Sensing camera intelligent deployment and control method and device based on spatial differential evolution

Abstract

The invention provides a sensing camera intelligent deployment and control method and device based on spatial differential evolution, and the method comprises the steps: creating a scene layer for a research region, generating ground target points which are not located on an obstacle according to a preset density based on the scene layer, selecting candidate points of a camera for the research region, constructing sight lines from all the candidate points to all the ground target points, performing intervisibility analysis on the sight lines to obtain primarily-selected visual points, screening out a final-selected visual point from the primarily-selected visual points as a visible ground target point based on the visual distance and the visual angle of the camera, setting a camera deployment and control number, and setting a camera deployment and control number; and based on the visible ground target point, selecting an optimal combination at all candidate points by adopting a spatial adaptive differential evolution algorithm, and taking the optimal combination as a camera deployment and control scheme of the research area. According to the invention, sampling points are selected at fixed intervals on the ground based on an actual scene, and camera deployment and control selection is carried out by adopting a spatial differential evolution algorithm, so that a blind area can be minimized, and a maximum monitoring coverage range can be realized.

Description

Technical Field

[0001] The present invention belongs to the field of geographic information science, and specifically relates to a method and device for intelligent deployment of perception cameras based on spatial differential evolution. Background Art

[0002] With the rapid development of information technology and network technology, monitoring systems are no longer only used in the communication, transportation, security and other industries. They are gradually developing towards other industries and the public, and occupying an increasingly important position in the construction of modern smart cities. The social security situation is becoming more and more complicated, public safety issues are constantly highlighted, urban crimes are prominent, and means are constantly updated and upgraded. These all urgently require the acceleration of the development of video surveillance systems that focus on active prevention. Therefore, the reasonable layout and installation location of the camera are issues that we need to give priority to. The traditional camera installation process is usually selected by experienced personnel based on experience, which leads to many problems, such as the existence of monitoring blind spots and excessive overlap of camera coverage.

[0003] With limited available resources, we always hope to obtain the largest range of actual monitoring effective areas, that is, to use a fixed number of cameras to achieve the maximum coverage of the target area, so as to maximize the utilization rate and avoid wasting resources. Although the traditional video surveillance installation method has certain rationality, it does not quantitatively analyze the possible problems such as blind spots in monitoring and overlapping coverage. In addition, there are many points to choose from. It is time-consuming and laborious to adjust the monitoring according to the effect after installation, which is very unrealistic. Therefore, this greatly reduces the utilization rate of the camera.

[0004] Most of the video sensor deployment optimization problems are aimed at meeting the maximum coverage requirements with the least number of video sensors. Previous research routes can be mainly divided into two types: discretization and continuation. Discretization, as the name implies, is to discretize the target area that needs to be covered by the camera into a target point set, and the location where the camera can be placed is also a point set of candidate points. Then analyze whether these target points are within the coverage range of these cameras, connect the target points with the candidate points, and finally perform combination optimization to select the best combination to achieve the maximum coverage requirement of the target area. The main discretization-related research includes Angella et al. using voting strategies for optimization in three-dimensional scenes, Conci and Zhao et al. using binary integer programming algorithms to optimize the planning problem of camera positions in two-dimensional scenes, Gupta et al. using non-dominated sorting genetic algorithms with elite strategies, Mini et al. using artificial bee colony algorithms and particle swarm optimization to solve sensor deployment problems, and then using heuristic methods for scheduling. Among them, the perception model for cameras is generally fan-shaped. Researchers Bairagi et al. designed a non-dominated sorting heuristic genetic algorithm based on II to solve the problem of minimizing energy consumption and maximizing coverage area of ​​3D video sensor nodes deployed in a 2D target area.

[0005] Continuation mainly treats the area that needs to be covered by the camera and the candidate locations where the camera can be placed as a continuous space, and adjusts the parameters of the video sensor through optimization methods. Among them, Ma used the simulated annealing algorithm to optimize the deployment in the case of continuity in three-dimensional space; Bouyagoub and other researchers also used the hill climbing algorithm in two-dimensional virtual space to deal with this problem.

[0006] Zhang Yanan summarized some common obstacles that block the line of sight, used grids to represent these obstacles, and used the A* search algorithm to solve the camera deployment optimization problem. Ugur and other researchers judged the key points in the area and classified the camera deployment optimization problem as a 01 optimization problem, which ignored many important practical factors that need to be considered. Such a processing method is obviously not practical enough. Compared with the above-mentioned methods, the data model used by ZHONG et al. is more realistic, and houses and other objects use 3D model data. Moreover, in terms of imaging models, it does not simply use 2D sectors or circles, but truncated pyramids, and also considers 5 influencing factors including architecture and landscape. However, the screening method for candidate points does not use intelligent algorithms, but a more traditional exhaustive method. The disadvantages of this method are also obvious. If there are many candidate points and many combinations, this exhaustive method is not practical enough.

[0007] The essence of camera deployment is to reduce the unmonitored areas as much as possible. In addition to reasonable deployment, other methods can be used to fill the blind spots. Researchers such as Milosavljevic used augmented reality (AR) and technology to integrate the three-dimensional (3D) geographic information system (GIS) and video surveillance system.

[0008] In summary, from the above analysis, it can be seen that the research on camera control planning has received attention from related fields, and relevant theoretical and methodological research has also made corresponding progress, but there are still some problems that cannot be ignored. For example, in recent domestic and foreign research, most researchers still focus on factors such as camera parameters, and pay insufficient attention to the joint operation of multiple cameras, and the actual coupling of cameras and geographic scenes and the application in actual geographic environments are not enough. Actual geographic data is often simply regarded as a two-dimensional geometric scene, and there is basically no real three-dimensional scene. The simulated obstacles are also simulated as simple squares to simulate buildings. In general, further development is still needed, and the current research still lacks application specificity and practicality. Summary of the invention

[0009] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method and device for intelligent deployment of perception cameras based on spatial differential evolution. Based on the actual scene, sampling points are selected at fixed intervals on the ground, and the spatial differential evolution algorithm is used to calculate the camera distribution point combination that can achieve the maximum monitoring coverage when the number of cameras is fixed among the candidate points where cameras can be installed.

[0010] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0011] A method for intelligent deployment of perception cameras based on spatial differential evolution, comprising:

[0012] Create a scene layer for the study area;

[0013] Generate ground target points that are not on obstacles based on the scene layer according to the preset density;

[0014] Select candidate camera points for the study area;

[0015] Construct the sight lines from all candidate points to all ground target points;

[0016] Conduct line of sight analysis to obtain preliminary visual points;

[0017] Based on the visible distance and viewing angle of the camera, the final selected visible points are selected from the preliminary selected visible points as visible ground target points;

[0018] The number of cameras to be deployed is set, and the spatial adaptive differential evolution algorithm is used to select the best combination among all candidate points based on visible ground target points. The best combination is used as the camera deployment plan for the study area.

[0019] Furthermore, the method for generating ground target points not located on obstacles according to the preset density is specifically as follows:

[0020] Generate two-dimensional data points on the scene layer according to the preset density, upgrade the two-dimensional data points to three-dimensional data points based on the ground data of the digital surface model of the study area, eliminate the three-dimensional data points on obstacles due to the dimension upgrade according to the size of the elevation value, and retain the three-dimensional data points that need to be observed on the ground as ground target points.

[0021] Furthermore, the method of selecting the best combination of all candidate points using the spatial adaptive differential evolution algorithm is as follows:

[0022] Construct a control population including several chromosomes, where the chromosomes are encoded by all candidate points;

[0023] Establish a fitness function to obtain the fitness function value of the chromosome;

[0024] Update the chromosomes and iteratively control the population through mutation operations and / or crossover selection;

[0025] The optimal chromosome is selected based on the fitness function value.

[0026] Furthermore, the method of establishing the fitness function is:

[0027] Get the coverage of ground target points. The calculation formula is:

[0028]

[0029] |C|=m

[0030] Among them, w i is the weight corresponding to the i-th visible ground target point, m is the total number of all visible ground target points, It is The weight corresponding to the ground target point, n is the total number of all ground target points, C is the set obtained by the union of all visible ground target points of the selected cameras, p1, p2, p3…p m For all visible ground target points, It is The set of visible ground target points corresponding to the cameras, q is the total number of cameras;

[0031] Get ground target point coverage repeatability The calculation formula is:

[0032]

[0033] Wherein, the set D is the set of all visible ground target points repeated in two or more cameras, |D| is the number of elements in the set D, is the kth visible ground target point repeated in two or more cameras, Used to determine visible ground target points that are repeated in two or more cameras Is it present in If it appears, it is 1, otherwise it is 0;

[0034] Get the standard deviation of ground target point coverage repetition, the calculation formula is:

[0035]

[0036] Get the maximum average angle Oritation of the ground target points covered repeatedly. The calculation formula is:

[0037]

[0038] Among them, θ r is the plane angle between any two cameras that repeatedly cover the same visible ground target point, and the angle value range must be [0°, 180°]. lat 、x lon The longitude and latitude of point x is converted to radians, y lat ,y lon The longitude and latitude of point y are converted to radians, z lat 、z lon is the coordinate of the latitude and longitude of point z in radians, point y is the ground target point, point z and point x are the position coordinates of two cameras covering point y at the same time, max(θ r ) function is to take the maximum value among all angles;

[0039] Based on the ground target point coverage, ground target point coverage repetition, the standard deviation of ground target point coverage repetition, and the average value of the maximum angle of the repeatedly covered ground target points, the fitness function is obtained. The calculation formula is:

[0040]

[0041] Among them, k1 and k2 are coverage and , k3 and k4 are the weights of (1-Oritation) and σ respectively.

[0042] Furthermore, the method for updating the chromosome through mutation operation is specifically as follows:

[0043] The mutation operation uses space adaptive differential evolution, and the formula is as follows:

[0044]

[0045] Among them, v l is the lth variant individual, chm best is the optimal solution for the current population, chm best2 is the suboptimal solution of the current population, r1 and r2 are two different random integers in the population size, chm r1 represents the random individual corresponding to r1, chm r2 represents the random individual corresponding to r2, rand p [0,1] represents the random real number taken by the mutation operation in round p, curround is the current iteration number, R m , R m2 is the expected starting probability of selecting the item, R n , R n2 The optimal probability selected for the expectation is the shrinkage parameter, ⊙ represents the element-wise multiplication, chm far is a distal term.

[0046] Furthermore, the method for updating the chromosome through the crossover operation is specifically as follows:

[0047] The adaptive crossover operation of random perturbation is introduced by referring to the harmony search algorithm. The formula is as follows:

[0048]

[0049] Among them, j is a gene fragment of a certain length, dim is the latitude, that is, the number of solutions to be found, NP is the population size, For the The gene fragments of the new individuals produced by the crossover, is the gene fragment of the current population individual, For the The gene fragments of the mutant individuals, rand(j) is the randomly obtained gene fragment, M R is the retention probability, C R is the crossover probability, Indicates The random real number taken by the round crossover operation;

[0050] The calculation formula for each gene in rand(j) is:

[0051] ran = randint[2, L-1];

[0052] Among them, randj (t) represents the t-th gene position in rand(j), chm j (t) represents the tth gene position in the gene fragment of the current population individual, randint[0,1] represents a random number of 0 or 1, and randint[2, L-1] represents a random number between 2 and L-1

[0053] The parameters are adjusted through adaptive adjustment. The specific formula is as follows:

[0054]

[0055] Among them, MR s (N), CR s (N) represents the retention probability and crossover probability of the sth individual in the Nth generation, ω s =(f avg -f s ) / f avg , α, β are the adjustment intervals of retention probability and crossover probability, respectively, f avg is the average fitness value of the current population, f s is the fitness value of the current s-th individual, and MR1(N) and CR1(N) of the Nth generation are obtained according to the Logistic function.

[0056] A perception camera intelligent deployment control device based on spatial differential evolution, comprising:

[0057] The scene layer creation module is used to create scene layers for the study area;

[0058] A ground target point acquisition module is used to generate ground target points that are not on obstacles based on a scene layer according to a preset density;

[0059] The candidate point acquisition module is used to select candidate points of the camera in the study area;

[0060] The sight line construction module is used to construct the sight lines from all candidate points to all ground target points;

[0061] A module for obtaining preliminary visual points is used to perform line of sight analysis to obtain preliminary visual points;

[0062] The final selected visible point acquisition module is used to select the final selected visible point from the preliminary selected visible points as the visible ground target point based on the visible distance and viewing angle of the camera;

[0063] The camera deployment scheme acquisition module is used to set the number of camera deployments, select the best combination among all candidate points based on visible ground target points using a spatial adaptive differential evolution algorithm, and use the best combination as the camera deployment scheme for the study area.

[0064] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned intelligent deployment control method of perception cameras based on spatial differential evolution when executing the program.

[0065] A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned intelligent deployment control method of perception cameras based on spatial differential evolution.

[0066] A computer program product includes a computer program, which, when executed by a processor, implements the above-mentioned intelligent deployment method of perception cameras based on spatial differential evolution.

[0067] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0068] The method for intelligent deployment of perception cameras based on spatial differential evolution provided by the present invention is based on the scene of the research area, comprehensively considers various limiting factors such as monitoring blind spots, obstacle occlusion, line of sight radiation range, coverage overlap, effective monitoring area, etc., and combines the specific parameter characteristics of the camera. Based on the visible distance and viewing angle of the camera, the final selected visual points are selected from the preliminary selected visual points as visible ground target points, the number of camera deployments is set, and the spatial adaptive differential evolution algorithm is used to select the best combination among all candidate points based on the visible ground target points. Compared with the traditional empirical method, this embodiment can select the relatively optimal combination from tens of millions of candidate point combinations, and is not easy to fall into the local optimum, providing reliable guarantees for the governance, development and emergency response of each region. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0070] Figure 1 It is a flow chart of a method for intelligent deployment and planning of urban surveillance cameras based on a binary-coded differential evolution algorithm of the present invention;

[0071] Figure 2 It is a three-dimensional data map of the study area of ​​the specific implementation method of the present invention;

[0072] Figure 3 It is a schematic diagram of the distribution of three-dimensional ground points after discretization;

[0073] Figure 4 It is a candidate location for camera installation that is manually selected;

[0074] Figure 5It is a schematic diagram of the result of constructing sight lines between each candidate point and the ground point;

[0075] Figure 6 is the DSM data map of the reference area;

[0076] Figure 7 is the result table of population coding;

[0077] Figure 8 This is a comparison chart of the convergence results of ten-pass manual point placement, genetic algorithm and spatial adaptive difference algorithm;

[0078] Fig. 9 This is the first time that the evolution curves of genetic algorithm and spatial adaptive algorithm are simultaneously performed;

[0079] Fig.10 This is the first time that the evolution curves of genetic algorithm and spatial adaptive algorithm are simultaneously performed;

[0080] Fig.11 is the point distribution diagram of the optimal convergence result;

[0081] Fig.12 It is a visual diagram of the optimal convergence result. DETAILED DESCRIPTION

[0082] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0083] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0084] First embodiment

[0085] In recent years, especially in urban areas, the normal operation and efficient management of social life cannot be separated from video surveillance. However, in the process of using cameras, it is easy to have problems such as blurred images, obstacles blocking the line of sight, and too large an ineffective area of ​​the monitoring area. The efficiency of the camera, which should have played a major role in bringing many conveniences, has been greatly reduced. Comprehensively analyzing factors such as space, cost, and placement, the ideal situation is to use the least number of cameras to achieve the largest effective area of ​​observation. However, if the location and placement of the camera are selected and placed based on pure human feeling, the error will be too large, and the effective monitoring rate of video surveillance will be very low.

[0086] Therefore, based on the above assumptions, the first embodiment proposes an improvement based on the intelligent optimization algorithm and combined with specific conditions, and uses a genetic algorithm to achieve the maximum monitoring coverage of the camera when the number of cameras is fixed, such as Figure 1 As shown, including:

[0087] Step S1: Create a scene layer for the study area;

[0088] Step S2: Generate ground target points that are not on obstacles based on the scene layer according to a preset density;

[0089] Step S3: Select candidate camera points for the study area;

[0090] Step S4: construct the sight lines from all candidate points to all ground target points;

[0091] Step S5: Perform line of sight analysis to obtain preliminary visible points;

[0092] Step S6: selecting a final selected visible point from the preliminary selected visible points as a visible ground target point based on the visible distance and viewing angle of the camera;

[0093] Step S7: Set the number of cameras to be deployed, and select the best combination among all candidate points using a spatial adaptive differential evolution algorithm based on visible ground target points, and use the best combination as the camera deployment plan for the study area.

[0094] The intelligent deployment method of perception cameras based on spatial differential evolution provided in this embodiment is based on the scene of the research area, comprehensively considers various limiting factors such as monitoring blind spots, obstacle occlusion, line of sight radiation range, coverage overlap, effective monitoring area, etc., and combines the specific parameter characteristics of the camera. Based on the visual distance and viewing angle of the camera, the final selected visual points are selected from the preliminary selected visual points as visible ground target points, the number of camera deployments is set, and the spatial adaptive differential evolution algorithm is used to select the best combination among all candidate points based on the visible ground target points. Compared with the traditional empirical method, this embodiment can select the relatively optimal combination from tens of millions of candidate point combinations, and is not easy to fall into the local optimum, providing reliable guarantees for the governance, development and emergency response of each region.

[0095] The above steps are described in detail below.

[0096] In step S1 of this embodiment, a scene layer is created: taking a part of a community as an example, the scene layer creation tool in the ArcGIS Pro software geoprocessing tool is used to convert the original 3D tile data into an integrated grid scene layer (slpk format), and the result is displayed in the local scene. The drone obtains the data of a part of the community, such as Figure 2 Shown is a three-dimensional data map of the study area of ​​a specific implementation method of the present invention.

[0097] In step S2 of this embodiment, the method for generating ground target points that are not on obstacles according to the preset density is specifically as follows:

[0098] Generate two-dimensional data points on the scene layer according to the preset density, upgrade the two-dimensional data points to three-dimensional data points based on the ground data in the digital surface model (DSM) format of the study area, eliminate the three-dimensional data points on the ground obstacles due to the dimension upgrade according to the size of the elevation value, and retain the three-dimensional data points on the ground that need to be observed as ground target points.

[0099] Specifically: Discretization of the monitoring target area: First, use Python to automatically generate two-dimensional data points of preset density on the two-dimensional surface corresponding to the scene layer, and then upgrade the two-dimensional data points to three dimensions based on the ground data in the digital surface model format, that is, attach the two-dimensional grid data points to the surface of the three-dimensional scene. The digital surface model format data is converted from osgb data by Supermap software, such as Figure 6 As shown in the figure, ground obstacles are usually houses, trees, etc. Then, according to the size of the z value, the points on the houses, trees, etc. are removed due to the dimension increase, and only the ground target points that need to be observed are retained, a total of 7783, as shown in the figure. Figure 3 shown.

[0100] In step S3 of this embodiment, camera candidate points are selected: appropriate camera control candidate points are selected based on the city-wide oblique photogrammetry DSM data and previous experience. For example, low-resolution camera candidate points are located on both sides of the road at a height of 3 meters. High-resolution cameras are generally distributed at locations such as the side edges of buildings.

[0101] In the specific implementation of this embodiment, the camera height is set to 3 meters. A total of 367 candidate points are selected here. The red points in the figure are the selected candidate points. Figure 4 shown.

[0102] In step S4 of this embodiment, constructing sight lines: using the construct sight line tool provided by the 3D Analyst toolbox in ArcGIS Pro to construct sight lines from all candidate points to the ground target point.

[0103] In step S5 of this embodiment, line of sight analysis: use the line of sight provided by the 3D Analyst toolbox in ArcGIS Pro to analyze all sight lines, determine whether a certain sight line can pass through obstacles, and whether the target point is visible, such as Figure 5 shown.

[0104] In step S6 of this embodiment, the visible points are screened: first, the actual distance between the candidate point corresponding to the camera and the ground target point is calculated in the attribute table, and the angle between the line of sight and the plumb line of the ground target point is calculated. Finally, the points that meet the conditions are screened according to the visible distance, viewing angle and other restrictive conditions of the specific camera. Specifically, when the actual distance between the candidate point and the ground target point is less than or equal to the visible distance of the camera and the angle between the line of sight and the plumb line of the ground target point is less than or equal to the viewing angle, it indicates that the ground target point is a qualified point, that is, the ground target point is a visible ground target point; when the actual distance between the candidate point and the ground target point is greater than the visible distance of the camera and the angle between the line of sight or the plumb line of the ground target point is greater than the viewing angle, it indicates that the ground target point is not a qualified point, that is, the ground target point is not a visible ground target point. In the specific implementation of this embodiment, the camera parameters used limit the visible distance to 50 meters and the viewing angle to 75°.

[0105] In the specific implementation manner of this embodiment, the preset density can be set according to actual needs, such as 0.5m / piece or 1m / piece.

[0106] In step S7 of this embodiment, the method of selecting the best combination of all candidate points using the spatial adaptive differential evolution algorithm based on the visible ground target points is specifically as follows:

[0107] Step S701: constructing a control population including a number of chromosomes, where the chromosomes are encoded by all candidate points;

[0108] Step S702: Establish a fitness function to obtain the fitness function value of the chromosome;

[0109] Step S703: updating the chromosomes and iteratively controlling the population through mutation operation of spatial adaptive differential evolution and / or adaptive crossover selection with random disturbance introduced by using harmony search algorithm;

[0110] Step S704: Select the optimal chromosome based on the fitness function value.

[0111] In step S701 of this embodiment, the encoding method adopted for the candidate points is binary encoding, forming a binary encoding matrix composed of 0 and 1, the matrix size is NP*M, NP is the population size, the size of M is maxL*n, maxL is the number of bits of the binary encoding of the maximum candidate point index number, and n is the number of available cameras. In the specific implementation method of this embodiment, the encoding method adopted for the candidate points is binary encoding, forming a binary encoding matrix composed of 0 and 1, the matrix size is NP*M, NP is the population size set to 30, the size of M is maxL*n, maxL is the number of bits of the binary encoding of the maximum candidate point index number 367 is 9, and n is the number of available cameras, which is set to 25 here. Figure 7 shown.

[0112] In S702 of this embodiment, in the method of establishing a fitness function, the design of the fitness function simultaneously considers four dimensions: coverage of the ground, coverage repetition of ground target points, standard deviation of coverage repetition of ground target points, and average value of maximum angles of repeatedly covered ground target points.

[0113] Get the coverage of ground target points. The calculation formula is:

[0114]

[0115] |C|=m

[0116] Among them, w i is the weight corresponding to the i-th visible ground target point, m is the total number of all visible ground target points, It is The weight corresponding to the ground target point, n is the total number of all ground target points, C is the set obtained by the union of all visible ground target points of the selected cameras, p1, p2, p3…p m For all visible ground target points, It is The set of visible ground target points corresponding to the cameras, q is the total number of cameras.

[0117] Get ground target point coverage repeatability That is, the average number of times a repeatedly covered ground target point is covered by different cameras is calculated as follows:

[0118]

[0119] Wherein, the set D is the set of all visible ground target points repeated in two or more cameras, |D| is the number of elements in the set D, is the kth visible ground target point repeated in two or more cameras, Used to determine visible ground target points that are repeated in two or more cameras Is it present in If it is present, it is 1, otherwise it is 0.

[0120] For example, the set of all visible ground target points repeated in two or more cameras includes the first visible target point, the second visible target point and the third visible target point, and the total number of selected cameras is 5, among which the first visible target point can be repeatedly seen by 2 cameras, the second visible target point can be repeatedly seen by 3 cameras, and the third visible target point can be repeatedly seen by 2 cameras, then

[0121] Obtain the standard deviation σ of the coverage repetition of ground target points, that is, calculate the standard deviation of the coverage of visible ground target points that are repeatedly covered by different cameras. The calculation formula is:

[0122]

[0123] Get the maximum average angle Oritation of the ground target points that are repeatedly covered. The calculation formula is:

[0124]

[0125] Among them, θ r is the plane angle between any two cameras that repeatedly cover the same visible ground target point, and the angle value range must be [0°, 180°]. lat 、x lon The longitude and latitude of point x is converted to radians, y lat ,y lon The longitude and latitude of point y are converted to radians, z lat 、z lon is the coordinates of the latitude and longitude of point z in radians, point y is the ground target point, z and x are the position coordinates of the two cameras covering point y at the same time, max(θ r) function takes the maximum value among all angles, so Oritation is the average value of the maximum angles between two or more cameras corresponding to the ground target points that are repeatedly covered.

[0126] Based on the ground target point coverage, ground target point coverage repetition, the standard deviation of ground target point coverage repetition, and the average value of the maximum angle of the repeatedly covered ground target points, the fitness function is obtained. The calculation formula is:

[0127]

[0128] Among them, k1 and k2 are coverage and The weight of k3 and k4 are the weights of (1-Oritation) and σ, respectively. are all positive attributes, that is, the evolution direction of the fitness function Fitness is to hope that coverage and Evolving in a larger direction, (1-Oritation) and σ are both negative attributes, and their increase will make the value of the fitness function Fitness smaller and the evaluation of the individual lower.

[0129] In S703 of this embodiment, the method for updating the chromosome through the mutation operation is specifically as follows:

[0130] The mutation operation uses the Spatial Adaptive Differential Evolution Algorithm (SA-DEA / best-and-rand / 2) method, and the formula is as follows:

[0131]

[0132] Among them, v l is the lth variant individual, chm best is the optimal solution for the current population, chm best2 is the suboptimal solution of the current population, r1 and r2 are two different random integers in the population size, chm r1 represents the random individual corresponding to r1, chm r2 represents the random individual corresponding to r2, rand p [0,1] represents the random real number taken by the mutation operation in round p, curround is the current iteration number, R m , R m2 is the expected starting probability of selecting the item, R n , R n2 The optimal probability of the expected selection is the shrinkage parameter, and the suboptimal probability of the expected selection ⊙ represents the multiplication of the elements. farIt is the distal term, that is, the distal individuals randomly selected from the tail with poor performance in the current population, usually the distal individuals randomly selected from the tail 1 / 3 with poor performance in the current population.

[0133] In step S703 of this embodiment, the method for updating the chromosome through the crossover operation is specifically as follows:

[0134] The adaptive crossover operation of random perturbation is introduced by referring to the harmony search algorithm. The formula is as follows:

[0135]

[0136] Among them, j is a gene fragment of a certain length, dim is the latitude, that is, the number of solutions to be found, NP is the population size, For the The gene fragments of the new individuals produced by the crossover, is the gene fragment of the current population individual, For the The gene fragments of the mutant individuals, rand(j) is the randomly obtained gene fragment, M R is the retention probability, C R is the crossover probability, Indicates The random real number taken by the round crossover operation.

[0137] Since the individuals in this paper are encoded by index numbers, crossover cannot be performed point by point, but rather a crossover operation is performed with each L bit as a gene point. Where L is the number of bits after the binary encoding index number. Each gene in rand(j) is:

[0138] ran = randint[2, L-1]

[0139] Among them, rand j (t) represents the t-th gene position in rand(j), chm j (t) represents the tth gene position in the gene fragment of the current population individual, randint[0,1] represents a random number of 0 or 1, and randint[2, L-1] represents a random number between 2 and L-1.

[0140] Through continuous cycles, a new crossover offspring population is eventually generated, whose size is the same as that of the parent.

[0141] Among them, the parameters are adjusted by a parameter adaptive adjustment method, and the specific formula is as follows:

[0142]

[0143] Among them, MRs (N), CR s (N) represents the retention probability and crossover probability of the sth individual in the Nth generation, ω s =(f avg -f s ) / f avg , α, β are the adjustment intervals of retention probability and crossover probability, respectively, f avg is the average fitness value of the current population, f s is the fitness value of the current sth individual, that is, the retention and crossover probability values ​​are adjusted by the fitness value. The calculation formulas of MR1(N) and CR1(N) of the Nth generation are based on the Logistic function, as follows:

[0144]

[0145] Among them, c MR 、c CR , α MR , α CR 、b MR 、b CR represents the control quantity, c MR 、c CR Used to control the maximum value of the Logistic function. The value of the Logistic function tends to c as n increases. MR / CR +d MR / Cr ; α MR / CR Controls the translation position of the Logistic function curve. Specifically, α MR / CR Determines the starting position of the curve moving on the n-axis. MR / CR When it increases, the curve moves to the left, otherwise it moves to the right; b MR / CR Determines the steepness of the Logistic function, in fact it controls the slope of the function curve. MR / CR When b is larger, the curve is steeper and the function changes faster. MR / CR When it is smaller, the curve is flatter and changes more slowly; MR / CR Controls the minimum value of the Logistic function.

[0146] Selection operation: Select the one with the largest fitness value from the generated new population and the old population to update the old population.

[0147] In summary, this embodiment is based on the three-dimensional DSM data of the research area produced by aerial oblique photogrammetry, comprehensively considers various limiting factors such as monitoring blind spots, tree and building occlusion, line of sight radiation range, coverage overlap, effective monitoring area, etc., and combines the specific parameter characteristics of video surveillance equipment to establish an intelligent comprehensive evaluation model for the network coverage of video surveillance equipment, and designs the spatial population coding and differential evolution operator of the spatial multi-mode differential evolution camera spatial deployment algorithm. In the specific evolution process, the candidate points are first binary-coded to initialize the population and parameters, the individual fitness function is calculated, the parent is selected layer by layer, hybridization and mutation operations are performed, and the parameters are adaptively optimized and adjusted at each layer to improve efficiency and continuously update the population until convergence. Compared with the traditional empirical method, this embodiment can select the relatively optimal combination from tens of millions of candidate point combinations, and also adopts an adaptive adjustment method in terms of parameters, so that the population can evolve naturally according to the S-shaped curve as a whole, and then the optimal solution can be obtained more quickly, and it is not easy to fall into the local optimum, providing reliable guarantees for urban governance, development and emergency response.

[0148] In order to verify the effectiveness of the present invention, this embodiment is based on the real three-dimensional community data (OSGB format) of a community, comprehensively considers various limiting factors such as monitoring blind spots, tree and building obstructions, line of sight radiation range, effective monitoring area, etc., combined with the specific parameters of the video surveillance equipment, and uses the genetic algorithm, a spatial intelligent calculation method, to construct the line of sight between the camera and the ground monitoring point. Combined with the line of sight analysis tool of ArcGIS Pro, it is determined whether the ground test points involved can be seen, and an intelligent optimization algorithm is designed. After multiple iterations, selection, hybridization, mutation and other processes are continuously performed to obtain the optimal placement position for video surveillance.

[0149] First, the parameters used in the parameter adjustment operation of the algorithm of the present invention are set to R m =0.5, R n =0.3, R m2 =0.8, R n2 =0.1, α MR =α CR =6, b MR =b CR =0.03, c CR =0.15, d CR =0.8, α=0.2, c MR =0.6, d MR=0.2, β=0.05, k1=0.8, k2=0.2, k3=2, k4=1. The population and iteration times are set to 30, 2000, and 30 populations respectively. The iteration times are set to 2000. The binary genetic algorithm and the binary improved differential evolution proposed in this patent are tested. The number of cameras is 25, the iteration times are 2000, and each method is run 10 times. The maximum coverage rate results are obtained each time. Figure 8 shown.

[0151] Depend on Fig. 9 and Fig.10 It can be seen that the ratio of the camera coverage area of ​​the best control position calculated by the method of the present invention is not easy to fall into the local optimum value compared with the result of the traditional genetic algorithm. Although the genetic algorithm can sometimes get better results, it is easy to stagnate in a search area for a long time, but the present invention is not easy to fall into the local optimum. At the same time, it can also be clearly seen that the time required for the differential evolution algorithm to converge is much less than that of the genetic algorithm.

[0152] Visualization and application: Finally, in ArcGIS Pro, this experiment screens the lines after line of sight analysis in their attribute tables according to the index numbers of the optimal individuals obtained in the above algorithm, and then exports them separately as a Shapefile file. Fig.12 As shown in the figure, the visualization in ArcGIS Pro is a visual diagram of the optimal convergence result. Fig.11 As shown in Figure 1, it is the point distribution diagram of the optimal convergence result.

[0153] Second embodiment

[0154] The second embodiment provides a perception camera intelligent deployment control device based on spatial differential evolution, including:

[0155] The scene layer creation module is used to create scene layers for the study area;

[0156] A ground target point acquisition module is used to generate ground target points that are not on obstacles based on a scene layer according to a preset density;

[0157] The candidate point acquisition module is used to select candidate points of the camera in the study area;

[0158] The sight line construction module is used to construct the sight lines from all candidate points to all ground target points;

[0159] A module for obtaining preliminary visual points is used to perform line of sight analysis to obtain preliminary visual points;

[0160] The final selected visible point acquisition module is used to select the final selected visible point from the preliminary selected visible points as the visible ground target point based on the visible distance and viewing angle of the camera;

[0161] The camera deployment scheme acquisition module is used to set the number of camera deployments, select the best combination among all candidate points based on visible ground target points using a spatial adaptive differential evolution algorithm, and use the best combination as the camera deployment scheme for the study area.

[0162] Third embodiment

[0163] The third embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned intelligent deployment method of perception cameras based on spatial differential evolution when executing the program.

[0164] Fourth embodiment

[0165] The fourth embodiment provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned intelligent deployment method of perception cameras based on spatial differential evolution is implemented.

[0166] The memory in the embodiment of the present invention is used to store various types of data to support the operation of the electronic device. Examples of such data include: any computer program used to operate on the electronic device.

[0167] Fifth embodiment

[0168] A fourth embodiment provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned method for intelligent deployment of perception cameras based on spatial differential evolution.

[0169] The method for intelligent control of perception cameras based on spatial differential evolution disclosed in the embodiment of the present invention can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the method for intelligent control of perception cameras based on spatial differential evolution can be completed by hardware integrated logic circuits or software instructions in the processor. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor can implement or execute the methods, steps and logic block diagrams disclosed in the embodiment of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. In combination with the steps of the method disclosed in the embodiment of the present invention, it can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium, which is located in a memory, and the processor reads the information in the memory, and combines its hardware to complete the steps of the method for intelligent control of perception cameras based on spatial differential evolution provided in the embodiment of the present invention.

[0170] In an exemplary embodiment, the electronic device may be implemented by one or more application specific integrated circuits (ASIC), DSP, programmable logic device (PLD), complex programmable logic device (CPLD), FPGA, general purpose processor, controller, microcontroller (MCU), microprocessor, or other electronic components to execute the aforementioned method.

[0171] It can be understood that the memory can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), and direct RAM bus random access memory (DRRAM, Direct Rambus Random Access Memory).The memories described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memories.

[0172] The above embodiments are merely examples of the technical solutions of the present invention. The methods involved in the present invention are not limited to the contents described in the above embodiments, but are subject to the scope defined in the claims. Any modification, supplement or equivalent replacement made by a person skilled in the art based on the embodiment is within the scope of protection required by the claims of the present invention.

Claims

1. A method for intelligent deployment of perception cameras based on spatial differential evolution, characterized in that: include: Create a scene layer for the study area; Generate ground target points that are not on obstacles based on the scene layer according to the preset density; Select candidate camera points for the study area; Construct the sight lines from all candidate points to all ground target points; Conduct line of sight analysis to obtain preliminary visual points; Based on the visible distance and viewing angle of the camera, the final selected visible points are selected from the preliminary selected visible points as visible ground target points; The number of cameras to be deployed is set, and the spatial adaptive differential evolution algorithm is used to select the best combination among all candidate points based on visible ground target points. The best combination is used as the camera deployment plan for the study area.

2. According to claim 1, the method for intelligent deployment of perception cameras based on spatial differential evolution is characterized in that: The method for generating ground target points that are not on obstacles according to the preset density is as follows: Generate two-dimensional data points on the scene layer according to the preset density, upgrade the two-dimensional data points to three-dimensional data points based on the ground data of the digital surface model of the study area, eliminate the three-dimensional data points on obstacles due to the dimension upgrade according to the size of the elevation value, and retain the three-dimensional data points that need to be observed on the ground as ground target points.

3. The method for intelligent deployment of perception cameras based on spatial differential evolution according to claim 1 is characterized in that: The specific method of selecting the best combination of all candidate points using the spatial adaptive differential evolution algorithm is as follows: Construct a control population including several chromosomes, where the chromosomes are encoded by all candidate points; Establish a fitness function to obtain the fitness function value of the chromosome; Update the chromosomes and iteratively control the population through mutation operations and / or crossover selection; The optimal chromosome is selected based on the fitness function value.

4. The method for intelligent deployment of perception cameras based on spatial differential evolution according to claim 3 is characterized in that: The method to establish the fitness function is: Get the coverage of ground target points. The calculation formula is: |C|=m Among them, w i is the weight corresponding to the i-th visible ground target point, m is the total number of all visible ground target points, It is The weight corresponding to the ground target point, n is the total number of all ground target points, C is the set obtained by the union of all visible ground target points of the selected cameras, p1, p2, p3…p m For all visible ground target points, It is The set of visible ground target points corresponding to the cameras, q is the total number of cameras; Get ground target point coverage repeatability The calculation formula is: Wherein, the set D is the set of all visible ground target points repeated in two or more cameras, |D| is the number of elements in the set D, is the kth visible ground target point repeated in two or more cameras, Used to determine visible ground target points that are repeated in two or more cameras Is it present in If it appears, it is 1, otherwise it is 0; Get the standard deviation of ground target point coverage repetition, the calculation formula is: Get the maximum average angle Oritation of the ground target points covered repeatedly. The calculation formula is: Among them, θ r is the plane angle between any two cameras that repeatedly cover the same visible ground target point, and the angle value range must be [0°, 180°]. lat 、x lon The longitude and latitude of point x is converted to radians, y lat ,y lon The longitude and latitude of point y are converted to radians, z lat 、z lon is the coordinate of the latitude and longitude of point z in radians, point y is the ground target point, point z and point x are the position coordinates of two cameras covering point y at the same time, max(θ r ) function is to take the maximum value among all angles; Based on the ground target point coverage, ground target point coverage repetition, the standard deviation of ground target point coverage repetition, and the average value of the maximum angle of the repeatedly covered ground target points, the fitness function is obtained. The calculation formula is: Among them, k1 and k2 are coverage and , k3 and k4 are the weights of (1-Oritation) and σ respectively.

5. The method for intelligent deployment of perception cameras based on spatial differential evolution according to claim 1 is characterized in that: The specific method for updating chromosomes through mutation operations is: The mutation operation uses space adaptive differential evolution, and the formula is as follows: Among them, v l is the lth variant individual, chm best is the optimal solution for the current population, chm best2 is the suboptimal solution of the current population, r1 and r2 are two different random integers in the population size, chm r1 represents the random individual corresponding to r1, chm r2 represents the random individual corresponding to r2, rand p [0,1] represents the random real number taken by the mutation operation in round p, curround is the current iteration number, R m , R m2 is the expected starting probability of selecting the item, R n , R n2 The optimal probability selected for the expectation is the shrinkage parameter, ⊙ represents the element-wise multiplication, chm far is a distal term.

6. The method for intelligent deployment of perception cameras based on spatial differential evolution according to claim 1 is characterized in that: The specific method for updating chromosomes through crossover operation is: The adaptive crossover operation of random perturbation is introduced by referring to the harmony search algorithm. The formula is as follows: Among them, j is a gene fragment of a certain length, dim is the latitude, that is, the number of solutions to be found, NP is the population size, For the The gene fragments of the new individuals produced by the crossover, is the gene fragment of the current population individual, For the The gene fragments of the mutant individuals, rnad(j) is the randomly obtained gene fragment, M R is the retention probability, C R is the crossover probability, Indicates The random real number taken by the round crossover operation; The calculation formula for each gene in rnad(j) is: rank[2,L-1]: Among them, rand j (t) represents the t-th gene position in rand(j), chm j (t) represents the tth gene position in the gene fragment of the current population individual, randint[0,1] represents a random number of 0 or 1, and randint[2, L-1] represents a random number between 2 and L-1 The parameters are adjusted through adaptive adjustment. The specific formula is as follows: Among them, MR s (N), CR s (N) represents the retention probability and crossover probability of the sth individual in the Nth generation, ω s =(f avg -f s ) / f avg , α, β are the adjustment intervals of retention probability and crossover probability, respectively, f avg is the average fitness value of the current population, f s is the fitness value of the current s-th individual, and MR1(N) and CR1(N) of the Nth generation are obtained according to the Logistic function.

7. A perception camera intelligent deployment control device based on spatial differential evolution, characterized in that: include: The scene layer creation module is used to create scene layers for the study area; A ground target point acquisition module is used to generate ground target points that are not on obstacles based on a scene layer according to a preset density; The candidate point acquisition module is used to select candidate points of the camera in the study area; The sight line construction module is used to construct the sight lines from all candidate points to all ground target points; A module for obtaining preliminary visual points is used to perform line of sight analysis to obtain preliminary visual points; The final selected visible point acquisition module is used to select the final selected visible point from the preliminary selected visible points as the visible ground target point based on the visible distance and viewing angle of the camera; The camera deployment scheme acquisition module is used to set the number of camera deployments, select the best combination among all candidate points based on visible ground target points using a spatial adaptive differential evolution algorithm, and use the best combination as the camera deployment scheme for the study area.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the intelligent deployment method of perception cameras based on spatial differential evolution as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the intelligent deployment control method of perception cameras based on spatial differential evolution as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the intelligent deployment control method of perception cameras based on spatial differential evolution as described in any one of claims 1 to 6 is implemented.

Images