Unmanned aerial vehicle fire extinguishing targeted interception method and system based on mobile fire source detection

Abstract

The invention provides an unmanned aerial vehicle fire extinguishing targeted interception method and system based on mobile fire source detection, and the technical scheme of the invention integrates a spectrum sensor and a data fusion algorithm, and can accurately recognize mobile kindling, such as kindling carried by animals; according to the invention, a three-dimensional dynamic positioning technology of GPS and visual SLAM fusion is realized, and fire source positioning and trajectory prediction are accurately carried out in combination with a motion model; and meanwhile, the risk levels are quantified by integrating multi-dimensional indexes such as kindling basic attributes, different targeted interception strategies are implemented for different risk levels, such as measures such as sound wave expelling, fire extinguishing bomb striking and dry ice cooling, and the timeliness and effectiveness of fire source processing are effectively improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of fire extinguishing technology, and in particular to a UAV fire extinguishing targeting interception method and system based on mobile fire source detection. BACKGROUND

[0002] In nature, the phenomenon of mobile fire sources, such as animals carrying fire sources to cause fire spread, has been confirmed by many studies. For example, in Australia, birds of prey such as black kites and whistling kites will actively pick up burning branches and throw them into unburned areas. In North American forest fires, rodents collecting hot branches for nesting can cause rekindling, and the body hair of animals such as African elephants can attach sparks to ignite new areas. With global warming, forest and grassland fires occur frequently, and the problem of fire spread caused by animals carrying fire sources is becoming increasingly prominent, posing a serious threat to the ecological environment and human life and property safety.

[0003] However, existing firefighting technology has obvious defects in dealing with such scenarios and cannot effectively solve the problem of fire prevention and disposal caused by animals carrying fire sources. Specifically, existing firefighting technology mainly focuses on fire monitoring and fire extinguishing bomb launching, and most UAVs rely on thermal imaging technology to detect open fires, without realizing early confirmation of fire sources. Risk assessment is mostly based on fire size and does not include mobile fire source characteristics such as animal attributes and behavior characteristics, making it difficult to achieve pre-intervention. Traditional two-dimensional positioning is used for positioning technology, which has large trajectory prediction errors in complex terrain and cannot meet the needs of dynamic interception. SUMMARY

[0004] The present disclosure at least provides a UAV fire extinguishing targeting interception method and system based on mobile fire source detection to overcome at least one of the above technical defects.

[0005] According to an aspect of the present disclosure, a UAV fire extinguishing targeting interception method based on mobile fire source detection is provided, comprising:

[0006] A multi-spectral detector is used to perform multi-spectral detection on a target detection area to obtain a multi-spectral signal. The multi-spectral signal includes ultraviolet signals, infrared signals, terahertz signals, and quantum magnetic signals. The multi-spectral signal is subjected to time delay integration processing to enhance the signal-to-noise ratio of the multi-spectral signal. The multi-spectral signal with enhanced signal-to-noise ratio is subjected to adaptive non-uniformity correction. The multi-spectral signal after adaptive non-uniformity correction is subjected to polarization encoding anti-interference processing to obtain a mobile fire source signal.

[0007] Based on the moving fire signal, a GPS signal corresponding to the moving fire is acquired; based on an environment map constructed by using a visual SLAM system, the GPS signal is solved and coordinate conversion is performed, to obtain a three-dimensional coordinate of the moving fire; a type of the moving fire is determined, and a motion model of the moving fire is constructed; based on the motion model and the three-dimensional coordinate, trajectory prediction is performed, to obtain a predicted motion trajectory of the moving fire; error correction is performed on the predicted motion trajectory, to obtain a target coordinate of the moving fire;

[0008] Based on the type of the moving fire, corresponding fire basic attributes, behavior characteristics, and historical risk records are acquired; and based on the acquired fire basic attributes, behavior characteristics, and historical risk records, a risk assessment model is used to perform risk assessment, to obtain a risk level;

[0009] For the moving fire with a high risk level, a drone is actively intercepted based on the target coordinate; for the moving fire with a medium risk level, a drone is tracked and monitored based on the target coordinate.

[0010] In a possible implementation, the time delay integration processing on the multi-spectrum signals comprises:

[0011] For each signal in the multi-spectrum signals, the signals of a same target region are subjected to multiple integration and superposition processes in the time dimension;

[0012] For the ultraviolet signal, a first preset number of integration and superposition processes are performed on a target ultraviolet wave band at a terminal stage of the integration and superposition processes, to enhance the flame free radical;

[0013] For the infrared signal, a second preset number of integration and superposition processes are performed on a target infrared wave band, to strengthen the characteristic of the CO2 characteristic absorption peak;

[0014] In the process of the integration and superposition processes, a motion compensation mechanism is used to adjust the phase or scanning direction of the charge transfer of the time delay integration processing, so that the integration path conforms to the actual motion.

[0015] In a possible implementation, the multi-spectrum detector comprises an ultraviolet sensor, a short-wave infrared sensor, a terahertz sensor, and a quantum magnetometer.

[0016] The adaptive non-uniformity correction on the multi-spectrum signals with the enhanced signal-to-noise ratio comprises:

[0017] For each multi-spectrum detector, a corresponding response characteristic dynamic model is established.

[0018] In a case where the temperature of the target detection region is greater than a first preset temperature, the signal corresponding to the multispectral detector in the multispectral signal enhanced in signal-to-noise ratio is corrected by using a response characteristic dynamic model and adopting scene-based non-uniformity correction.

[0019] In a case where the temperature of the target detection region is less than a second preset temperature, the signal corresponding to the multispectral detector in the multispectral signal enhanced in signal-to-noise ratio is corrected by using a two-step correction method; the two-step correction method includes two-point correction and multi-point correction; the first preset temperature is higher than the second preset temperature.

[0020] The corrected multispectral signal is processed by using an inter-frame correlation constraint to correct inter-frame smear.

[0021] In a possible implementation, the multispectral signal after adaptive non-uniformity correction is subjected to polarization encoding anti-interference processing to obtain a moving fire signal, including:

[0022] An interference feature library is obtained.

[0023] The multispectral signal after adaptive non-uniformity correction is subjected to polarization encoding to obtain a polar code.

[0024] The moving fire signal is filtered from the multispectral signal after adaptive non-uniformity correction based on a likelihood ratio of the polar code and a feature in the interference feature library.

[0025] In a possible implementation, the moving fire type is determined, and a motion model of the moving fire is constructed, including:

[0026] Based on the moving fire signal, behavior data corresponding to the moving fire is obtained.

[0027] Based on the behavior data, the fire type of the moving fire is determined, and a motion model of the moving fire is constructed.

[0028] The fire type includes at least one of the following: flying animals; tree-dwelling animals; and ground-dwelling animals.

[0029] In a possible implementation, the motion trajectory of the moving fire is predicted based on the motion model and the three-dimensional coordinates.

[0030] A fire physical propagation model, a wind field influence model, and a turbulent diffusion model are constructed.

[0031] Based on the three-dimensional coordinates, the fire physical propagation model, and the turbulent diffusion model, a fire state vector at a current time is determined; the fire state vector includes a fire position, a fire propagation speed, and a fire temperature.

[0032] determine a wind field parameter based on the wind field influence model;

[0033] determine a moving fire containment parameter based on the motion model;

[0034] perform trajectory prediction based on the fire state vector, the wind field parameter and the moving fire containment parameter to obtain a predicted motion trajectory of the moving fire.

[0035] In a possible implementation, the error correction on the predicted motion trajectory to obtain the target coordinates of the moving fire includes:

[0036] filter processing the predicted motion trajectory using a federated Kalman filter framework;

[0037] performing, in sequence, continuity constraint processing, loop detection correction cumulative error processing, terrain constraint processing, and obstacle avoidance collision detection on the data obtained through the filter processing to obtain the target coordinates of the moving fire; wherein the terrain constraint processing is used for height verification.

[0038] In a possible implementation, for the moving fire with a high risk level, the active interception by the unmanned aerial vehicle based on the target coordinates includes:

[0039] For the moving fire with a high risk level, performing unmanned aerial vehicle sound wave expelling processing, unmanned aerial vehicle fire extinguishing bomb striking processing, and unmanned aerial vehicle cooling processing based on the target coordinates.

[0040] In a possible implementation, the unmanned aerial vehicle fire extinguishing targeted interception method based on moving fire source detection further includes:

[0041] For the moving fire with a low risk level, generating a dry ice fog model and constructing an isolation belt based on the target coordinates.

[0042] Based on the isolation belt, a low-temperature barrier for the moving fire is generated using the dry ice fog model.

[0043] According to another aspect of the present disclosure, an unmanned aerial vehicle fire extinguishing targeted interception system based on moving fire source detection is provided, comprising:

[0044] The multispectral detection and fire source confirmation module is configured to perform multispectral detection on a target detection area by using a multispectral detector to obtain a multispectral signal, wherein the multispectral signal includes ultraviolet signals, infrared signals, terahertz signals, and quantum magnetic force signals; perform time delay integration processing on the multispectral signal to enhance the signal-to-noise ratio of the multispectral signal; perform adaptive non-uniformity correction on the multispectral signal after the signal-to-noise ratio is enhanced; and perform polarization encoding anti-interference processing on the multispectral signal after the adaptive non-uniformity correction to obtain a mobile fire source signal.

[0045] The three-dimensional fire source coordinate acquisition module is configured to obtain a GPS signal corresponding to the mobile fire source based on the mobile fire source signal; and perform calculation and coordinate conversion on the GPS signal based on an environment map constructed by using a visual SLAM system to obtain three-dimensional coordinates of the mobile fire source.

[0046] The diffusion point optimization and risk assessment module is configured to determine the type of the mobile fire source and construct a motion model of the mobile fire source; perform trajectory prediction based on the motion model and the three-dimensional coordinates to obtain a predicted motion trajectory of the mobile fire source; and perform error correction on the predicted motion trajectory to obtain a target coordinate of the mobile fire source.

[0047] The diffusion point optimization and risk assessment module is configured to determine the type of the mobile fire source and construct a motion model of the mobile fire source; perform trajectory prediction based on the motion model and the three-dimensional coordinates to obtain a predicted motion trajectory of the mobile fire source; and perform error correction on the predicted motion trajectory to obtain a target coordinate of the mobile fire source.

[0048] The hierarchical targeted interception module is configured to perform active interception of a UAV based on the target coordinate for a mobile fire source with a high risk level; and perform tracking and monitoring of a UAV based on the target coordinate for a mobile fire source with a medium risk level.

[0049] The unmanned aerial vehicle fire extinguishing targeted interception method and system based on mobile fire source detection of the present disclosure utilizes a multispectral detector to perform multispectral detection, then enhances the signal-to-noise ratio through time delay integration, improves the image quality through adaptive non-uniformity correction, and excludes external interference through polarization encoding anti-interference technology, so as to exclude pseudo fire sources such as high-temperature rocks and environmental electromagnetic interference, and make the fire source identification accuracy rate accurate; then, the differential GPS and visual SLAM fusion technology are adopted to simulate the fire source propagation path in combination with the motion model of the mobile fire source, and the trajectory is predicted; the predicted motion trajectory is corrected for errors through the federal Kalman filter architecture, so as to provide accurate coordinates for interception, that is, the target coordinates of the mobile fire source are obtained; then, based on the type of the mobile fire source, a risk assessment model is used to perform risk assessment; finally, for the mobile fire source with a high risk level, the unmanned aerial vehicle is actively intercepted based on the target coordinates; for the mobile fire source with a medium risk level, the unmanned aerial vehicle is tracked and monitored based on the target coordinates. The technical solution of the present disclosure integrates a spectral sensor and a data fusion algorithm, and can accurately identify a mobile fire source such as an animal-carrying fire source; the present disclosure realizes a GPS and visual SLAM fused three-dimensional dynamic positioning technology, accurately locates the fire source and predicts the trajectory in combination with the motion model; at the same time, the present disclosure quantifies the risk level based on multi-dimensional indicators such as the basic attributes of the fire source, and implements different targeted interception strategies for different risk levels, such as implementing sound wave driving, fire extinguishing bomb attack, dry ice cooling and other measures, which effectively improves the timeliness and effectiveness of fire source processing; specifically, the fire source can be intervened 5-10 minutes in advance, the response time is shortened by more than 83% compared with the traditional scheme, and hierarchical dynamic interception is realized, and the interception success rate in a high-risk scene reaches 94%.

[0050] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0051] The accompanying drawings are used to better understand the present scheme and do not limit the present disclosure. Among them:

[0052] Figure 1 is a flowchart of the unmanned aerial vehicle fire extinguishing targeted interception method based on mobile fire source detection in an embodiment of the present disclosure;

[0053] Figure 2 is a flowchart of determining a fire source in an embodiment of the present disclosure;

[0054] Figure 3 is a flowchart of locating a fire source in an embodiment of the present disclosure;

[0055] Figure 4 is a flowchart of the unmanned aerial vehicle fire extinguishing targeted interception method based on mobile fire source detection in another embodiment of the present disclosure;

[0056] Figure 5 This is a schematic diagram of the structure of the drone-based fire suppression and targeted interception system based on mobile fire source detection in this embodiment of the present disclosure. Detailed Implementation

[0057] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0058] This disclosure addresses the significant shortcomings of existing technologies in dealing with fires caused by animals carrying fire sources. Specifically, conventional technologies lack specificity for the particular scenario of "animals carrying fire sources"; most drones rely on single thermal imaging technology, which cannot accurately identify fire sources when they are first carried, leading to frequent omissions; risk assessments do not incorporate animal attributes and behavioral characteristics; and traditional two-dimensional positioning technology has large trajectory prediction errors in complex terrain. This disclosure proposes a drone-based targeted interception method and system for fire suppression based on mobile fire source detection. The technical solution of this disclosure can accurately identify fire sources carried by animals, dynamically predict the fire source spread path, and implement graded targeted interception, solving the problems of slow response, inaccurate identification, and inefficient handling in existing technologies.

[0059] The technical solution of this disclosure will be described below through specific embodiments.

[0060] like Figure 1 The diagram shown is a flowchart of the drone-based targeted interception method for fire suppression based on mobile fire source detection in this embodiment. The executing entity in this embodiment is a computing device, component, or system with data processing capabilities. Specifically, the method in this embodiment may include the following steps:

[0061] S110. A multispectral detector is used to perform multispectral detection on the target detection area to obtain a multispectral signal; wherein the multispectral signal includes ultraviolet signal, infrared signal, terahertz signal, and quantum magnetic force signal; the multispectral signal is subjected to time delay integration processing to enhance the signal-to-noise ratio of the multispectral signal; the multispectral signal with enhanced signal-to-noise ratio is subjected to adaptive non-uniformity correction; the multispectral signal after adaptive non-uniformity correction is subjected to polarization coding anti-interference processing to obtain the moving fire signal.

[0062] The multispectral detector includes an ultraviolet sensor (185-260nm), a short-wave infrared sensor (1.4-3μm), a terahertz sensor (0.1-1THz), and a quantum magnetometer.

[0063] Wherein, the ultraviolet sensor multi-spectral resolution is 0.1nm, which is used for flame free radical feature capture; the short-wave infrared sensor thermal sensitivity is 10mK@300K, which is suitable for underground ember three-dimensional imaging; the terahertz sensor penetration depth is 1.2m@wood, which can be used to confirm whether the animal and the floating object carry fire; the quantum magnetometer sensitivity is 0.1nT@1Hz, which can be used to confirm whether the animal and the floating object carry fire. For surface hidden fire tracking.

[0064] This step enhances the signal-to-noise ratio by time delay integration (TDI), improves the image quality by adaptive non-uniformity correction (NUC), and excludes external interference by polarization encoding anti-interference technology, so as to exclude false fire sources such as high-temperature rocks and environmental electromagnetic interference, and improve the fire identification accuracy to 85%-92% and the false alarm rate to less than 2 times / hour.

[0065] S120, based on the moving fire signal, acquiring the GPS signal corresponding to the moving fire; based on the environment map constructed by the visual SLAM system, solving and coordinate converting the GPS signal to obtain the three-dimensional coordinates of the moving fire; determining the type of the moving fire and constructing the motion model of the moving fire; based on the motion model and the three-dimensional coordinates, performing trajectory prediction to obtain the predicted motion trajectory of the moving fire; performing error correction on the predicted motion trajectory to obtain the target coordinates of the moving fire.

[0066] Positioning technology: differential GPS and visual SLAM fusion technology, combined with motion model (such as hawk flight trajectory, squirrel ground movement mode) and machine learning algorithm, realize three-dimensional dynamic modeling. The vertical positioning accuracy of this technology is ±2m, and the trajectory prediction error of 10s prediction is 8-12m, which is much better than traditional two-dimensional positioning technology.

[0067] Dynamic updating mechanism: integrate fluid mechanics, heat conduction model and meteorological data such as wind speed and direction to simulate the fire spread path in real time and provide accurate coordinates (target coordinates) for interception.

[0068] In specific implementation, the GPS signal can be collected by using the differential GPS subsystem, wherein the differential GPS subsystem: adopts RTK (real-time dynamic) technology, accepts Beidou+GPS dual constellation signals; when the base station spacing is less than or equal to 10KM, the plane positioning accuracy is ±10cm and the height accuracy is ±20cm.

[0069] The visual SLAM system is configured with a binocular vision camera, which realizes weak light environment mapping in combination with infrared structured light.

[0070] In addition, 9-axis IMU and laser radar can also be used as auxiliary sensors to collect related data.

[0071] The above coordinate conversion is used to unify the coordinates of the moving fire, and specifically can be realized by the following unified WGS84 geocentric coordinate to local east-north-up (ENU) coordinate system conversion model:

[0072]

[0073] In the formula, (x enu ,y enu ,z enu ) represents the east-north-up coordinate system coordinate vector, (x,y,z) represents the WGS84 geocentric coordinate system coordinate vector, and (x0,y0,z0) represents the WGS84 coordinates of the reference point.

[0074] The above visual SLAM system uses the PTAM (Parallel Tracking and Mapping) algorithm based on feature points to improve the visual SLAM front-end odometry:

[0075] ORB feature improvement: F = {f1, f2,..., f n}, where each f i = (p i , d i )

[0076] The symbols have the following meanings.

[0077]

[0078]

[0079] Dynamic feature filtering:

[0080] Introduce the optical flow method to calculate the motion vector of the feature point: When ||v i || > v th , it is determined as a dynamic feature.

[0081]

[0082] Camera pose estimation: use PnP (perspective n-point) algorithm to solve the camera external parameter:

[0083]

[0084]

[0085] Iterative optimization: use the practical Levenberg-Marquardt algorithm to minimize the re-projection error

[0086]

[0087]

[0088] Projection function:

[0089]

[0090]

[0091] The movement model for moving fire is based on animal behavior and can be divided into three types, as shown in the table below:

[0092]

[0093]

[0094] Hidden Markov Model (HMM) Modeling:

[0095] State space S = {stationary, moving, accelerating, turning, throwing} Observation space O = {position, velocity, acceleration, attitude}

[0096] Transition probability matrix A = [a ij ], where: a ij =P(q) t =j|q t-1 =i) Observation probability matrix B = [b j (k)], where: b j (k)=P(o t =u k |q t =j) Initial state probability π = [π i ], where: π i =P(q) l =i)

[0097]

[0098]

[0099] S130. Based on the type of mobile tinder, obtain the corresponding basic attributes, behavioral characteristics, and historical risk records of the tinder; and, based on the obtained basic attributes, behavioral characteristics, and historical risk records of the tinder, use a risk assessment model to conduct a risk assessment and obtain a risk level.

[0100] like Figure 4 As shown, the basic attributes of a fire starter include at least one of the following: species, weight, and fur type. Behavioral characteristics include movement speed patterns, nesting material collection frequency, and history of exposure to fire sources. Historical risk records include the number of fires caused and the effectiveness of behavioral correction. The type of fire starter can be represented by an animal ID. The basic attributes, behavioral characteristics, and historical risk records of the fire starter are pre-stored in a database, and the data can be retrieved from the database using the animal ID.

[0101] The analytic hierarchy process can be used to calculate the weight in the risk assessment process, specifically:

[0102] Risk = [R s ,R w , R h ,R f ,R his ,R ter ]

[0103] R s : species risk coefficient, raptor = 0.38, rodent 0.27, ungulate = 0.18

[0104] R w : body weight impact factor,

[0105] R h : movement speed factor,

[0106] R f : fur flammability index, range [0, 1]

[0107] R his : historical fire frequency factor,

[0108] R ter : terrain impact index, determined by slope and vegetation

[0109] The risk assessment model is as follows: forward propagation from the input layer to the hidden layer

[0110] Hidden layer neuron input:

[0111] ReLU activation function:

[0112] Output of the jth neuron in the hidden layer after ReLU activation Original input (unactivated value) of the jth neuron in the hidden layer

[0113] Forward propagation from the hidden layer to the output layer

[0114] Output layer neuron input:

[0115]

[0116] Softmax activation function (risk level probability): Softmax activation function, compute the prediction probability of k-class risk level (high / medium / low), z (m) is the model output layer unactivated value, k / m is the class index, corresponding to three risk levels.

[0117] Cross-entropy loss function:

[0118] Measure the error of the predicted probability and the true value y i，k (0 or 1 represents whether it belongs to the kth class), N is the number of samples. Among them, y i，k is the true risk level label (0 or 1), y i,k is the model prediction probability.

[0119] Backpropagation gradient calculation

[0120] Output layer error:

[0121]

[0122] Weight gradient:

[0123] Pre-layer weight gradient, calculate the gradient of weight for updating, is the hidden layer activation value

[0124] Post-layer weight gradient, x i is the input layer data

[0125] Bias gradient:

[0126] Pre-bias gradient, is the output layer bias

[0127] Post-bias gradient, is the hidden layer bias

[0128] Adam optimization algorithm:

[0129] First moment and second moment estimation:

[0130]

[0131] Training optimization of UAV risk assessment model (Adam algorithm updates parameters) and inference decision (determines the risk level)

[0132] Bias correction:

[0133]

[0134]

[0135]

[0136] Parameter update( for the gradient):

[0137]

[0138] θ t ]]> Model parameters at time t η Learning rate (controls parameter update step size) ∈ Small constant (to prevent denominator from being 0)

[0139] Risk level determination

[0140]

[0141] wherein, represent the probabilities of low, medium, and high risk levels.

[0142] S140, for the mobile fire brand with a high risk level, performing active interception by the unmanned aerial vehicle based on the target coordinates; and for the mobile fire brand with a medium risk level, performing tracking and monitoring by the unmanned aerial vehicle based on the target coordinates.

[0143] For the mobile fire brand with a high risk level, performing active interception by the unmanned aerial vehicle based on the target coordinates, specifically including: for the mobile fire brand with a high risk level, performing unmanned aerial vehicle sound wave repelling processing, unmanned aerial vehicle fire extinguishing bomb striking processing, and unmanned aerial vehicle cooling processing based on the target coordinates.

[0144] The sound wave repelling technology is specifically as follows:

[0145] Frequency optimization model: f opt = f0+k·ΔT

[0146]

[0147]

[0148] Directivity sound wave equation:

[0149] wherein, I(r, θ): sound intensity at distance r, angle θ; I0 is the initial intensity of the sound source, D(θ) is the directivity function (describing the energy distribution of sound waves at different angles θ), and r is the distance from the sound source to the target.

[0150] The fire extinguishing bomb striking technology is as follows:

[0151] The trajectory model is x = v0cosθ·t

[0152] x, y Horizontal and vertical positions of fire extinguishing bomb [v0] Initial velocity of fire extinguishing bomb θ Launch angle (angle with horizontal direction) t Flight time g Gravity acceleration

[0153] Prediction landing point correction:

[0154] Delta x = k1 * v w + k2 * delta T

[0155]

[0156]

[0157] Delta y = k3 * p + k4 * h

[0158]

[0159] The technical parameters of the cooling device (dry ice spraying device) in the unmanned aerial vehicle cooling process are as follows:

[0160] Q = m * delta H

[0161]

[0162]

[0163]

[0164] For medium-risk unmanned aerial vehicle tracking and monitoring, the specific scheme is as follows: trajectory tracking algorithm in unmanned aerial vehicle tracking control:

[0165]

[0166] Where e(t) is the tracking error, k p , k d is the control gain

[0167] Obstacle avoidance model:

[0168]

[0169]

[0170] For low-risk mobile fire, based on the target coordinates, a dry ice fog model is generated and an isolation zone is constructed; based on the isolation zone, a low-temperature barrier is generated for the mobile fire using the dry ice fog model.

[0171] Dry ice fog model:

[0172] Phase change process calculation:

[0173]

[0174] Fog droplet diffusion model: sigma y = sigma z = a * z b

[0175]

[0176]

[0177] The algorithm for constructing the isolation belt is as follows:

[0178] Optimal width calculation: W=k1·v+k2·FVI+k3·T

[0179]

[0180] Coverage efficiency model:

[0181]

[0182] In some embodiments, the time delay integration processing on the multi-spectral signals can include the following steps:

[0183] For each signal in the multi-spectral signals, the signals of the same target region are integrated and superimposed multiple times in the time dimension.

[0184] Among them, for the ultraviolet signal, the target ultraviolet band is integrated and superimposed for a first preset number of times at the end stage of the integration and superimposition processing to enhance the flame free radical; for the infrared signal, the target infrared band is integrated and superimposed for a second preset number of times to strengthen the characteristic of the CO2 characteristic absorption peak.

[0185] This embodiment realizes adaptive integration of wave bands: n=16 (first preset number) for ultraviolet band, and n=64 (second preset number) for short-wave infrared band, to enhance the 308nm characteristic radiation of flame free radical (OH*) at the end stage and to strengthen the 4.26μm CO2 characteristic absorption peak detection.

[0186] During the integration and superimposition processing, the phase or scanning direction of the charge transfer of the time delay integration processing is adjusted by using a motion compensation mechanism to make the integration path conform to the actual motion.

[0187] The optical flow method is used to estimate the target motion vector, and dynamic target TDI is realized through phase offset correction.

[0188] The technical principle of this embodiment is as follows:

[0189] By exposing and integrating the same target region multiple times in the time dimension, the n times of sampling signals are superimposed as:

[0190]

[0191] S TDI (TDI) Signal after multiple exposure integrations n Sampling times S(t-iT) Original signal of i-th sampling T Sampling period W(i) Weighting coefficient of i-th sampling

[0192] where T is a sampling period, w(i) is a Gaussian weighting coefficient, and the effective signal-to-noise ratio (SNR) is enhanced to that of traditional sampling times.

[0193] In some embodiments, the adaptive non-uniformity correction on the multi-spectral signal with enhanced SNR can include the following steps:

[0194] For each multi-spectral detector, a corresponding response characteristic dynamic model is established; in the case that the temperature of the target detection area is greater than a first preset temperature, the response characteristic dynamic model is used to correct the signal corresponding to the multi-spectral detector in the multi-spectral signal with enhanced SNR by using scene-based non-uniformity correction; in the case that the temperature of the target detection area is less than a second preset temperature, the signal corresponding to the multi-spectral detector in the multi-spectral signal with enhanced SNR is corrected by using a two-step correction method; wherein the two-step correction method includes two-point correction and multi-point correction; wherein the first preset temperature is higher than the second preset temperature; the corrected multi-spectral signal is processed by using inter-frame correlation constraints to correct inter-frame smearing.

[0195] In this embodiment, the established response characteristic dynamic model is as follows:

[0196] S real (x,y)=k(x,y)·S raw (x,y)+b(x,y)

[0197] where S real (x,y) is the corrected signal; S raw (x,y) is the original signal of the sensor, k(x,y) is the gain coefficient, and b(x,y) is the offset coefficient, which is realized by an online updating mechanism:

[0198] k t+1 (x,y)=k t (x,y)+α·[R-k t (x,y)·S t (x,y)-b t (x,y)]·S t (x,y)

[0199] b t+1 (x,y)=b t (x,y)+α·[R-k t (x,y)·S t (x,y)-b t (x,y)

[0200] k t+1 (x,y) Gain coefficient at next time k t (x,y) Gain coefficient at current time α Learning rate R Reference radiation value S t (x,y)]]> Original signal at current time b t (x,y)]]> Offset coefficient at current time b t+1 (x,y)]]> Offset coefficient at next time

[0201] α is the adaptive learning rate (dynamically adjusted from 0.001 to 0.1), and R is the reference radiation value.

[0202] This embodiment implements a scene adaptation strategy:

[0203] High temperature scene (>500℃): Scene-based non-uniformity correction (SBNUC) is adopted, and the parameters are updated every 10 frames; Low temperature scene (<100℃): Switch to two-step correction (two-point correction + multi-point correction) to reduce computational complexity.

[0204] At the same time, inter-frame correlation constraints are introduced to prevent correction motion blur caused by fast-moving targets.

[0205] In some embodiments, performing polarization coding anti-interference processing on the multispectral signal after adaptive non-uniformity correction to obtain the moving fire signal may include:

[0206] Obtain the interference feature library; perform polar coding on the multispectral signal after adaptive non-uniformity correction to obtain the polar code; based on the likelihood ratio between the polar code and the features in the interference feature library, select the moving fire signal from the multispectral signal after adaptive non-uniformity correction.

[0207] Specifically, a (512,256) polarization structure is adopted, which is achieved through the channel polarization effect: information bits: carry fire source characteristic parameters (temperature, radiation intensity, spectral distribution); freeze bits: preset typical interference characteristics (high temperature rocks, electromagnetic pulses, etc.).

[0208] Encoder Generate Matrix

[0209] Where B_N is the bit flipping matrix.

[0210] In this embodiment, a database of interference features is constructed:

[0211] I = {I temp I spectrum I frequency I polarization}

[0212] Interference identification is achieved by calculating the likelihood ratio (LLR) of polar codes:

[0213]

[0214] When LLR(x)≥3, it is determined to be a real fire source signal.

[0215] In summary, as Figure 2As shown, the multispectral detector collects multispectral signals, and then performs TDI signal enhancement, NUC response correction, polarization encoding, construction of a feature vector of the polar code, calculation of a likelihood ratio, preliminary screening through LLR, and then enters D-S fusion optimization determination to perform fire confirmation operations.

[0216] As shown in Figure 2 , P is the probability value of the post-fusion determination "fire existence confidence", which is used to quantify the degree of evidence support and guide the process branching (such as P≥0.8 strong support for fire → "fire confirmation"; 0.5≤P<0.8 requires supplementary evidence → "secondary weighted detection"; P<0.5 is determined as interference → "interference exclusion"). Fire confirmation: when the confidence probability P output by the D-S fusion is less than 0.8, it is determined as a real fire, triggering the "fire confirmation" process. At this time, the system marks the region as a fire risk point and outputs positioning coordinates, spectral characteristics and other information for subsequent fire extinguishing decision. Secondary weighted detection: if (0.5

[0217] In some embodiments, the above determining a moving fire type and constructing a motion model of the moving fire can include the following steps:

[0218] Based on the moving fire signal, behavior data corresponding to the moving fire is obtained; based on the behavior data, a fire type of the moving fire is determined, and a motion model of the moving fire is constructed; wherein the fire type includes at least one of the following: flying animals; arboreal animals; terrestrial animals.

[0219] Specifically, as shown in Figure 3 , the above behavior data can be collected by an animal behavior sensor, for example, pictures of moving pictures collected by a camera as behavior data.

[0220] In some embodiments, as shown in Figure 3 , the trajectory prediction based on the motion model and the three-dimensional coordinates obtains a predicted motion trajectory of the moving fire. Specifically:

[0221] A physical propagation model of the ignition source, a wind field influence model, and a turbulent diffusion model are constructed. Based on the three-dimensional coordinates, the physical propagation model, and the turbulent diffusion model, the current state vector of the ignition source is determined. The ignition source state vector includes the ignition source position, the ignition source propagation speed, and the ignition source temperature. Based on the wind field influence model, wind field parameters are determined. Based on the motion model, control parameters for the moving ignition source are determined. Based on the ignition source state vector, the wind field parameters, and the control parameters for the moving ignition source, trajectory prediction is performed to obtain the predicted trajectory of the moving ignition source.

[0222] The above embodiments are trajectory prediction based on fluid dynamics coupling.

[0223] The specific physical propagation model of the spark is as follows:

[0224] Establish the three-dimensional heat conduction equation:

[0225] Where α is the thermal diffusivity and Q is the heat source intensity.

[0226] Boundary conditions: Temperature normal derivative

[0227] K is the thermal conductivity, h is the convective heat transfer coefficient, ε is the emissivity, σ is the Stefan-Boltzmann constant, T is the surface temperature, and T is the surface temperature. ∞ Ambient temperature, Temperature normal derivative

[0228] The wind field impact model is as follows:

[0229] Modified Log wind profile model:

[0230] Where u(z) is the wind speed at height z, u ref For reference height z ref The wind speed at the location is z0, and the surface roughness length is z0.

[0231] The turbulent diffusion model is as follows: σ y =σ z =a zb

[0232] Where σ y , σ z denoted as , where is the horizontal and vertical diffusion coefficient, and a and b are stability parameters.

[0233] The trajectory prediction algorithm for performing trajectory prediction is as follows:

[0234] Joint prediction model: X t+1 =f(X) t u t w t ,θ)

[0235] where X t is the fire state vector (position, velocity, temperature); u t is the mobile fire control parameter; w t is the wind field parameter; and theta is the model parameter

[0236] Finally, filtering is performed, and the specific particle filtering error calculation is as follows:

[0237]

[0238]

[0239] In some embodiments, the error correction of the predicted motion trajectory to obtain the target coordinates of the mobile fire can specifically include:

[0240] The predicted motion trajectory is filtered using a federated Kalman filter architecture; the data obtained by filtering is sequentially subjected to continuity constraint processing, loop detection correction cumulative error processing, terrain constraint processing, and obstacle avoidance collision detection to obtain the target coordinates of the mobile fire; wherein the terrain constraint processing is used for height verification.

[0241] The embodiment realizes a dynamic updating and error correction mechanism.

[0242] The embodiment realizes multi-source data fusion filtering using a federated Kalman filter architecture, specifically:

[0243]

[0244] where X k is the fusion state, P k is the fusion covariance, w i is the sub-filter weight.

[0245] The error correction strategy of the embodiment includes time dimension correction, wherein the time dimension correction includes short-term correction and long-term correction.

[0246] Short-term correction (≤10S): motion continuity constraint based on IMU data

[0247]

[0248] Long-term correction (>10S): use loop detection to correct cumulative error

[0249]

[0250] The above terrain constraint uses DEM data for height verification:

[0251] z valid= h(x, y) ± ξ h

[0252] ξ h = 0.5m (forest scene), ξ h = 0.2m (grassland scene)

[0253] Obstacle avoidance collision detection: Collision detection based on laser radar point cloud:

[0254] min d ||X pred -O||≥ d saje

[0255] d safe = 0.8m (flying animals)

[0256] d safe = 0.8m (terrestrial animals)

[0257] In the technical solution of the present disclosure, the integrated configuration of the multispectral detector and the data fusion algorithm are used to accurately identify the fire carried by animals. The three-dimensional dynamic positioning technology of GPS and visual SLAM fusion, combined with animal behavior models and machine learning algorithms, predicts the fire source trajectory. The diffusion point optimization and risk assessment model of genetic algorithm and analytic hierarchy process / neural network quantifies the risk level by integrating multi-dimensional indicators. The hierarchical targeted interception strategy for different risk levels includes the combined application of sound wave repulsion, fire extinguishing bomb attack, dry ice cooling and other measures.

[0258] In terms of accuracy: the multispectral fusion detection improves the identification accuracy by 21-27% compared to single thermal imaging, and can identify the early stage of fire carrying.

[0259] In terms of proactivity: through trajectory prediction and diffusion point optimization (and teaching steps), the intervention of fire movement is advanced by 5-10 minutes, and the response time is shortened by more than 83% compared to traditional solutions.

[0260] In terms of intelligence: the risk assessment model realizes hierarchical dynamic interception, and the combined interception success rate of "sound wave + fire extinguishing bomb" in high-risk scenarios reaches 94%.

[0261] In addition, in terms of multispectral detector configuration, microwave radar can be considered to replace part of the sensors to assist in positioning the position of the fire carried by animals in strong smoke environment.

[0262] For the risk assessment model, in addition to the analytic hierarchy process and neural network, support vector machine (SVM) algorithm can also be used for risk level quantification to improve the generalization ability of the model.

[0263] In the hierarchical interception measure, for the high-risk level scene, a water-based extinguishing agent spraying system can be used instead, and a high-pressure water gun is used to cool and extinguish the fire source, which is used in cooperation with the sound wave driving.

[0264] The three-dimensional positioning technology can consider introducing a laser radar (LiDAR) to improve the positioning accuracy and environmental perception ability in complex terrain, especially in dense vegetation areas.

[0265] Based on the same inventive concept, the present disclosure provides a UAV fire extinguishing targeted interception system based on mobile fire source detection. The components of the system perform the same or similar steps as the above method, and therefore similar parts will not be described again. As shown in Figure 5 The UAV fire extinguishing targeted interception system based on mobile fire source detection of the embodiment includes:

[0266] The multispectral detection and fire source confirmation module 510 is configured to perform multispectral detection on a target detection area by using a multispectral detector to obtain a multispectral signal. The multispectral signal includes ultraviolet signals, infrared signals, terahertz signals, and quantum magnetic signals. The multispectral signal is subjected to time delay integration processing to enhance the signal-to-noise ratio of the multispectral signal. The multispectral signal with the enhanced signal-to-noise ratio is subjected to adaptive non-uniformity correction. The multispectral signal after the adaptive non-uniformity correction is subjected to polarization encoding anti-interference processing to obtain a mobile fire source signal.

[0267] The three-dimensional fire source coordinate acquisition module 520 is configured to acquire a GPS signal corresponding to the mobile fire source based on the mobile fire source signal. The GPS signal is solved and coordinate-converted based on an environment map constructed by using a visual SLAM system to obtain a three-dimensional coordinate of the mobile fire source.

[0268] The diffusion point optimization and risk assessment module 530 is configured to determine the type of the mobile fire source and construct a motion model of the mobile fire source. The motion model and the three-dimensional coordinate are used to perform trajectory prediction to obtain a predicted motion trajectory of the mobile fire source. The predicted motion trajectory is subjected to error correction to obtain a target coordinate of the mobile fire source. The corresponding fire source basic attribute, behavior feature, and historical risk record are acquired based on the type of the mobile fire source. A risk assessment model is used to perform risk assessment based on the acquired fire source basic attribute, behavior feature, and historical risk record to obtain a risk level.

[0269] The hierarchical targeted interception module 540 is configured to perform UAV active interception based on the target coordinate for a mobile fire source with a high risk level. For a mobile fire source with a medium risk level, the UAV is tracked and monitored based on the target coordinate.

[0270] Various implementations of the techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0271] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, or entirely on a remote machine or server.

[0272] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0273] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0274] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0275] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server is generally established by computer programs running on the respective computers and having a client-server relationship to each other. The servers can be cloud servers, servers of a distributed system, or servers combined with a blockchain.

[0276] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the scope of the present disclosure. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, without departing from the desired results of the technology disclosed herein, which are not limited herein.

[0277] The specific embodiments described above are not intended to be limiting, and persons skilled in the art will appreciate that various modifications, combinations, sub-combinations and alternatives can be made to the specific embodiments without departing from the spirit and principles of the disclosure. Accordingly, the disclosure is not limited to the specific embodiments described above.

Claims

1. A method for unmanned aerial vehicle fire extinguishing targeting interception based on mobile fire source detection, characterized in that, The method comprises: a multi-spectral detector is used to perform multi-spectral detection on a target detection area to obtain a multi-spectral signal, wherein the multi-spectral signal comprises an ultraviolet signal, an infrared signal, a terahertz signal and a quantum magnetic force signal; time delay integration processing is performed on the multi-spectral signal to enhance the signal-to-noise ratio of the multi-spectral signal; adaptive non-uniformity correction is performed on the multi-spectral signal after the signal-to-noise ratio is enhanced; polarization coding anti-interference processing is performed on the multi-spectral signal after the adaptive non-uniformity correction to obtain a mobile fire signal; based on the mobile fire signal, a GPS signal corresponding to the mobile fire is obtained; the GPS signal is solved and coordinate-converted based on an environment map constructed by a visual SLAM system to obtain a three-dimensional coordinate of the mobile fire; the type of the mobile fire is determined, and a motion model of the mobile fire is constructed; trajectory prediction is performed based on the motion model and the three-dimensional coordinate to obtain a predicted motion trajectory of the mobile fire; error correction is performed on the predicted motion trajectory to obtain a target coordinate of the mobile fire; based on the type of the mobile fire, corresponding fire basic attributes, behavior characteristics and historical risk records are obtained; and based on the obtained fire basic attributes, behavior characteristics and historical risk records, a risk assessment model is used to perform risk assessment to obtain a risk level; for a mobile fire with a high risk level, a UAV is actively intercepted based on the target coordinate; for a mobile fire with a medium risk level, a UAV is tracked and monitored based on the target coordinate.

2. The method of claim 1, wherein, The time delay integration processing on the multi-spectral signal comprises: for each signal in the multi-spectral signal, the signals of the same target area are integrated and superimposed multiple times in the time dimension; wherein, for the ultraviolet signal, a first preset number of integration and superimposition processes are performed on the target ultraviolet band at the end stage of the integration and superimposition process to enhance the flame free radical; for the infrared signal, a second preset number of integration and superimposition processes are performed on the target infrared band to strengthen the characteristic absorption peak of CO2; in the process of the integration and superimposition process, a motion compensation mechanism is used to adjust the phase or scanning direction of the time delay integration processing charge transfer, so that the integration path conforms to the actual motion.

3. The method of claim 1, wherein, The multi-spectral detector comprises an ultraviolet sensor, a short-wave infrared sensor, a terahertz sensor and a quantum magnetometer. The adaptive non-uniformity correction on the multi-spectral signal after the signal-to-noise ratio is enhanced comprises: for each multi-spectral detector, a corresponding response characteristic dynamic model is established; in the case that the temperature of the target detection area is greater than a first preset temperature, the response characteristic dynamic model is used, and scene-based non-uniformity correction is adopted to correct the signal corresponding to the multi-spectral detector in the multi-spectral signal after the signal-to-noise ratio is enhanced; in the case that the temperature of the target detection area is less than a second preset temperature, a two-step correction is adopted to correct the signal corresponding to the multi-spectral detector in the multi-spectral signal after the signal-to-noise ratio is enhanced; wherein, the two-step correction comprises two-point correction and multi-point correction; wherein, the first preset temperature is higher than the second preset temperature. The corrected multi-spectral signal is processed by using an inter-frame correlation constraint to correct inter-frame smear.

4. The method of claim 1, wherein, The multi-spectral signal after adaptive non-uniformity correction is subjected to polarization encoding anti-interference processing to obtain a mobile fire signal, including: Obtaining an interference feature library; Polarization encoding is performed on the multi-spectral signal after adaptive non-uniformity correction to obtain a polar code; Based on the likelihood ratio of the polar code and the features in the interference feature library, the mobile fire signal is screened from the multi-spectral signal after adaptive non-uniformity correction.

5. The method of claim 1, wherein, The mobile fire type is determined, and a motion model of the mobile fire is constructed, including: Based on the mobile fire signal, behavior data corresponding to the mobile fire is obtained; Based on the behavior data, the fire type of the mobile fire is determined, and a motion model of the mobile fire is constructed; The fire type includes at least one of the following: flying animals; arboreal animals; terrestrial animals.

6. The method of claim 1, wherein, Based on the motion model and the three-dimensional coordinates, a trajectory prediction is performed to obtain a predicted motion trajectory of the mobile fire; A fire physical propagation model, a wind field influence model, and a turbulent diffusion model are constructed; Based on the three-dimensional coordinates, the fire physical propagation model, and the turbulent diffusion model, a fire state vector at the current time is determined; wherein the fire state vector includes fire position, fire propagation speed, and fire temperature; Based on the wind field influence model, a wind field parameter is determined; Based on the motion model, a mobile fire control parameter is determined; Based on the fire state vector, the wind field parameter, and the mobile fire control parameter, a trajectory prediction is performed to obtain a predicted motion trajectory of the mobile fire.

7. The method of claim 1, wherein, The predicted motion trajectory is subjected to error correction to obtain a target coordinate of the mobile fire, including: The predicted motion trajectory is filtered by using a federal Kalman filter architecture; The data obtained by filtering is subjected to successive continuity constraint processing, loop detection correction cumulative error processing, terrain constraint processing, and obstacle avoidance collision detection to obtain a target coordinate of the mobile fire; wherein the terrain constraint processing is used for height verification.

8. The method of claim 1, wherein, For the mobile fire with a high risk level, the target coordinate is used for active interception by a UAV, including: For the mobile fire with a high risk level, the target coordinate is used for sound wave repelling processing by a UAV, fire extinguishing bomb striking processing by a UAV, and cooling processing by a UAV.

9. The method of claim 1, wherein, Further including: For the mobile fire with a low risk level, based on the target coordinate, a dry ice fog model is generated, and an isolation belt is constructed; Based on the isolation belt, a low-temperature barrier for the mobile fire is generated by using the dry ice fog model.

10. A UAV fire extinguishing targeted interception system based on mobile fire source detection, characterized in that, Including: The multispectral detection and fire source confirmation module is configured to perform multispectral detection on a target detection area by using a multispectral detector to obtain a multispectral signal, wherein the multispectral signal includes ultraviolet signals, infrared signals, terahertz signals, and quantum magnetic force signals; perform time delay integration processing on the multispectral signal to enhance the signal-to-noise ratio of the multispectral signal; perform adaptive non-uniformity correction on the multispectral signal after the signal-to-noise ratio is enhanced; and perform polarization coding anti-interference processing on the multispectral signal after the adaptive non-uniformity correction to obtain a mobile fire source signal; The three-dimensional fire source coordinate acquisition module is configured to acquire a GPS signal corresponding to the mobile fire source based on the mobile fire source signal; and perform calculation and coordinate conversion on the GPS signal based on an environment map constructed by using a visual SLAM system to obtain three-dimensional coordinates of the mobile fire source; The diffusion point optimization and risk assessment module is configured to determine the type of the mobile fire source and construct a motion model of the mobile fire source; perform trajectory prediction based on the motion model and the three-dimensional coordinates to obtain a predicted motion trajectory of the mobile fire source; and perform error correction on the predicted motion trajectory to obtain target coordinates of the mobile fire source; The corresponding fire source basic attributes, behavior characteristics, and historical risk records are obtained based on the type of the mobile fire source; and a risk assessment model is used to perform risk assessment based on the obtained fire source basic attributes, behavior characteristics, and historical risk records to obtain a risk level; The hierarchical targeted interception module is configured to perform active interception of a UAV based on the target coordinates for a mobile fire source with a high risk level; and perform tracking and monitoring of a UAV based on the target coordinates for a mobile fire source with a medium risk level.

Images