A Method for Optimizing Coverage of a Directed Visual Sensor Network

Abstract

The present invention discloses a method for optimizing the coverage of a directed vision sensor network, which includes: establishing a two-dimensional model according to the spatial structure of the target monitoring area, arranging vision sensor nodes in the established two-dimensional model, and establishing a vision sensor position model; constructing a perception model for each vision sensor according to the performance of each vision sensor to obtain the coverage result of each vision sensor; obtaining a vision sensor network coverage system for the entire target monitoring area according to the vision sensor position model and the perception model of the sensor; optimizing the vision sensor network coverage system so that the coverage intensity value of any target point in the target monitoring area is the optimal value in the vision sensor network coverage result. The present invention improves the problems of manpower consumption, weak timeliness and weak pertinence existing in the monitoring technology, reduces manual on-site operations, and greatly saves manpower and material resources.

Description

Technical Field

[0001] The present invention belongs to the field of visual sensor networks, and particularly relates to a method for optimizing the coverage of a directed visual sensor network. Background Technique

[0002] In recent years, the continuous development of technologies such as computers, mobile robots, artificial intelligence, and intelligent control has greatly promoted the intelligence of human social production and life. At the same time, the concepts of smart cities, smart factories, etc. have been put forward, making it urgent for people to apply relevant technologies to actual industrial production. Drones have been favored by a large number of researchers due to their low manufacturing cost, wide range of uses, and other advantages. Multiple drones equipped with visual sensors can achieve their own positioning through the images captured by the cameras and cover the target monitoring scene to complete specific tasks.

[0003] Visual sensors have been widely used due to their advantages such as low cost, light weight, non-contact sensing, and rich information. In recent years, in the face of increasingly complex working environments and high-difficulty tasks, single visual sensors are often unable to cope due to their limited working ability and sensing range. For example, in monitoring tasks, the working angle of a single visual sensor is limited, and it cannot monitor the target area from multiple angles, easily resulting in the omission of important information; the intelligent agents equipped with visual sensors are restricted by the load capacity and can carry limited fuel, so the endurance time is difficult to guarantee; in addition, during the execution of tasks by a single visual sensor, once there are situations such as energy shortage or equipment failure, it may directly lead to the failure of the task. To solve the above-mentioned problems, researchers have been inspired by the activities of biological clusters and combined multiple visual sensors into a visual sensor network (VSN) to complete many tasks, making VSN a hot research topic in recent years. Compared with single visual sensors, visual sensor networks can complete concurrent and comprehensive complex tasks, and their advantages are as follows: 1) In practical applications such as resource exploration, forest fire prevention, and geological disaster monitoring, visual sensor networks observe the target area from multiple directions and angles, and then achieve comprehensive coverage of the target area; 2) Visual sensor networks can improve the success rate of task execution. Even when a certain sensor fails and becomes ineffective, the remaining sensors can still complete the task, improving the reliability and robustness of the system; 3) It can achieve precise positioning through information exchange between the "neighbors" of visual sensors in the case of limited communication equipment or network, and improve the survival rate of intelligent agents in complex and dangerous environments.

[0004] Current visual sensor network coverage technologies are applied in various aspects, but most existing methods focus on isotropic sensing networks. However, for field sensors such as visual sensors, they capture scene information within a certain area in their field of view rather than a single point. Therefore, visual sensor networks have the characteristics of area sensing and directional sensing, which makes their coverage problem more challenging. Currently, most research is based on intelligent algorithms to solve the coverage optimization problem. At the same time, single intelligent algorithms have problems such as premature convergence, slow convergence, and being easily trapped in local optima. Summary of the Invention

[0005] In view of this, the present invention proposes a coverage optimization method for a directed visual sensor network, including the following steps:

[0006] S1. According to the two-dimensional structure of the target monitoring area, establish a two-dimensional model, arrange visual sensor nodes in the established two-dimensional model, and establish a visual sensor position model;

[0007] S2. According to the performance of each visual sensor, construct a perception model for each visual sensor to obtain the coverage result of each visual sensor;

[0008] S3. According to the visual sensor position model and the perception model of the sensor, obtain the visual sensor network coverage system of the entire target monitoring area;

[0009] S4. Optimize the visual sensor network coverage system so that the coverage intensity value of any target point in the target monitoring area is the optimal value in the visual sensor network coverage result.

[0010] Further, step S1 is specifically as follows:

[0011] Construct a coordinate system for the target monitoring area, and the pose coordinates of the sensor node c i are where p i =[x i y i T is the coordinate of the visual sensor c i on the xy coordinate system, θ i is the direction angle, and θ i is the angle between the direction vector i of the visual sensor c and the x direction of the coordinate system.

[0012] Further, step S2 is specifically as follows:

[0013] S21. According to the performance of each visual sensor, set the condition that the point p0 in the target monitoring area can be covered by the sensor node c i : The coordinate p of the sensor node c i ​​i The Euclidean distance to point p0 does not exceed the sensor radius r, and the included angle β with the sensor direction vector is not greater than α / 2;

[0014]

[0015] where the perception radius r and the perception angle α of the vision sensor are fixed performance parameters of the vision sensor;

[0016] S22. Construct a single sensor perception model:

[0017]

[0018] where C s (c i , q k ) represents the coverage performance of sensor c i for point q k , σ is a constant, d v (c i , q k ) represents the distance from sensor c i to the monitoring point q k , and Ω i represents the monitoring range of sensor c i .

[0019] Furthermore, step S3 is specifically as follows:

[0020] For a vision sensor network coverage system composed of N mobile vision sensors, there is a point q in the monitoring area, and it is desired that the perception intensity of the vision sensor for monitoring point q reaches the maximum value. Integrate the importance weight function Φ in the environment with the coverage performance f m of the vision sensor network system to construct the objective function J(s1, s2, …, s N ):

[0021] J(s1, s2, K, s N ) = ∫ Q f m φ(q)dq

[0022] where f m = max{C s (c1, q m ), C s (c2, q m ), …, C s (c N , q m )} represents the perception of the target point q in the monitoring area by the vision sensor network mThe maximum value of the coverage performance, Q represents the entire monitoring environment, Φ(q) represents the importance of point q, and s N represents the pose coordinates of the Nth sensor.

[0023] Furthermore, step S4 is specifically as follows:

[0024] S41. Construct the dynamic equation of the vision sensor c i is the partial derivative of J(s) along the gradient ascent direction:

[0025]

[0026] where u i is the control input of the vision sensor c i including two parts: position and orientation angle, and α i is the control gain;

[0027] S42. Construct the Lyapunov function of J(s):

[0028]

[0029] where s * is the value of the objective function J(s) when it evolves to the maximum along the gradient ascent direction;

[0030] S43. Take the derivative of the Lyapunov function V(s) with respect to time:

[0031]

[0032] When the derivative of the Lyapunov function V(s) with respect to time is less than zero, the dynamic system of the vision sensor c i is in a stable state;

[0033] S44. The gradient control rate of the vision sensor c i is expressed as:

[0034]

[0035] where v i =[v xi v yi T is the first derivative of the position in the pose of the vision sensor c i representing the velocity, and ω i is the first derivative of the angle in the pose of the vision sensor c i representing the angular velocity, then:

[0036]

[0037] ​Sensor c i The kinematics of is described by the following discrete-time system:

[0038] p i (a + 1) = p i (a) + v i (a)

[0039] θ i (a + 1) = θ i (a) + ω i (a)

[0040] where a is the iteration index, v i (a) and ω i (a) are the velocity and angular velocity at the a-th iteration, respectively.

[0041] The beneficial effects brought by the technical solution provided by the present invention are:

[0042] This article will propose an optimization method for the coverage of a directed visual sensor network for a visual sensor network with anisotropic perception to achieve the maximum coverage of the network. Monitoring the environment using visual sensors reduces manual on-site operations and greatly saves manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flowchart of an optimization method for the coverage of a directed visual sensor network according to the present invention;

[0044] Figure 2 is a position model diagram of a visual sensor in an embodiment of the present invention;

[0045] Figure 3 is the initial position distribution of a visual sensor network in an embodiment of the present invention;

[0046] Figure 4 is a diagram of the optimized coverage result of a visual sensor network based on GABA in an embodiment of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the following will further describe the embodiments of the present invention with reference to the accompanying drawings.

[0048] An embodiment of the present invention proposes an optimization method for the coverage of a visual sensor network based on the Gradient Ascent Based Algorithm (GABA). The flowchart of the method is as shown in Figure 1 and includes the following steps:

[0049] S1. Based on the two-dimensional structure of the target monitoring area, establish a two-dimensional model. Arrange visual sensor nodes in the established two-dimensional model and establish a visual sensor position model. Refer to Figure 2 , Figure 2 is the visual sensor position model diagram of the embodiment of the present invention.

[0050] Construct a coordinate system for the target monitoring area. The pose coordinates of sensor node c i are where p i = [x i y i T is the coordinate of visual sensor c i on the xy coordinate system, θ i is the direction angle, and θ i is the direction vector of visual sensor c i and the included angle between the x direction of the coordinate system. and the x direction of the coordinate system.

[0051] S2. According to the performance of each visual sensor, such as the perception radius and perception angle of view, etc., construct the perception model of each visual sensor to obtain the coverage result of each visual sensor.

[0052] S21. First, according to the performance of each visual sensor, set the condition for point p0 in the target monitoring area to be covered by sensor node c i as:

[0053] (1) The Euclidean distance from the coordinate p i of sensor node c i to point p0 does not exceed the sensor radius r:

[0054]

[0055] (2) The included angle β with the direction vector i of sensor c is not greater than α / 2:

[0056]

[0057] Among them, the visual sensor perception radius r and perception angle of view α are fixed performance parameters of the visual sensor.

[0058] S22. Due to the anisotropic perception characteristics of the visual sensor, that is, the perception performance of the sensor not only decays with the distance from the target point to the sensor, but also decays with the direction from the target point to the sensor. Therefore, construct a single sensor perception model:

[0059]

[0060] Among them, C s (c i , q k ) is a monotonically decreasing differentiable function, representing the coverage performance of sensor c i with respect to point q k . κ is a negative constant, and d v (c i , q k ) is the distance from sensor c i to the monitoring point q k . Ω i represents the monitoring range of sensor c i .

[0061] S3. According to the vision sensor position model and the perception model of a single vision sensor, a vision sensor network coverage system for the entire target monitoring area is obtained. Refer to Figure 3 , Figure 3 which is the initial position distribution of the vision sensor network in the embodiment of the present invention.

[0062] For a vision sensor network coverage system composed of N mobile vision sensors, there is a point q in the monitoring area. It is desired that the perception intensity of the vision sensor for monitoring point q reaches the maximum value. Integrate the importance weight function Φ in the environment with the coverage performance f m of the vision sensor network system to construct the objective function J(s1, s2, …, s N ):

[0063] J(s1, s2, K, s N ) = ∫ Q f m φ(q) dq

[0064] Among them, f m = max{C s (c1, q m ), C s (c2, q m ), …, C s (c N , q m )} represents the coverage performance of the vision sensor network. Q represents the entire monitoring environment, that is, the coverage intensity value of the target point q m in the set monitoring area is the optimal value in the vision sensor network coverage result. Φ(q) represents the importance of point q, N is the total number of sensor networks, and s N represents the pose coordinates of the Nth sensor.

[0065] S4. Optimize the vision sensor network coverage system so that the coverage intensity value of any target point in the target monitoring area is the optimal value in the vision sensor network coverage result. Refer to Figure 4 ,Figure 4 This is the optimization result graph of the vision sensor network based on GABA in the embodiment of the present invention.

[0066] S41. Construct the dynamic equation of vision sensor c i as the partial derivative of J(s) in the gradient ascent direction: is the partial derivative of J(s) along the gradient ascent direction:

[0067]

[0068] where u i is the control input of vision sensor c i including two parts: position and direction angle, and α i is the control gain;

[0069] S42. Construct the Lyapunov function of J(s):

[0070]

[0071] where s * is the value when the objective function J(s) evolves to the maximum along the gradient ascent direction;

[0072] S43. Take the derivative of the Lyapunov function V(s) with respect to time:

[0073]

[0074] When the derivative of the Lyapunov function V(s) with respect to time is less than zero, the dynamic system of vision sensor c i is in a stable state;

[0075] S44. The gradient control rate of vision sensor c i is expressed as:

[0076]

[0077] where v i =[v xi v yi T is the first-order derivative of the position in the pose of vision sensor c i representing the velocity, and ω i is the first-order derivative of the angle in the pose of vision sensor c i representing the angular velocity, then:

[0078]

[0079] The kinematics of sensor c i is described by the following discrete-time system:

[0080] pi (a + 1) = p i (a) + v i (a)

[0081] θ i (a + 1) = θ i (a) + ω i (a)

[0082] where a is the number of iteration indices, v i (a) and ω i (a) are the speed and angular speed of the a-th iteration respectively.

[0083] The examples of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit of the present invention and the scope protected by the claims, and all of them fall within the protection scope of the present invention.

Claims

1. A method for optimizing the coverage of a directed visual sensor network, characterized in that, It includes the following steps: S1. Establish a two-dimensional model according to the two-dimensional structure of the target monitoring area. Arrange visual sensor nodes in the established two-dimensional model and establish a visual sensor position model; S2. Construct a perception model for each visual sensor according to the performance of each visual sensor to obtain the coverage result of each visual sensor; S3. Obtain a visual sensor network coverage system for the entire target monitoring area according to the visual sensor position model and the perception model of the sensor; S4. Optimize the visual sensor network coverage system so that the coverage intensity value of any target point in the target monitoring area is the optimal value in the visual sensor network coverage result; Step S1 is specifically as follows: Construct the coordinate system of the target monitoring area, and the visual sensor c i has a pose coordinate of where p i = [x i y i T is the coordinate of the visual sensor c i on the xy coordinate system, θ i is the direction angle, and θ i is the direction vector of the visual sensor and the included angle with the x direction of the coordinate system;​ Step S2 is specifically as follows: S21. Set the condition for point p0 in the target monitoring area to be covered by vision sensor c according to the performance of each vision sensor: i The Euclidean distance from the coordinate p i of vision sensor c i to point p0 does not exceed the radius r of the vision sensor, and the included angle β with the direction vector i of vision sensor c is not greater than α / 2; Among them, the perception radius r and perception angle α of the visual sensor are fixed performance parameters of the visual sensor; S22. Construct a single visual sensor perception model: Among them, C s (c i , q k ) represents the coverage performance of the vision sensor c i for the point q k . σ is a constant, and d v (c i , q k ) represents the distance from the vision sensor c i to the monitoring point q k . Ω i represents the monitoring range of the vision sensor c i ; Step S3 is specifically as follows: Integrate the importance weight function Φ in the environment with the coverage performance f of the visual sensor network system m to construct the objective function J(s1, s2, …, s N ): J(s1,s2,K,s N ) = ∫ Q f m φ(q)dq where, f m = max{C s (c1, q m ), C s (c2, q m ), …, C s (c N , q m )} represents the maximum value of the coverage performance of the visual sensor network for the target point q m in the monitoring area, Q represents the entire monitoring environment, Φ(q) represents the importance of point q, and s N represents the pose coordinates of the Nth visual sensor; Step S4 is specifically as follows: S41. Construct the vision sensor c i The dynamic equation of is the partial derivative of J(s) along the gradient ascent direction: where u i is the control input of the vision sensor c i , including two parts: position and orientation angle, and α i is the control gain; S42. Construct a Lyapunov function of J(s): where s * is the value obtained when the objective function J(s) evolves along the gradient ascent direction to reach the maximum value; S43. Take the derivative of the Lyapunov function V(s) with respect to time: When the derivative of the Lyapunov function V(s) with respect to time is less than zero, the dynamic system of the vision sensor c i is in a stable state; S44. Visual sensor c i The gradient control rate is expressed as: where v i = [v xi v yi T is the first derivative of the position in the pose of the visual sensor c i , representing the velocity, and ω i is the first derivative of the angle in the pose of the visual sensor c i , representing the angular velocity, then:​ Vision sensor c i The kinematics of which is described by the following discrete-time system: p i (a + 1) = p i (a) + v i (a) θ i (a + 1) = θ i (a) + ω i (a) where a is the iteration index number, v i (a) and ω i (a) are the velocity and angular velocity of the a-th iteration, respectively.

Images