiphysics gradient sensing module continuously monitors environmental parameters, acquiring temperature, air pressure, and humidity data according to a set sampling frequency. For each physics field, its spatial gradient vector and rate of change over time are calculated. When a gradient anomaly is detected, the degree of anomaly is assessed by calculating the Mahalanobis distance. in: For data points The Mahalanobis distance to the nearest Gaussian component; Let be the mean vector of the j-th Gaussian component; Let be the inverse matrix of the covariance matrix of the j-th Gaussian component; Indicates matrix transpose; This indicates taking the minimum value; It represents the square root.

[0084] Different levels of trigger signals are generated based on the magnitude of the Mahalanobis distance: Level 1 trigger is generated when the Mahalanobis distance of a single parameter exceeds the threshold; Level 2 trigger is generated when two parameters exceed the threshold simultaneously; and Level 3 trigger is generated when all three parameters exceed the threshold.

[0085] S3. Volatile Organic Compound (VOC) Concentration Gradient Detection: Upon receiving a trigger signal, the VOC concentration gradient detection module immediately activates. This step first activates the VOC sensor array to rapidly detect terpenes, aldehydes, and ketones. After acquiring multi-point concentration data, a three-dimensional concentration field is constructed using the Kriging interpolation algorithm, and the concentration gradient vector field is calculated and its divergence analyzed. When the divergence is negative and its absolute value is greater than a set threshold, gradient convergence is identified, potentially indicating the presence of an ignition source. By tracing the convergence direction of the gradient lines, the approximate location of the ignition source is estimated.

[0086] S4. Ion Wind Sensing Confirmation: After abnormal VOCs concentration is detected in step S3, the ion wind sensing module automatically starts ion field detection. This step measures the ion concentration using positive and negative ion detectors respectively, and calculates the concentration difference. The time evolution characteristics of the ion concentration difference are analyzed, and the time constant is extracted by fitting an exponential growth model. When the time constant is less than the set value and the concentration difference continues to increase, an ion field disturbance with fire-like characteristics is confirmed.

[0087] S5. Fire Electromagnetic Feature Identification: Simultaneously with steps S3 and S4, the fire electromagnetic feature identification module performs electromagnetic field detection. This step collects environmental electromagnetic signals using electric and magnetic field sensors, performs a Fast Fourier Transform (FFT) on the signals to obtain a frequency domain representation, calculates the power spectral density, and performs correlation analysis with the fire electromagnetic feature template. When the correlation coefficient exceeds a set threshold, a fire electromagnetic feature is detected, and an electromagnetic anomaly signal is output as an auxiliary basis for judgment.

[0088] S6. Multi-source Anomaly Signal Fusion Processing: The multi-source anomaly signal fusion processing module receives all detection signals generated in steps S2 to S5 and performs Bayesian inference for comprehensive judgment. This step first aligns the signals in time to ensure their synchronization. Then, it calculates the posterior probability of the fire based on Bayes' theorem and generates a graded judgment result based on the posterior probability value. The output of this step determines whether subsequent collaborative verification needs to be initiated.

[0089] S7. Collaborative Verification Communication: When step S6 determines a suspected fire, the collaborative verification communication module automatically executes multi-node collaborative verification. This step first broadcasts a verification request to surrounding nodes, including a summary of local detection data and a preliminary judgment result. Neighboring nodes receiving the request independently perform fire detection and feed back the results to the requesting node. Consensus is reached through a weighted voting mechanism; when the consensus threshold is reached, the fire event is confirmed and the process proceeds to the next step.

[0090] S8. Hierarchical Self-Starting Control Optimization: The hierarchical self-starting control module records the execution status of each step throughout the detection process, including trigger time, response delay, detection results, and resource consumption. This step analyzes the effectiveness of different triggering strategies and identifies the optimal startup sequence. It adjusts the trigger thresholds at each level based on environmental characteristics to achieve dynamic optimization. Simultaneously, this step is also responsible for device state management, determining state transitions based on the current detection results.

[0091] After a fire is confirmed, the device performs precise fire source location and continuous tracking. Utilizing gradient vector information from multiple nodes, the fire source location is estimated using the least squares method. Once located, the device maintains high-frequency monitoring, continuously tracking the fire's development and updating fire boundary information to provide real-time support for firefighting decisions.

[0092] The method of this invention also includes intelligent resource scheduling, a process that runs throughout the entire monitoring process, dynamically allocating and assessing the resource requirements of each module in real time. CPU time slices and memory resources are allocated according to priority ratios to ensure that critical detection functions receive sufficient computing resources. In high-threat situations, the device automatically increases the resource quotas for fire detection-related modules; in low-threat situations, it reduces the resource consumption of unnecessary functions, achieving efficient resource utilization.

[0093] The hierarchical data management system is responsible for classifying and processing the data generated during the detection process. Data is categorized into three levels based on importance: critical data, important data, and routine data. Critical data includes fire alarm information, fire source location, and fire spread rate; this type of data has the highest priority and is reliably delivered using a redundant transmission mechanism. Important data includes the detection results of each module and abnormal signal characteristics, and is transmitted after compression. Routine data includes environmental background parameters and operational logs, and is uploaded in batches during periods of network downtime.

[0094] Data compression employs a combination of differential coding and Huffman coding. For time series data, the difference between adjacent sampling points is first calculated: in: This is the i-th difference value; This is the i-th sampled value; This is the (i-1)th sampled value.

[0095] Differential values ​​typically cluster near zero, exhibiting better compression characteristics. The device statistically analyzes the probability distribution of these differential values, constructs a Huffman coding tree, and assigns short codewords to high-frequency differential values ​​and long codewords to low-frequency differential values, achieving variable-length coding. The compression ratio is defined as: Where: CR is the compression ratio; The length of the original data; is the length of the compressed data; n is the number of sampling points; b is the number of bits per sample value; m is the number of different difference values; The frequency of the i-th difference value; is the encoding length of the i-th difference value.

[0096] Through the coordinated efforts of the above steps, this invention achieves intelligent, graded, and self-starting monitoring of forest fires. The collaboration between the gradient sensing module and the VOCs detection module enables the device to quickly transition from physical field anomalies to chemical detection, shortening the fire detection time. The combination of VOCs detection and ion sensing provides a dual verification mechanism, reducing the false alarm rate. Electromagnetic feature recognition serves as an auxiliary means, providing additional judgment criteria when the results from other detection modules are ambiguous. The multi-source signal fusion processing module integrates all detection information and derives a scientific judgment result through Bayesian inference. The collaborative verification communication module introduces spatial dimension verification, further improving the device's reliability through independent observation of multiple nodes.

[0097] This invention fully considers the complexity of the forest environment and the dynamic characteristics of fire development. Through a comprehensive monitoring strategy involving multiple parameters, levels, and nodes, it constructs an intelligent, reliable, and efficient forest fire monitoring device. The device's hierarchical self-starting mechanism solves the problem of resource waste caused by the continuous operation of traditional monitoring equipment. Multi-source signal fusion technology improves the accuracy of fire detection, and collaborative verification methods enhance the device's reliability, providing strong technical support for forest fire prevention.

[0098] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. A self-starting forest fire monitoring device, characterized in that, include: The multi-physics gradient sensing module is used to collect temperature, air pressure and humidity data through spatially distributed sensing nodes, calculate the spatial gradient vector and time change rate of each physical field, and generate a graded trigger signal when the gradient change rate exceeds the dynamic threshold determined based on the statistical characteristics of historical data. The trigger signal output is connected to the trigger input of the volatile organic compound concentration gradient detection module, and the gradient anomaly signal output is connected to the first input of the multi-source anomaly signal fusion processing module. The volatile organic compound concentration gradient detection module is activated upon receiving the trigger signal to detect the concentration distribution of plant pyrolysis products, construct a three-dimensional concentration field through a spatial interpolation algorithm, calculate the divergence of the concentration gradient vector field, and determine the presence of a fire source based on the negative value characteristic of the divergence. Its VOCs abnormal signal output terminal is connected to the second input terminal of the multi-source abnormal signal fusion processing module, and the abnormal confirmation output terminal is connected to the start input terminal of the ion wind sensing module. The ion wind sensing module is activated after an anomaly is detected in volatile organic compounds. It measures the concentration difference between positive and negative ions and analyzes the ion field disturbance characteristics through a time evolution model. Its ion anomaly signal output terminal is connected to the third input terminal of the multi-source anomaly signal fusion processing module. The electromagnetic feature identification module for fire has its electromagnetic anomaly signal output terminal connected to the fourth input terminal of the multi-source anomaly signal fusion processing module. The multi-source abnormal signal fusion processing module is used to use the output of each detection module as observation evidence in a Bayesian inference framework to calculate the posterior probability of fire and generate a graded judgment result. Its suspected fire output end is connected to the verification request input end of the collaborative verification communication module, and the judgment result output end is connected to the judgment input end of the graded self-starting control module. The collaborative verification communication module is used to request verification from neighboring nodes when a fire is suspected. It forms a distributed consensus through a weighted voting mechanism. Its verification result output is connected to the verification input of the multi-source abnormal signal fusion processing module, and the consensus result output is connected to the confirmation input of the hierarchical self-starting control module. The graded self-starting control module is used to manage the working state transition of the device according to the trigger signal level and detection results, and control the start-up sequence of each functional module. Its status control output terminal is connected to the control input terminal of the multi-physics gradient sensing module, and the first, second and third start-up control terminals are respectively connected to the start-up input terminals of the volatile organic compound concentration gradient detection module, the ion wind sensing module and the fire electromagnetic feature identification module.

2. The self-starting forest fire monitoring device according to claim 1, characterized in that, The multi-physics gradient sensing module calculates the gradient vector using the central difference method. The dynamic threshold is determined by multiplying the mean and standard deviation of the historical gradient change rate by the sensitivity adjustment coefficient. Based on single, two, or three physical field anomalies, it generates first-level, second-level, or third-level trigger signals, respectively.

3. The self-starting forest fire monitoring device according to claim 1, characterized in that, The volatile organic compound concentration gradient detection module uses the Kriging interpolation algorithm to construct the concentration field. It determines the interpolation weight coefficients by solving a system of linear equations containing a semi-variogram matrix. When the concentration gradient vector field shows centripetal convergence and the divergence is negative, it is determined that there is a fire source.

4. The self-starting forest fire monitoring device according to claim 1, characterized in that, The multi-source abnormal signal fusion processing module decomposes the likelihood probability of each piece of evidence appearing under fire conditions into the product of the conditional probabilities of each independent detection module, calculates the marginal probability using the total probability formula, and classifies the evidence into three judgment levels: normal, suspected, and confirmed based on the posterior probability value.

5. The self-starting forest fire monitoring device according to claim 1, characterized in that, The voting weight of the collaborative verification communication module is determined by the weighted sum of the exponential decay function of the node distance and the historical accuracy. The fire source location is estimated by the least squares method using multi-node gradient vector information.

6. A self-starting forest fire monitoring method, characterized in that, Includes the following steps: S1. Environmental baseline learning: Collect environmental data and establish the statistical distribution of environmental parameters based on Gaussian mixture model; S2. Multi-physics gradient sensing and triggering: Calculate the time change rate of the physical field gradient. When the change rate exceeds the dynamic threshold, a graded trigger signal is generated. The trigger signal is transmitted to the volatile organic compound concentration gradient detection step, and the gradient anomaly signal is transmitted to the multi-source anomaly signal fusion processing step. S3, Volatile Organic Compound Concentration Gradient Detection: After receiving the trigger signal, a three-dimensional concentration field is constructed, and the fire source is identified through divergence analysis. The abnormal VOCs signal is transmitted to the multi-source abnormal signal fusion processing step, and the abnormal confirmation signal triggers the ion wind sensing confirmation step. S4. Ion wind sensing confirmation: Measure the ion concentration difference and analyze its time evolution characteristics. Transmit the ion anomaly signal to the multi-source anomaly signal fusion processing step. S5. Fire Electromagnetic Feature Identification: Detect electromagnetic field disturbances and extract features; transmit electromagnetic abnormal signals to the multi-source abnormal signal fusion processing step. S6. Multi-source abnormal signal fusion processing: Bayesian inference is used to calculate the posterior probability of fire, generate a graded judgment, and the suspected fire result triggers the collaborative verification communication step. S7. Collaborative verification communication: Fire events are confirmed through distributed weighted voting, and the verification results are fed back to the multi-source abnormal signal fusion processing step. S8. Hierarchical self-starting control: Based on the detection results, manage the device state transitions and control the execution sequence of each step.

7. The self-starting forest fire monitoring method according to claim 6, characterized in that, In S2, the degree of anomaly is assessed by calculating the Mahalanobis distance from the data point to each component of the Gaussian mixture model. The Mahalanobis distance is the square root of the product of the transpose matrix, the inverse covariance matrix, and the difference matrix of the difference between the data point and the mean of the Gaussian component.

8. The self-starting forest fire monitoring method according to claim 6, characterized in that, In S3, a spherical semivariogram model is used to describe spatial correlation. When the distance is less than the range, the semivariogram value is the nugget value plus the product of the sill value and the distance correlation term; when the distance is greater than the range, the semivariogram value is the sum of the nugget value and the sill value.

9. The self-starting forest fire monitoring method according to claim 6, characterized in that, It also includes a fire source location step: estimating the fire source location by minimizing the sum of squared distances from spatial points to each node gradient line, where the distance from a point to a gradient line is calculated by dividing the magnitude of the cross product of the position vector difference and the gradient vector by the magnitude of the gradient vector.

10. The self-starting forest fire monitoring method according to claim 6, characterized in that, The time evolution of the ion concentration difference in S4 is described by an exponential growth model. The model includes three parameters: initial concentration difference, maximum concentration difference, and time constant. The time constant is extracted by fitting and used as the basis for fire characteristic determination.

Images