Forest fire early warning method and system based on artificial intelligence

Abstract

The invention provides a forest fire early warning method and system based on artificial intelligence, and relates to the technical field of forest fire early warning, and the method comprises the steps: collecting forest fire data in real time, including environment, vegetation and external data, and obtaining historical data; analyzing the forest fire data by using a forward-backward algorithm to obtain fire factor data, performing classified statistics on historical data to obtain comburent data, and calculating fire associated factor data by means of an association rule algorithm; and calculating an ignition association combination value by adopting an aurora optimization algorithm, and determining a fire triggering threshold value through a threshold value setting method according to the fire triggering factor data. According to the fire association combination value, whether the fire association factor data exceeds a fire initiation threshold value is judged, if yes, a risk elimination mechanism is triggered, and if not, forest fire data are continuously collected. By integrating the forward-backward algorithm, the association rule algorithm and the aurora optimization algorithm, the advantages of all the algorithms are exerted, the complex data relation is precisely processed, and the reliability of the forest fire data is improved. And the occurrence of fire is restrained from the source.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire early warning, and in particular to an artificial intelligence-based forest fire early warning method and system. Background Art

[0002] Early fire warnings mainly rely on manual monitoring and experience judgment. Usually, forest firefighters and monitoring stations discover the source of fire through visual observation, patrols, etc. The limitations of this method are slow response speed, limited coverage, and high reliance on personnel experience and judgment. With the development of aerospace technology, remote sensing technology has been introduced into forest fire monitoring. Satellite imaging and aerial photography technology make large-scale monitoring possible, helping to identify fire risk areas. The wide range and accuracy of monitoring have been improved, but the efficiency of data processing and early warning is still limited. With the development of sensor technology and the Internet of Things, real-time data collection has become possible. By deploying environmental sensors (such as temperature, humidity, wind speed, etc.), fire risk factors can be monitored in real time, which greatly improves the efficiency of data collection and transmission, and can provide timely feedback on environmental changes, providing a basis for early warning.

[0003] Forest fire early warning methods refer to a series of technologies and means to identify and predict the possibility of forest fires in advance so as to take preventive measures or respond in time. These methods integrate knowledge and technologies from multiple disciplines, including meteorology, ecology, geographic information systems (GIS), remote sensing technology, the Internet of Things (IoT), artificial intelligence (AI), etc., and predict and prevent forest fires through real-time monitoring, data analysis and intelligent decision support. These systems rely on a variety of advanced technologies and tools to improve the accuracy and response time of fire warnings.

[0004] In the existing technology, only a single or simple algorithm is used to analyze the data, and multiple potential triggering factors are ignored. It is difficult to comprehensively consider the complex and multiple fire factors and early warning systems only take action after detecting high-risk signals. There is a lack of proactive intervention measures, and it is impossible to curb the occurrence of fires at the source. Summary of the invention

[0005] The present invention provides a forest fire warning method and system based on artificial intelligence, which are used to solve the defects in the prior art that multiple potential triggering factors are ignored when analyzing data, and it is difficult to comprehensively consider complex and multiple fire factors and the warning system only takes action after detecting high-risk signals, lacking active intervention measures.

[0006] On the one hand, the present invention provides a forest fire early warning method based on artificial intelligence, comprising: S1: Obtain forest fire data of the target forest area in real time, and obtain historical data based on databases and literature. Forest fire data includes environmental data, vegetation data and external data.

[0007] S2: Analyze forest fire data through the forward-backward algorithm to obtain data on fire-causing factors, classify and count historical data according to different dimensions to obtain combustible material data, and use the association rule algorithm to calculate the combustible material data and fire-causing factor data to obtain fire-related factor data.

[0008] S3: The fire-related combination value is obtained by calculating the environmental data, vegetation data and external data through the Aurora optimization algorithm, and the fire-causing threshold is obtained from the fire-causing factor data according to the threshold setting method.

[0009] S4: Determine whether the fire-related factor data exceeds the fire-causing threshold based on the fire-related combination value. If so, trigger the risk elimination mechanism; otherwise, continue to collect forest fire data in real time.

[0010] According to a forest fire early warning method based on artificial intelligence provided by the present invention, the step of obtaining data on fire-causing factors includes: Missing values, error values, and outliers in the forest fire data are removed to obtain cleaned data, and the fire intensity of the forest fire is divided into multiple fire stages, which are divided into the first stage, the second stage, the third stage, and the fourth stage.

[0011] Initialize the initial probability distribution of the fire stage, the state transition probability matrix, and the observation probability matrix, and calculate the forward probability and backward probability of the sequence in the cleaned data through the forward-backward algorithm.

[0012] The time probability and moment probability are calculated through forward probability and backward probability, and the initial probability distribution, state transition probability matrix and observation probability matrix are updated to obtain the key factor identification data.

[0013] The fire-causing factor data were obtained by analyzing the updated key factor identification data under multiple fire stages and observation values, including humidity, temperature and human activities.

[0014] According to a forest fire early warning method based on artificial intelligence provided by the present invention, the forward probability and the backward probability are expressed as follows:

[0015] In the formula is the number of fire stages, It's time The observed value of is a moment in the time series, is the index of the fire stage, Indicates transition from state To status The probability of The system is in the fire stage The probability of is time, based on the previous observations, the system is in the fire stage probability.

[0016] According to an artificial intelligence-based forest fire early warning method provided by the present invention, in step S2, the step of obtaining combustion data includes: Collect information on combustion residues during the fire stage and conduct preliminary classification into vegetation residues and non-vegetation residues. For vegetation residues, combine on-site investigation records and vegetation distribution maps to determine the plant species and plant parts. For non-vegetation residues, identify whether they contain flammable substances.

[0017] For vegetation residues, the burnt area and vegetation density are used to calculate the biomass of the burnt materials. For non-vegetation burnt materials, the non-biomass is calculated based on the volume or area of ​​the debris on site. The biomass and non-biomass are combined to obtain the total amount of burnt materials.

[0018] The completeness of combustion can be determined by inspecting the remaining unburned materials on site, and the intensity and speed of combustion can be determined by analyzing the temperature change curve of the recorder.

[0019] According to a forest fire early warning method based on artificial intelligence provided by the present invention, the step of obtaining fire-related factor data includes: S21: Discretize the combustible material data and the fire-causing factor data and convert them into the form used for the association rule algorithm. Suppose the combustible material data is B (b1, b2, ..., b m ), where m is the number of elements in the combustion material data, and the fire-causing factor data is F (f1, f2, ..., f n ), where n is the number of elements in the fire-causing factor data.

[0020] S22: Define the minimum support and minimum confidence as the thresholds of the frequent itemsets that need to be retained, find all frequent itemsets, and calculate the support G of each frequent item set.

[0021] S23: Generate all possible binomial combinations from frequent itemsets to construct candidate itemsets.

[0022] S24: Calculate the support T of each candidate item set and retain the item sets that meet the minimum support.

[0023] S25: Repeat steps S22 to S24 until the preset number of iterations is reached, generate corresponding association rules from the frequent item set, and calculate whether the confidence of the association rule is greater than the minimum confidence. If so, retain the association rule, and extract the factors closely related to the fire from the association rule as the fire-related factor data, otherwise delete the association rule.

[0024] According to an artificial intelligence-based forest fire early warning method provided by the present invention, in step S3, the step of obtaining the fire association combination value includes: S31: Aurora emission stage: Process missing values ​​and outliers in environmental data, vegetation data, and external data, and perform standardization to obtain comprehensive data. Let the environmental data be U (U1, U2, ..., U Q ), where Q represents the number of environmental data, vegetation data O (O1, O2, ..., O W ), where W represents the number of vegetation data, and the external data P (P1, P2, ..., P E ), where E represents the number of external data.

[0025] S32: Extract key features from the comprehensive data as optimization targets, randomly generate a set of aurora particles as ignition factor values, and define the fitness function. The aurora particles X include (c1, c2, ..., c j ), where j is the number of auroral particles.

[0026] S33: Radiation attenuation stage: Introduce the attenuation factor, and adjust the position update formula according to the attenuation factor to obtain the aurora position update formula.

[0027] S34: Update the position of the aurora particles using the aurora position update formula.

[0028] S35: Aurora scattering optimization stage: The fitness of each aurora particle is calculated according to the fitness function, the updated aurora particle position is determined according to the fitness, and the individual best particle and the global best particle are iteratively updated.

[0029] S36: Repeat steps S33 to S35 until the fitness change is less than a preset threshold, and select the ignition factor value with the highest fitness as the ignition association combination value.

[0030] According to an artificial intelligence-based forest fire early warning method provided by the present invention, in step S34, the aurora position update formula is expressed as:

[0031] In the formula, is the updated aurora particle position, is the current position of the aurora particle, is the current best position of the aurora particles, is the global best position, and is a random number between (0,1), is the attenuation factor.

[0032] According to an artificial intelligence-based forest fire early warning method provided by the present invention, in step S3, the step of obtaining a fire triggering threshold comprises: The data on fire-causing factors are organized into a table format, with each row representing a fire event and each column corresponding to a fire-causing factor.

[0033] The mean and standard deviation of each fire-causing factor are calculated, and the threshold range is determined based on the preset K value to obtain the fire-causing threshold.

[0034] The threshold range is determined based on the preset value and is divided into a lower threshold and an upper threshold.

[0035] According to a forest fire early warning method based on artificial intelligence provided by the present invention, the formulas of the lower threshold and the upper threshold are respectively:

[0036] In the formula, L is the lower threshold, U is the upper threshold, K is the preset constant, and u is the mean. is the standard deviation.

[0037] On the other hand, the present invention also provides a forest fire early warning system based on artificial intelligence, which adopts any of the above-mentioned forest fire early warning methods based on artificial intelligence, and the early warning system comprises: Data collection module: obtain forest fire data of the target forest area in real time, and obtain historical data based on database and literature. Forest fire data includes environmental data, vegetation data and external data.

[0038] Analysis module: Analyze forest fire data through forward-backward algorithm to obtain data on fire-causing factors, classify and count historical data according to different dimensions to obtain combustion material data, and use association rule algorithm to calculate combustion material data and fire-causing factor data to obtain fire-related factor data.

[0039] Risk judgment module: The fire-related combination value is calculated by the Aurora optimization algorithm on environmental data, vegetation data and external data, and the fire-causing threshold is obtained from the fire-causing factor data according to the threshold setting method.

[0040] Prevention and control module: Determine whether the fire-related factor data exceeds the fire-causing threshold based on the fire-related combination value. If so, trigger the risk elimination mechanism; otherwise, continue to collect forest fire data in real time.

[0041] The present invention provides an artificial intelligence-based forest fire warning method and system, which integrates the forward-backward algorithm, the association rule algorithm and the Aurora optimization algorithm, gives full play to the advantages of each algorithm, accurately processes complex data relationships, improves the accuracy and reliability of risk assessment, solves the limitations of single or simple combination algorithms and reduces the problems of false alarms and missed alarms in the warning system, accurately predicts the possibility and development trend of fire, discovers potential fire risks in advance, provides strong support for timely preventive measures, and effectively reduces the losses caused by fire.

[0042] The present invention provides an artificial intelligence-based forest fire early warning method and system, which sets dynamic thresholds based on fire-causing factor data, and quickly triggers prevention and control mechanisms or continuous monitoring by comparing the fire-related combination value with the threshold in real time, thereby optimizing the accuracy and timeliness of risk judgment and prevention and control, and solving the defects of fixed thresholds and delayed feedback in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] 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.

[0044] Figure 1 This is one of the flow charts of a forest fire early warning method based on artificial intelligence provided by an embodiment of the present invention; Figure 2 This is the second flow chart of a forest fire early warning method based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] 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.

[0046] Combine the following Figure 1-Figure 2 The present invention describes an artificial intelligence-based forest fire early warning method and system.

[0047] Figure 1 This is one of the flow charts of a forest fire early warning method based on artificial intelligence provided by an embodiment of the present invention.

[0048] like Figure 1 As shown, an embodiment of the present invention provides a forest fire early warning method based on artificial intelligence, and the method mainly includes the following steps: S1: Obtain forest fire data of the target forest area in real time, and obtain historical data based on databases and literature. Forest fire data includes environmental data, vegetation data and external data.

[0049] S2: Analyze forest fire data through the forward-backward algorithm to obtain data on fire-causing factors, classify and count historical data according to different dimensions to obtain combustible material data, and use the association rule algorithm to calculate the combustible material data and fire-causing factor data to obtain fire-related factor data.

[0050] In step S2, the step of obtaining fire-causing factor data includes: Missing values, error values, and outliers in the forest fire data are removed to obtain cleaned data, and the fire intensity of the forest fire is divided into multiple fire stages, which are divided into the first stage, the second stage, the third stage, and the fourth stage.

[0051] Initialize the initial probability distribution of the fire stage, the state transition probability matrix and the observation probability matrix, and calculate the forward probability and backward probability of the sequence in the cleaned data through the forward-backward algorithm. The formula is expressed as:

[0052] Where N is the number of fire stages, It's time The observed value of , t is the moment in the time series, is the index of the fire stage, Indicates transition from state To status The probability of The system is in the fire stage The probability of is time, based on the previous observations, the system is in the fire stage probability.

[0053] The time probability and moment probability are calculated through forward probability and backward probability, and the initial probability distribution, state transition probability matrix and observation probability matrix are updated to obtain the key factor identification data.

[0054] The fire-causing factor data were obtained by analyzing the updated key factor identification data under multiple fire stages and observation values, including humidity, temperature and human activities.

[0055] The steps to obtain combustion data include: Collect information on combustion residues during the fire stage and conduct preliminary classification into vegetation residues and non-vegetation residues. For vegetation residues, combine on-site investigation records and vegetation distribution maps to determine the plant species and plant parts. For non-vegetation residues, identify whether they contain flammable substances.

[0056] For vegetation residues, the burnt area and vegetation density are used to calculate the biomass of the burnt materials. For non-vegetation burnt materials, the non-biomass is calculated based on the volume or area of ​​the debris on site. The biomass and non-biomass are combined to obtain the total amount of burnt materials.

[0057] The completeness of combustion can be determined by inspecting the remaining unburned materials on site, and the intensity and speed of combustion can be determined by analyzing the temperature change curve of the recorder.

[0058] Figure 2 This is the second flow chart of a forest fire early warning method based on artificial intelligence provided by an embodiment of the present invention.

[0059] like Figure 2 As shown, the steps of obtaining fire-related factor data include: S21: Discretize the combustible material data and the fire-causing factor data and convert them into the form used for the association rule algorithm. Suppose the combustible material data is B (b1, b2, ..., b m ), where m is the number of elements in the combustion material data, and the fire-causing factor data is F (f1, f2, ..., f n ), where n is the number of elements in the fire-causing factor data.

[0060] S22: Define the minimum support and minimum confidence as the thresholds of the frequent itemsets that need to be retained, find all frequent itemsets, and calculate the support G of each frequent item set.

[0061] S23: Generate all possible binomial combinations from frequent itemsets to construct candidate itemsets.

[0062] S24: Calculate the support T of each candidate item set and retain the item sets that meet the minimum support.

[0063] S25: Repeat steps S22 to S24 until the preset number of iterations is reached, generate corresponding association rules from the frequent item set, and calculate whether the confidence of the association rule is greater than the minimum confidence. If so, retain the association rule, and extract the factors closely related to the fire from the association rule as the fire-related factor data, otherwise delete the association rule.

[0064] S3: The fire-related combination value is obtained by calculating the environmental data, vegetation data and external data through the Aurora optimization algorithm, and the fire-causing threshold is obtained from the fire-causing factor data according to the threshold setting method.

[0065] The steps to obtain the fire-related combination value include: S31: Aurora emission stage: Process missing values ​​and outliers in environmental data, vegetation data, and external data, and perform standardization to obtain comprehensive data. Let the environmental data be U (U1, U2, ..., U Q ), where Q represents the number of environmental data, vegetation data O (O1, O2, ..., O W ), where W represents the number of vegetation data, and the external data P (P1, P2, ..., P E ), where E represents the number of external data.

[0066] S32: Extract key features from the comprehensive data as optimization targets, randomly generate a set of aurora particles as ignition factor values, and define the fitness function. The aurora particles X include (c1, c2, ..., c j ), where j is the number of auroral particles.

[0067] The fitness function is expressed as:

[0068] In the formula, is the classification accuracy of a specific combination, is the support of the combination in the data set, is the confidence of the combination, , , is the weight coefficient.

[0069] S33: Radiation attenuation stage: Introduce the attenuation factor, and adjust the position update formula according to the attenuation factor to obtain the aurora position update formula.

[0070] S34: Update the position of the aurora particles using the aurora position update formula.

[0071] In step S34, the aurora position update formula is expressed as:

[0072] In the formula, is the updated aurora particle position, is the current position of the aurora particle, is the current best position of the aurora particles, is the global best position, and is a random number between (0,1), is the attenuation factor.

[0073] S35: Aurora scattering optimization stage: The fitness of each aurora particle is calculated according to the fitness function, the updated aurora particle position is determined according to the fitness, and the individual best particle and the global best particle are iteratively updated.

[0074] S36: Repeat steps S33 to S35 until the fitness change is less than a preset threshold, and select the ignition factor value with the highest fitness as the ignition association combination value.

[0075] The steps to obtain the fire triggering threshold include: The data on fire-causing factors are organized into a table format, with each row representing a fire event and each column corresponding to a fire-causing factor.

[0076] The mean and standard deviation of each fire-causing factor are calculated, and the threshold range is determined based on the preset K value to obtain the fire-causing threshold.

[0077] The formulas for mean and standard deviation are:

[0078] In the formula, It is observations of factors that cause fires, n is the number of data samples, u is the mean, is the standard deviation and u is the mean.

[0079] The threshold range is determined based on the preset value and is divided into a lower threshold and an upper threshold.

[0080] The formulas for the lower threshold value and the upper threshold value are:

[0081] In the formula, L is the lower threshold, U is the upper threshold, K is the preset constant, and u is the mean. is the standard deviation.

[0082] S4: Determine whether the fire-related factor data exceeds the fire-causing threshold based on the fire-related combination value. If yes, trigger the risk elimination mechanism, otherwise continue to collect forest fire data in real time. The system will continuously and accurately compare and deeply analyze the calculated fire-related combination value with the established fire-causing threshold, and use rigorous logic to determine whether the fire-related factor data exceeds the corresponding threshold range. Once it is determined that the data of fire-related factors exceeds the set fire-causing threshold, the system will quickly and decisively automatically trigger a complete set of comprehensive and efficient risk elimination mechanisms. This mechanism covers such as timely and accurate push of detailed fire risk alarm information to relevant management departments and personnel, and rapid launch of pre-set fire-fighting plans, including the deployment of fire-fighting resources, organizing professional fire-fighting teams to rush to possible fire scenes and other specific actions; and if, after careful judgment, the data of fire-related factors does not exceed the fire-causing threshold, the system will continue to perform the key task of real-time collection of forest fire data in the target forest area in an orderly manner, to ensure that the real-time status of the forest area can be continuously, dynamically and accurately grasped, thereby providing the latest and reliable data support for subsequent fire risk assessment and early warning, and maintaining the efficient operation and keen monitoring capabilities of the entire forest fire early warning system.

[0083] The present invention obtains data on factors causing fires by analyzing forest fire data using a forward-backward algorithm, collects data on combustible materials based on historical data, and uses an association rule algorithm to find data on ignition-related factors. In addition, the Aurora optimization algorithm is used to calculate ignition-related combination values, and a threshold value for causing fires is determined based on a threshold setting method. This method can more accurately identify fire risk factors and potential ignition associations, improve the accuracy and reliability of the model, reduce false alarms and missed alarms, and more accurately predict the possibility and development trend of fires, discover potential fire risks in advance, provide strong support for timely preventive measures, and effectively reduce the losses caused by fires.

[0084] Based on the same general inventive concept, the present invention also protects a forest fire early warning system based on artificial intelligence, which can adopt the above-mentioned forest fire early warning method based on artificial intelligence. The early warning system includes: Data acquisition module: real-time acquisition of forest fire data in the target forest area, and acquisition of historical data based on databases and literature, the forest fire data including environmental data, vegetation data and external data; Analysis module: Analyze the forest fire data by forward-backward algorithm to obtain fire-causing factor data, classify and count the historical data according to different dimensions to obtain combustion material data, and use association rule algorithm to calculate the combustion material data and the fire-causing factor data to obtain fire-related factor data; Risk judgment module: calculates the environmental data, the vegetation data and the external data through the Aurora optimization algorithm to obtain a fire-related combination value, and obtains a fire-causing threshold from the fire-causing factor data according to a threshold setting method; Prevention and control module: determine whether the fire-related factor data exceeds the fire-causing threshold according to the fire-related combination value, and if so, trigger the risk elimination mechanism; otherwise, continue to collect the forest fire data in real time.

[0085] The present invention integrates the forward-backward algorithm, the association rule algorithm and the Aurora optimization algorithm, gives full play to the advantages of each algorithm, accurately processes complex data relationships, improves the accuracy and reliability of risk assessment, and makes up for the limitations of single or simple combination algorithms. Dynamic thresholds are set based on fire-causing factor data, and the prevention and control mechanism is quickly triggered or continuous monitoring is carried out by comparing the fire-related combination value and the threshold in real time, thereby optimizing the accuracy and timeliness of risk judgment and prevention and control, and solving the defects of fixed thresholds and delayed feedback in the prior art.

[0086] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0087] 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.

[0088] 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 forest fire early warning method based on artificial intelligence, characterized in that: include: S1: Acquire forest fire data of the target forest area in real time, and acquire historical data based on databases and literature, wherein the forest fire data includes environmental data, vegetation data, and external data; S2: Analyze the forest fire data by a forward-backward algorithm to obtain fire-causing factor data, classify and count the historical data according to different dimensions to obtain combustion material data, and use an association rule algorithm to calculate the combustion material data and the fire-causing factor data to obtain fire-related factor data; S3: calculating the environmental data, the vegetation data and the external data by using the Aurora optimization algorithm to obtain a fire-related combination value, and obtaining a fire-causing threshold value from the fire-causing factor data according to a threshold setting method; S4: judging whether the fire-related factor data exceeds the fire-causing threshold value according to the fire-related combination value, if yes, triggering the risk elimination mechanism, otherwise continuing to collect the forest fire data in real time.

2. The artificial intelligence-based forest fire early warning method according to claim 1, characterized in that: In step S2, the step of obtaining the fire-causing factor data includes: Remove missing values, error values ​​and outliers in the forest fire data to obtain cleaned data, and divide the fire intensity of the forest fire into multiple fire stages, wherein the fire stages are divided into the first stage, the second stage, the third stage and the fourth stage; Initializing the initial probability distribution, state transition probability matrix and observation probability matrix of the fire stage, and calculating the forward probability and backward probability of the sequence in the cleaned data by the forward-backward algorithm; Calculating the time probability and the moment probability through the forward probability and the backward probability and updating the initial probability distribution, the state transition probability matrix and the observation probability matrix to obtain key factor identification data; The fire-causing factor data is obtained by analyzing the updated key factor identification data under multiple fire stages and observation values, wherein the observation values ​​include humidity, temperature and human activities.

3. The artificial intelligence-based forest fire early warning method according to claim 2, characterized in that: The forward probability and the backward probability are expressed as follows: ; In the formula is the number of fire stages, It's time The observed value of is a moment in the time series, is the index of the fire stage, Indicates transition from state To status The probability of The system is in the fire stage The probability of is time, based on the previous observations, the system is in the fire stage probability.

4. The artificial intelligence-based forest fire early warning method according to claim 2, characterized in that: In step S2, the step of obtaining the combustion material data includes: Collect information on combustion residues at the fire stage and perform preliminary classification into vegetation residues and non-vegetation residues. For vegetation residues, determine the plant species and plant parts in combination with on-site survey records and vegetation distribution maps; for non-vegetation residues, identify whether they contain flammable substances; For the vegetation residues, the burnt area and vegetation density are used to calculate the biomass of the burnt material. For the non-vegetation burnt material, the non-biomass is calculated according to the volume or area of ​​the on-site debris. The biomass and the non-biomass are combined to obtain the total amount of burnt material information. The completeness of combustion can be determined by inspecting the remaining unburned materials on site, and the intensity and speed of combustion can be determined by analyzing the temperature change curve of the recorder.

5. The artificial intelligence-based forest fire early warning method according to claim 1, characterized in that: In step S2, the step of obtaining the fire-related factor data includes: S21: Discretize the combustion material data and the fire-causing factor data and convert them into a form for association rule algorithm. Suppose the combustion material data is B (b1, b2, ..., b m ), where m is the number of elements in the combustion material data, and the fire-causing factor data is F (f1, f2, ..., f n ), where n is the number of elements of the fire-causing factor data; S22: define the minimum support and the minimum confidence as the thresholds of the frequent itemsets that need to be retained, find all the frequent itemsets, and calculate the support G of each of the frequent itemsets; S23: generating all possible binomial combinations from the frequent itemsets to construct candidate itemsets; S24: Calculate the support T of each candidate item set, and retain the item sets that meet the minimum support; S25: Repeat steps S22 to S24 until a preset number of iterations is reached, generate corresponding association rules from the frequent item set, and calculate whether the confidence of the association rule is greater than the minimum confidence. If so, retain the association rule, and extract the factors closely related to the fire from the association rule as the fire-related factor data; otherwise, delete the association rule.

6. The artificial intelligence-based forest fire early warning method according to claim 1, characterized in that: In step S3, the step of obtaining the ignition-related combination value includes: S31: Aurora emission stage: Process the missing values ​​and abnormal values ​​in the environmental data, the vegetation data and the external data, and perform standardization to obtain comprehensive data. Suppose the environmental data is U (U1, U2, ..., U Q ), where Q represents the number of the environmental data, the vegetation data O (O1, O2, ..., O W ), where W represents the number of vegetation data, the external data P (P1, P2, ..., P E ), where E represents the number of the external data; S32: extract key features from the comprehensive data as optimization targets, randomly generate a set of aurora particles as ignition factor values, and define a fitness function, wherein the aurora particles X include (c1, c2, ..., c j ), where j is the number of auroral particles; S33: Radiation attenuation stage: introducing an attenuation factor, and adjusting the position update formula according to the attenuation factor to obtain the aurora position update formula; S34: using the aurora position update formula to update the position of the aurora particle; S35: Aurora scattering optimization stage: calculating the fitness of each aurora particle according to the fitness function, determining the updated aurora particle position according to the fitness, and iteratively updating the individual best particle and the global best particle; S36: Repeat steps S33 to S35 until the fitness change is less than a preset threshold, and select the ignition factor value with the highest fitness as the ignition association combination value.

7. The artificial intelligence-based forest fire early warning method according to claim 6, characterized in that: In step S34, the aurora position update formula is expressed as: ; In the formula, is the updated aurora particle position, is the current position of the aurora particle, is the current best position of the aurora particles, is the global best position, and is a random number between (0,1), is the attenuation factor.

8. The artificial intelligence-based forest fire early warning method according to claim 1, characterized in that: In step S3, the step of obtaining the fire triggering threshold comprises: Arrange the fire-causing factor data into a table, where each row represents a fire event and each column corresponds to a fire-causing factor; Calculating the mean and standard deviation of each of the fire-causing factors, and determining a threshold range based on a preset K value to obtain a fire-causing threshold; The threshold range is determined based on the preset value and is divided into a lower threshold and an upper threshold.

9. The artificial intelligence-based forest fire early warning method according to claim 8, characterized in that: The formulas for the lower threshold and the upper threshold are respectively: ; In the formula, is the lower threshold, is the upper threshold, is a preset constant, is the mean, is the standard deviation.

10. An artificial intelligence-based forest fire early warning system, which adopts an artificial intelligence-based forest fire early warning method as claimed in any one of claims 1 to 9, characterized in that: The early warning system includes: Data acquisition module: real-time acquisition of forest fire data in the target forest area, and acquisition of historical data based on databases and literature, the forest fire data including environmental data, vegetation data and external data; Analysis module: Analyze the forest fire data by forward-backward algorithm to obtain fire-causing factor data, classify and count the historical data according to different dimensions to obtain combustion material data, and use association rule algorithm to calculate the combustion material data and the fire-causing factor data to obtain fire-related factor data; Risk judgment module: calculates the environmental data, the vegetation data and the external data through the Aurora optimization algorithm to obtain a fire-related combination value, and obtains a fire-causing threshold from the fire-causing factor data according to a threshold setting method; Prevention and control module: determine whether the fire-related factor data exceeds the fire-causing threshold according to the fire-related combination value, and if so, trigger the risk elimination mechanism; otherwise, continue to collect the forest fire data in real time.

Images