A safety prevention and control design system and method based on artificial intelligence

Abstract

The present invention belongs to the technical field of rodent control, and discloses a safety control design system and method based on artificial intelligence; the system comprises: arranging n sensor nodes in an area to be controlled to build an Internet of Things sensor network, and collecting environmental data and rodent activity data in real time based on the Internet of Things sensor network; fusing the environmental data and the rodent activity data to obtain comprehensive rodent data; building a rodent detection model, and using the rodent detection model to analyze the comprehensive rodent data to obtain location activity data of rodent activities; building a rodent virtual model, and using the rodent virtual model to predict change trend data based on environmental data and rodent activity data; using an optimization algorithm to solve the optimal rodent poison placement strategy according to the location activity data and the change trend data; and sending the optimal rodent poison placement strategy to an on-site execution terminal to place the rodent poison, thereby improving the control efficiency and reducing the labor cost.

Description

Technical Field

[0001] The present invention relates to the technical field of rodent control, and more specifically, to an artificial intelligence-based safety control design system and method. Background Art

[0002] The patent application publication number CN116092006A discloses a digital monitoring and intelligent rodent collection and identification system for agricultural and forestry pests, including: a box body, a first camera and a second camera. The box body is buried underground, and the top plate of the box body is provided with a transparent window. The first camera is placed above the top plate to capture the overhead angle image of the detection target, and the second camera is placed in the box body and below the top plate to capture the upward angle image of the detection target. It can reduce the difficulty of accurate identification of pests and rodents, and is conducive to correctly grasping the development trend of the rodent situation.

[0003] However, traditional rodent monitoring methods often rely on manual inspections or simple trap counting, which lacks real-time and comprehensiveness, making it difficult to accurately grasp the rodent situation in a timely manner. For example, in large granaries or farmlands, manual inspections are often time-consuming and labor-intensive, and rodents in hidden areas are easily missed. Secondly, existing systems generally lack intelligence and automation, and are unable to dynamically adjust prevention and control strategies based on environmental changes and rodent activities, resulting in delayed or inappropriate prevention and control measures. For example, when seasons change or the weather changes drastically, the activity patterns of rodents may change significantly, and traditional methods are difficult to respond quickly. Furthermore, it is difficult for existing systems to fully cover and effectively monitor large areas, and there are monitoring blind spots. This problem is particularly prominent in complex environments such as urban communities or large factories, which often makes it difficult to control the spread of rodent infestations. In addition, it is often time-consuming, labor-intensive and inefficient to take prevention and control measures in advance or to blindly carry out prevention and control without knowing the dynamics of rodents. At the same time, it is difficult for existing systems to achieve optimal allocation of rodent control resources, which often leads to waste of resources or poor control effects, restricting the scientific nature and effectiveness of prevention and control strategies. These problems are intertwined in actual applications, seriously affecting the overall effect of rodent control.

[0004] In view of this, the present invention proposes a safety prevention design system and method based on artificial intelligence to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a safety prevention design system based on artificial intelligence, comprising: a network construction module, which is used to arrange n sensor nodes in the area to be controlled to build an Internet of Things sensor network, and collect environmental data and rodent activity data in real time based on the Internet of Things sensor network; n is a positive integer greater than 1;

[0006] The data fusion module is used to fuse the environmental data and rodent activity data to obtain comprehensive rodent damage data;

[0007] The preliminary model building module is used to build a rodent damage detection model, and use the rodent damage detection model to analyze the comprehensive rodent damage data to obtain the location and activity level data of rodent damage activities;

[0008] The virtual simulation module is used to build a virtual model of rodent damage. Based on the environmental data and rodent damage activity data, the virtual model of rodent damage is used to predict the change trend data of rodent damage activities in the future r time period;

[0009] The strategy optimization module is used to use the optimization algorithm to solve the optimal rodent poison placement strategy based on the location activity data and change trend data of rodent activities; the optimal rodent poison placement strategy is sent to the on-site execution terminal for the placement of rodent poison; and each module is connected by wired and / or wireless means.

[0010] Furthermore, the construction method of the IoT sensor network includes:

[0011] The area to be controlled is randomly divided into N sub-areas, and the decision variables are defined as the boundary coordinates of each sub-area and the node type corresponding to each sub-area; the objective function of regional division is defined And constraints; the constraints are that the area of ​​each sub-region is within the range of [A_min, A_max], the coverage radius of each node is R, and the total cost does not exceed the budget C; where A_min is the preset lower limit of the area, and A_max is the preset upper limit of the area;

[0012] Encoding the decision variables, using real number encoding or binary encoding; randomly generating an initial sponge population according to the encoding method, each sponge in the sponge population corresponds to a regional division scheme;

[0013] Calculate the objective function of each sponge in the region division The function value on is recorded as the comprehensive motion value corresponding to each sponge; based on the calculated comprehensive motion value, an iterative comprehensive optimization operation is performed, and the iterative comprehensive optimization operation includes a displacement operation, a deformation operation, and a fusion operation;

[0014] Repeat the iterative comprehensive optimization operation until the preset maximum number of iterations is reached, and decode the sponge with the smallest comprehensive motion value in the sponge population at this time, and the obtained regional division scheme is the optimal regional division scheme;

[0015] The area to be controlled is re-divided into sub-areas using the optimal area division scheme to obtain several sub-areas, and nodes of corresponding node types are set corresponding to the sub-areas based on the optimal area division scheme; that is, the construction of the Internet of Things sensor network is completed.

[0016] Furthermore, the environmental data includes environmental temperature data, environmental humidity data, environmental light intensity data and environmental CO 2 Concentration data; rodent activity data include sound data, distributed infrared data, on-site image / video data and movement status; distributed infrared data include the average number of hot spots per unit time, the rate of temperature change per unit time and the duration of abnormally high temperature; movement status includes movement intensity, movement frequency and movement duration.

[0017] Furthermore, the region division objective function ; in, , , and is the weight of the corresponding item, For the The weight coefficient of each sub-region, For the The number of nodes in each sub-region; For the The coverage quality score of each sub-area, For the The coverage area of ​​the sub-regions, is the total area of ​​the area to be controlled, For the Unit cost of each node type; For the The number of nodes of the node type; For the The average energy consumption of the node types, is the weight of network lifetime; For network life;

[0018] Coverage Quality Score ;in, For the The area coverage weight of each sub-region is and is an adjustable weight parameter; For the The average coverage radius of the sensor nodes deployed in each sub-area; It is the largest coverage radius among all sensor nodes; For the Rodent infestation density index for each sub-area;

[0019] ;in, For the The price of the node type, The longest expected service life of all node types. For the The expected service life of each node type, Adjust the weight for life to balance the importance of life. For the The coefficients corresponding to the cost influencing factors of each node are: For the Factors affecting node cost.

[0020] Furthermore, the displacement operation includes:

[0021] For each sponge , according to its comprehensive motion value and other spongy bodies Location , calculate its displacement vector ;

[0022] ; in, is a mixing coefficient, is a random number, is a positive number, For other spongy bodies The comprehensive sports value, Spongy original position; is a Gaussian noise term, is the noise intensity; It is the sponge with the smallest comprehensive motion value in the current sponge population;

[0023] Spongy body Along its displacement vector Displace to get a new position ;

[0024] The deformation operation includes:

[0025] For each sponge , randomly select one of the dimensions to deform and generate a deformation intensity factor ; in the In one dimension, the spongy Perform the transformation and keep the remaining dimensions unchanged;

[0026] The formula for the transformation is:

[0027] ;in, Spongy In the The new position in the dimension, Spongy In the The original position in the dimension; For the current sponge population, The location of the cavernous body with the largest integrated motion value in the dimension; is the nonlinear deformation index parameter; is the sign function; deformation intensity factor ;in, is the maximum value of the preset deformation strength, is the minimum value of the preset deformation strength, is the current iteration number, is the preset maximum number of iterations, is a positive adjustment parameter;

[0028] The fusion operation includes:

[0029] Randomly select two different sponges and Fusion is performed to generate a fusion scale factor ; Generate a new sponge by linear interpolation ;in, Spongy original position; Replace the sponge with the largest comprehensive movement value in the current sponge population.

[0030] Furthermore, the fusion processing method includes:

[0031] Standardize or normalize the environmental data and rodent activity data, and unify the format of heterogeneous data to obtain standard environmental data and standard rodent activity data;

[0032] The standard environmental data and standard rodent pest activity data are respectively feature encoded using autoencoders to obtain environmental feature vectors and rodent pest activity feature vectors; the environmental feature vectors and rodent pest activity feature vectors are vector-level concatenated to obtain a comprehensive high-dimensional feature vector, which is the comprehensive rodent pest data.

[0033] Furthermore, the method of constructing the rodent damage detection model includes:

[0034] Using convolutional neural network, recurrent neural network or a combination of convolutional neural network and recurrent neural network as the basic framework of rodent detection model;

[0035] The input of the rodent detection model is defined as comprehensive rodent damage data, and the output is the rodent damage activity score of each node. Several sets of annotated comprehensive rodent damage data are collected within a fixed period of time as training sets. The annotations include the rodent damage activity of each node. The rodent detection model is trained using the training set. During the training process, the loss function is defined. ; and use optimization algorithms such as stochastic gradient descent to update parameters; until the function value of the loss function is minimum and does not change for P consecutive times, the parameters of the rodent detection model at this time are fixed, that is, the rodent detection model is built;

[0036] Loss Function ; in, is the total number of nodes, For the The rodent infestation activity score predicted by the rodent infestation detection model for each node, For the The rodent activity level of each node; For the The weight of each node; is the weight of the spatial smoothing term, is the spatial smoothness index, is the time smoothing term weight, For the Nodes adjacent to a node The rodent infestation activity score predicted by the rodent infestation detection model, is the weight of the distribution consistency term, For the The distribution of rodent infestation activity predicted by the rodent infestation detection model for each node; For the The rodent infestation activity score predicted by the rodent infestation detection model for each node at the previous time step; For the The actual rodent activity distribution of each node; is the KL divergence.

[0037] Furthermore, the construction method of the rodent damage virtual model includes:

[0038] The rodent damage virtual model is a joint simulation of the cell sub-model and the rodent behavior sub-model; the cell sub-model is: discretizing the area to be controlled into M1 cell grids, M1 is a positive integer greater than 1; defining the cell state of each cell grid to be described by state variables, which include the number of rodents, food quantity, ambient temperature and ambient humidity;

[0039] The mouse behavior sub-model is: defining the update rules of cell states; the update rules include reproduction rules, foraging rules, activity rules, and migration rules;

[0040] The breeding rule is: define the breeding age threshold ag of mice, and regard mice older than the breeding age threshold ag as adult mice;

[0041] The reproduction probability of mice is calculated based on the reproduction function , the formula of the reproduction function is:

[0042] ; in, is the number of rodents per unit area, is the amount of food per unit area, is the ambient temperature, is the ambient humidity; , and To adjust the parameters, is the temperature suitability function, is the humidity suitability function at each time step, for each adult mouse, Generate new rats for probability;

[0043] The foraging rule is: define the perception radius of food for mice as Rf; mice perceive food within the circular range formed by the perception radius Rf; mice tend to move towards food, and the step length of movement is defined as ;in, The area with the largest amount of food within the Rf range. is the amount of food in the area where the rodent is currently located, To The distance of the area, is the rate coefficient, is a positive constant; is the crowding influence coefficient; and is the adjustment index;

[0044] The activity rule is: define the daytime activity intensity of rodents as and night activity intensity ; then the activity intensity of rodents per unit time ;in is the preset basic activity intensity. is the light intensity suitability function, is the light intensity;

[0045] Migration rules are: Define environmental quality score ;when When the environmental quality score is lower than the preset lower threshold, the rodents will migrate; the migration direction is toward the adjacent area where the environmental quality score is greater than the preset upper threshold. ;in The area with the highest environmental quality score among the neighboring areas of the current area where the rodents are located. Score the environmental quality of the current area where the rodent is located. The current area where the rodent is located The distance of the area, is the migration rate coefficient;

[0046] Collect historical data and actual monitoring data, assign initial values ​​to various parameters in the virtual model of rodent damage, use historical data as input, simulate the dynamics of rodent populations in each cell grid in the virtual model of rodent damage, compare the dynamics of rodent populations with the actual monitoring data, and continuously adjust the parameters of the virtual model of rodent damage to maximize the degree of fit between the dynamics of rodent populations and the actual monitoring data; and obtain the optimal combination of parameters of the virtual model of rodent damage;

[0047] Using the combination of optimal parameters of the rodent pest virtual model and taking the current real-time collected environmental data and rodent pest activity data as initial conditions, the rodent population dynamics in the future r time period are simulated in the rodent pest virtual model, and the rodent change trend data of each cell grid are extracted from the rodent population dynamics. The rodent change trend data is the growth rate of the rodent population; the rodent change trend data of the cell grid is used as the change trend data of rodent pest activity.

[0048] Furthermore, the solution to the optimal rat poison placement strategy includes:

[0049] Construct strategy variables and strategy objective functions; strategy variables include the type of rat poison, dosage of rat poison and location of rat poison;

[0050] For each cell grid, determine the corresponding strategy variable, that is, Line The type of rat poison required for the cell grid corresponding to the column , No. Line The dose of rat poison required for the cell grid corresponding to the column and Line Whether rat poison is placed in the cell grid corresponding to the column ;

[0051] Strategy objective function ;in, For the Line The rodent activity intensity of the cell grid corresponding to the column, For the Line The cost of placing rat poison per unit in the cell grid corresponding to the column, , and Balance weights for strategies; For the Line The mouse change trend data of the cell grid corresponding to the column, For the Line The sum of the rodent activity scores of the n1 nodes in the neighborhood of the cell grid corresponding to the column;

[0052] Encode the strategy variable into the gene sequence of an individual; randomly generate N3 individuals as the initial population, decode each individual, and obtain the corresponding rat poison placement strategy; substitute the rat poison placement strategy into the strategy objective function, and the calculated value is used as the individual's fitness score;

[0053] According to the size of the fitness score, the selection operator is used to select excellent individuals. Two parent individuals are randomly selected, and some genes are exchanged at a certain intersection to generate new offspring individuals. Some genes of the individuals are mutated with a fixed probability to generate a new generation of population until the iterative algebra reaches the preset upper limit to obtain the final population. The individual with the highest fitness score is selected from the final population and decoded to obtain the optimal rat poison placement strategy.

[0054] A safety prevention design method based on artificial intelligence is implemented based on the safety prevention design system based on artificial intelligence, comprising: S1, arranging n sensor nodes in the area to be controlled to build an Internet of Things sensor network, and collecting environmental data and rodent activity data in real time based on the Internet of Things sensor network; n is a positive integer greater than 1;

[0055] S2, integrating environmental data and rodent activity data to obtain comprehensive rodent damage data;

[0056] S3. Construct a rodent damage detection model, use the rodent damage detection model to analyze the comprehensive rodent damage data, and obtain the location and activity level data of rodent damage activities;

[0057] S4, constructing a virtual model of rodent damage, based on environmental data and rodent damage activity data, using the virtual model of rodent damage, predicting the change trend data of rodent damage activities in the future r time period;

[0058] S5. According to the location activity level data and change trend data of rodent infestation activities, the optimal rodent poison placement strategy is solved by using an optimization algorithm; the optimal rodent poison placement strategy is sent to the on-site execution terminal for the placement of rodent poison.

[0059] The technical effects and advantages of the artificial intelligence-based safety prevention design system and method of the present invention are as follows:

[0060] By constructing an Internet of Things sensor network, the present invention realizes comprehensive real-time monitoring of the prevention and control area, eliminates monitoring blind spots, greatly improves the comprehensiveness and timeliness of data collection, realizes the intelligence, automation and precision of rodent control, and greatly improves the prevention and control effect and efficiency; by constructing a virtual model of rodent damage, it can accurately predict future rodent damage activities, realize early prevention and control, effectively reduce the risk of rodent damage, actively prevent the occurrence and spread of rodent damage, and shift from passive response to active prevention and control; secondly, the optimization algorithm is used to automatically generate the optimal rodent poison placement strategy, which not only improves the prevention and control effect, but also realizes the efficient use of resources; it has strong adaptive capabilities, and can dynamically adjust the prevention and control strategy according to environmental changes and rodent damage activities to ensure that the prevention and control measures are always in the best state; in addition, the prevention and control efficiency is greatly improved and the labor cost is reduced; through simulation, the system also reduces the risks and costs in actual operations, reduces environmental pollution and ecological risks; at the same time, the adaptive capability enables it to continuously optimize the prevention and control strategy and maintain high efficiency for a long time. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of a safety prevention and control design system based on artificial intelligence of the present invention;

[0062] Figure 2 This is a schematic diagram of a safety prevention and control design method based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0064] Example 1

[0065] See also Figure 1 As shown, the artificial intelligence-based safety prevention and control design system described in this embodiment includes:

[0066] A network construction module is used to deploy n sensor nodes in the area to be controlled to build an Internet of Things sensor network, and collect environmental data and rodent activity data in real time based on the Internet of Things sensor network; n is a positive integer greater than 1;

[0067] The data fusion module is used to fuse the environmental data and rodent activity data to obtain comprehensive rodent damage data;

[0068] The preliminary model building module is used to build a rodent damage detection model, and use the rodent damage detection model to analyze the comprehensive rodent damage data to obtain the location and activity level data of rodent damage activities;

[0069] The virtual simulation module is used to build a virtual model of rodent damage. Based on the environmental data and rodent damage activity data, the virtual model of rodent damage is used to predict the change trend data of rodent damage activities in the future r time period;

[0070] The strategy optimization module is used to use the optimization algorithm to solve the optimal rodent poison placement strategy based on the location activity data and change trend data of rodent activities; the optimal rodent poison placement strategy is sent to the on-site execution terminal for the placement of rodent poison; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0071] The construction methods of IoT sensor network include:

[0072] The area to be controlled is randomly divided into N sub-areas, and the decision variables are defined as the boundary coordinates of each sub-area and the node type (sensor type) corresponding to each sub-area; the objective function of area division is defined And constraints; the constraints are that the area of ​​each sub-region is within the range of [A_min, A_max], the coverage radius of each node is R, and the total cost does not exceed the budget C; where A_min is the preset lower limit of the area, and A_max is the preset upper limit of the area.

[0073] ; in, , , and is the weight of the corresponding item, For the The weight coefficient of each sub-region, For the The number of nodes in each sub-region; For the The coverage quality score of each sub-area, For the The coverage area of ​​the sub-regions, is the total area of ​​the area to be controlled, For the Unit cost of each node type; For the The number of nodes of the node type; For the The average energy consumption of the node types, is the weight of network lifetime; is the network lifetime (the length of time that the IoT sensor network can function normally).

[0074] Coverage Quality Score ;in, For the The area coverage weight of each sub-region is and It is an adjustable weight parameter used to balance the importance of different factors; For the The average coverage radius of the sensor nodes deployed in each sub-area; It is the largest coverage radius among all sensor nodes; For the The rodent infestation density index of each sub-area is obtained based on historical data or on-site investigations. The rodent infestation records of the area in the past period of time (such as recent years) are collected and analyzed, and data such as the number of captured rodents, sighting frequency, and rodent damage reports are counted. Based on these historical data, a rodent infestation density index is calculated (weighted summation or other methods) for each sub-area.

[0075] ;in, For the The price of the node type, The longest expected service life of all node types. For the The expected service life of each node type, Adjust the weight for life to balance the importance of life. For the The coefficients corresponding to the cost influencing factors of each node are: For the The factors affecting node cost include the difficulty of node maintenance (measured by the mean time between failures), the environmental adaptability of the node (a percentage score is given based on the range of environmental conditions that the node can adapt to), the data transmission capacity of the node (a benchmark transmission capacity is set, and the node is scored based on its performance relative to the benchmark), and the anti-interference ability of the node (scored by the percentage of normal working time under interference of different intensities).

[0076] Encode the decision variables using real number encoding or binary encoding; according to the encoding method, randomly generate the initial sponge population, each sponge in the sponge population corresponds to a possible regional division scheme (sub-region division and corresponding node type); each sponge is represented by a corresponding F-dimensional vector, where F is also the dimension of the decision variable. Calculate the objective function of each sponge in the regional division The function value on is recorded as the comprehensive motion value corresponding to each sponge; based on the calculated comprehensive motion value, an iterative comprehensive optimization operation is performed, and the iterative comprehensive optimization operation includes a displacement operation, a deformation operation, and a fusion operation.

[0077] The displacement operations include:

[0078] For each sponge , according to its comprehensive motion value and other spongy bodies Location , calculate its displacement vector ;

[0079] ;in, is a mixing coefficient, is a random number, is a small positive number used to avoid the denominator being zero. For other spongy bodies The comprehensive sports value, Spongy original position; is a Gaussian noise term, is the noise intensity; It is the sponge with the smallest comprehensive motion value in the current sponge population (global optimal solution).

[0080] Spongy body Along its displacement vector Displace to get a new position .

[0081] The deformation operation methods include:

[0082] For each sponge , randomly select one of the dimensions to deform and generate a deformation intensity factor ; in the In one dimension, the spongy Transforms the shape while leaving the remaining dimensions unchanged.

[0083] The formula for the transformation is:

[0084] ;in, Spongy In the The new position in the dimension, Spongy In the The original position in the dimension; For the current sponge population, The location of the cavernous body with the largest integrated motion value in the dimension; is the nonlinear deformation index parameter, which is used to control the nonlinear degree of deformation; is the sign function used to determine the direction of the deformation.

[0085] Deformation intensity factor ;in, is the maximum value of the preset deformation strength, is the minimum value of the preset deformation strength, is the current iteration number, is the preset maximum number of iterations, It is a positive adjustment parameter used to control the decay rate of deformation intensity.

[0086] The fusion operation methods include:

[0087] Randomly select two different sponges and Fusion is performed to generate a fusion scale factor , Obey the uniform distribution of [0, 1]; generate a new sponge-like body by linear interpolation ;in, Spongy Original location.

[0088] use Replace the sponge with the largest comprehensive motion value in the current sponge population;

[0089] The iterative comprehensive optimization operation is repeated until the preset maximum number of iterations is reached, and the sponge with the smallest comprehensive motion value in the sponge population is decoded, and the obtained region division scheme is the optimal region division scheme.

[0090] The area to be controlled is re-divided into sub-areas using the optimal area division scheme to obtain several sub-areas, and nodes of corresponding node types are set corresponding to the sub-areas based on the optimal area division scheme; that is, the construction of the Internet of Things sensor network is completed.

[0091] Node types include temperature sensors, humidity sensors, light sensors, sound sensors, infrared sensors, image sensors, gas sensors, and motion sensors.

[0092] The temperature sensor and humidity sensor collect ambient temperature data and ambient humidity data respectively; the light sensor is used to collect ambient light intensity data; the sound sensor is used to collect sound data for detecting the sound of rodent activities; the infrared sensor is used to collect infrared data (two-dimensional temperature distribution of the entire field of view, i.e. thermal image) for detecting rodent body temperature; the image sensor is used to collect on-site image / video data for detecting rodent morphology; the gas sensor is used to collect ambient CO 2 Concentration data; motion sensors are used to detect motion in the environment and to assist in detecting rodent activity.

[0093] Continuous sampling is used for temperature sensors, humidity sensors and light sensors, and the frequency of continuous sampling can be set to every minute, etc. When the sound sensor, infrared sensor and motion sensor detect suspicious rodent activity, the image sensor is triggered to take samples; for gas sensors, timed sampling is set, such as sampling every hour.

[0094] Environmental data includes ambient temperature data, ambient humidity data, ambient light intensity data and ambient CO 2 Concentration data; rodent activity data include sound data, distributed infrared data, on-site image / video data and movement status; distributed infrared data include the average number of hot spots per unit time, the rate of temperature change per unit time and the duration of abnormally high temperature.

[0095] The average number of hot spots per unit time, that is, the number of points with abnormally high temperatures in the thermal image per unit time; the temperature change rate per unit time, the rate of change of points with abnormally high temperatures in the thermal image per unit time; the length of time that points with abnormally high temperatures in the thermal image continue to appear.

[0096] The motion state includes motion intensity (the magnitude of the synthetic acceleration obtained by synthesizing the acceleration values ​​in three directions, where the three directions are the three axes of the three-dimensional coordinate system), motion frequency (by performing spectral analysis (such as FFT transformation) on the motion intensity, the energy of different frequency components is obtained; the proportion of energy in certain specific frequency ranges is counted) and motion duration (by setting the threshold conditions for the start and end of the motion, the duration of each motion event is calculated).

[0097] Further, the fusion processing method includes:

[0098] The environmental data and rodent pest activity data are standardized or normalized, and the format of heterogeneous data is unified to obtain corresponding standard data, namely standard environmental data and standard rodent pest activity data.

[0099] The standard environmental data and standard rodent pest activity data are respectively feature encoded using autoencoders to obtain corresponding feature vectors, namely the environmental feature vector and the rodent pest activity feature vector; the environmental feature vector and the rodent pest activity feature vector are vector-level spliced ​​to obtain a comprehensive high-dimensional feature vector, namely the comprehensive rodent pest data.

[0100] Ways to build a rodent detection model include:

[0101] A convolutional neural network, a recurrent neural network, or a combination of a convolutional neural network and a recurrent neural network is used as the basic framework of the rodent pest detection model.

[0102] The input of the rodent infestation detection model is defined as comprehensive rodent infestation data (comprehensive high-dimensional feature vector), and the output is the rodent infestation activity score of each node.

[0103] Several sets of annotated comprehensive rodent infestation data are collected as training sets within a fixed period of time in history; the annotation includes the rodent infestation activity level (e.g., 0-5 points) of each node location; the rodent infestation detection model is trained using the training set, and during the training process, the output value of the rodent infestation detection model is gradually approached to the annotated rodent infestation activity level by adjusting the parameters of the rodent infestation detection model; specifically, the loss function is defined ; and use optimization algorithms such as stochastic gradient descent to update parameters; until the function value of the loss function is minimized and does not change for P consecutive times, the parameters of the rodent detection model at this time are fixed, that is, the rodent detection model is completed.

[0104] Loss Function ; in, is the total number of nodes, For the The rodent infestation activity score predicted by the rodent infestation detection model for each node, For the The rodent activity level of each node; For the The weight of a node is used to indicate the importance of the node; is the weight of the spatial smoothing term, is the spatial smoothness index, is the time smoothing term weight, For the Nodes adjacent to a node The rodent infestation activity score predicted by the rodent infestation detection model, is the weight of the distribution consistency term, For the The distribution of rodent infestation activity predicted by the rodent infestation detection model for each node (spatial distribution of rodent infestation activity scores); For the The rodent infestation activity score predicted by the rodent infestation detection model for each node at the previous time step; For the The actual rodent infestation activity distribution of each node (the actual rodent infestation activity distribution obtained); is the KL divergence, which measures the difference between two distributions.

[0105] The constructed rodent infestation detection model is used to perform forward propagation calculations on the comprehensive rodent infestation data obtained in real time, and the rodent infestation activity score corresponding to the location of each node is output; the rodent infestation activity score corresponding to the location of all nodes is the location activity data of rodent infestation activities.

[0106] The construction methods of the rodent damage virtual model include:

[0107] The virtual model of rodent damage is a joint simulation of the cell sub-model and the rodent behavior sub-model; the cell sub-model is: the area to be controlled is discretized into M1 cell grids, M1 is a positive integer greater than 1; the cell state of each cell grid is defined as described by state variables, and the state variables include the number of rodents, the amount of food, the ambient temperature and the ambient humidity.

[0108] The mouse behavior sub-model is as follows: defining the update rules of cell states, describing how the state of each cell at the next moment is determined by the states of the current cell and its neighboring cells; the update rules include reproduction rules, foraging rules, activity rules, and migration rules.

[0109] The breeding rule is: define the breeding age threshold ag of mice, and treat mice older than the breeding age threshold ag as adult mice. Only adult mice can reproduce;

[0110] The reproduction probability of mice is calculated based on the reproduction function , the formula of the reproduction function is:

[0111] ; in, is the number of rodents per unit area, is the amount of food per unit area, is the ambient temperature, is the ambient humidity; , and To adjust the parameters, is the temperature suitability function, is the humidity suitability function. The temperature suitability function and humidity suitability function adopt the form of Gaussian distribution function. At each time step, for each adult mouse, Generate new rats for probability.

[0112] The foraging rule is: define the perception radius of food for mice as Rf; within the circular range formed by the perception radius Rf, mice perceive food; mice tend to move towards food, and the step length of movement is defined as ;

[0113] ;in, The area with the largest amount of food within the Rf range. is the amount of food in the area where the rodent is currently located, To The distance of the area, is the rate coefficient, Be a very small positive constant to avoid the denominator being 0; is the crowding influence coefficient. When the rats in the area are crowded, the moving step length will be reduced; and It is an adjustment index used to adjust the impact of food quantity difference and distance on moving step length.

[0114] The activity rule is: define the daytime activity intensity of rodents as and night activity intensity ; then the activity intensity of rodents per unit time ;in is the preset basic activity intensity. is the light intensity suitability function (obtained by experimentally observing the activities of mice under different light intensities, obtaining a certain amount of data samples, and then using the regression analysis data fitting method), is the light intensity.

[0115] Migration rules are: Define environmental quality score , which is determined by factors such as temperature, humidity, food quantity and disturbance intensity; When the environmental quality score is lower than the preset lower threshold, the current environmental conditions are considered to have deteriorated and the rodents will migrate. The migration direction is toward the adjacent area where the environmental quality score is greater than the preset upper threshold. ;in The area with the highest environmental quality score among the neighboring areas of the current area where the rodents are located. Score the environmental quality of the current area where the rodent is located. The current area where the rodent is located The distance of the area, is the migration rate coefficient.

[0116] Environmental Quality Rating ;in, is the temperature weight, is the humidity weight, is the food weight, is the interference intensity weight, and ; is the food regulation parameter, determining the growth rate of the function, Adjust the parameters for the interference intensity and determine the decay rate of the function; is the amount of food per unit area, is the intensity of disturbance per unit area; the disturbance intensity is evaluated and described in a quantitative or semi-quantitative way; for example, according to the intensity of capture activities, the number of natural enemies, and the frequency of extreme weather, the corresponding disturbance intensity scores are given respectively, and then the weighted sum of each score is obtained to obtain the comprehensive disturbance intensity value.

[0117] Collect historical data and actual monitoring data (population growth rate and population Gini coefficient) for a sufficiently long period of time (such as more than 1 year). Historical data include environmental data (temperature, humidity, light, etc.) and rodent activity data (quantity, distribution, activity intensity, etc.). Conduct field surveys in the monitoring areas to obtain basic information on rodents, including population size, food sources, and habitats. Organize and standardize all data.

[0118] Assign initial values ​​to various parameters in the virtual model of rodent damage, use historical data as input, simulate the dynamics of rodent population in each cell grid in the virtual model (the data types are also population growth rate and population Gini coefficient), compare the dynamics of rodent population with the actual monitoring data, and continuously adjust the parameters of the virtual model of rodent damage to maximize the degree of fit between the dynamics of rodent population and the actual monitoring data; obtain the optimal parameter combination of the virtual model of rodent damage.

[0119] Using the combination of optimal parameters of the rodent pest virtual model and taking the current real-time collected environmental data and rodent pest activity data as initial conditions, the rodent population dynamics in the future r time period (such as 1 month) are simulated in the rodent pest virtual model, and the rodent change trend data of each cell grid are extracted from the rodent population dynamics. The rodent change trend data is the rodent population growth rate (the growth rate of the rodent population in each cell grid); the rodent change trend data of the cell grid is used as the change trend data of rodent pest activity.

[0120] The solution to the optimal rat poison placement strategy includes:

[0121] Construct strategy variables and strategy objective functions; strategy variables include the type of rat poison, dosage of rat poison and location of rat poison;

[0122] For each cell grid, determine the corresponding strategy variable, that is, (Rat poison type), (rat poison dosage) and (whether to place, 0 or 1); For the Line The type of rat poison required for the cell grid corresponding to the column, For the Line The required dose of rat poison for the cell grid corresponding to the column, No. Line Whether rat poison is placed in the cell grid corresponding to the column.

[0123] Strategy objective function ; in, For the Line The rodent activity intensity of the cell grid corresponding to the column, For the Line The cost of placing rat poison per unit in the cell grid corresponding to the column, , and Balance weights for strategies; For the Line The mouse change trend data of the cell grid corresponding to the column, For the Line The sum of the rodent activity scores of the n1 nodes in the neighborhood (preset neighborhood radius) of the cell grid corresponding to the column.

[0124] Encode the strategy variable as the gene sequence of an individual (chromosome), for example, Using binary encoding, Encoded with real numbers, Encoded with 0 / 1.

[0125] Randomly generate N3 individuals as the initial population, decode each individual, and obtain the corresponding rat poison placement strategy; substitute the rat poison placement strategy into the strategy objective function, and calculate the value as the individual's fitness score.

[0126] According to the size of the fitness score, selection operators such as roulette are used to select excellent individuals, and individuals of two parent generations are randomly selected. Some genes are exchanged at a certain intersection to generate new offspring individuals. Some genes of the individuals are mutated with a fixed probability to increase population diversity. According to the results of selection, crossover and mutation, a new generation of population is generated until the iterative algebra reaches the preset upper limit to obtain the final population. The individual with the highest fitness score is selected from the final population, decoded, and the optimal rat poison placement strategy is obtained.

[0127] In this embodiment, by constructing an Internet of Things sensor network, comprehensive real-time monitoring of the prevention and control area is achieved, monitoring blind spots are eliminated, the comprehensiveness and timeliness of data collection are greatly improved, the intelligent, automated and precise rodent control is achieved, and the prevention and control effect and efficiency are greatly improved; by constructing a virtual model of rodent damage, it is possible to accurately predict future rodent damage activities, achieve early prevention and control, effectively reduce the risk of rodent damage, actively prevent the occurrence and spread of rodent damage, and shift from passive response to active prevention and control; secondly, the optimization algorithm is used to automatically generate the optimal rodent poison placement strategy, which not only improves the prevention and control effect, but also realizes the efficient use of resources; it has strong adaptive capabilities and can dynamically adjust the prevention and control strategy according to environmental changes and rodent damage activities to ensure that the prevention and control measures are always in the best state; in addition, the prevention and control efficiency is greatly improved and the labor cost is reduced; through simulation, the system also reduces the risks and costs in actual operations, reduces environmental pollution and ecological risks; at the same time, the adaptive capability enables it to continuously optimize the prevention and control strategy and maintain high efficiency for a long time.

[0128] Example 2

[0129] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a security prevention and control design method based on artificial intelligence, including:

[0130] S1. n sensor nodes are arranged in the area to be controlled to build an Internet of Things sensor network, and environmental data and rodent activity data are collected in real time based on the Internet of Things sensor network; n is a positive integer greater than 1;

[0131] S2, integrating environmental data and rodent activity data to obtain comprehensive rodent damage data;

[0132] S3. Construct a rodent damage detection model, use the rodent damage detection model to analyze the comprehensive rodent damage data, and obtain the location and activity level data of rodent damage activities;

[0133] S4, constructing a virtual model of rodent damage, based on environmental data and rodent damage activity data, using the virtual model of rodent damage, predicting the change trend data of rodent damage activities in the future r time period;

[0134] S5. According to the location activity level data and change trend data of rodent infestation activities, the optimal rodent poison placement strategy is solved by using an optimization algorithm; the optimal rodent poison placement strategy is sent to the on-site execution terminal for the placement of rodent poison.

[0135] Example 3

[0136] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the artificial intelligence-based security prevention and control design method provided above is implemented.

[0137] Since the electronic device introduced in this embodiment is an electronic device used to implement a security prevention and control design method based on artificial intelligence in the embodiment of this application, based on the security prevention and control design method based on artificial intelligence introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the security prevention and control design method based on artificial intelligence in the embodiment of this application, it belongs to the scope of protection of this application.

[0138] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0139] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A safety prevention and control design system based on artificial intelligence, characterized in that: include: A network construction module is used to deploy n sensor nodes in the area to be controlled to build an Internet of Things sensor network, and collect environmental data and rodent activity data in real time based on the Internet of Things sensor network; The data fusion module is used to fuse the environmental data and rodent activity data to obtain comprehensive rodent damage data; The preliminary model building module is used to build a rodent damage detection model, and use the rodent damage detection model to analyze the comprehensive rodent damage data to obtain the location and activity level data of rodent damage activities; The virtual simulation module is used to build a virtual model of rodent damage. Based on the environmental data and rodent damage activity data, the virtual model of rodent damage is used to predict the change trend data of rodent damage activities in the future r time period; The construction methods of the rodent damage virtual model include: The rodent damage virtual model is a joint simulation of the cell sub-model and the rodent behavior sub-model; the cell sub-model is: discretizing the area to be controlled into M1 cell grids, defining the cell state of each cell grid to be described by state variables, and the state variables include the number of rodents, the amount of food, the ambient temperature and the ambient humidity; the rodent behavior sub-model is: defining the update rules of the cell state; the update rules include reproduction rules, foraging rules, activity rules and migration rules; The strategy optimization module is used to obtain the optimal rodent poison placement strategy using an optimization algorithm based on the location activity data and change trend data of rodent pest activities; the optimal rodent poison placement strategy is sent to the on-site execution terminal for the placement of rodent poison; each module is connected by wired and / or wireless means; The solution method of the optimal rat poison placement strategy includes: Construct strategy variables and strategy objective functions; strategy variables include the type of rat poison, dosage of rat poison and location of rat poison; For each cell grid, determine the corresponding strategy variables, namely, the type of rat poison x_eu required for the cell grid corresponding to the e-th row and the u-th column, the dose of rat poison d_eu required for the cell grid corresponding to the e-th row and the u-th column, and whether rat poison Y_eu is placed in the cell grid corresponding to the e-th row and the u-th column; Encode the strategy variable into the gene sequence of an individual; randomly generate N3 individuals as the initial population, decode each individual, and obtain the corresponding rat poison placement strategy; substitute the rat poison placement strategy into the strategy objective function, and the calculated value is used as the individual's fitness score; According to the size of the fitness score, the selection operator is used to select excellent individuals. Two parent individuals are randomly selected, and some genes are exchanged at a certain intersection to generate new offspring individuals. Some genes of the individuals are mutated with a fixed probability to generate a new generation of population until the iterative algebra reaches the preset upper limit to obtain the final population. The individual with the highest fitness score is selected from the final population and decoded to obtain the optimal rat poison placement strategy.

2. The artificial intelligence-based safety prevention design system according to claim 1 is characterized in that: The construction method of the Internet of Things sensor network includes: The area to be controlled is randomly divided into N sub-areas, and the decision variables are defined as the boundary coordinates of each sub-area and the node type corresponding to each sub-area; the regional division objective function FH and constraint conditions are defined; the constraint conditions are that the area of ​​each sub-area is within the range of [A_min, A_max], the coverage radius of each node is R, and the total cost does not exceed the budget C; where A_min is the preset lower limit of the area, and A_max is the preset upper limit of the area; Encoding the decision variables, using real number encoding or binary encoding; randomly generating an initial sponge population according to the encoding method, each sponge in the sponge population corresponds to a regional division scheme; Calculate the function value of each sponge on the region division objective function FH, and record it as the comprehensive motion value corresponding to each sponge; perform iterative comprehensive optimization operation based on the calculated comprehensive motion value, and the iterative comprehensive optimization operation includes displacement operation, deformation operation and fusion operation; Repeat the iterative comprehensive optimization operation until the preset maximum number of iterations is reached, and decode the sponge with the smallest comprehensive motion value in the sponge population at this time, and the obtained regional division scheme is the optimal regional division scheme; The area to be controlled is re-divided into sub-areas using the optimal area division scheme to obtain several sub-areas, and nodes of corresponding node types are set corresponding to the sub-areas based on the optimal area division scheme; that is, the construction of the Internet of Things sensor network is completed.

3. The artificial intelligence-based safety prevention design system according to claim 2 is characterized in that: The environmental data include environmental temperature data, environmental humidity data, environmental light intensity data and environmental CO2 concentration data; rodent activity data include sound data, distributed infrared data, on-site image / video data and movement status; distributed infrared data include the average number of hot spots per unit time, the temperature change rate per unit time and the duration of abnormal high temperature; Exercise status includes exercise intensity, exercise frequency and exercise duration.

4. The artificial intelligence-based safety prevention design system according to claim 3 is characterized in that: The region partition objective function Among them, w1, w2, w3 and w4 are the weights of the corresponding items, w_i is the weight coefficient of the i-th sub-area, n_i is the number of nodes in the i-th sub-area; Q_i is the coverage quality score of the i-th sub-area, S_i is the coverage area of ​​the i-th sub-area, SZ is the total area of ​​the area to be controlled, C_k is the unit cost of the k-th node type; m_k is the number of nodes of the k-th node type; E_k is the average energy consumption of the k-th node type, α is the weight of the network life; WS is the network life; Coverage Quality Score Among them, α_i is the area coverage weight of the ith sub-area, α1 and β are adjustable weight parameters; R_i is the average coverage radius of the sensor nodes deployed in the ith sub-area; Rmax is the maximum coverage radius of all sensor nodes; D_i is the rodent density index of the ith sub-area; Among them, P_k is the price of the kth node type, Lmax is the longest expected service life among all node types, L_k is the expected service life of the kth node type, γ is the life adjustment weight, K_j is the coefficient corresponding to the jth node cost influencing factor, and X_j is the jth node cost influencing factor.

5. The artificial intelligence-based safety prevention design system according to claim 4 is characterized in that: The displacement operation includes: For each spongy body I, calculate its displacement vector M_I according to its comprehensive motion value FH(I) and the position x_J of other spongy bodies J; Among them, δ is a mixing coefficient, ra is a random number, ε is a positive number, FH(J) is the comprehensive motion value of other sponges J, x_I is the original position of sponge I; N(0,σ) is a Gaussian noise term, σ is the noise intensity; Xb is the sponge with the smallest comprehensive motion value in the current sponge population; Displace the sponge body I along its displacement vector M_I to obtain a new position x_I_new=x_I+M_I; The deformation operation includes: For each sponge I, randomly select one dimension to deform and generate a deformation intensity factor α(t); in the vth dimension, deform the sponge I and keep the other dimensions unchanged; The formula for the transformation is: x_I_new[v]=x_I[v]+α(t)×sign(Xb[v]-x_I[v])×|Xb[v]-x_I[v]| τ ; Where x_I_new[v] is the new position of sponge I in the vth dimension, x_I[v] is the original position of sponge I in the vth dimension; Xb[v] is the position of the sponge with the largest comprehensive motion value in the vth dimension in the current sponge population; τ is the nonlinear deformation index parameter; sign() is the sign function; deformation intensity factor Among them, αmax is the preset maximum value of the deformation strength, αmin is the preset minimum value of the deformation strength, t is the current iteration number, T is the preset maximum iteration number, and a is a positive adjustment parameter; The fusion operation includes: Randomly select two different sponges I and I′ for fusion, and generate a fusion ratio factor ε1; generate a new sponge Xnew = ε1×x_I+(1-ε1)×x_I′ by linear interpolation; where x_I′ is the original position of sponge I′; replace the sponge with the largest comprehensive motion value in the current sponge population with Xnew.

6. The artificial intelligence-based safety prevention design system according to claim 5 is characterized in that: The method of performing the fusion processing includes: Standardize or normalize the environmental data and rodent activity data, and unify the format of heterogeneous data to obtain standard environmental data and standard rodent activity data; The standard environmental data and standard rodent pest activity data are respectively feature encoded using autoencoders to obtain environmental feature vectors and rodent pest activity feature vectors; the environmental feature vectors and rodent pest activity feature vectors are vector-level concatenated to obtain a comprehensive high-dimensional feature vector, which is the comprehensive rodent pest data.

7. The artificial intelligence-based safety prevention design system according to claim 6 is characterized in that: The method of constructing the rodent damage detection model includes: Using convolutional neural network, recurrent neural network or a combination of convolutional neural network and recurrent neural network as the basic framework of rodent detection model; The input of the rodent detection model is defined as comprehensive rodent data, and the output is the rodent activity score of each node. Several sets of annotated comprehensive rodent data are collected as training sets within a fixed historical period. The annotations include the rodent activity of each node. The rodent detection model is trained using the training set. During the training process, the loss function Loss is defined. The parameters are updated using the stochastic gradient descent optimization algorithm. The parameters of the rodent detection model are fixed until the function value of the loss function is minimized and does not change for P consecutive times. The construction of the rodent detection model is completed. Loss Function Among them, N1 is the total number of nodes, y_s is the rodent activity score predicted by the rodent detection model of the s-th node, t_s is the labeled rodent activity of the s-th node; β_s is the weight of the s-th node; γ1 is the weight of the spatial smoothing term, γ2 is the spatial smoothing term index, δ2 is the weight of the temporal smoothing term, y_s′ is the rodent activity score predicted by the rodent detection model of the node s′ adjacent to the s-th node, θ is the weight of the distribution consistency term, p_s is the distribution of rodent activity predicted by the rodent detection model of the s-th node; y_s_prev is the rodent activity score predicted by the rodent detection model of the s-th node in the previous time step; q_s is the actual rodent activity distribution of the s-th node; KL() is the KL divergence.

8. The artificial intelligence-based safety prevention design system according to claim 7 is characterized in that: The breeding rule is: define a breeding age threshold ag of rodents, and regard rodents older than the breeding age threshold ag as adult rodents; The reproduction probability Pb ​​of rodents is calculated based on the reproduction function. The formula of the reproduction function is: Pb = a1×exp(-z×N_mo)×(1-exp(-c1×Food))×g(Tenv)×h(Henv); where N_mo is the number of rodents per unit area, Food is the amount of food per unit area, Tenv is the ambient temperature, and Hev is the ambient humidity; a1, z, and c1 are adjustment parameters, g(Tenv) is the temperature suitability function, and h(Henv) is the humidity suitability function. At each time step, for each adult rodent, a new rodent is generated with a probability of Pb; The foraging rule is: define the perception radius of food for mice as Rf; mice perceive food within the circular range formed by the perception radius Rf; mice tend to move towards food, and the step length of movement is defined as Among them, Fmax is the area with the largest amount of food within the Rf range, Fcur is the amount of food in the area where the rodent is currently located, dmax is the distance to the Fmax area, k1 is the rate coefficient, ω1 is a positive constant; ω2 is the crowding influence coefficient; a2 and b2 are adjustment indexes; The activity rule is: define the daytime activity intensity of rodents as Aday and the nighttime activity intensity as Anight; then the rodent activity intensity per unit time AP = A_0×(Aday×g(Tenv)×h(Henv)+Anight×f(Lp)); where A_0 is the preset basic activity intensity, f(Lp) is the light intensity suitability function, and Lp is the light intensity; The migration rules are as follows: define the environmental quality score Qw; when Qw is lower than the preset lower limit of environmental quality, the rodents migrate; the migration direction is toward the adjacent area where the environmental quality score is greater than the preset upper limit of environmental quality, and the migration step length is Where Qmax is the area with the highest environmental quality score among the adjacent areas of the current area where the rodent is located, Qcur is the environmental quality score of the current area where the rodent is located, Dmax is the distance from the current area where the rodent is located to the Qmax area, and h1 is the migration rate coefficient; Collect historical data and actual monitoring data, assign initial values ​​to various parameters in the virtual model of rodent damage, use historical data as input, simulate the dynamics of rodent populations in each cell grid in the virtual model of rodent damage, compare the dynamics of rodent populations with the actual monitoring data, and continuously adjust the parameters of the virtual model of rodent damage to maximize the degree of fit between the dynamics of rodent populations and the actual monitoring data; and obtain the optimal combination of parameters of the virtual model of rodent damage; Using the combination of optimal parameters of the rodent pest virtual model and taking the current real-time collected environmental data and rodent pest activity data as initial conditions, the rodent population dynamics in the future r time period are simulated in the rodent pest virtual model, and the rodent change trend data of each cell grid are extracted from the rodent population dynamics. The rodent change trend data is the growth rate of the rodent population; the rodent change trend data of the cell grid is used as the change trend data of rodent pest activity.

9. The artificial intelligence-based safety prevention design system according to claim 8 is characterized in that: The strategy objective function FU=W1×∑ e,u (A_eu×Y_eu)+W2×∑ e,u (C_eu×d_eu×Y_eu)+W3×∑ e,u (G_eu+S_n1); where A_eu is the rodent activity intensity of the cell grid corresponding to the e-th row and the u-th column, C_eu is the cost of placing unit rat poison in the cell grid corresponding to the e-th row and the u-th column, W1, W2 and W3 are strategy balance weights; G_eu is the rodent change trend data of the cell grid corresponding to the e-th row and the u-th column, and S_n1 is the sum of the rodent activity scores of the n1 nodes in the neighborhood of the cell grid corresponding to the e-th row and the u-th column.

10. A safety prevention and control design method based on artificial intelligence, which is implemented based on the safety prevention and control design system based on artificial intelligence according to any one of claims 1 to 9, characterized in that: include: S1. Arrange n sensor nodes in the area to be controlled to build an Internet of Things sensor network, and collect environmental data and rodent activity data in real time based on the Internet of Things sensor network; S2, integrating environmental data and rodent activity data to obtain comprehensive rodent damage data; S3. Construct a rodent damage detection model, use the rodent damage detection model to analyze the comprehensive rodent damage data, and obtain the location and activity level data of rodent damage activities; S4, constructing a virtual model of rodent damage, based on environmental data and rodent damage activity data, using the virtual model of rodent damage, predicting the change trend data of rodent damage activities in the future r time period; S5. Based on the location activity data and change trend data of rodent activities, an optimization algorithm is used to solve the optimal rodent poison placement strategy; The optimal rat poison placement strategy is sent to the on-site execution terminal for the placement of rat poison.

Images