A security guard scheduling method and system based on an optimization algorithm

Abstract

This invention discloses a security personnel dispatching method and system based on optimization algorithms, relating to the technical field of path planning. The method involves acquiring a topological undirected graph and constraints of a target area, determining an optimization objective based on these constraints, optimizing the target path according to the objective and constraints, sending the target path to a drone, and enabling the drone to execute the target path and obtain an image of the area. The area image is then used to identify the event level, and security personnel are dispatched based on the event level. Constructing and optimizing the topological undirected graph of the target area ensures that drones can inspect key areas and acquire high-quality area images. Event levels are determined through image recognition, and the on-site situation is assessed, providing a basis for subsequent personnel dispatch and improving the accuracy and response speed of security personnel dispatch.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, specifically to a security personnel scheduling method and system based on optimization algorithms. Background Technology

[0002] With the continuous advancement of technology, security personnel scheduling methods based on optimization algorithms have emerged, aiming to achieve efficient and accurate scheduling of security personnel through intelligent means, combined with real-time data and advanced algorithm models.

[0003] Publication No. CN118627686A discloses a property security personnel dispatch optimization system; a task receiving module receives security task requests from the property management system; a task analysis module analyzes and evaluates the nature, urgency, and required resources of the received task requests; a dispatch decision module automatically allocates tasks and makes dispatch decisions based on the task analysis results, combined with the current location, status, and capabilities of the security personnel; a task execution monitoring module monitors the task execution status of the security personnel in real time; and a task evaluation feedback module evaluates the completed tasks to optimize subsequent dispatch decisions.

[0004] Traditional methods have many limitations. For example, manual inspections are not real-time and it is difficult to make accurate personnel allocation decisions quickly in the face of emergencies, resulting in a waste of security resources or insufficient security in key areas. Manual dispatch often lacks quantitative analysis of the real-time situation and cannot adjust the distribution of security personnel in a timely manner according to the dynamic situation on site. Summary of the Invention

[0005] The purpose of this invention is to address the problems mentioned in the background art, such as the poor real-time performance of manual inspections, the difficulty in making accurate personnel allocation decisions in the face of emergencies, the waste of security resources or insufficient security in key areas, and the lack of quantitative analysis of real-time situations in manual dispatching, which makes it impossible to adjust the distribution of security personnel in a timely manner according to the dynamic situation on site. Therefore, this invention proposes a security personnel dispatching method and system based on optimization algorithms.

[0006] A first aspect of this invention provides a security personnel scheduling method based on an optimization algorithm, the method comprising: Obtain the topological undirected graph and constraints of the target region, and determine the optimization objective based on the topological undirected graph and constraints; The target path is obtained by optimizing according to the optimization objective and constraints, and the target path is sent to the UAV so that the UAV can execute the target path to obtain the region image; The event level is determined by identifying the area image, and security personnel are dispatched based on the event level.

[0007] Optionally, the constraints include a first constraint, a second constraint, and a third constraint, including: First constraint: Where t is the total flight time of the UAV, v is the real-time flight speed of the UAV, n is the number of nodes inspected, and t k Let l be the detection time for the k-th node. i t represents the flight distance between adjacent nodes. max This refers to the maximum flight time of the drone; Second constraint: Among them, v α-max v is the maximum horizontal movement speed of the drone. β-max v is the maximum ascent speed of the drone. γ-max This represents the maximum descent speed of the drone; The third constraint is: s' = s + ε; where s' is the safe distance between the UAV and the node, s is the preset safe distance, and ε is the hovering error compensation value.

[0008] Optionally, the target path is obtained by optimizing according to the optimization objective and constraints, including: Step 1: Calculate the Euclidean distance from the drone to the starting point as the first distance, and calculate the Euclidean distance from the drone to each node to obtain the distance set; Step 2: Filter the second distance in the distance set to obtain the target second distance, and calculate the weighted fusion distance based on the first distance and the target second distance; Step 3: Linearly normalize the weighted fusion distance to obtain normalized data, and invert the normalized data to obtain the initial Q value; Step 4: Initialize the training parameters according to the optimization objective and the constraints, determine the state s of the UAV, select action a according to the initial Q value, and send action a to the UAV to make the UAV execute action a; Step 5: Obtain the state s' of the drone and calculate the reward R. Update the initial Q value according to the reward R to obtain the Q' value. Update the Q table according to the Q' value. The Q table contains multiple Q values. Step 6: Repeat steps 4 and 5 until the termination condition is met, then output the target path; the termination condition is reaching a preset number of iterations or the algorithm converges.

[0009] Optionally, the initial Q value is updated according to the reward R to obtain the Q' value, including: Q-value update formula: ];Q=(s,a), Q'=(s',a'), Q''=(s'',a''), R=(s,a); where Q=(s,a) is the expected score of the drone performing action a in state s, s is the current state of the drone, a is the action that the drone can perform in the current state s, ← is the update symbol, α is the learning rate, R=(s,a) is the reward score of the drone after performing action a in state s, γ is the discount factor, ω is the deep learning factor, A is the set of actions of the drone, Q'=(s',a') is the expected score of the drone performing action a' in state s', and Q''=(s'',a'') is the expected score of the drone performing action a'' in state s''.

[0010] Optionally, the event level is obtained by identifying the region image, and the event level includes low, medium and high levels, including: The area image is input into the target model for identification to obtain the number of people, and the event level is determined based on the number of people. If the number of people is less than or equal to the first threshold, it is classified as low-level; if the first threshold is less than the number of people but less than or equal to the second threshold, it is classified as medium-level; and if the second threshold is less than the number of people, it is classified as high-level.

[0011] A second aspect of this invention provides a security personnel dispatching system based on an optimization algorithm, the system comprising: The target optimization module is used to obtain the topological undirected graph and constraints of the target region, and determine the optimization target based on the topological undirected graph and constraints. The path optimization module is used to optimize the target path according to the optimization objective and constraints, and send the target path to the UAV so that the UAV can execute the target path to obtain the region image; The personnel dispatch module is used to identify the event level from the area image and dispatch security personnel according to the event level.

[0012] Optionally, the constraints include a first constraint, a second constraint, and a third constraint, including: First constraint: Where t is the total flight time of the UAV, v is the real-time flight speed of the UAV, n is the number of nodes inspected, and t k Let l be the detection time for the k-th node. i t represents the flight distance between adjacent nodes. max This refers to the maximum flight time of the drone; Second constraint: ; where v α-max v is the maximum horizontal movement speed of the drone. β-max v is the maximum ascent speed of the drone. γ-max This represents the maximum descent speed of the drone; The third constraint is: s' = s + ε; where s' is the safe distance between the UAV and the node, s is the preset safe distance, and ε is the hovering error compensation value.

[0013] Optionally, the path optimization module includes: The first operation module is used to calculate the Euclidean distance from the drone to the starting point as the first distance, and to calculate the Euclidean distance from the drone to each node to obtain the distance set; The second operation module is used to filter the second distance in the distance set to obtain the target second distance, and to calculate the weighted fusion distance based on the first distance and the target second distance; The third operation module is used to linearly normalize the weighted fusion distance to obtain normalized data, and to invert the normalized data to obtain an initial Q value. The fourth operation module is used to initialize training parameters according to the optimization objective and the constraints, determine the state s of the UAV, select action a according to the initial Q value, and send action a to the UAV to make the UAV execute action a. The fifth operation module is used to acquire the state s' of the UAV, calculate the reward R, update the initial Q value according to the reward R to obtain the Q' value, and update the Q table according to the Q' value; the Q table contains multiple Q values; The sixth operation module is used to repeatedly execute the fourth and fifth operation modules until the termination condition is met, at which point the target path is output; the termination condition is reaching a preset number of iterations or the algorithm converges.

[0014] Optionally, the fifth operation module also includes: Q-value update formula: ];Q=(s,a), Q'=(s',a'), Q''=(s'',a''), R=(s,a); where Q=(s,a) is the expected score of the drone performing action a in state s, s is the current state of the drone, a is the action that the drone can perform in the current state s, ← is the update symbol, α is the learning rate, R=(s,a) is the reward score of the drone after performing action a in state s, γ is the discount factor, ω is the deep learning factor, A is the set of actions of the drone, Q'=(s',a') is the expected score of the drone performing action a' in state s', and Q''=(s'',a'') is the expected score of the drone performing action a'' in state s''.

[0015] Optionally, the event levels include low, medium, and high, and the personnel scheduling module includes: The level determination module is used to input the area image into the target model to identify the number of people and determine the event level based on the number of people. The judgment execution module is used to determine the level as low if the number of people is less than or equal to the first threshold, as medium if the first threshold is less than or equal to the number of people and less than the second threshold, and as high if the second threshold is less than the number of people.

[0016] The beneficial effects of this invention are: This invention proposes a security personnel dispatching method based on an optimization algorithm. It obtains the topological undirected graph and constraints of the target area, determines the optimization objective based on these conditions, optimizes the target path according to the objective and constraints, sends the target path to a drone, and enables the drone to execute the target path and obtain area images. The area images are then used to identify event levels, and security personnel are dispatched based on these event levels. Constructing and optimizing the topological undirected graph of the target area ensures that drones can inspect key areas and acquire high-quality area images. Event levels are determined through image recognition, and the on-site situation is assessed, providing a basis for subsequent personnel dispatch and improving the accuracy and response speed of security personnel dispatch. Attached Figure Description

[0017] Figure 1 A flowchart of a security personnel scheduling method based on an optimization algorithm is provided for an embodiment of the present invention; Figure 2 This invention provides a framework diagram of a security personnel dispatching system based on an optimization algorithm. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0019] This invention provides a security personnel scheduling method based on an optimization algorithm. See also... Figure 1 , Figure 1 A flowchart illustrating a security personnel scheduling method based on an optimization algorithm, provided in an embodiment of the present invention. The method includes the following steps: S101, Obtain the topological undirected graph and constraints of the target region, and determine the optimization objective based on the topological undirected graph and constraints; S102, optimize according to the optimization objective and constraints to obtain the target path, and send the target path to the UAV so that the UAV can execute the target path to obtain the area image; S103: Identify the event level from the regional image and dispatch security personnel based on the event level.

[0020] The security personnel dispatching method based on optimization algorithms provided in this invention constructs and optimizes the topological undirected graph of the target area to ensure that drones can inspect key areas and acquire high-quality regional images. By identifying the event level through image recognition and assessing the on-site situation, it provides a basis for subsequent personnel dispatching, thereby improving the accuracy and response speed of security personnel dispatching.

[0021] In one implementation, by acquiring the topological undirected graph of the target area and determining the optimization objective (shortest inspection path) and constraints, a precise basis for UAV path planning is provided. The topological undirected graph shows the connection relationship of the distribution of buildings in public places, which enables path planning to fully consider the actual layout within the area, avoid invalid or repeated flight paths, and thus improve the efficiency and accuracy of UAV inspection.

[0022] In one implementation, an undirected topological graph is constructed by: collecting a detailed building distribution map of the target area, including the location, shape, size of buildings, and the relative distances and connections between them; abstracting each building or key area as a node in the graph; and abstracting the connections between buildings, such as passageways, roads, or line-of-sight paths, as edges; and constructing an undirected graph by analyzing the topological structure of the nodes and edges. During the construction process, weights are added to the nodes and edges according to actual needs. For example, the weight of a node can represent the importance or population density of the area, and the weight of an edge can represent the length of the connecting passage or the difficulty of passage. The resulting undirected topological graph can reflect the layout and connection relationships of the target area, providing basic data support for subsequent drone path planning and security personnel dispatch.

[0023] In one implementation, the target path is obtained by optimizing based on the optimization objective and constraints, and the UAV is executed along the path to acquire regional images, thus achieving efficient inspection of the target area. The optimized path enables the UAV to complete the inspection task of the key area within a limited time and energy, following the optimal route, ensuring that the acquired regional images have high coverage and representativeness, and providing more comprehensive and accurate image data for subsequent personnel identification and event level classification.

[0024] In one implementation, the event level (low, medium, and high) is obtained by identifying the regional image, which enables the dispatch of security personnel to be accurately allocated according to the severity of the event. By identifying and analyzing the personnel and related situations in the image, the danger level of the event can be determined, preventing the blind dispatch of security personnel and ensuring that a sufficient number of security personnel can be quickly and reasonably deployed to the scene to handle the event, thereby improving the pertinence and effectiveness of security work.

[0025] In one implementation, when determining the dispatch of security personnel, the events are classified according to the event levels obtained from regional image recognition. Event levels are typically divided into three levels: low, medium, and high. Low-level events are handled by a small number of security personnel for routine processing, such as general crowd gatherings or minor disorder. Medium-level events are handled by a larger number of security personnel for on-site control and guidance, such as small-scale conflicts or equipment malfunctions. High-level events are handled by a large number of security personnel for rapid response, emergency handling, and personnel evacuation, such as fires and violent incidents.

[0026] In one embodiment, the constraints include a first constraint, a second constraint, and a third constraint, including: First constraint: Where t is the total flight time of the UAV, v is the real-time flight speed of the UAV, n is the number of nodes inspected, and t k Let l be the detection time for the k-th node. i t represents the flight distance between adjacent nodes. max This refers to the maximum flight time of the drone; Second constraint: ; where v α-max v is the maximum horizontal movement speed of the drone. β-max v is the maximum ascent speed of the drone. γ-max This represents the maximum descent speed of the drone; Third constraint: s' = s + ε; Where s' is the safe distance between the drone and the node, s is the preset safe distance, and ε is the hovering error compensation value. In one implementation, t is the total flight time of the drone, that is, the entire process time from starting from the charging pile, completing the inspection task (or returning midway) and returning to the charging pile, which needs to be calculated in real time and compared with the maximum flight time; The objective function value is the total distance of the inspection path, l i It is the flight distance between two adjacent inspection targets (nodes). It is the sum of the distances between all adjacent nodes; v is the real-time flight speed of the UAV, i.e., the real-time flight speed of the UAV during the inspection process; n is the number of nodes inspected, the total number of inspection targets; t k The detection time for the k-th node is the time the drone hovers after arriving at the k-th target point to complete the detection (such as taking pictures or collecting data).

[0027] In one implementation, t max This refers to the maximum flight time of the drone, also known as its endurance. It is determined by the capacity of the lithium battery onboard the drone and represents the longest continuous flight time the drone can fly on a single full charge. It is a fixed performance parameter. α-maxThis refers to the drone's maximum horizontal movement speed, i.e., the maximum horizontal flight speed displayed on the drone's dashboard; it is the upper speed limit that the drone cannot exceed when flying horizontally. β-max This is the maximum ascent speed of the drone, i.e., the upper limit of speed that the drone cannot exceed during vertical ascent (to prevent excessive motor load or decreased climbing efficiency); v γ-max This is the maximum descent speed of the drone, which is the upper limit of the speed that the drone cannot exceed during its vertical descent (to avoid losing control or ground impact due to excessive descent).

[0028] In one implementation, s' is the safe distance between the UAV and the node, that is, the actual safe distance that the UAV and the inspection target (node) need to maintain; s is the preset safe distance, the preset safe distance of the inspection target, which is determined by the experimenters to prevent the UAV from getting too close to the inspection target; ε is the hovering error compensation value. When the UAV hovers at the target point, there may be a slight positional deviation, such as drift caused by airflow. An additional compensation distance needs to be added to ensure that the UAV has a hovering error, that is, the UAV can still maintain a safe distance from the inspection target. The magnitude of ε is determined by the hovering accuracy performance of the UAV.

[0029] In one implementation, the first constraint limits the total flight time of the UAV to ensure that it completes the task within its maximum endurance, while optimizing the path distance and improving inspection efficiency. The second constraint limits the speed of the UAV to ensure its flight safety during horizontal, ascent, and descent processes, avoiding loss of control or damage due to excessive speed. The third constraint, through safety distance and hovering error compensation, ensures that the UAV maintains a safe distance from the target during inspection, avoiding collision risks and improving the reliability and stability of UAV inspection.

[0030] In one embodiment, optimizing based on the optimization objective and constraints to obtain the target path includes: Step 1: Calculate the Euclidean distance from the drone to the starting point as the first distance, and calculate the Euclidean distance from the drone to each node to obtain the distance set; Step 2: Filter the second distance in the distance set to obtain the target second distance, and calculate the weighted fusion distance based on the first distance and the target second distance; Step 3: Linearly normalize the weighted fusion distance to obtain normalized data, and invert the normalized data to obtain the initial Q value; Step 4: Initialize the training parameters according to the optimization objective and constraints, determine the state s of the UAV, select action a according to the initial Q value, and send action a to the UAV to make the UAV execute action a; Step 5: Obtain the drone's state s' and calculate the reward R. Update the initial Q value based on the reward R to obtain the Q' value. Update the Q table based on the Q' value. The Q table contains multiple Q values. Step 6: Repeat steps 4 and 5 until the termination condition is met, then output the target path; the termination condition is reaching the preset number of iterations or the algorithm converges.

[0031] In one implementation, the training parameters are as follows: the initial position of the drone is set to a charging station, the initial battery is fully charged, the inspection target list is all nodes, the algorithm learning rate is 0.1, the discount factor is 0.9, and the deep learning factor is 0.8.

[0032] In one implementation, the Euclidean distances from the UAV to the starting point and each node are calculated, for example: UAV coordinates s i =(x i ,y i ), node coordinates s g =(x g ,y g Euclidean distance The second distance to the target is obtained by filtering (the filtering rule is to select the second distance with the smallest value to ensure that the drone will move towards the nearest undetected node first, avoiding blind exploration caused by multiple targets being scattered), and then the weighted fusion distance is calculated (e.g., D(s)). i )=(1−η)*D(s i ,s g )+η*D(s i ,s s ), D(s i ) represents the weighted fusion distance, η represents the weight, and D(s) represents the weighted fusion distance. i ,s g ) represents the second distance to the target, D(s) i ,s s (The first distance) can provide a reasonable initial direction for the UAV's path planning, enabling the UAV to choose actions closer to the target node in the initial stage, avoiding a large number of invalid paths caused by initial random exploration in standard Q-learning, and improving the initial efficiency of path planning.

[0033] In one implementation, the weighted fusion distance is linearly normalized and the values ​​are reversed (for example, GY is normalized data, GY=1-GY, that is, the current GY is equal to 1 minus the previous GY) to obtain an initial Q value. This allows the initial Q value to better reflect the distance relationship between the UAV and the target node. In the subsequent Q learning process, the reasonable setting of the initial Q value enables the algorithm to converge to the optimal path more quickly.

[0034] In one implementation, during the update phase, a deep learning factor is introduced to expand the search dimension of Q-value updates. At the same time, the next state and the next-next state after the current action are evaluated, which realizes global path optimization. The two-step state search method enables the algorithm to better weigh the cumulative advantages and disadvantages of the path, avoid falling into the trap of local optima, and thus plan an inspection path that is more in line with the shortest path goal, thereby improving the global optimization capability of path planning.

[0035] In one embodiment, updating the initial Q value according to the reward R to obtain the Q' value includes: Q-value update formula: ];Q=(s,a), Q'=(s',a'), Q''=(s'',a''), R=(s,a); where Q=(s,a) is the expected score of the drone performing action a in state s, s is the current state of the drone, a is the action that the drone can perform in the current state s, ← is the update symbol, α is the learning rate, R=(s,a) is the reward score of the drone after performing action a in state s, γ is the discount factor, ω is the deep learning factor, A is the set of actions of the drone, Q'=(s',a') is the expected score of the drone performing action a' in state s', and Q''=(s'',a'') is the expected score of the drone performing action a'' in state s''.

[0036] In one implementation, s represents the current state of the drone, which includes environmental information (such as obstacle positions and target distribution) and its own state (such as current position, remaining battery power, and flight speed); a represents the actions that the drone can perform in the current state s, such as moving in a certain direction, adjusting flight speed, and returning to home to recharge; ← represents the update symbol, for example, assigning the calculation result on the right side of the formula to Q on the left side, thus achieving iterative optimization of the Q value; α represents the learning rate, which is a positive number less than 1, controlling the update step size of each Q value (the smaller α is, the smoother the update; the larger α is, the more aggressive the update). In one implementation, R=(s,a) is the reward score of the drone after performing action a in state s, which refers to the immediate gain / penalty of the environment feedback after the drone performs action a in state s (calculated by a reward function, such as a positive reward for reaching a new target point and a negative reward for approaching a collision; the reward function is based on existing technology); γ is a discount factor, ranging from 0 to 1, used to weigh the Q value of the immediate reward R against the future reward (s' and s''). The closer γ is to 1, the more emphasis is placed on the future reward; the closer γ is to 0, the more emphasis is placed on the immediate reward; ω is a deep learning factor, ranging from (0.5,1), used to weight the Q value of state (s') and state (s''). Since s' is closer to the current state, the weight w is greater than (1-w); A is the set of actions of the drone, which refers to the set of all possible actions that the drone can perform in any state, such as: moving up, down, left, and right, hovering, and charging. To take the maximum value of all possible actions a' in action space A, representing the maximum expected score that the UAV can obtain in the next state s' after the current action a; the next state s' refers to the new state formed by the change of the environment and its own state after the UAV performs action a in the current state s; the actions that can be executed in state s' (belonging to action space A) are used to derive the next state s''. The maximum value of a' is taken for all possible actions in action space A, representing the maximum expected future score that the UAV can obtain in the next next state s''; s'' is the next next state, which refers to the new state obtained by the UAV after performing action a' in state s', and is the core used to expand the search dimensions; a'' is the action that can be performed in state s'' (belonging to action space A).

[0037] In one implementation, by introducing a deep learning factor ω and a two-step state search mechanism, the formula not only considers the immediate reward R and the maximum Q value of the next state s', but also further incorporates the maximum Q value of the next-next state s''. This allows for a more comprehensive evaluation of the long-term merits of the path, enabling the UAV to effectively avoid local optima traps during path planning and to plan inspection paths that are closer to the global optimum. By setting parameters such as the learning rate α and the discount factor γ, the learning step size of the algorithm and the degree of emphasis on future rewards can be flexibly controlled, further improving the adaptability and reliability of path planning.

[0038] In one embodiment, an event level is obtained by identifying the region image. The event levels include low, medium, and high levels, including: The area image is input into the target model for identification to obtain the number of people, and the event level is determined based on the number of people. If the number of people is less than or equal to the first threshold, it is classified as low-level; if the first threshold is less than the number of people but less than or equal to the second threshold, it is classified as medium-level; and if the second threshold is less than the number of people, it is classified as high-level.

[0039] In one implementation, the target model is the YOLO series model, which classifies events into three levels—low, medium, and high—based on the number of people involved. This allows security personnel to take appropriate measures according to the severity of the event. Low-level events can be handled by a small number of security personnel for routine processing, medium-level events can be handled by a moderate number of security personnel for on-site control, and high-level events can be handled by a large number of security personnel for emergency response. For example, if the first threshold is 3 people and the second threshold is 5 people, then events with 2 people are classified as low-level, events with 4 people as medium-level, and events with more than 5 people as high-level.

[0040] Based on the same inventive concept, this invention also provides a security personnel dispatching system based on an optimization algorithm. See also... Figure 2 , Figure 2A framework diagram of a security personnel dispatching system based on an optimization algorithm provided in this embodiment of the invention includes: The target optimization module is used to obtain the topological undirected graph and constraints of the target region, and to determine the optimization target based on the topological undirected graph and constraints. The path optimization module is used to optimize the target path according to the optimization objective and constraints, and send the target path to the UAV so that the UAV can execute the target path to obtain the area image; The personnel dispatch module is used to identify the event level from the regional images and dispatch security personnel according to the event level.

[0041] The security personnel dispatching system based on optimization algorithms provided in this invention constructs and optimizes the topological undirected graph of the target area to ensure that drones can inspect key areas and acquire high-quality regional images. By identifying the event level through image recognition and assessing the on-site situation, it provides a basis for subsequent personnel dispatching, thereby improving the accuracy and response speed of security personnel dispatching.

[0042] In one embodiment, the constraints include a first constraint, a second constraint, and a third constraint, including: First constraint: Where t is the total flight time of the UAV, v is the real-time flight speed of the UAV, n is the number of nodes inspected, and t k Let l be the detection time for the k-th node. i t represents the flight distance between adjacent nodes. max This refers to the maximum flight time of the drone; Second constraint: ; where v α-max v is the maximum horizontal movement speed of the drone. β-max v is the maximum ascent speed of the drone. γ-max This represents the maximum descent speed of the drone; Third constraint: s' = s + ε; Where s' is the safe distance between the drone and the node, s is the preset safe distance, and ε is the hovering error compensation value.

[0043] In one embodiment, the path optimization module includes: The first operation module is used to calculate the Euclidean distance from the drone to the starting point as the first distance, and to calculate the Euclidean distance from the drone to each node to obtain the distance set; The second operation module is used to filter the second distance in the distance set to obtain the target second distance, and calculate the weighted fusion distance based on the first distance and the target second distance; The third operation module is used to linearly normalize the weighted fusion distance to obtain normalized data, and to invert the normalized data to obtain the initial Q value. The fourth operation module is used to initialize training parameters according to the optimization objective and constraints, determine the state s of the UAV, select action a according to the initial Q value, and send action a to the UAV so that the UAV can execute action a. The fifth operation module is used to obtain the state s' of the UAV, calculate the reward R, update the initial Q value according to the reward R to obtain the Q' value, and update the Q table according to the Q' value; the Q table contains multiple Q values; The sixth operation module is used to repeatedly execute the fourth and fifth operation modules until the termination condition is met, at which point the target path is output; the termination condition is reaching the preset number of iterations or the algorithm converges.

[0044] In one embodiment, the fifth operation module further includes: Q-value update formula: ]; Q=(s,a), Q'=(s',a'), Q''=(s'',a''), R=(s,a); Where Q=(s,a) is the expected score of the drone performing action a in state s, s is the current state of the drone, a is the action that the drone can perform in the current state s, ← is the update symbol, α is the learning rate, R=(s,a) is the reward score of the drone after performing action a in state s, γ is the discount factor, ω is the deep learning factor, A is the set of actions of the drone, Q'=(s',a') is the expected score of the drone performing action a' in state s', and Q''=(s'',a'') is the expected score of the drone performing action a'' in state s''.

[0045] In one embodiment, the event levels include low, medium, and high, and the personnel dispatch module includes: The level judgment module is used to input the area image into the target model to identify the number of people and judge the event level based on the number of people. The judgment execution module is used to determine the level as low if the number of people is less than or equal to the first threshold, as medium if the first threshold is less than or equal to the number of people and less than the second threshold, and as high if the second threshold is less than the number of people.

[0046] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A security personnel scheduling method based on an optimization algorithm, characterized in that, The method includes: Obtain the topological undirected graph and constraints of the target region, and determine the optimization objective based on the topological undirected graph and constraints; The target path is obtained by optimizing according to the optimization objective and constraints, and the target path is sent to the UAV so that the UAV can execute the target path to obtain the region image; The event level is determined by identifying the area image, and security personnel are dispatched based on the event level.

2. The security personnel scheduling method based on an optimization algorithm according to claim 1, characterized in that, The constraints include a first constraint, a second constraint, and a third constraint, including: First constraint: Where t is the total flight time of the UAV, v is the real-time flight speed of the UAV, n is the number of nodes inspected, and t k Let l be the detection time for the k-th node. i t represents the flight distance between adjacent nodes. max This refers to the maximum flight time of the drone; Second constraint: ; where v α-max v is the maximum horizontal movement speed of the drone. β-max v is the maximum ascent speed of the drone. γ-max This represents the maximum descent speed of the drone; The third constraint is: s' = s + ε; where s' is the safe distance between the UAV and the node, s is the preset safe distance, and ε is the hovering error compensation value.

3. The security personnel scheduling method based on an optimization algorithm according to claim 1, characterized in that, The target path is obtained by optimizing according to the optimization objective and constraints, including: Step 1: Calculate the Euclidean distance from the drone to the starting point as the first distance, and calculate the Euclidean distance from the drone to each node to obtain the distance set; Step 2: Filter the second distance in the distance set to obtain the target second distance, and calculate the weighted fusion distance based on the first distance and the target second distance; Step 3: Linearly normalize the weighted fusion distance to obtain normalized data, and invert the normalized data to obtain the initial Q value; Step 4: Initialize the training parameters according to the optimization objective and the constraints, determine the state s of the UAV, select action a according to the initial Q value, and send action a to the UAV to make the UAV execute action a; Step 5: Obtain the state s' of the drone and calculate the reward R. Update the initial Q value according to the reward R to obtain the Q' value. Update the Q table according to the Q' value. The Q table contains multiple Q values. Step 6: Repeat steps 4 and 5 until the termination condition is met, then output the target path; the termination condition is reaching a preset number of iterations or the algorithm converges.

4. The security personnel scheduling method based on an optimization algorithm according to claim 3, characterized in that, The initial Q value is updated based on the reward R to obtain the Q' value. Includes: Q-value update formula: ];Q=(s,a), Q'=(s',a'), Q''=(s'',a''), R=(s,a); where Q=(s,a) is the expected score of the drone performing action a in state s, s is the current state of the drone, a is the action that the drone can perform in the current state s, ← is the update symbol, α is the learning rate, R=(s,a) is the reward score of the drone after performing action a in state s, γ is the discount factor, ω is the deep learning factor, A is the set of actions of the drone, Q'=(s',a') is the expected score of the drone performing action a' in state s', and Q''=(s'',a'') is the expected score of the drone performing action a'' in state s''.

5. The security personnel scheduling method based on an optimization algorithm according to claim 1, characterized in that, The event level is obtained by identifying the region image, and the event level includes low, medium and high levels, including: The area image is input into the target model for identification to obtain the number of people, and the event level is determined based on the number of people. If the number of people is less than or equal to the first threshold, it is classified as low-level; if the first threshold is less than the number of people but less than or equal to the second threshold, it is classified as medium-level; and if the second threshold is less than the number of people, it is classified as high-level.

6. A security personnel dispatching system based on an optimization algorithm, characterized in that, The system includes: The target optimization module is used to obtain the topological undirected graph and constraints of the target region, and determine the optimization target based on the topological undirected graph and constraints. The path optimization module is used to optimize the target path according to the optimization objective and constraints, and send the target path to the UAV so that the UAV can execute the target path to obtain the region image; The personnel dispatch module is used to identify the event level from the area image and dispatch security personnel according to the event level.

7. A security personnel dispatching system based on an optimization algorithm according to claim 6, characterized in that, The constraints include a first constraint, a second constraint, and a third constraint, including: First constraint: Where t is the total flight time of the UAV, v is the real-time flight speed of the UAV, n is the number of nodes inspected, and t k Let l be the detection time for the k-th node. i t represents the flight distance between adjacent nodes. max This refers to the maximum flight time of the drone; Second constraint: ; where v α-max v is the maximum horizontal movement speed of the drone. β-max v is the maximum ascent speed of the drone. γ-max This represents the maximum descent speed of the drone; The third constraint is: s' = s + ε; where s' is the safe distance between the UAV and the node, s is the preset safe distance, and ε is the hovering error compensation value.

8. A security personnel dispatching system based on an optimization algorithm according to claim 6, characterized in that, The path optimization module includes: The first operation module is used to calculate the Euclidean distance from the drone to the starting point as the first distance, and to calculate the Euclidean distance from the drone to each node to obtain the distance set; The second operation module is used to filter the second distance in the distance set to obtain the target second distance, and to calculate the weighted fusion distance based on the first distance and the target second distance; The third operation module is used to linearly normalize the weighted fusion distance to obtain normalized data, and to invert the normalized data to obtain an initial Q value. The fourth operation module is used to initialize training parameters according to the optimization objective and the constraints, determine the state s of the UAV, select action a according to the initial Q value, and send action a to the UAV to make the UAV execute action a. The fifth operation module is used to acquire the state s' of the UAV, calculate the reward R, update the initial Q value according to the reward R to obtain the Q' value, and update the Q table according to the Q' value; the Q table contains multiple Q values; The sixth operation module is used to repeatedly execute the fourth and fifth operation modules until the termination condition is met, at which point the target path is output; the termination condition is reaching a preset number of iterations or the algorithm converges.

9. A security personnel dispatching system based on an optimization algorithm according to claim 8, characterized in that, The fifth operation module also includes: Q-value update formula: ]; Q=(s,a), Q'=(s',a'), Q''=(s'',a''), R=(s,a); Where Q=(s,a) is the expected score of the drone performing action a in state s, s is the current state of the drone, a is the action that the drone can perform in the current state s, ← is the update symbol, α is the learning rate, R=(s,a) is the reward score of the drone after performing action a in state s, γ is the discount factor, ω is the deep learning factor, A is the set of actions of the drone, Q'=(s',a') is the expected score of the drone performing action a' in state s', and Q''=(s'',a'') is the expected score of the drone performing action a'' in state s''.

10. A security personnel dispatching system based on an optimization algorithm according to claim 6, characterized in that, The event levels include low, medium, and high, and the personnel scheduling module includes: The level determination module is used to input the area image into the target model to identify the number of people and determine the event level based on the number of people. The judgment execution module is used to determine the level as low if the number of people is less than or equal to the first threshold, as medium if the first threshold is less than or equal to the number of people and less than the second threshold, and as high if the second threshold is less than the number of people.

Images