A fire risk dynamic assessment and fire patrol path optimization method and system

Abstract

This application belongs to the field of fire patrol route optimization technology, specifically providing a method and system for dynamic fire risk assessment and fire patrol route optimization. The method mainly includes: acquiring multi-source observation data of the assessment unit; establishing a hidden Markov model based on historical multi-source observation data; generating a first probability value for the assessment unit being in a preset first risk state and a second probability value for being in a preset second risk state based on real-time multi-source observation data using the hidden Markov model; adjusting the probability parameters in the hidden Markov model according to real-time weather data meeting preset weather conditions; and regenerating the second probability value to obtain the optimized patrol route. This application enables the proactive identification of areas where risk increases sharply due to the superposition of weather and anomalies, generating patrol routes that prioritize coverage of such areas. This allows the deployment of patrol forces to match the dynamically changing high-risk distribution, significantly improving the accuracy and timeliness of proactive prevention and control.

Description

Technical Field

[0001] This application belongs to the field of fire patrol route optimization technology, and specifically relates to a method and system for dynamic fire risk assessment and fire patrol route optimization. Background Technology

[0002] In the field of urban fire safety management, in order to improve the effectiveness of preventive patrols, patrol routes covering multiple areas are usually planned for fire patrol forces.

[0003] The current patrol routes are mainly planned in advance based on relatively fixed factors such as the geographical location, building density, and population of each area, or patrol each area in turn according to a fixed schedule. During the patrol, the patrol personnel mainly rely on their personal experience to observe whether there are any fire hazards.

[0004] However, fire risk is influenced by both real-time environmental observation and immediate weather conditions. Using the patrol route planning method described above, it is difficult to proactively identify areas with abnormal signs and rapidly increasing risks in dry, windy weather conditions that are prone to fires. Consequently, it is difficult to adjust patrol priorities in a timely manner. High-risk areas, due to lack of focus, are highly likely to become breeding grounds for fire hazards. This makes it difficult to effectively realize the preventative value of fire patrols, hindering the formation of early warnings and effective interception of sudden fires, and significantly reducing the overall prevention and control effectiveness. Summary of the Invention

[0005] This application provides a method and system for dynamic fire risk assessment and fire patrol route optimization, which effectively solves the problem that existing patrol route planning methods are unable to actively identify areas with abnormal signs and rapidly increasing risks in dry, windy, and other fire-prone weather conditions. This enables the deployment of patrol forces to match the dynamically changing high-risk distribution, significantly improving the accuracy and timeliness of proactive prevention and control.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] Firstly, this application provides a method for dynamic fire risk assessment and fire patrol route optimization, including:

[0008] The process involves acquiring multi-source observation data for an assessment unit, which is a pre-defined area based on geographic information. This multi-source data includes real-time environmental sound data, real-time image data, and real-time weather data collected from patrol vehicles. A Hidden Markov Model (HMM) is established based on historical multi-source observation data. Based on real-time multi-source observation data, the HMM generates a first probability value indicating the assessment unit is in a preset first risk state and a second probability value indicating it is in a preset second risk state. The process determines whether the real-time weather data meets preset weather conditions and whether the second probability value exceeds a preset statistical threshold. If so, the probability parameter for transitioning from the first risk state to the second risk state in the HMM is adjusted upwards based on the second probability value, and the second probability value is regenerated based on the adjusted HMM. Finally, path planning is performed based on the second probability value to obtain an optimized patrol route.

[0009] Furthermore, determining whether the real-time weather data meets preset weather conditions includes: determining whether the real-time weather data meets a first weather condition or a second weather condition; determining whether the real-time weather data meets the first weather condition includes: determining whether the temperature value in the real-time weather data is higher than a preset first temperature threshold and the humidity value is lower than a preset humidity threshold; if so, the preset weather condition is met; determining whether the real-time weather data meets the second weather condition includes: determining whether the temperature value in the real-time weather data is lower than a preset second temperature threshold, the humidity value is lower than a humidity threshold, and the wind speed value is higher than a preset wind speed threshold; if so, the preset weather condition is met.

[0010] Furthermore, the statistical threshold is determined through the following steps: using a hidden Markov model, calculate the historical probability statistical baseline value of the assessment unit being in the second risk state; based on real-time weather data, calculate a dynamic adjustment value for adjusting the historical probability statistical baseline value; subtract the absolute value of the dynamic adjustment value from the historical probability statistical baseline value to obtain the dynamic historical statistical threshold.

[0011] Furthermore, based on real-time weather data, a dynamic adjustment value is calculated to adjust the historical probability statistical benchmark value, including: if the real-time weather data meets the first weather condition, the historical average temperature and standard deviation of the assessment unit are obtained, and the dynamic adjustment value is calculated based on the temperature value in the real-time weather data and the historical average temperature and standard deviation; if the real-time weather data meets the second weather condition, the historical average wind speed and standard deviation of the assessment unit are obtained, and the dynamic adjustment value is calculated based on the wind speed value in the real-time weather data and the historical average wind speed and standard deviation.

[0012] Furthermore, the probability parameter of the transition from the first risk state to the second risk state in the Hidden Markov Model is adjusted upward based on the second probability value, including: determining the adjustment coefficient based on the weather conditions satisfied by real-time weather data; calculating the upward adjustment amount of the probability parameter of the transition from the first risk state to the second risk state in the Hidden Markov Model based on the adjustment coefficient and the second probability value; and adding the upward adjustment amount to the probability parameter to obtain the adjusted probability parameter.

[0013] Furthermore, the adjustment coefficient is determined based on the weather conditions met by the real-time weather data, including: if the real-time weather data meets the first weather condition, then the preset first coefficient is determined as the adjustment coefficient; if the real-time weather data meets the second weather condition, then the preset second coefficient is determined as the adjustment coefficient; wherein, the first coefficient is greater than the second coefficient.

[0014] Furthermore, the adjustment amount of the probability parameter for the transition from the first risk state to the second risk state in the Hidden Markov Model is calculated based on the adjustment coefficient and the second probability value, including: using the product of the adjustment coefficient and the second probability value as the adjustment amount of the probability parameter for the transition from the first risk state to the second risk state in the Hidden Markov Model.

[0015] Furthermore, path planning is performed based on the second probability value to obtain an optimized patrol route, including: obtaining the length of time that the evaluation unit has not been patrolled; generating the priority of the evaluation unit based on the second probability value and the length of time it has not been patrolled; and generating an optimized patrol route based on the priority of each evaluation unit and preset patrol constraints.

[0016] Furthermore, based on the second probability value and the length of time without patrolling, the priority of the evaluation unit is generated, including: calculating a time penalty factor based on the length of time without patrolling, the time penalty factor being proportional to the length of time without patrolling; multiplying the second probability value by the time penalty factor to obtain the priority score of the evaluation unit; and sorting the priority scores of each evaluation unit from high to low to obtain the priority of each evaluation unit.

[0017] Secondly, this application provides a dynamic fire risk assessment and fire patrol route optimization system, including:

[0018] Data acquisition module: used to acquire multi-source observation data of the evaluation unit. The evaluation unit is a region pre-divided based on geographic information. The multi-source observation data includes real-time environmental sound data, real-time image data and real-time weather data collected from the patrol vehicle.

[0019] State assessment module: It is used to build a hidden Markov model based on historical multi-source observation data, and based on real-time multi-source observation data, it uses the hidden Markov model to generate a first probability value of the assessment unit being in a preset first risk state and a second probability value of the unit being in a preset second risk state.

[0020] Dynamic adjustment module: Used to determine whether the real-time weather data meets the preset weather conditions and whether the second probability value exceeds the preset statistical threshold. If so, the probability parameter of the transition from the first risk state to the second risk state in the Hidden Markov Model is increased according to the second probability value, and the second probability value is regenerated according to the adjusted Hidden Markov Model.

[0021] Path planning module: Used to plan paths based on the second probability value to obtain optimized patrol routes.

[0022] Thirdly, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the steps of a dynamic fire risk assessment and fire patrol route optimization method.

[0023] The beneficial effects of this application are:

[0024] This application employs a hidden Markov model-based approach to fuse multi-source observation data for risk assessment. It also proactively increases the risk state transition probability under severe weather conditions to dynamically enhance risk assessment, thereby optimizing patrol routes. This effectively addresses the problem that existing patrol route planning methods struggle to proactively identify areas with abnormal signs and rapidly increasing risks during dry, windy, and fire-prone weather conditions. It enables proactive identification of areas where risk surges due to the combined effects of weather and anomalies, generating patrol routes that prioritize these areas. This allows patrol force deployment to match dynamically changing high-risk distributions, significantly improving the accuracy and timeliness of proactive prevention and control.

[0025] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description and the accompanying drawings. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0027] Figure 1 A flowchart illustrating the dynamic fire risk assessment and fire patrol route optimization method of this application is shown.

[0028] Figure 2This paper illustrates a flowchart of the process for determining whether real-time weather data meets preset weather conditions.

[0029] Figure 3 A schematic diagram of the calculation process for the dynamic adjustment value in this application is shown;

[0030] Figure 4 A schematic diagram of the path planning process based on the second probability value of this application is shown. Detailed Implementation

[0031] To address the problems raised in the background technology, this application employs a hidden Markov model-based approach to fuse multi-source observation data to assess risk, and proactively increases the risk state transition probability under severe weather conditions to dynamically strengthen risk assessment. Based on this, patrol routes are optimized, significantly improving the accuracy and timeliness of proactive prevention and control.

[0032] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] In some embodiments, such as Figure 1 As shown, this application provides a method for dynamic fire risk assessment and fire patrol route optimization, including:

[0034] S1. Acquire multi-source observation data for the evaluation unit. The multi-source observation data includes real-time environmental sound data, real-time image data, and real-time weather data collected from the patrol vehicle.

[0035] The assessment unit is a pre-defined area based on geographic information, such as a city block, an industrial park, or a residential community.

[0036] Real-time ambient sound data represents the raw audio signals collected from the environment by a high-sensitivity directional microphone array installed on a fire patrol vehicle. After acquiring the raw audio signals, a fast Fourier transform can be performed on the raw audio signals to extract the energy intensity of specific frequency bands as sound feature values, which are used to quantify abnormal noises in the environment, such as continuous electric arcing, wood cracking, or the roar of construction machinery.

[0037] Real-time image data represents raw video streams of streets captured by panoramic cameras mounted on fire patrol vehicles. Image frames can be extracted from the raw video stream at fixed time intervals, and pre-trained convolutional neural network models are used to analyze the image frames, identify and quantify image feature values ​​such as the degree of occupancy of fire lanes and the proportion of debris accumulation in the image.

[0038] Real-time weather data represents the current temperature, humidity, and wind speed values ​​of the corresponding assessment unit obtained by accessing meteorological data services through an application programming interface (API).

[0039] The sound feature values, image feature values, temperature values, humidity values, and wind speed values ​​are aligned according to a unified timestamp to form a multi-source observation data vector. .

[0040] S2. Establish a hidden Markov model based on historical multi-source observation data. Based on real-time multi-source observation data, use the hidden Markov model to generate the first probability value of the evaluation unit being in a preset first risk state and the second probability value of it being in a preset second risk state.

[0041] Collect multi-source observation data vectors for this evaluation unit over a relatively long period, such as the past year, denoted as . T represents the total number of historical moments.

[0042] Hidden Markov models assume that the internal risk state of the assessment unit is an invisible, discrete random variable, i.e., a hidden state. Two hidden states are presupposed. This represents the highest risk level, indicating the presence of potential hazards. This represents a second risk level, such as a high-risk level, and The risk level is higher than The parameters of a Hidden Markov Model are defined as follows: A represents the state transition probability matrix, and B represents the observation probability matrix. This represents the probability distribution of the initial state.

[0043] The state transition probability matrix A is 2 2 matrices, , This represents the probability of transitioning from hidden state i to hidden state j, where i and j represent indices, i=1,2, j=1,2. , Let represent the hidden states of the evaluation unit at time t and t+1 in the discrete time series, respectively. The observation probability matrix B contains the probability of observing a certain multi-source observation data vector o in each hidden state. For each state... Its observation probability density function is denoted as Implemented using Gaussian mixture model , M represents a preset positive integer, indicating the number of Gaussian components in the Gaussian mixture model. The mixing weights represent the m-th Gaussian component, indicating the state. The probability that an observation is generated by the m-th component is given below. Let represent the mean vector of the m-th Gaussian component, defining the center position of this component in the feature space. Let represent the covariance matrix of the m-th Gaussian component, describing the variance of this component in each dimension of the feature space and the correlation between them. The representative mean is The covariance matrix is The probability density function value of the multivariate Gaussian distribution at point o, and the initial state probability distribution. For a two-dimensional vector [ , ], This represents the state at the beginning of the sequence. The prior probability.

[0044] The forward-backward algorithm is used to learn the model parameters. The goal is to find a set of parameters λ such that the probability of generating a historical observation sequence O is given by these parameters. To maximize the convergence, the algorithm iteratively executes the following two steps:

[0045] With the current parameter λ fixed, and given the entire sequence O, calculate the state at any historical time t. The probability of is denoted as . Calculate the state at any two adjacent times t and t+1. and The probability of is denoted as . The state at time t is calculated. And the multi-source observation data vector at that moment The probability of being generated by the m-th Gaussian component is denoted as . .

[0046] use , , To update the model parameter λ:

[0047] Update state transition probability : , representing from state Transferred to The number of times accounted for from An estimated proportion of the total number of departures.

[0048] Update the initial state distribution : , representing the probability of the sequence being in each state at the beginning.

[0049] Update observation probability density function The parameters, i.e., the parameters for updating the Gaussian mixture model { , , }, during the update, with As a weighting coefficient: , , , .

[0050] After the iteration is completed, the optimal model parameters are obtained. The model training is complete.

[0051] Using a trained model and the latest observational data, determine the current risk status, including the following steps:

[0052] Maintain a sliding time window of fixed length L. At the current time t, extract the multi-source observation data vectors from the L most recent times within the window to form a real-time observation sequence. .

[0053] will sequence and the trained model The input is fed into the forward algorithm, which calculates the probability of being in each hidden state at the final time step, given the observation sequence. Forward variables are defined. , representing the middle moment So far, we have seen a partial observation sequence and are currently in a certain state. The joint probability density is calculated recursively: sequence From the starting moment Initially, using the initial state probability and in state The first data observed probability density Initialize the forward variable as Then, for each subsequent moment... Recursively calculate the forward variables for the next time step: The recursive calculation continues until the final time step. ,get and .

[0054] Finally, the state of the evaluation unit at the current time t is calculated through normalization. Posterior probability: And output the first probability value accordingly. Second probability value , This reflects the statistical confidence level that the assessment unit is currently in a second-risk state based on all recent multi-source observation data.

[0055] S3. Determine whether the real-time weather data meets the preset weather conditions and whether the second probability value exceeds the preset statistical threshold. If so, adjust the probability parameter of the transition from the first risk state to the second risk state in the Hidden Markov Model according to the second probability value, and regenerate the second probability value according to the adjusted Hidden Markov Model.

[0056] The system checks whether both conditions are true simultaneously. If they are true, it triggers model parameter adjustment, resulting in an updated model. Using the model For the same real-time observation sequence Perform the online real-time inference again to obtain an updated second probability value. If neither of the two conditions is true at the same time, then execute S4 directly.

[0057] S4. Path planning is performed based on the second probability value to obtain the optimized patrol route.

[0058] In some embodiments, such as Figure 2 As shown, determining whether real-time weather data meets preset weather conditions includes:

[0059] Sa3.1. Determine whether the real-time weather data meets the first weather condition or the second weather condition. The first weather condition can represent high temperature and dry conditions, while the second weather condition can represent low temperature, dry conditions and strong wind conditions.

[0060] The determination of whether the real-time weather data meets the first weather condition includes: determining whether the temperature value in the real-time weather data is higher than a preset first temperature threshold. And the humidity value is lower than the preset humidity threshold. If so, then the preset weather conditions are met. First temperature threshold. Temperature can be 35℃, humidity threshold It could be 30%.

[0061] Determining whether real-time weather data meets the second weather condition includes: determining whether the temperature value in the real-time weather data is lower than a preset second temperature threshold. Humidity value is below the humidity threshold And the wind speed value is higher than the preset wind speed threshold. If so, then the preset weather conditions are met. Second temperature threshold. It can be 5℃, wind speed threshold. It can be 10.8 m / s.

[0062] In some embodiments, the statistical threshold is determined by the following steps:

[0063] Sb3.1. Using a hidden Markov model, calculate the historical probability statistical baseline value of the assessment unit being in the second risk state.

[0064] Specifically, from the historical database, M sets of historical observation sequences for the same period in history are retrieved for this assessment unit, each set of sequences Represents a certain period in history Data at any given moment This represents the length of the historical sequence. For each historical sequence... Using the pre-trained original model Calculate its corresponding second probability value according to the steps in S2, denoted as... For all M histories The arithmetic mean is calculated to obtain the historical probability statistical benchmark value B, which represents the average risk probability level of the assessment unit in the typical historical period.

[0065] Sb3.2. Based on real-time weather data, calculate a dynamic adjustment value D for adjusting the historical probability statistics baseline.

[0066] Sb3.3. Subtract the absolute value of the dynamic adjustment value D from the historical probability statistical baseline value B to obtain the dynamic historical statistical threshold. ,Right now This means that the worse the weather, the lower the judgment threshold, and the more sensitive the system is to risk.

[0067] In some embodiments, such as Figure 3 As shown, based on real-time weather data, a dynamic adjustment value D is calculated to adjust the historical probability statistics baseline value, including: if the real-time weather data meets the first weather condition, then the temperature value... Exceeding the first temperature threshold The difference is used as the dynamic adjustment value D, referring to the formula: If the real-time weather data meets the weather conditions for the second day, then the wind speed value will be... Exceeding the first wind speed threshold The difference is used as a dynamic adjustment value, referring to the formula: .

[0068] Based on real-time weather data, a dynamic adjustment value D is calculated to adjust the historical probability statistical benchmark value. This includes: if the real-time weather data meets the first weather condition, obtaining the historical average and standard deviation of the temperature for the same period of the evaluation unit, and adjusting the temperature value in the real-time weather data. The dynamic adjustment value is calculated based on the historical average temperature and standard deviation for the same period, using the following formula: , This represents obtaining the historical average temperature for the same period in the evaluation unit. This represents the standard deviation of temperature for the same historical period in this assessment unit; if the real-time weather data meets the conditions for the second day, then the historical mean and standard deviation of wind speed for the same historical period in this assessment unit are obtained, based on the wind speed values ​​in the real-time weather data. The dynamic adjustment value is calculated based on the historical average wind speed and standard deviation for the same period, using the following formula: (Refer to the formula below) , This represents obtaining the historical average wind speed for the same period in the evaluation unit. This represents the standard deviation of wind speed for the same historical period in the assessment unit.

[0069] In some embodiments, adjusting the probability parameter of the transition from the first risk state to the second risk state in the hidden Markov model based on the second probability value includes:

[0070] Sc3.1. Determine the adjustment coefficient based on the weather conditions met by the real-time weather data.

[0071] Adjustment coefficient include , If the first weather condition is met, then the adjustment coefficient is set to... If the second weather condition is met, then set the adjustment coefficient. According to basic fire science, high temperature and dryness generally have a stronger promoting effect on fires than low temperature and strong winds. ,For example 1.5 is acceptable. 1.2 is acceptable.

[0072] Sc3.2. Calculate the up-adjustment of the probability parameter for the transition from the first risk state to the second risk state in the Hidden Markov Model based on the adjustment coefficient and the second probability value.

[0073] The formula for calculating the upward adjustment is as follows: , This represents an upward adjustment in volume and a higher probability of risk. The higher the elevation, the worse the weather, and the greater the upward adjustment. The larger.

[0074] Sc3.3. Add the adjustment amount to the probability parameter to obtain the adjusted probability parameter.

[0075] Read the state transition probability matrix A of the Hidden Markov Model of the evaluation unit from storage and locate its elements. , indicating from state Transition to state The probability of is denoted as . ,Will With upward adjustment Add them together to get the initial adjusted result. ,Right now Since the probability values ​​have a domain of [0,1] and the sum of probabilities in each row is 1, after boundary clipping, we obtain... , ,because ,exist After being modified, it needs to be recalculated. To ensure the sum is 1, the adjusted... ,use and Replacing the old values ​​in matrix A yields a new state transition matrix. The updated model is .

[0076] In some embodiments, such as Figure 4 As shown, path planning is performed based on the second probability value to obtain the optimized patrol route, including:

[0077] S4.1. Obtain the length of time the evaluation unit has not been patrolled.

[0078] Specifically, maintain a database table to record a timestamp field for each evaluation unit k. Each time a patrol vehicle completes an on-site inspection of an assessment unit, it writes the current time into that unit's data. Read the current time. For each evaluation unit k within the target area, its value is read from the database. Calculate the time difference , , This refers to the length of time that the evaluation unit k has not been patrolled.

[0079] S4.2. Based on the second probability value and the length of time the area has not been patrolled. Generate the priority of the evaluation unit, and use the priority score of the evaluation unit. express.

[0080] S4.3. Generate an optimized patrol path based on the priority of each evaluation unit and the preset patrol constraints.

[0081] Obtain the priority order list and the geographical coordinates of each evaluation unit. And preset patrol constraints, such as the coordinates of the patrol vehicle's starting point. The maximum driving distance allowed for a single patrol mission Looking for a way from Depart, visit several evaluation unit points, and then return. Given a travel path R, the optimization objective is to maximize the sum of priority scores of the units visited along the path. The constraint is the total travel distance of the route. .

[0082] First, initialize the current path. The current cumulative path distance d=0, the initialized visited set is empty, and each evaluation unit that has not yet been added to the path is evaluated in descending order of priority. Perform a greedy insertion operation, for the current evaluation unit to be inserted. Try to insert it into every possible position in the current path R except the starting point, and calculate the additional travel distance required for each insertion. The total distance after insertion from all satisfied values Among the insertion schemes, select the one that makes The largest insertion position will be the evaluation unit. Insert at that position, update path R and cumulative distance, and mark this evaluation cell as visited, until all cells have been traversed, or no cell can be found without violating the maximum distance constraint. Insert the current path under the premise of setting the starting point Add it to the end of the path to form a complete closed-loop patrol route.

[0083] In some embodiments, the priority of the evaluation unit is generated based on the second probability value and the length of time it has not been patrolled, including:

[0084] S4.2.1. Based on the length of time not patrolled Calculate the time penalty factor Time penalty factor The length of time not patrolled Proportional Possible forms: , This represents a small, positive coefficient used to control the intensity of the time effect, for example... 0.1 is acceptable.

[0085] S4.2.2. Combine the second probability value with the time penalty factor Multiply by the product to obtain the priority score of the evaluation unit. Reference formula: , This represents the second probability value of the evaluation unit.

[0086] S4.2.3. Sort the priority scores of each evaluation unit from high to low to obtain the priority of each evaluation unit.

[0087] Collect the priority scores of all K units to be planned and evaluated. , , , The global patrol priority of each assessment unit is obtained by sorting them in descending order from high to low.

[0088] In some embodiments, a fire risk dynamic assessment and fire patrol route optimization system includes:

[0089] Data acquisition module: used to acquire multi-source observation data of the evaluation unit. The evaluation unit is a region pre-divided based on geographic information. The multi-source observation data includes real-time environmental sound data, real-time image data and real-time weather data collected from the patrol vehicle.

[0090] State assessment module: It is used to build a hidden Markov model based on historical multi-source observation data, and based on real-time multi-source observation data, it uses the hidden Markov model to generate a first probability value of the assessment unit being in a preset first risk state and a second probability value of the unit being in a preset second risk state.

[0091] Dynamic adjustment module: Used to determine whether the real-time weather data meets the preset weather conditions and whether the second probability value exceeds the preset statistical threshold. If so, the probability parameter of the transition from the first risk state to the second risk state in the Hidden Markov Model is increased according to the second probability value, and the second probability value is regenerated according to the adjusted Hidden Markov Model.

[0092] Path planning module: Used to plan paths based on the second probability value to obtain optimized patrol routes.

[0093] In some embodiments, this application provides a readable storage medium storing computer program instructions, which are read and executed by a processor to perform the steps of a dynamic fire risk assessment and fire patrol route optimization method.

[0094] It should be noted that, in this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0095] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory.

[0096] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for dynamic assessment of fire risk and optimization of fire patrol path, characterized in that, include: Acquire multi-source observation data for the evaluation unit, which is a region pre-divided based on geographic information. The multi-source observation data includes real-time environmental sound data, real-time image data, and real-time weather data collected from the patrol vehicle. A hidden Markov model is established based on historical multi-source observation data. Based on real-time multi-source observation data, the hidden Markov model is used to generate a first probability value of the evaluation unit being in a preset first risk state and a second probability value of being in a preset second risk state. Determine whether the real-time weather data meets the preset weather conditions and whether the second probability value exceeds the preset statistical threshold. If so, adjust the probability parameter of the transition from the first risk state to the second risk state in the Hidden Markov Model according to the second probability value, and regenerate the second probability value according to the adjusted Hidden Markov Model. Path planning is performed based on the second probability value to obtain an optimized patrol route.

2. The method of claim 1, wherein, Determining whether the real-time weather data meets preset weather conditions includes: Determine whether the real-time weather data meets the first or second weather condition; Determining whether real-time weather data meets the first weather condition includes: determining whether the temperature value in the real-time weather data is higher than the preset first temperature threshold and the humidity value is lower than the preset humidity threshold; if so, the preset weather condition is met. Determining whether real-time weather data meets the second weather condition includes: determining whether the temperature value in the real-time weather data is lower than the preset second temperature threshold, the humidity value is lower than the humidity threshold, and the wind speed value is higher than the preset wind speed threshold. If so, the preset weather condition is met.

3. The method of claim 2, wherein, The statistical threshold is determined through the following steps: Using the Hidden Markov Model, the historical probability statistical baseline value of the assessment unit being in the second risk state is calculated; Based on the real-time weather data, a dynamic adjustment value is calculated to adjust the historical probability statistical benchmark value. The dynamic historical statistical threshold is obtained by subtracting the absolute value of the dynamic adjustment value from the historical probability statistical benchmark value.

4. The method of claim 3, wherein, Based on the real-time weather data, a dynamic adjustment value is calculated to adjust the historical probability statistical benchmark value, including: If the real-time weather data meets the first weather condition, the historical average temperature and standard deviation of the same period for the assessment unit are obtained, and the dynamic adjustment value is calculated based on the temperature value in the real-time weather data and the historical average temperature and standard deviation. If the real-time weather data meets the conditions for the second day of the year, the historical average wind speed and standard deviation for the same period of the year for that assessment unit are obtained, and the dynamic adjustment value is calculated based on the wind speed value in the real-time weather data and the historical average wind speed and standard deviation for the same period of the year.

5. The method of claim 2, wherein, The probability parameters for transitioning from the first risk state to the second risk state in the Hidden Markov Model are adjusted upwards based on the second probability value, including: The adjustment coefficient is determined based on the weather conditions met by real-time weather data; The adjustment amount of the probability parameter for transitioning from the first risk state to the second risk state in the Hidden Markov Model is calculated based on the adjustment coefficient and the second probability value. The probability parameter is adjusted by adding an upward adjustment amount to obtain the adjusted probability parameter.

6. The method of claim 5, wherein, Adjustment factors are determined based on weather conditions met by real-time weather data, including: If the real-time weather data meets the first weather condition, then the preset first coefficient is determined as the adjustment coefficient; If the real-time weather data meets the second weather condition, then the preset second coefficient will be determined as the adjustment coefficient; The first coefficient is greater than the second coefficient.

7. The method of claim 5, wherein, The adjustment amount of the probability parameter for transitioning from the first risk state to the second risk state in the Hidden Markov Model is calculated based on the adjustment coefficient and the second probability value, including: The product of the adjustment coefficient and the second probability value is used as the upward adjustment of the probability parameter for the transition from the first risk state to the second risk state in the Hidden Markov Model.

8. The method of claim 1, wherein, Based on the second probability value, path planning is performed to obtain an optimized patrol route, including: Obtain the length of time the assessment unit has not been patrolled; Based on the second probability value and the length of time the unit has not been patrolled, a priority for the evaluation unit is generated. Based on the priority of each evaluation unit and the preset patrol constraints, an optimized patrol path is generated.

9. The method according to claim 8, characterized in that, Based on the second probability value and the length of time the unit has not been patrolled, a priority for the evaluation unit is generated, including: A time penalty factor is calculated based on the length of time the patrol is not conducted, and the time penalty factor is directly proportional to the length of time the patrol is not conducted. Multiply the second probability value by the time penalty factor to obtain the priority score of the evaluation unit; The priority of each evaluation unit is obtained by sorting the priority scores of each evaluation unit from high to low.

10. A dynamic fire risk assessment and fire patrol route optimization system, characterized in that, include: Data acquisition module: used to acquire multi-source observation data of the evaluation unit, which is a region pre-divided based on geographic information. The multi-source observation data includes real-time environmental sound data, real-time image data and real-time weather data collected from the patrol vehicle. State assessment module: used to establish a hidden Markov model based on historical multi-source observation data, and based on real-time multi-source observation data, use the hidden Markov model to generate a first probability value of the assessment unit being in a preset first risk state and a second probability value of being in a preset second risk state. Dynamic adjustment module: used to determine whether the real-time weather data meets the preset weather conditions and the second probability value exceeds the preset statistical threshold. If so, the probability parameter of the transition from the first risk state to the second risk state in the hidden Markov model is increased according to the second probability value, and the second probability value is regenerated according to the adjusted hidden Markov model. Path planning module: Used to plan the path based on the second probability value to obtain an optimized patrol route.

Images