Radar monitoring intelligent early warning method and system suitable for forest fire prevention

Abstract

The present invention provides a radar monitoring intelligent early warning method and system suitable for forest fire prevention, and relates to the technical field of forest fire prevention. The method includes: establishing a three-dimensional map, dividing the area to be detected; collecting forest fire factor data; constructing a forest fire risk assessment model based on a convolutional neural network, and using the Hippo algorithm to optimize the hyperparameters of the model, dividing the risk level of the area to be detected; performing fixed radar monitoring on the high-risk level area; generating a mobile monitoring path for the low-risk level area; and constructing a prediction model for fire early warning. The present invention greatly reduces the number of fixed radar deployments and reduces the cost of forest fire early warning by finely dividing the detection area, using fixed monitoring radars to continuously monitor the high-risk level areas, and using mobile radars to conduct regular inspections of the low-risk areas. At the same time, it realizes comprehensive monitoring of the entire target forest, making up for the limitations of a single monitoring method.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire prevention, and in particular to a radar monitoring intelligent early warning method suitable for forest fire prevention, and a radar monitoring intelligent early warning system suitable for forest fire prevention. Background Art

[0002] Forest fire is a natural disaster, which is mainly caused by the burning of trees, shrubs and other vegetation due to human or natural reasons. Forest fire not only causes serious damage to the forest ecosystem, but also threatens nearby human settlements and infrastructure. Forest fires are usually caused by the following main factors: dry and hot climate, dry vegetation and human activities such as smoking, picnics, construction, etc. Once it occurs, forest fires may spread rapidly and cause large-scale destruction. It will not only cause a lot of economic losses, but also damage the ecological environment, threaten the survival of animals, and have a serious impact on human life. Therefore, it is particularly important to achieve forest fire early warning.

[0003] Among the existing technologies for forest fire prevention, forest fire early warning based on radar monitoring can greatly improve the timeliness and accuracy of forest fire prevention, and provide timely and effective fire prevention and emergency support. Radar monitoring for forest fire early warning is not limited by weather conditions, and can work in various severe weather conditions. It can monitor a large area at the same time, which helps to detect signs of large-scale fires at an early stage. It can detect targets at a long distance and buy time for responding to fires in advance. It can obtain monitoring data in real time and issue early warnings. It can penetrate certain obstacles and obtain fire information hidden in them.

[0004] When using radar monitoring to monitor and warn of forest fires, the forest area is generally large and the effective monitoring range of radar is limited. If comprehensive monitoring is carried out, the cost investment is too high. If regional monitoring is carried out, the comprehensiveness of monitoring cannot be ensured, and blind spots are likely to occur, causing huge losses. Summary of the invention

[0005] The present invention provides a radar monitoring intelligent early warning method and system suitable for forest fire prevention, which is used to solve the defects of the prior art that the cost of comprehensive monitoring is too high and regional monitoring is prone to monitoring blind spots.

[0006] On the one hand, the present invention provides a radar monitoring intelligent early warning method applicable to forest fire prevention, comprising:

[0007] S1. Create a three-dimensional map of the target forest, mark the trajectory and frequency of human activities on the three-dimensional map, and divide the target forest into N areas to be detected.

[0008] S2. Collect forest fire factor data of each area to be detected, the forest fire factor data includes vegetation data, meteorological data and external data, and standardize the factor data to obtain standardized data.

[0009] S3. Build a forest fire risk assessment model based on convolutional neural network, use Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model, input standardized data, and output the fire risk level of each area to be detected. The fire risk level includes high risk level and low risk level.

[0010] S4. Carry out fixed radar monitoring of high-risk areas. According to the monitoring radius of the fixed monitoring radar, obtain the deployment location of the fixed monitoring radar required to monitor all high-risk areas, and optimize the required number of fixed monitoring radars.

[0011] S5. Merge all low-risk areas that are not monitored by fixed monitoring radars into mobile radar monitoring areas, obtain the boundary curve of the mobile radar monitoring area, plan the mobile monitoring path of the mobile radar monitoring area, and use the mobile monitoring radar for mobile monitoring according to the preset period.

[0012] S6. Receive the monitoring data from fixed monitoring radars and mobile monitoring radars in real time, and build a prediction model based on support vector machines. Input the monitoring data, output the forest fire prediction value, and compare it with the preset forest fire threshold. If the forest fire prediction value is greater than the forest fire threshold, issue a forest fire warning in a timely manner.

[0013] According to a radar monitoring intelligent early warning method suitable for forest fire prevention provided by the present invention, vegetation data includes plant type data and plant density data. Meteorological data includes ambient temperature data and ambient humidity data. External data includes human activity data.

[0014] According to a radar monitoring intelligent early warning method suitable for forest fire prevention provided by the present invention, in step S3, the process of constructing a forest fire risk assessment model based on a convolutional neural network includes:

[0015] Collect historical forest fire factor data and standardize them to obtain standardized historical data, which are then divided into training set and validation set.

[0016] Feature extraction is performed on the standardized data to obtain vegetation characteristics, meteorological characteristics and human activity characteristics.

[0017] A convolutional neural network model was constructed, with vegetation characteristics, meteorological characteristics, and human activity characteristics as input and fire risk level as output. The convolutional neural network model was trained, and the model parameters that met the test accuracy were retained to obtain a forest fire risk assessment model.

[0018] According to a radar monitoring intelligent early warning method suitable for forest fire prevention provided by the present invention, the process of using the Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model includes:

[0019] A set of initial hippo positions are randomly generated, and each hippo individual represents a hyperparameter combination, which includes the number of convolution kernels, learning rate, and number of layers.

[0020] For each hippopotamus individual, the accuracy is calculated on the validation set and used as the fitness value.

[0021] Update the hippopotamus position, the update formula is:

[0022]

[0023] in, and are the positions of the i-th hippopotamus individual in the j-th dimension at the t+1th and tth iterations, respectively. is the weight that controls the approach to the optimal individual, is the weight controlling random exploration. is the position of the best hippopotamus in the current iteration in the jth dimension, is the position of a hippopotamus randomly selected from the population in the jth dimension.

[0024] Repeatedly update the hippo position and calculate the fitness value until the preset number of iterations is reached.

[0025] The hyperparameter combination represented by the hippopotamus position with the highest accuracy was selected as the hyperparameter combination for the forest fire risk assessment model.

[0026] According to a radar monitoring intelligent early warning method applicable to forest fire prevention provided by the present invention, in step S4, the process of obtaining the deployment position of each fixed monitoring radar includes:

[0027] Set the set of center coordinates of all high-risk areas to be P = {p1, p2, ..., p T}, the effective monitoring radius of the fixed monitoring radar is R.

[0028] S411. Initialize an empty set C, and initialize the set U of all high-risk areas to set P.

[0029] S412: Select a high-risk level area i in the set U, and calculate the distance between the high-risk level area i and the center coordinates of all other high-risk level areas in the set U.

[0030] S413. Calculate the maximum distance between the high-risk level area i and the center coordinates of all other high-risk level areas in the set U. The formula is:

[0031]

[0032] Where, d ij Represents the distance from the center coordinate of the i-th high-risk level area to the center coordinate of the j-th high-risk level area.

[0033] S414, if d imax ≤2R, then high-risk area i is taken as the placement location of a fixed monitoring radar and added to set C, and high-risk area i and all high-risk areas whose distance to high-risk area i is less than R are removed from set U.

[0034] S415, if d imax >2R, then select the high-risk area k closest to the high-risk area i in set U, calculate the midpoint coordinates of high-risk area i and high-risk area k, add it to set C as the placement position of a fixed monitoring radar, and remove all high-risk areas whose distance to the midpoint coordinates does not exceed R from set U.

[0035] S416, repeat steps S412 to S415 until the set U is empty, the set C is the deployment location of each fixed monitoring radar, and the number of fixed monitoring radars is .

[0036] According to a radar monitoring intelligent early warning method applicable to forest fire prevention provided by the present invention, the process of optimizing the required number of fixed monitoring radars includes:

[0037] S421. Set the set C = {q1, q2, ..., q M}, the position coordinates of each fixed monitoring radar are (x m ,y m , z m ). Calculate the distance between the position of each fixed monitoring radar and the center coordinates of the high-risk level area. And set the set of high-risk level areas that can be monitored by the fixed monitoring radar q, and the formula is expressed as:

[0038]

[0039] Where, d qp Indicates the distance between each fixed monitoring radar position and the center coordinates of the high risk level area.

[0040] S422. Initialize a flag set flag={1, 1, ..., 1}, where the number of elements 1 is M, and 1 indicates that each fixed monitoring radar is in an unremoved state.

[0041] S423, traverse the positions of M fixed monitoring radars, and calculate the position of the fixed monitoring radar q after removing the ath fixed monitoring radar a The remaining high-risk level areas that can be monitored by the fixed monitoring radar are expressed as follows:

[0042]

[0043] In the formula, q b Indicates the bth fixed monitoring radar. If C1=C, set flag (a) to 0, indicating that the fixed monitoring radar q can be removed. a , and update M1=M-1 until for all a, when flag(a)=1, there are .

[0044] According to a radar monitoring intelligent early warning method for forest fire prevention provided by the present invention, in step S5, the process of obtaining the boundary curve of the mobile radar monitoring area includes:

[0045] S51. Set the set of center coordinates of all low-risk areas to be H = {h1, h2, ..., h S}.

[0046] S52: Calculate the distance d between the center coordinates of each low-risk level area and the position of each fixed monitoring radar. sq , if d sq <R, then remove the low risk level area from the set H.

[0047] S53, merge all low-risk areas that are not monitored by fixed monitoring radar into area G, collect n sampling points at the boundary of area G, and the coordinates of the sampling points are (x n ,y n ).

[0048] S54, in the horizontal coordinate interval The interval boundary curve is established inside, and the formula is expressed as:

[0049]

[0050] In the formula, x represents the horizontal coordinate of the current point, Indicates The horizontal coordinate of the node, , , and is the coefficient to be determined.

[0051] Set the boundary condition to natural boundary condition, at the node At the point where two adjacent boundary curves have the formula and Its first-order derivative and second-order derivative are all continuous.

[0052] S55. Solve the coefficients using linear algebra , , and The value of is used to obtain the boundary curve of the mobile radar monitoring area.

[0053] According to a radar monitoring intelligent early warning method suitable for forest fire prevention provided by the present invention, the process of planning a monitoring path for a mobile radar monitoring area includes:

[0054] According to the boundary curve, the minimum rectangle that can completely cover the mobile radar monitoring area is obtained.

[0055] The effective width of mobile radar monitoring is set to L, and the minimum rectangle is filled with squares with a side length of L until the squares completely cover the minimum rectangle.

[0056] The squares are deleted and the squares that overlap with the mobile monitoring area are retained.

[0057] The center point of the reserved square is used as the necessary point for mobile radar monitoring. The greedy algorithm is used to generate the initial mobile monitoring path, which is then optimized using the genetic algorithm to obtain the mobile radar monitoring path.

[0058] According to a radar monitoring intelligent early warning method suitable for forest fire prevention provided by the present invention, in step S6, the monitoring data includes echo intensity data, Doppler velocity data, temperature data and particulate matter concentration data.

[0059] The process of building a prediction model based on support vector machines includes:

[0060] Collect historical radar monitoring data and perform feature extraction to obtain echo intensity characteristics, Doppler velocity characteristics, temperature characteristics and particle concentration characteristics.

[0061] A support vector machine model was constructed, and the echo intensity characteristics, Doppler velocity characteristics, temperature characteristics and particulate matter concentration characteristics were taken as input, and the forest fire prediction value was taken as output. The support vector machine model was trained, and the model parameters that met the test accuracy were retained to obtain the prediction model.

[0062] Input the current monitoring data into the prediction model and output the current forest fire prediction value.

[0063] The current forest fire prediction value is compared with the preset forest fire threshold. If the forest fire prediction value is greater than the forest fire threshold, a forest fire warning is issued in a timely manner.

[0064] On the other hand, the present invention also provides a radar monitoring intelligent early warning system suitable for forest fire prevention, comprising:

[0065] The three-dimensional geographic information modeling module is used to establish a three-dimensional geographic information model of the target forest, mark the trajectory and frequency of human activities, and divide the target forest into N areas to be detected.

[0066] The data collection and preprocessing module is used to collect forest fire factor data of each area to be detected, and to perform standardization on the forest fire factor data to obtain standardized data.

[0067] The fire risk assessment module is used to build a forest fire risk assessment model based on convolutional neural networks, and use the Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model. It inputs standardized data and outputs the fire risk level of each area to be detected. The fire risk level includes high risk level and low risk level.

[0068] The radar monitoring layout optimization module is used to conduct radar monitoring of the target forest, including using fixed monitoring radars to monitor high-risk areas, obtaining the locations of fixed monitoring radars and optimizing their number. Mobile monitoring is performed on low-risk areas according to preset cycles, and mobile monitoring paths are planned.

[0069] The intelligent early warning module is used to receive monitoring data from fixed monitoring radars and mobile monitoring radars in real time, and to build a prediction model based on support vector machines. It inputs monitoring data, outputs forest fire prediction values, and compares them with preset forest fire thresholds. If the forest fire prediction value is greater than the forest fire threshold, a forest fire warning will be issued in a timely manner.

[0070] The radar monitoring intelligent early warning method and system for forest fire prevention provided by the present invention, by constructing a forest fire assessment model based on a convolutional neural network and optimizing it using the Hippo algorithm, the performance of the model is further improved, and intelligent regional fire risk assessment is performed in combination with multi-source data, and the detection area is finely divided, so that targeted monitoring measures can be taken to improve monitoring efficiency. Fixed monitoring radars can monitor high-risk areas around the clock with high time resolution to ensure continuous monitoring of high-risk level areas. Regular inspections of low-risk areas using mobile radars can greatly reduce the number of fixed radar deployments, thereby reducing overall monitoring investment and reducing forest fire early warning costs. The combination of the two monitoring methods makes up for the limitations of a single monitoring method, can make full use of their respective advantages, and achieve comprehensive and efficient monitoring of the target forest area. At the same time, by fusing the monitoring data of fixed radars and mobile radars, more abundant input information can be provided for the intelligent prediction model, which can more accurately identify fire hazards and improve the accuracy and reliability of early warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0072] Figure 1 It is a flow chart of a radar monitoring intelligent early warning method applicable to forest fire prevention provided by an embodiment of the present invention;

[0073] Figure 2 It is a structural diagram of a radar monitoring intelligent early warning system suitable for forest fire prevention provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] Combine the following Figure 1-Figure 2 The present invention describes the radar monitoring intelligent early warning method and system suitable for forest fire prevention.

[0076] Figure 1 It is a flow chart of a radar monitoring intelligent early warning method suitable for forest fire prevention provided in an embodiment of the present invention.

[0077] like Figure 1 As shown, the radar monitoring intelligent early warning method and system for forest fire prevention provided in the embodiment of the present invention can be executed by the radar monitoring intelligent early warning method for forest fire prevention, and the method mainly includes the following steps:

[0078] S1. Build a three-dimensional map of the target forest, mark the trajectory and frequency of human activities, and divide the target forest into N areas to be detected.

[0079] In this embodiment, laser radar is used to comprehensively scan and map the target forest to obtain high-precision three-dimensional terrain data. Monitoring equipment is deployed inside the forest to collect spatial trajectory and frequency data of human activities.

[0080] The collected 3D terrain data is imported into GIS to construct a 3D digital map of the target forest. The trajectory and frequency distribution of human activities are marked on the 3D map to intuitively reflect the degree of human interference in different areas.

[0081] Use the grid analysis tool of GIS software to grid the target forest according to the set N value, and obtain the center coordinates of each area to be detected.

[0082] S2. Collect forest fire factor data of each area to be detected, the forest fire factor data includes vegetation data, meteorological data and external data, and standardize the factor data to obtain standardized data.

[0083] Vegetation data includes plant type data and plant density data, meteorological data includes ambient temperature data and ambient humidity data, and external data includes human activity data.

[0084] S3. Build a forest fire risk assessment model based on convolutional neural network, use Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model, input standardized data, and output the fire risk level of each area to be detected. The fire risk level includes high risk level and low risk level.

[0085] The process of building a forest fire risk assessment model based on convolutional neural networks includes:

[0086] Collect historical forest fire factor data and standardize them to obtain standardized historical data, which are then divided into training set and validation set.

[0087] Feature extraction is performed on the standardized data to obtain vegetation characteristics, meteorological characteristics and human activity characteristics.

[0088] A convolutional neural network model was constructed, with vegetation characteristics, meteorological characteristics, and human activity characteristics as input and fire risk level as output. The convolutional neural network model was trained, and the model parameters that met the test accuracy were retained to obtain a forest fire risk assessment model.

[0089] The process of optimizing the hyperparameters of the forest fire risk assessment model using the Hippo algorithm includes:

[0090] A set of initial hippo positions are randomly generated, and each hippo individual represents a hyperparameter combination, which includes the number of convolution kernels, learning rate, and number of layers.

[0091] For each hippopotamus individual, the accuracy is calculated on the validation set and used as the fitness value.

[0092] Update the hippopotamus position, the update formula is:

[0093]

[0094] in, and are the positions of the i-th hippopotamus individual in the j-th dimension at the t+1th and tth iterations, respectively. is the weight that controls the approach to the optimal individual, is the weight controlling random exploration. is the position of the best hippopotamus in the current iteration in the jth dimension, is the position of a hippopotamus randomly selected from the population in the jth dimension.

[0095] Repeatedly update the hippo position and calculate the fitness value until the preset number of iterations is reached.

[0096] The hyperparameter combination represented by the hippopotamus position with the highest accuracy was selected as the hyperparameter combination for the forest fire risk assessment model.

[0097] S4. Carry out fixed radar monitoring of high-risk areas. According to the monitoring radius of the fixed monitoring radar, obtain the deployment location of the fixed monitoring radar required to monitor all high-risk areas, and optimize the required number of fixed monitoring radars.

[0098] The process of obtaining the location of each fixed monitoring radar includes:

[0099] Set the set of center coordinates of all high-risk areas to be P = {p1, p2, ..., p T}, including T high-risk areas to be detected. The effective monitoring radius of the fixed monitoring radar is R.

[0100] S411, initialize an empty set C for storing the positions of the placed fixed monitoring radars, and initialize the unmonitored high-risk level area set U to the set P.

[0101] S412: Select a high-risk level area i in the set U, and calculate the distance d from the high-risk level area i to the center coordinates of all other high-risk level areas in the set U. ij , high-risk area j traverses all areas in set U except i, with a distance d ij The calculation formula is expressed as:

[0102]

[0103] In the formula, (x i ,y i , z i ) represents the center coordinate of the ith high-risk area, (x j ,y j , z j ) represents the center coordinates of the jth high-risk level area.

[0104] S413. Calculate the maximum distance between the high-risk level area i and the center coordinates of all other high-risk level areas in the set U. The formula is:

[0105]

[0106] Where, d ij Represents the distance from the center coordinate of the i-th high-risk level area to the center coordinate of the j-th high-risk level area.

[0107] S414, if d imax ≤2R, then high-risk area i is taken as the placement location of a fixed monitoring radar and added to set C, and high-risk area i and all high-risk areas whose distance to high-risk area i is less than R are removed from set U.

[0108] S415, if d imax >2R, then select the high-risk area k closest to the high-risk area i in the set U, and calculate the midpoint coordinates O of the high-risk area i and the high-risk area k. ik , as a fixed monitoring radar placement location, join the set C, and change the distance to the midpoint coordinate O ik All high risk level areas not exceeding R are removed from the set U.

[0109] S416, repeat steps S412 to S415 until the set U is empty, the set C is the deployment location of each fixed monitoring radar, and the number of fixed monitoring radars is .

[0110] The process of optimizing the required number of fixed surveillance radars involves:

[0111] S421. Set the set C = {q1, q2, ..., q M}, the position coordinates of each fixed monitoring radar are (x m ,y m , z m ). Calculate the distance between each fixed monitoring radar position and the center coordinates of the high-risk level area. The formula is expressed as:

[0112]

[0113] In the formula, (x q ,y q , z q ) represents the position coordinates of the qth fixed monitoring radar, (x p ,y p , z p ) represents the center coordinates of the pth high-risk level area.

[0114] Set the set of high-risk areas that can be monitored by the fixed monitoring radar q, and the formula is expressed as:

[0115]

[0116] Where, d qp Indicates the distance between each fixed monitoring radar position and the center coordinates of the high risk level area.

[0117] S422. Initialize a flag set flag={1, 1, ..., 1}, where the number of elements 1 is M, and 1 indicates that each fixed monitoring radar is in an unremoved state.

[0118] S423, traverse the positions of M fixed monitoring radars, and calculate the position of the fixed monitoring radar q after removing the ath fixed monitoring radar a The remaining high-risk level areas that can be monitored by the fixed monitoring radar are expressed as follows:

[0119]

[0120] In the formula, q b Represents the bth fixed monitoring radar.

[0121] If C1=C, the fixed monitoring radar q is removed a If all high-risk areas can still be fully covered after q, flag (a) is set to 0, indicating that the fixed monitoring radar q can be removed a , and update M1=M-1. Until for all a, when flag(a)=1, there are , which means that it is no longer possible to remove any fixed monitoring radar that is not removed and maintain monitoring of all high-risk level areas.

[0122] S5. Merge all low-risk areas that are not monitored by fixed monitoring radars into mobile radar monitoring areas, obtain the boundary curve of the mobile radar monitoring area, plan the mobile monitoring path of the mobile radar monitoring area, and use the mobile monitoring radar for mobile monitoring according to the preset period.

[0123] The process of merging all low-risk areas that are not monitored by fixed monitoring radars to obtain a mobile radar monitoring area and obtaining a boundary curve of the mobile radar monitoring area includes:

[0124] S51. Set the set of center coordinates of all low-risk areas to be H = {h1, h2, ..., h S}, including S low-risk areas to be detected.

[0125] S52: Calculate the distance d between the center coordinates of each low-risk level area and the position of each fixed monitoring radar. sq , if d sq <R, it means that the low risk level area is within the monitoring range of the fixed monitoring radar and is removed from the set H.

[0126] S53, merge all low-risk areas that are not monitored by fixed monitoring radars to form area G, perform multi-point sampling at the boundary of area G, and obtain n sampling points with coordinates (x n ,y n ).

[0127] S54, in the horizontal coordinate interval The boundary curve is established inside, and the formula is expressed as:

[0128]

[0129] In the formula, x represents the horizontal coordinate of the current point, Indicates The horizontal coordinate of the node, , , and is the coefficient to be determined.

[0130] Set the boundary condition to natural boundary condition, that is, the second-order derivative of the boundary curve at both ends is zero. At the point where two adjacent boundary curves have the formula and Its first-order derivative and second-order derivative are all continuous.

[0131] S55. Solve the coefficients using linear algebra , , and The value of is used to obtain the boundary curve of the mobile radar monitoring area.

[0132] Solving for the coefficients using linear algebra , , and The process includes:

[0133] Convert the interval boundary curve into a matrix, expressed as:

[0134]

[0135] Combine the interval boundary curves into a system of linear equations:

[0136]

[0137] In the formula, Representation interval The submatrix of the upper interval boundary curve coefficient constraints;

[0138] According to the interval boundary curve formula, we get the equation group:

[0139]

[0140] In the formula, , m n It is a quantity calculated based on boundary conditions and continuity conditions.

[0141] The matrix is ​​constructed by the equation group, and the matrix formula is expressed as:

[0142]

[0143] Solve the linear equations and get the coefficients , , and The value of is used to obtain the boundary curve of the mobile radar monitoring area.

[0144] The process of planning a mobile monitoring path for a mobile radar monitoring area includes:

[0145] According to the boundary curve, the minimum rectangle that can completely cover the mobile radar monitoring area is obtained.

[0146] The effective width of mobile radar monitoring is set to L, and the minimum rectangle is filled with squares with a side length of L until the squares completely cover the minimum rectangle.

[0147] The squares are deleted and the squares that overlap with the mobile monitoring area are retained.

[0148] The center point of the reserved square is used as the necessary point for mobile radar monitoring. The greedy algorithm is used to generate the initial mobile monitoring path, which is then optimized using the genetic algorithm to obtain the mobile radar monitoring path.

[0149] In this embodiment, mobile radar monitoring can be implemented by using drones, helicopters or forestry vehicles carrying radar equipment. The mobile monitoring cycle can be flexibly set according to seasonal changes. The climate and environmental conditions in different seasons will significantly affect the probability of forest fires and the intensity of their spread. Spring and summer are often the peak seasons for forest fires, and dry and high temperatures increase the risk of fires. Winter and rainy seasons are relatively low because the climate is relatively humid and the risk of vegetation burning is reduced.

[0150] According to the characteristics of fires in different seasons, the working cycle of the monitoring system needs to be adjusted accordingly. In high-risk seasons, the monitoring frequency needs to be increased, while in low-risk seasons, the monitoring frequency can be appropriately reduced to save energy and management costs.

[0151] S6. Receive the monitoring data from fixed monitoring radars and mobile monitoring radars in real time, and build a prediction model based on support vector machines. Input the monitoring data, output the forest fire prediction value, and compare it with the preset forest fire threshold. If the forest fire prediction value is greater than the forest fire threshold, issue a forest fire warning in a timely manner.

[0152] The monitoring data include echo intensity data, Doppler velocity data, temperature data and particle concentration data.

[0153] The process of building a prediction model based on support vector machines includes:

[0154] Collect historical radar monitoring data and perform feature extraction to obtain echo intensity characteristics, Doppler velocity characteristics, temperature characteristics and particle concentration characteristics.

[0155] A support vector machine model was constructed, and the echo intensity characteristics, Doppler velocity characteristics, temperature characteristics and particulate matter concentration characteristics were taken as input, and the forest fire prediction value was taken as output. The support vector machine model was trained, and the model parameters that met the test accuracy were retained to obtain the prediction model.

[0156] Input the current monitoring data into the prediction model and output the current forest fire prediction value.

[0157] The current forest fire prediction value is compared with the preset forest fire threshold. If the forest fire prediction value is greater than the forest fire threshold, a forest fire warning is issued in a timely manner.

[0158] The echo intensity reflects the amount of energy reflected by the object encountered by the radar beam. A stronger echo may indicate the presence of a larger object or dense material, such as thick smoke, hot gas masses, etc. Doppler velocity determines the moving speed of the target relative to the radar by measuring the frequency change of the echo signal. It helps to monitor wind direction and wind speed because the rising hot air flow generated by the fire and the flow of the surrounding air will produce specific speed characteristics. High temperature areas may indicate that a fire has occurred or is about to occur. The smoke produced by the fire contains a large amount of particulate matter, and the radar can monitor its concentration to a certain extent.

[0159] In this embodiment, according to seasonal changes, the threshold parameters of forest fire warning need to be dynamically adjusted. In the high-incidence season of fire, the threshold is appropriately lowered to improve the monitoring sensitivity. In the low-risk season, the threshold is appropriately increased to reduce the probability of false alarms. In the high-risk season, all units need to work closely together to prepare emergency plans and linkage mechanisms. Once the prediction model finds abnormal signs, the emergency response should be initiated quickly to shorten the disposal time.

[0160] In summary, this embodiment provides a radar monitoring intelligent early warning method and system suitable for forest fire prevention. By constructing a forest fire assessment model based on a convolutional neural network and optimizing it using the Hippo algorithm, the performance of the model is further improved. By combining multi-source data for intelligent regional fire risk assessment, the detection area is finely divided, and targeted monitoring measures can be taken to improve monitoring efficiency. Fixed monitoring radars can monitor high-risk areas around the clock with high temporal resolution to ensure continuous monitoring of high-risk areas. For low-risk areas, regular inspections using mobile radars can greatly reduce the number of fixed radar deployments, thereby reducing overall monitoring investment and reducing forest fire early warning costs. Fixed monitoring radars can provide real-time data with high temporal resolution, which is conducive to timely discovery of fire hazards and shortening early warning response time. The flexibility of mobile monitoring radars can adjust the monitoring path and frequency in a timely manner according to actual conditions, thereby improving the temporal and spatial coverage of monitoring. The combination of the two monitoring methods realizes comprehensive three-dimensional monitoring of the entire target forest area, makes up for the limitations of a single monitoring method, and can make full use of their respective advantages to achieve comprehensive and efficient monitoring of the target forest area. At the same time, by fusing fixed radar and mobile radar monitoring data, richer input information can be provided for the intelligent prediction model, which can more accurately identify fire hazards and improve the accuracy and reliability of early warning.

[0161] Based on the same general inventive concept, the present invention also protects a radar monitoring intelligent early warning system suitable for forest fire prevention. The radar monitoring intelligent early warning system suitable for forest fire prevention provided by the present invention is described below. The radar monitoring intelligent early warning system suitable for forest fire prevention described below and the radar monitoring intelligent early warning method suitable for forest fire prevention described above can be referenced to each other.

[0162] Figure 2 It is a structural diagram of a radar monitoring intelligent early warning system suitable for forest fire prevention provided by an embodiment of the present invention.

[0163] like Figure 2 As shown in the figure, the radar monitoring intelligent early warning system suitable for forest fire prevention includes:

[0164] The three-dimensional geographic information modeling module is used to establish a three-dimensional geographic information model of the target forest, mark the trajectory and frequency of human activities, and divide the target forest into N areas to be detected.

[0165] The data collection and preprocessing module is used to collect forest fire factor data of each area to be detected, and to perform standardization on the forest fire factor data to obtain standardized data.

[0166] The fire risk assessment module is used to build a forest fire risk assessment model based on convolutional neural networks, and use the Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model. It inputs standardized data and outputs the fire risk level of each area to be detected. The fire risk level includes high risk level and low risk level.

[0167] The radar monitoring layout optimization module is used to conduct radar monitoring of the target forest, including using fixed monitoring radars to monitor high-risk areas, obtaining the locations of fixed monitoring radars and optimizing their number. Mobile monitoring is performed on low-risk areas according to preset cycles, and mobile monitoring paths are planned.

[0168] The intelligent early warning module is used to receive monitoring data from fixed monitoring radars and mobile monitoring radars in real time, and to build a prediction model based on support vector machines. It inputs monitoring data, outputs forest fire prediction values, and compares them with preset forest fire thresholds. If the forest fire prediction value is greater than the forest fire threshold, a forest fire warning will be issued in a timely manner.

[0169] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A radar monitoring intelligent early warning method suitable for forest fire prevention, characterized in that: include: S1. Establish a three-dimensional map of the target forest, mark the trajectory and frequency of human activities on the three-dimensional map, and divide the target forest into N areas to be detected; S2, collecting forest fire factor data of each of the areas to be detected, wherein the forest fire factor data includes vegetation data, meteorological data and external data, and performing standardization processing on the factor data to obtain standardized data; S3, constructing a forest fire risk assessment model based on a convolutional neural network, and using the Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model, inputting the standardized data, and outputting the fire risk level of each of the areas to be detected, wherein the fire risk level includes a high risk level and a low risk level; S4. Perform fixed radar monitoring on the high-risk level areas, obtain the deployment positions of the fixed monitoring radars required to monitor all the high-risk level areas according to the monitoring radius of the fixed monitoring radars, and optimize the required number of fixed monitoring radars; The process of obtaining the location of each fixed monitoring radar includes: Set the set of center coordinates of all high-risk areas to be P = {p1, p2, ..., p T }, the effective monitoring radius of the fixed monitoring radar is R; S411, initializing an empty set C, and initializing the set U of all high-risk level areas to the set P; S412, selecting a high-risk level area i in the set U, and calculating the distance between the high-risk level area i and the center coordinates of all other high-risk level areas in the set U; S413, calculating the maximum distance between the high-risk level area i and the center coordinates of all other high-risk level areas in the set U, which is expressed as: ; Where, d ij represents the distance from the center coordinate of the i-th high-risk level area to the center coordinate of the j-th high-risk level area; S414, if d imax ≤2R, the high-risk area i is taken as the placement location of a fixed monitoring radar and added to the set C, and the high-risk area i and all high-risk areas whose distance to the high-risk area i is less than R are removed from the set U; S415, if d imax >2R, then select the high-risk area k closest to the high-risk area i in the set U, calculate the midpoint coordinates of the high-risk area i and the high-risk area k, add it to the set C as the placement position of a fixed monitoring radar, and remove all high-risk areas whose distance from the midpoint coordinates is no more than R from the set U; S416, repeat steps S412 to S415 until the set U is empty, the set C is the deployment location of each fixed monitoring radar, and the number of fixed monitoring radars is ; The process of optimizing the required number of fixed surveillance radars involves: S421. Set the set C = {q1, q2, ..., q M }, the position coordinates of each fixed monitoring radar are (x m ,y m , z m ); calculate the distance between the position of each fixed monitoring radar and the center coordinates of the high-risk level area; and set the set of high-risk level areas that can be monitored by the fixed monitoring radar q, the formula is expressed as: ; Where, d qp Indicates the distance between each fixed monitoring radar position and the center coordinates of the high risk level area; S422, initializing a flag set flag={1, 1, ..., 1}, where the number of elements 1 is M, and 1 indicates that each fixed monitoring radar is in an unremoved state; S423, traverse the positions of M fixed monitoring radars, and calculate the position of the fixed monitoring radar q after removing the ath fixed monitoring radar a The remaining high-risk level areas that can be monitored by the fixed monitoring radar are expressed as follows: ; In the formula, q b Indicates the bth fixed monitoring radar. If C1=C, set flag (a) to 0, indicating that the fixed monitoring radar q can be removed. a , and update M1=M-1 until for all a, when flag(a)=1, there are ; S5. Merge all low-risk areas that are not monitored by fixed monitoring radars into mobile radar monitoring areas, obtain the boundary curve of the mobile radar monitoring area, plan the mobile monitoring path of the mobile radar monitoring area, and use the mobile monitoring radar for mobile monitoring according to a preset period; S6. Receive the monitoring data of the fixed monitoring radar and the mobile monitoring radar in real time, and build a prediction model based on a support vector machine, input the monitoring data, output a forest fire prediction value, and compare it with a preset forest fire threshold. If the forest fire prediction value is greater than the forest fire threshold, issue a forest fire warning in a timely manner.

2. The radar monitoring intelligent early warning method for forest fire prevention according to claim 1 is characterized in that: The vegetation data includes plant type data and plant density data; the meteorological data includes ambient temperature data and ambient humidity data; and the external data includes human activity data.

3. The radar monitoring intelligent early warning method for forest fire prevention according to claim 1 is characterized in that: In step S3, the process of constructing a forest fire risk assessment model based on a convolutional neural network includes: Collect historical forest fire factor data, perform standardization processing, obtain standardized historical data, and divide the standardized data into a training set and a validation set; Extracting features from the standardized data to obtain vegetation features, meteorological features, and human activity features; A convolutional neural network model is constructed, the vegetation characteristics, meteorological characteristics and human activity characteristics are taken as input, and the fire risk level is taken as output. The convolutional neural network model is trained, and the model parameters that meet the test accuracy are retained to obtain a forest fire risk assessment model.

4. The radar monitoring intelligent early warning method for forest fire prevention according to claim 3 is characterized in that: The process of optimizing the hyperparameters of the forest fire risk assessment model using the Hippo algorithm includes: Randomly generate a set of initial hippo positions, each hippo individual represents a hyperparameter combination, and the hyperparameter combination includes the number of convolution kernels, learning rate and number of layers; For each hippopotamus individual, the accuracy is calculated on the validation set, and the accuracy is used as the fitness value; Update the hippopotamus position, the update formula is: ; in, and are the positions of the i-th hippopotamus individual in the j-th dimension at the t+1th and tth iterations respectively; is the weight that controls the approach to the optimal individual, is the weight controlling random exploration; is the position of the best hippopotamus in the current iteration in the jth dimension, is the position of a hippopotamus randomly selected from the population in the jth dimension; Repeatedly update the hippopotamus position and calculate the fitness value until the preset number of iterations is reached; The hyperparameter combination represented by the hippopotamus position with the highest accuracy was selected as the hyperparameter combination for the forest fire risk assessment model.

5. The radar monitoring intelligent early warning method for forest fire prevention according to claim 1 is characterized in that: In step S5, the process of obtaining the boundary curve of the mobile radar monitoring area includes: S51, set the set of center coordinates of all low-risk level areas to H = {h1, h2, ..., h S }; S52: Calculate the distance d between the center coordinates of each low-risk level area and the position of each fixed monitoring radar. sq , if d sq <R, then remove the low risk level area from the set H; S53, merge all low-risk areas that are not monitored by fixed monitoring radar into area G, collect n sampling points at the boundary of area G, and the coordinates of the sampling points are (x n ,y n ); S54, in the horizontal coordinate interval The interval boundary curve is established inside, and the formula is expressed as: ; In the formula, x represents the horizontal coordinate of the current point, Indicates The horizontal coordinate of the node, , , and is the coefficient to be determined; Set the boundary condition to natural boundary condition, at the node At the point where two adjacent boundary curves have the formula and and its first-order derivative and second-order derivative are continuous; S55. Solve the coefficients using linear algebra , , and The value of is used to obtain the boundary curve of the mobile radar monitoring area.

6. The radar monitoring intelligent early warning method for forest fire prevention according to claim 1 is characterized in that: The process of planning a monitoring path for the mobile radar monitoring area includes: According to the boundary curve, a minimum rectangle capable of completely covering the mobile radar monitoring area is obtained; The effective width of the mobile radar monitoring is set to L, and the minimum rectangle is filled with a square with a side length of L until the square completely covers the minimum rectangle; The squares are deleted, and the squares having overlapping areas with the mobile monitoring area are retained; The center point of the reserved square is used as the necessary point for mobile radar monitoring. The greedy algorithm is used to generate the initial mobile monitoring path, which is then optimized using the genetic algorithm to obtain the mobile radar monitoring path.

7. The radar monitoring intelligent early warning method for forest fire prevention according to claim 1 is characterized in that: In step S6, the monitoring data includes echo intensity data, Doppler velocity data, temperature data and particle concentration data; The process of building a prediction model based on support vector machines includes: Collect historical radar monitoring data and perform feature extraction to obtain echo intensity features, Doppler velocity features, temperature features, and particle concentration features; Constructing a support vector machine model, taking the echo intensity feature, Doppler velocity feature, temperature feature and particle concentration feature as input, taking the forest fire prediction value as output, training the support vector machine model, retaining model parameters that meet the test accuracy, and obtaining a prediction model; Inputting current monitoring data into the prediction model and outputting current forest fire prediction values; The current forest fire prediction value is compared with a preset forest fire threshold value. If the forest fire prediction value is greater than the forest fire threshold value, a forest fire warning is issued in a timely manner.

8. The radar monitoring intelligent early warning system applicable to forest fire prevention adopts the radar monitoring intelligent early warning method applicable to forest fire prevention as described in any one of claims 1 to 7, characterized in that: The radar monitoring intelligent early warning system suitable for forest fire prevention includes: The three-dimensional geographic information modeling module is used to establish a three-dimensional geographic information model of the target forest, mark the trajectory and frequency of human activities, and divide the target forest into N areas to be detected; The data collection and preprocessing module is used to collect the forest fire factor data of each area to be detected, and to perform standardization processing on the forest fire factor data to obtain standardized data; The fire risk assessment module is used to build a forest fire risk assessment model based on a convolutional neural network, and use the Hippo algorithm to optimize the hyperparameters of the forest fire risk assessment model. It inputs standardized data and outputs the fire risk level of each area to be detected. The fire risk level includes high risk level and low risk level. The radar monitoring layout optimization module is used to conduct radar monitoring of the target forest, including using fixed monitoring radars to monitor high-risk areas, obtaining the locations of fixed monitoring radars and optimizing their number; conducting mobile monitoring of low-risk areas according to preset cycles and planning mobile monitoring routes; The intelligent early warning module is used to receive monitoring data from fixed monitoring radars and mobile monitoring radars in real time, and to build a prediction model based on support vector machines. It inputs monitoring data, outputs forest fire prediction values, and compares them with preset forest fire thresholds. If the forest fire prediction value is greater than the forest fire threshold, a forest fire warning will be issued in a timely manner.

Images