Video sensor deployment optimization method and device, equipment and medium

Abstract

The invention discloses a video sensor deployment optimization method, device and equipment and a medium, and relates to the technical field of video surveillance, and the method comprises the steps: simulating the movement and cooperation of a plurality of video sensors in a solution space through employing a particle swarm optimization (PSO) algorithm, finding an initial global optimal deployment scheme, and carrying out the optimization of the video sensor deployment. The interaction between adjacent sensors is simulated by introducing a virtual force algorithm VFA during particle swarm optimization (PSO) simulation, and the virtual force algorithm VFA generates a mutually exclusive virtual force between the adjacent sensors according to the overlapping degree of a sensor coverage area so as to push the sensors to move towards an uncovered area, so that the sensor coverage area is not covered. Therefore, the mutual exclusive relation between the adjacent video sensors in the global situation is converted into optimization factors for guiding the particle swarm optimization (PSO) algorithm, the displacement of each video sensor can be continuously adjusted to achieve global optimization, the phenomenon of node redundancy is eliminated, and the optimal deployment effect of the video sensors is achieved.

Description

Technical Field

[0001] The present invention relates to the field of video surveillance technology, and in particular to a video sensor deployment optimization method, device, equipment and medium. Background Art

[0002] With the construction of smart grids and the advancement of automation technology, video sensors are increasingly used in substations. As an important means of monitoring, video sensors can capture the operating status of equipment in substations in real time, monitor the safety of key areas, and even trigger alarms when abnormal conditions occur. These sensors can not only monitor equipment in substations, such as transformers and circuit breakers, but also identify personnel activities and potential safety hazards. In intelligent management, the application of video sensor networks has greatly improved the operation and maintenance efficiency of substations, reduced the frequency and time of manual inspections, and ensured the safety and stability of the power grid system. Because substations are the core nodes of the power system, their safety is crucial to the reliable operation of the entire power grid. Therefore, it is of great significance to achieve all-weather and all-round monitoring of equipment through video monitoring.

[0003] Because the monitoring area in the substation is wide and complex, a large number of sensors are often deployed to ensure that every corner is covered by video sensors; however, this "large-scale deployment" method can easily lead to node redundancy, increase hardware costs and the complexity of system maintenance; excessive intensive deployment of sensors will not only cause waste of resources, but may also lead to the emergence of monitoring blind spots and overlapping areas; how to optimize the deployment location and viewing angle of sensors while reducing the number of sensors to ensure the coverage of the entire monitoring area has become an urgent problem to be solved.

[0004] In response to this phenomenon, researchers have proposed different algorithms and technical means to optimize the deployment of video sensors; some researchers have proposed heuristic-based algorithms that simulate the optimization process in nature to find the optimal deployment plan for sensors; the particle swarm optimization (PSO) algorithm continuously optimizes the position of each particle (i.e., sensor) by simulating the collaborative behavior of a particle group in the search space; although the traditional PSO algorithm can optimize the deployment of sensors to a certain extent, it does not take into account the interactions between sensors, resulting in a lack of pertinence in the sensor position adjustment process and prone to random fluctuations; at the same time, this type of algorithm still has room for improvement in convergence speed and computational efficiency, especially when dealing with large-scale sensor networks, too many iterations will lead to increased computational complexity, making it difficult to quickly obtain an ideal deployment plan; and this type of algorithm lacks effective processing of the overlapping relationship between sensors, cannot significantly reduce redundant nodes, and has limited optimization effects.

[0005] To address the problems in the above algorithms, a sensor deployment optimization method based on a virtual force model is usually used. The virtual force model generates attraction or repulsion by simulating the interaction between sensor nodes, pushing the sensor nodes to move to uncovered areas. However, this method usually only considers local mutual exclusion relationships and ignores the global coverage of the entire network, which can easily cause the sensor to fall into a local optimum, resulting in node redundancy and poor optimization effect on video sensor deployment. Summary of the invention

[0006] The embodiments of the present invention provide a method, device, equipment and medium for optimizing the deployment of a video sensor, which can solve the problem in the prior art that the current methods usually only consider local mutual exclusion relationships and ignore the global coverage of the entire network, which easily causes the sensor to fall into a local optimum, resulting in node redundancy and poor optimization effect on the deployment of video sensors.

[0007] An embodiment of the present invention provides a video sensor deployment optimization method, comprising the following steps:

[0008] Construct a video sensor network model, which includes multiple randomly deployed video sensors;

[0009] Based on the particle swarm optimization algorithm, multiple video sensors are simulated as particles, the updated position and speed of the particles are simulated as the rotation angle of the video sensor, and the objective function of the particles is simulated as the coverage rate of the video sensor to determine the initial fitness value of each video sensor; the individual extreme value of the video sensor and the group extreme value of the video sensor are determined according to the initial fitness value, and each video sensor continuously updates the initial speed and position of the corresponding video sensor using the corresponding individual extreme value of the video sensor and the group extreme value of the video sensor, and obtains the updated speed and position of each video sensor;

[0010] Based on the virtual force algorithm, multiple video sensors are simulated as virtual charged particles. According to the virtual force guidance factor, the virtual force of each video sensor on the overlapping area with the adjacent video sensor is determined to update the coverage of each video sensor. According to the updated coverage, the updated speed and position of each video sensor are optimized, and the final speed and position of each video sensor are obtained by combining the dynamic weight of the virtual force.

[0011] According to the final speed and position of each video sensor, the rotation angle and rotation direction of the corresponding video sensor are determined to achieve deployment optimization of the video sensors.

[0012] Preferably, the constructing of the video sensor network model includes:

[0013] Randomly deploy N video sensors and M obstacles in any monitoring area ω;

[0014] The direction angle of the video sensor is defined as η, R is defined as the farthest visible distance, γ is defined as the maximum field of view, and the maximum area that can be monitored is called FOV, which is a fan-shaped area;

[0015] Define θ' as the overlapping area of ​​the FOV area of ​​the video sensor and the FOV area of ​​another video sensor, and define θ' as the blocked area of ​​the FOV area of ​​the video sensor to form a video sensor network model.

[0016] Preferably, the continuously updating the initial speed and position of the corresponding video sensor includes:

[0017] According to each particle X in the particle swarm optimization algorithm PSO i Corresponding to the initial coverage of the video sensor, determine the initial fitness value of each video sensor, and determine the individual extreme value P of the video sensor according to the initial fitness value i =(P i1 ,P i2 ,...,P iD ) T and the video sensor population extreme value P g =(P g1 ,P g2 ,...,P gD ) T ;

[0018] Each video sensor continuously updates the initial speed and position of the corresponding video sensor using the corresponding video sensor individual extreme value and the video sensor group extreme value;

[0019] The initial speed and position of the video sensor during the continuous update are respectively expressed as:

[0020]

[0021] Where: ω represents the inertia weight; d = 1, 2, ..., N; k represents the current iteration number; V id represents the speed of the particle; c 1 and c 2 represents a non-negative constant; r 1 and r 2 Represents a random number distributed in the interval [0,1].

[0022] Preferably, obtaining the updated speed and position of each video sensor includes:

[0023] Determine each particle X in the particle swarm optimization algorithm PSO iCorresponding to the initial coverage of the video sensor, the optimal position of the i-th particle is defined as P i , then P i It is expressed as:

[0024]

[0025] The optimal position experienced by all groups in the particle swarm is expressed as: P g (t) = Min (f (x 1 ), f(x 1 ), ···, f(x 2 ), f(x 1 )), then the updated speed and position of the particle from time to time are:

[0026] V i (t+1)=δ(t)·V i (t)+C 1 ·r 1 (t)·(P i (t)-X i (t))+C 2 ·r 2 (t)·(P g (t)-X i (t))

[0027] X i (t+1)=X i (t)+V i (t+1)

[0028] Where: δ(t) represents the convergence factor; C 1 represents the local acceleration factor; C 2 represents the global acceleration factor; r 1 (t) and r 2 (t) represent random functions that generate random values ​​between [0, 1].

[0029] Preferably, updating the coverage of each video sensor includes:

[0030] In the video sensor network model, θ” represents the overlapping area of ​​the FOV area of ​​video sensor 1 and the FOV area of ​​video sensor 2, and θ’ represents the occluded area of ​​the FOV area of ​​video sensor 1;

[0031] In order to reduce the area of ​​θ' and θ", the repulsive forces from the virtual θ' and θ" areas to the center point of the video sensor 1 are virtualized by the virtual force method VFA, which are F 1 and F 2 ; F 1 and F 2Decomposed into two components of the direction angle η parallel to and perpendicular to the video sensor 1, the parallel component is F' 1 'and F' 2 ', the vertical component is F' 1 and F' 2 ; The vertical component is the main component that determines the rotation of the video sensor 1;

[0032] Assume that the center point of the FOV of the i-th video sensor is The center point of the jth overlapping region is Then the resultant force on video sensor i is expressed as The corresponding video sensor is rotated according to the resultant force received by the video sensor to update the initial coverage of each video sensor.

[0033] Preferably, obtaining the final speed and position of each video sensor includes:

[0034] In the virtual force algorithm VFA, the dynamic weight of the virtual force is expressed as 1-ψ(t), and ψ(t) is expressed as:

[0035]

[0036] Among them: MaxStep represents the maximum step of the particle swarm optimization algorithm PSO;

[0037] According to the dynamic weight of the virtual force, the final speed and position of each video sensor are obtained as follows:

[0038]

[0039] X i (t+1)=X i (t)+V i (t+1)

[0040] in: represents the virtual force guidance factor; δ(t) represents the convergence factor; c 1 and c 2 represents a non-negative constant; r 1 and r 2 Represents a random number distributed in the interval [0,1].

[0041] Preferably, when determining the rotation angle of the corresponding video sensor, the current overlapping area θ'+θ'' of each video sensor is recorded, and the current overlapping area is compared with the minimum value experienced. If the current overlapping area is less than the minimum value, it is updated to the minimum value, and the direction of the current video sensor is recorded as the minimum value direction;

[0042] Compare the sum of the overlapping areas θ'+θ'' of all current video sensors with the historical θ'+θ''. When the current θ'+θ'' is less than the historical minimum θ'+θ'', update the current value to the historical value and record the set of direction angles of all current video sensors [η 1 ,η 2 ,···,η N ].

[0043] The embodiment of the present invention further provides a video sensor deployment optimization device, including:

[0044] A deployment module is used to build a video sensor network model, which includes multiple randomly deployed video sensors;

[0045] The particle swarm simulation module is used to simulate multiple video sensors as particles based on the particle swarm optimization algorithm, simulate the updated position and speed of the particles as the rotation angle of the video sensor, and simulate the objective function of the particles as the coverage rate of the video sensor to determine the initial fitness value of each video sensor; determine the individual extreme value of the video sensor and the group extreme value of the video sensor according to the initial fitness value, and each video sensor continuously updates the initial speed and position of the corresponding video sensor using the corresponding individual extreme value of the video sensor and the group extreme value of the video sensor to obtain the updated speed and position of each video sensor;

[0046] A virtual force simulation module is used to simulate multiple video sensors as virtual charged particles based on a virtual force algorithm; determine the virtual force of each video sensor on the video sensor in the overlapping area with the adjacent video sensor according to the virtual force guidance factor, so as to update the coverage of each video sensor; optimize the updated speed and position of each video sensor according to the updated coverage, and obtain the final speed and position of each video sensor in combination with the dynamic weight of the virtual force;

[0047] The deployment optimization module is used to determine the rotation angle and rotation direction of the corresponding video sensor according to the final speed and position of each video sensor, so as to optimize the deployment of the video sensor.

[0048] An embodiment of the present invention further provides an electronic device, including a memory and a processor;

[0049] The memory is used to store computer programs;

[0050] The processor is used to implement the steps of the above-mentioned video sensor deployment optimization method when executing the computer program stored in the memory.

[0051] An embodiment of the present invention further provides a computer-readable storage medium for storing a computer program, wherein when the computer program is executed by a processor, the steps of the video sensor deployment optimization method described above are implemented.

[0052] The embodiments of the present invention provide a video sensor deployment optimization method, device, equipment and medium. Compared with the prior art, the advantages thereof are as follows:

[0053] The present invention utilizes a particle swarm optimization algorithm (PSO) to simulate the movement and cooperation of multiple video sensors in a solution space, finds an initial global optimal deployment scheme, and introduces a virtual force algorithm (VFA) to simulate the interaction between adjacent sensors while simulating the particle swarm optimization algorithm (PSO). The virtual force algorithm (VFA) generates a mutually exclusive virtual force between adjacent sensors according to the degree of overlap of the sensor coverage areas, so as to push the sensors to move to uncovered areas, thereby converting the mutually exclusive relationship between adjacent video sensors in the global space into an optimization factor guiding the particle swarm optimization algorithm (PSO), and can continuously adjust the displacement of each video sensor to achieve global optimality, eliminate the phenomenon of node redundancy, and enable the video sensors to achieve optimal deployment effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of the overall process of a video sensor deployment optimization method provided by an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of a video sensor model of a video sensor deployment optimization method provided by an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of the effect of the direction adjustment of a video sensor on the coverage rate of a video sensor deployment optimization method provided in an embodiment of the present invention; (a) is a schematic diagram of the original sensor coverage; (b) is a schematic diagram of the coverage after the sensor is optimized;

[0057] Figure 4 A schematic diagram of a function extreme value optimization algorithm flow based on a PSO algorithm for a video sensor deployment optimization method provided by an embodiment of the present invention;

[0058] Figure 5 A schematic diagram of force analysis of a video sensor of a video sensor deployment optimization method provided by an embodiment of the present invention;

[0059] Figure 6 A schematic diagram of initial node coverage distribution of a video sensor deployment optimization method provided by an embodiment of the present invention;

[0060] Figure 7A schematic diagram of an algorithm training process of a video sensor deployment optimization method provided by an embodiment of the present invention;

[0061] Figure 8 A schematic diagram of node coverage distribution after optimization of a video sensor deployment optimization method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

[0063] See also Figure 1 , an embodiment of the present invention provides a video sensor deployment optimization method, comprising the following steps:

[0064] Step 1: Randomly deploy M obstacles with random positions and shapes within W, and randomly generate N video sensors, whose positions and initial angles are all random values. N represents the number of video sensors, W represents the area to be monitored, and M represents the number of obstacles.

[0065] Step 2: For each camera, i ranges from 1 to N; including:

[0066] Step 1: Get the parameters of i [x i ,y i , η i ], and detect the adjacent video sensors of the current video sensor and the obstacle set S i .

[0067] Step 2: For each region belonging to S.

[0068] ① Calculate the overlapping area and the centroid points θ' and θ" of the overlapping area.

[0069] ②When θ' and θ" are not empty.

[0070] ③Calculate the current repulsive force F 1 and F 2 , and the resultant force F.

[0071] ④Determine the rotation direction by virtual force analysis of the video sensor.

[0072] Step 3: Determine the rotation angle using the rotation angle definition formula

[0073] Step 4: Adjust the current direction angle using the updated speed and position formula.

[0074] Step 5: For each video sensor, record the current overlapping area θ'+θ'' and compare it with the minimum value experienced. If the current value is less than the minimum value pbest, update the minimum value pbest and record the direction of the current video sensor as the minimum direction.

[0075] Step 6: Compare the sum of the overlapping areas θ'+θ" of all current video sensors with the historical θ'+θ". When the current θ'+θ" is less than the historical minimum θ'+θ", update the current value to the historical value g. best , and record the set of all direction angles at this time [η 1 ,η 2 ,···,η N ].

[0076] Step 3: Update the number of iterations.

[0077] Step 4: Use the globally optimal direction angle as the target direction angle output.

[0078] Specifically:

[0079] 1. Network model of video sensor.

[0080] Assume that N video sensors and M obstacles are randomly deployed in any monitoring area ω; the direction angle of the video sensor is defined as η, R is the farthest visible distance, γ is the maximum field of view angle, and the maximum area that can be monitored is called FOV, which is a fan-shaped area; the FOV of the video sensor may overlap with the FOV of other sensors, and θ" is defined as the overlapping area; in addition, the FOV of the video sensor may also be blocked by obstacles, and θ' is defined as the blocked area; Figure 2 The figure shows a simple monitoring scenario consisting of two video sensors, and the definitions of relevant parameters are marked in the figure.

[0081] Compared with traditional sensor networks, video sensor networks are different in that they are direction-aware. Each video sensor can only capture events within the FOV, including the size and position of overlapping and occluded areas, and transmit event information within the coverage area to the control center. The direction angle of the video sensor is adjustable. The initial direction angle η of the i-th video sensor is iis a random value in the range of [0-2π]. After deployment, the position is fixed, but the direction angle can be adjusted independently, and the range is also [0-2π]. Once the monitoring area is determined, its coverage depends on the number of video sensors and the FOV parameter. However, due to the existence of overlapping areas θ” and θ', the coverage area cannot reach the sum of the FOVs of all sensors. When the total number N is determined, the total coverage area ω' in the monitoring area ω is obtained by the following formula.

[0082]

[0083] The overlapping areas θ” and θ’ jointly determine the constants N, R and γ, so the optimization goal is to minimize the overlapping areas θ” and θ’ to maximize the monitoring area ω’; the coverage of the video sensor network is expressed by the following formula.

[0084] f(η 1 ,η 2 ,···,η N )=Min(∑θ'+∑θ”-∪θ”)

[0085] That is, the optimization goal is to find a set of angles [η 1 ,η 2 ,···,η N ] maximizes the monitoring area; [η 1 ,η 2 ,···,η N ] not only determines f(η 1 ,η 2 ,···,η N ), and due to the different rotation angles, [η 1 ,η 2 ,···,η N ] is still the determining variable of θ" and θ', but the rotation angle can be simplified into a discrete value; if there are K rotatable angles, this problem can be transformed into an optimization problem of selecting the best solution from N×K solutions; however, due to the large number of video sensors, calculating the coverage of all solutions requires a huge amount of calculation, so an approximate calculation method is urgently needed to quickly obtain a better adjustment solution; therefore, the present invention adopts a particle swarm optimization method to obtain an approximate optimization result.

[0086] There are many types of optimization, such as Figure 3 As shown in the figure, if the position of the video sensor cannot be changed, but after adjusting the direction angle, Figure 3 The coverage shown in (b) is significantly higher than that of random deployment. Figure 3As shown in (a) in the figure, this optimization is used as an optimization scheme; Particle Swarm Optimization (PSO) is an optimization algorithm used to find approximate solutions to complex problems. It is simple, efficient and easy to implement. It usually provides better solutions to NP-hard problems.

[0087] 2. Particle swarm optimization (PSO) algorithm.

[0088] 1. The basic model of PSO algorithm.

[0089] Particle Swarm Optimization (PSO) is an optimization algorithm derived from swarm intelligence, first proposed by Kennedy and Eberhart in 1995. The algorithm is inspired by the predatory behavior of birds. When birds look for food, the most effective strategy is to search the area around the bird closest to the food. The PSO algorithm uses this biological behavior characteristic to solve optimization problems, in which each particle represents a potential solution, and the fitness value is determined by the fitness function. The speed of the particle determines its moving direction and distance, and the speed is dynamically adjusted according to the movement experience of itself and other particles, thereby achieving optimization in the solvable space.

[0090] Assume a D-dimensional search space, a population X consisting of n particles = (X 1 ,X 2 ,...,X n ), where the i-th particle is represented by a D-dimensional vector X i =(X i1 ,X i2 ,...,X iD ) T , represents the position of the ith particle in the D-dimensional search space, which is a potential solution to the problem; the position X of each particle can be calculated according to the objective function i The corresponding fitness value; the speed of the i-th particle is V = (V i1 ,V i2 ,...,V iD ) T , whose individual optimal extreme value is P i =(P i1 ,P i2 ,...,P iD ) T , the optimal extreme value of the population is P g =(P g1 ,P g2 ,...,P gD ) T .

[0091] In each iteration, particles update their own speed and position through individual extrema and group extrema, as shown in the following two formulas.

[0092]

[0093] Where: ω represents the inertia weight; d = 1, 2, ..., N; k represents the current iteration number; V id represents the speed of the particle; c 1 and c 2 represents a non-negative constant, called the acceleration factor; r 1 and r 2 represents a random number distributed in the interval [0,1]. To prevent blind searching of particles, the position and speed are generally limited to a certain interval [-X max ,X max ]、[-V max ,V max ].

[0094] The process of the function extreme value optimization algorithm based on the PSO algorithm is as follows Figure 4 As shown. Among them, particle and speed initialization is to randomly initialize particle speed and particle position; calculate particle fitness value as needed; determine individual extreme value and group extreme value according to initial particle fitness value; update particle speed and position according to the above two formulas; update individual extreme value and group extreme value according to particle fitness value in the new population.

[0095] 2. Particle swarm optimization and its application in video sensor networks.

[0096] Particle swarm optimization is generally used to solve approximate solutions to complex problems. It is simple, efficient and easy to implement. Typical NP-hard problems are generally solved by particle swarm optimization. PSO introduces the best position of each particle and the best position of the entire particle swarm, and uses two random quantities to adjust the position. Assume that there are N particles, and their current position is X(x 1 ,x 2 ,···,x n ), the moving speed is V(v 1 ,v 2 ,···,v n ); the optimal position of the i-th particle is defined as P i , so the minimization objective function is as follows.

[0097]

[0098] The best position experienced by all groups of particle swarm is: P g (t) = Min (f (x 1 ), f(x 1 ), ···, f(x 2 ), f(x 1 )), the particle update position and velocity from time t to time f+1 are determined by the following two formulas.

[0099] V i (t+1)=δ(t)·V i (t)+C 1 ·r 1 (t)·(P i (t)-X i (t))+C 2 ·r 2 (t)·(P g (t)-X i (t))

[0100] X i (t+1)=X i (t)+V i (t+1)

[0101] Where: δ(t) represents the convergence factor, which is used to adjust the local and overall convergence; when δ(t) is large, the local convergence is strong; when δ(t) is small, the global convergence is good; C 1 represents the local acceleration factor, C 2 represents the global acceleration factor, which is used to adjust the weights of local optimum and global optimum respectively; r 1 (t) and r 2 (t) Generate random functions with random values ​​between [0, 1] for evolution; From the above, we can see that the movement speed of a particle is determined by the current position of the particle, its own best position and the global best position.

[0102] If the rotation angle of the video sensor [η 1 ,η 2 ,···,η N ] is used as the update position of the particle swarm, and the coverage rate ω' is used as the objective optimization function of the particle swarm. The particle swarm algorithm can be used for coverage optimization problems.

[0103] However, the particle swarm optimization algorithm moves randomly within the movable range of the particles, and each particle needs multiple attempts to converge to a better position to obtain a globally optimized deployment solution; but in wireless video sensor networks, battery life is an important consideration, and such repeated attempts not only increase deployment time but also waste electricity; therefore, the present invention improves the standard particle swarm algorithm and introduces a virtual force-directed factor.

[0104] 3. Virtual force algorithm.

[0105] The virtual force algorithm (VFA) regards mobile sensor nodes as virtual charged particles; when the distance between nodes is less than a certain threshold, repulsion will occur between the nodes; when the distance is greater than the threshold, attraction will occur; the virtual force on a node is the resultant force of all neighboring nodes on it.

[0106] 1. Virtual force analysis of video sensors.

[0107] Although each video sensor is relatively independent, they are not completely unconnected. Within their respective FOVs, they can sense their adjacent sensors. The existence of this positional relationship enables the virtual force mechanism to be utilized. The purpose of introducing this mechanism in the present invention is to make adjacent video sensors repel each other and expand the coverage. Traditional sensor networks also have cases of using virtual force to optimize deployment, but they are mainly suitable for non-directional sensors, and the effect is not good for directional video sensors. Therefore, it is necessary to perform specific force analysis on video sensors.

[0108] The present invention takes video sensor 1 as an example to calculate the virtual force condition of a video sensor, such as Figure 5 As shown in the figure, the θ' area of ​​the video sensor 1 is blocked by obstacles, while the θ" area overlaps with the FOV of the video sensor 2. In order to reduce the area of ​​θ' and θ", the repulsive force from the center point of the θ' and θ" area to the center point of the video sensor 1 can be virtualized, denoted as F 1 and F 2 According to the video sensor network attribute 1, that is, the position of the video sensor is fixed but the direction angle can be adjusted, firstly, F 1 and F 2 Decomposed into two components parallel to and perpendicular to the video sensor direction angle η, where the parallel component is F' in this example 1 'and F' 2 ', and the vertical component is F' 1 and F' 2 ;like Figure 5 As shown, the vertical component F' 1 and F' 2 is the main component that determines the rotation of the video sensor, while the parallel component F' 1 'and F' 2 'The video sensor is fixed and does not work; when the center of mass is acted upon by multiple vertical components, the resultant force determines the direction of rotation, and the magnitude of the resultant force determines the angle of rotation; Figure 5 In, F' 1 >F' 2 , so the rotation direction of video sensor 1 is clockwise. Similarly, the rotation direction of video sensor 2 is also clockwise. Without loss of generality, let the center point of the FOV of the i-th video sensor be The center point of the jth overlapping region is The strength of the virtual force of the jth overlapping area on the video sensor i is determined by the size of the overlapping area, and the rotation angle It is expressed as:

[0109]

[0110] Among them: λ represents a positive constant, called the adjustment coefficient, which is set according to the FOV area; the coverage area is positively correlated with the repulsive force on the video sensor, which facilitates the video sensor to rotate quickly to reduce the repulsive force; when there is no overlapping area, the repulsive force on the video sensor is zero, that is, the video sensor has been adjusted to its optimal position.

[0111] The resultant force on the video sensor i is expressed as: The vertical component of Determines the rotation direction of the video sensor i; after obtaining a virtual force analysis of the video sensor FOV, substitute the virtual force rotation angle into the formula for updating the particle position and velocity, that is, introduce a virtual force guidance factor into the rotation angle when When it is larger, it indicates that the current video sensor is subject to a larger virtual force, so it needs to rotate as quickly as possible to quickly get rid of local constraints. When it is small, the global optimization objective function of the particle swarm determines whether to rotate the angle.

[0112] 2. Virtual force-guiding factor.

[0113] There are certain requirements for convergence in the PSO optimization algorithm. The addition of new guiding factors should not affect the convergence of PSO. The dynamic weight of the virtual force guiding factor is dynamically adjusted according to time. The weight of the virtual force is 1-ψ(t), and ψ(t) is expressed as:

[0114]

[0115] Among them: MaxStep represents the maximum step of the algorithm, which is set as a constant; in the early stage of the algorithm, the weight of the virtual force factor is large, which can make the video sensor with large force rotate quickly and escape from the local force constraint, thereby optimizing the initial value of the particle swarm and improving the optimization performance; when When the weight is small, it indicates that the video sensor is close to the local optimal solution. At this time, the optimization is determined by the global optimal solution and does not affect the convergence of the PSO algorithm. Therefore, after adding the virtual force guidance factor, the formula for updating the position and velocity of the particle from t to f+1 will be updated to the following formula.

[0116]

[0117] X i (t+1)=X i (t)+V i (t+1).

[0118] Specific test:

[0119] like Figure 6 As shown in the figure, in the simulation experiment, the number of randomly generated initial nodes is 20, and the initial coverage rate is 0.61. It can be seen from the figure that some nodes have high redundancy and low coverage rate. Through algorithm improvement and iteration, the optimal solution is found, that is, the coverage rate of the nodes is optimized, as shown in the figure. Figure 7 The figure shows a schematic diagram of the algorithm training process.

[0120] like Figure 8 As shown, by optimizing the completely random video sensor node deployment scheme, the number of nodes is 16, which is reduced by 20%, the coverage rate is increased to 86%, and the fitness of the optimized algorithm is improved to 80.3.

[0121] The present invention proposes a virtual force-guided particle swarm optimization (PSO) algorithm, which is used to solve the node redundancy problem in video sensor deployment. The main purpose is to reduce the number of sensor nodes, while improving monitoring coverage and optimizing resource allocation. The algorithm simulates the interaction between adjacent sensors by introducing a virtual force mechanism, and transforms their mutually exclusive relationship into an optimization factor guiding the particle swarm algorithm. In a specific implementation, the virtual force mechanism can generate a mutually exclusive virtual force according to the overlapping degree of the sensor coverage area, and push the sensor to move to the uncovered area, thereby reducing the overlapping coverage of the sensor. In addition, the present invention uses the virtual force as a guiding factor for particle swarm optimization, and guides the particle swarm to adjust its position in a more targeted manner in the search space, thereby avoiding the random adjustment process in the traditional PSO algorithm. The improvement effectively accelerates the convergence speed of the algorithm and significantly reduces the number of sensors required.

[0122] As an improved optimization algorithm, the present invention combines the advantages of the virtual force model and the PSO algorithm, and guides the movement of the particle swarm through the virtual force, making the sensor position adjustment more efficient and targeted; it not only accelerates the convergence speed of the algorithm, but also reduces the number of sensor nodes, optimizes the coverage, and avoids node redundancy; at the same time, the improved algorithm can effectively jump out of the local optimal solution and further improve the global monitoring effect of the network.

[0123] The present invention has a significant optimization effect in the optimal deployment of video sensors; firstly, by introducing a virtual force-guided factor, the coverage rate of the sensor network is improved, monitoring blind spots and repeated coverage are effectively avoided, and comprehensive monitoring of key areas is ensured; secondly, by optimizing the deployment scheme of the sensor, the present invention significantly reduces the number of nodes, thereby reducing hardware costs and maintenance costs, which not only reduces the initial investment burden, but also reduces long-term operating expenses; in addition, the improved particle swarm optimization algorithm combined with the virtual force-guided factor makes the algorithm more efficient and converges quickly when searching for the optimal solution, reduces the number of iterations, and thus improves the overall optimization efficiency; the efficient algorithm formed by the present invention enables the monitoring system to respond more quickly to monitoring needs in different environments, and enhances the flexibility and adaptability of the system.

[0124] The sensor node layout optimization based on the PSO algorithm in the present invention is to optimize the sensor node layout through the improved PSO algorithm, so as to ensure that the high coverage rate is maintained while reducing the number of nodes; for the virtual force guidance factor, specifically, the virtual force guidance factor is introduced into the deployment optimization of the video sensor, and the overlapping area is effectively reduced by simulating the mutual repulsion between sensors, and the coverage rate of the sensor is improved, thereby improving the overall monitoring effect; for the fitness function design, the fitness function combines the weighted combination of coverage rate, redundant coverage and the number of sensors, and provides an effective optimization target for the PSO algorithm. For the node redundancy minimization strategy, specifically, through the design of the fitness function and the iterative optimization of the algorithm, the redundant coverage of the sensor nodes is effectively reduced, and the deployment cost and energy consumption are reduced.

[0125] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A video sensor deployment optimization method, characterized in that: The following steps are involved: Constructing a video sensor network model, wherein the video sensor network model includes a plurality of randomly deployed video sensors; Based on the particle swarm optimization algorithm, multiple video sensors are simulated as particles, the updated position and speed of the particles are simulated as the rotation angle of the video sensor, and the objective function of the particles is simulated as the coverage rate of the video sensor to determine the initial fitness value of each video sensor; the individual extreme value of the video sensor and the group extreme value of the video sensor are determined according to the initial fitness value, and each video sensor continuously updates the initial speed and position of the corresponding video sensor using the corresponding individual extreme value of the video sensor and the group extreme value of the video sensor, and obtains the updated speed and position of each video sensor; Based on the virtual force algorithm, multiple video sensors are simulated as virtual charged particles; According to the virtual force guidance factor, the virtual force of each video sensor on the video sensor in the overlapping area with the adjacent video sensor is determined to update the coverage rate of each video sensor; according to the updated coverage rate, the updated speed and position of each video sensor are optimized, and the final speed and position of each video sensor are obtained in combination with the dynamic weight of the virtual force; According to the final speed and position of each video sensor, the rotation angle and rotation direction of the corresponding video sensor are determined to achieve deployment optimization of the video sensors.

2. A video sensor deployment optimization method according to claim 1, characterized in that: The video sensor network model is constructed, including: Randomly deploy N video sensors and M obstacles in any monitoring area ω; The direction angle of the video sensor is defined as η, R is defined as the farthest visible distance, γ is defined as the maximum field of view, and the maximum area that can be monitored is called FOV, which is a fan-shaped area; Define θ' as the overlapping area of ​​the FOV area of ​​the video sensor and the FOV area of ​​another video sensor, and define θ' as the occluded area of ​​the FOV area of ​​the video sensor, to form a video sensor network model.

3. The video sensor deployment optimization method according to claim 1, characterized in that: The continuously updating the initial speed and position of the corresponding video sensor includes: According to each particle X in the particle swarm optimization algorithm PSO i Corresponding to the initial coverage of the video sensor, determine the initial fitness value of each video sensor, and determine the individual extreme value P of the video sensor according to the initial fitness value i =(P i1 ,P i2 ,...,P iD ) T and the video sensor population extreme value P g =(P g1 ,P g2 ,...,P gD ) T ; Each video sensor continuously updates the initial speed and position of the corresponding video sensor using the corresponding video sensor individual extreme value and the video sensor group extreme value; The initial speed and position of the video sensor during the continuous update are respectively expressed as: Where: ω represents the inertia weight; d = 1, 2, ..., N; k represents the current iteration number; V id represents the velocity of the particle; c1 and c2 represent non-negative constants; r1 and r2 represent random numbers distributed in the interval [0,1].

4. The video sensor deployment optimization method according to claim 1, characterized in that: The obtaining of the updated speed and position of each video sensor includes: Determine each particle X in the particle swarm optimization algorithm PSO i Corresponding to the initial coverage of the video sensor, the optimal position of the i-th particle is defined as P i , then P i It is expressed as: The optimal position experienced by all groups in the particle swarm is expressed as: P g (t) = Min (f(x1), f(x1), ..., f(x2), f(x1)), then the updated speed and position of the particle from time to time are: V i (t+1)=δ(t)·V i (t)+C1·r1(t)·(P i (t)-X i (t))+C2·r2(t)·(P g (t)-X i (t))X i (t+1)=X i (t)+V i (t+1) Where: δ(t) represents the convergence factor; C1 represents the local acceleration factor; C2 represents the global acceleration factor; r1(t) and r2(t) represent random functions that generate random values ​​between [0, 1] respectively.

5. The video sensor deployment optimization method according to claim 1, characterized in that: The updating of the coverage of each video sensor includes: In the video sensor network model, θ” represents the overlapping area of ​​the FOV area of ​​video sensor 1 and the FOV area of ​​video sensor 2, and θ’ represents the occluded area of ​​the FOV area of ​​video sensor 1; In order to reduce the area of ​​θ' and θ", the repulsive forces from the virtual θ' and θ" areas to the center point of the video sensor 1 are F1 and F2 respectively according to the virtual force method VFA; F1 and F2 are decomposed into two components parallel to and perpendicular to the direction angle η of the video sensor 1, the parallel components are F'1' and F'2', and the perpendicular components are F'1 and F'2; the perpendicular component is the main component that determines the rotation of the video sensor 1; Assume that the center point of the FOV of the i-th video sensor is The center point of the jth overlapping region is Then the resultant force on video sensor i is expressed as The corresponding video sensor is rotated according to the resultant force received by the video sensor to update the coverage rate of each video sensor.

6. The video sensor deployment optimization method according to claim 1, characterized in that: The obtaining of the final speed and position of each video sensor comprises: In the virtual force algorithm VFA, the dynamic weight of the virtual force is expressed as 1-ψ(t), and ψ(t) is expressed as: Among them: MaxStep represents the maximum step of the particle swarm optimization algorithm PSO; According to the dynamic weight of the virtual force, the final speed and position of each video sensor are obtained as follows: X i (t+1)=X i (t)+V i (t+1) in: represents the virtual force guidance factor; δ(t) represents the convergence factor; c1 and c2 represent non-negative constants; r1 and r2 represent random numbers distributed in the interval [0,1].

7. The video sensor deployment optimization method according to claim 5, characterized in that: When determining the rotation angle of the corresponding video sensor, record the current overlapping area θ'+θ'' of each video sensor, compare the current overlapping area with the minimum value experienced, and if the current overlapping area is less than the minimum value, update it to the minimum value, and record the direction of the current video sensor as the minimum value direction; Compare the sum of the overlapping areas θ'+θ'' of all current video sensors with the historical θ'+θ''. When the current θ'+θ'' is less than the historical minimum θ'+θ'', update the current value to the historical value and record the set of all current video sensor direction angles [η1,η2,···,η N ].

8. A video sensor deployment optimization device, characterized in that: include: A deployment module, used to construct a video sensor network model, wherein the video sensor network model includes a plurality of randomly deployed video sensors; The particle swarm simulation module is used to simulate multiple video sensors as particles based on the particle swarm optimization algorithm, simulate the updated position and speed of the particles as the rotation angle of the video sensor, and simulate the objective function of the particles as the coverage rate of the video sensor to determine the initial fitness value of each video sensor; determine the individual extreme value of the video sensor and the group extreme value of the video sensor according to the initial fitness value, and each video sensor continuously updates the initial speed and position of the corresponding video sensor using the corresponding individual extreme value of the video sensor and the group extreme value of the video sensor to obtain the updated speed and position of each video sensor; A virtual force simulation module, used for simulating multiple video sensors as virtual charged particles based on a virtual force algorithm; According to the virtual force guidance factor, the virtual force of each video sensor on the video sensor in the overlapping area with the adjacent video sensor is determined to update the coverage rate of each video sensor; according to the updated coverage rate, the updated speed and position of each video sensor are optimized, and the final speed and position of each video sensor are obtained in combination with the dynamic weight of the virtual force; The deployment optimization module is used to determine the rotation angle and rotation direction of the corresponding video sensor according to the final speed and position of each video sensor, so as to optimize the deployment of the video sensor.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer programs; The processor is used to implement the steps of a video sensor deployment optimization method as described in any one of claims 1 to 7 when executing the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, which, when executed by a processor, implements the steps of a video sensor deployment optimization method as described in any one of claims 1 to 7.

Images