A heterogeneous multi-sensor joint deployment method for low-altitude target detection

Abstract

The application relates to a heterogeneous multi-sensor joint deployment method for low-altitude target detection, comprising the following steps: constructing three-dimensional monitoring grid units covering the ground and low-altitude airspace of a region to be monitored, and determining a multi-sensor set to be deployed and continuous state variables and discrete state variables thereof; judging the line-of-sight unobstruction of each three-dimensional monitoring grid unit by using a digital elevation model, and calculating the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative targets and non-cooperative targets under multi-sensor cooperative detection according to the result of the line-of-sight unobstruction judgment; constructing a multi-objective optimization function with the minimum total cost and the maximum monitoring coverage rate as the targets, wherein the monitoring coverage rate is determined based on the comprehensive detection probability; and solving the multi-objective optimization function by using a hybrid evolutionary algorithm to output a multi-sensor deployment scheme. The scheme can improve the global search capability and convergence efficiency of the algorithm, and obtain a sensor station deployment scheme considering the network construction cost and the three-dimensional coverage rate.

Description

Technical Field

[0001] This application relates to the field of sensor co-deployment technology, and in particular to a heterogeneous multi-sensor co-deployment method, system, computer device and computer-readable storage medium for low-altitude target detection. Background Technology

[0002] Low-altitude security equipment, as an infrastructure ensuring the safe operation of the low-altitude economy, has wide application value in airport defense, urban security, and key area surveillance. Sensor deployment planning aims to achieve full coverage of low-altitude intrusion targets by rationally deploying low-altitude security equipment such as radar and photoelectric devices. Traditional deployment planning methods are mainly based on line-of-sight analysis or simple geometric coverage models, determining equipment locations by calculating terrain visibility. However, as application scenarios expand to complex mountainous areas, urban buildings, and other non-ideal environments, traditional methods face challenges such as severe terrain obstruction, inaccurate blind zone calculations, and low equipment utilization.

[0003] In recent years, the introduction of intelligent optimization algorithms such as particle swarm optimization and genetic algorithms has provided new solutions for site deployment planning in complex environments. Compared with manual experience-based site selection, intelligent algorithms can automatically search for site deployment schemes with higher coverage in a large-scale solution space by simulating the iterative optimization of biological populations, and are especially suitable for collaborative site deployment planning of multiple devices.

[0004] However, most existing intelligent site deployment methods are based on two-dimensional planar assumptions or single targets, ignoring the three-dimensional line-of-sight occlusion problem caused by real terrain undulations, resulting in false coverage phenomena, and failing to distinguish the differentiated detection needs of cooperative and non-cooperative targets. In addition, existing algorithms use a single evolutionary mechanism to handle all variables. When faced with high-dimensional mixed-integer programming problems that include equipment location and equipment usage status, traditional algorithms are prone to getting trapped in local optima, making it difficult to balance economic cost and coverage in the planning results. Summary of the Invention

[0005] This application provides a heterogeneous multi-sensor joint deployment method, system, computer device, and computer-readable storage medium for low-altitude target detection, in order to at least solve the problem in related technologies where it is difficult to achieve an optimal balance between detection coverage and deployment cost in sensor deployment schemes.

[0006] In a first aspect, embodiments of this application provide a heterogeneous multi-sensor joint deployment method for low-altitude target detection, the method comprising: A three-dimensional monitoring grid cell covering the ground and low-altitude airspace of the area to be monitored is constructed, and the set of multiple sensors to be deployed and their continuous and discrete state variables are determined. The continuous state variables include spatial coordinates and line-of-sight angles, and the discrete state variables are the sensor on / off states. The line-of-sight unobstructed judgment is performed on each three-dimensional monitoring grid unit using a digital elevation model, and based on the result of the line-of-sight unobstructed judgment, the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative and non-cooperative targets under multi-sensor collaborative detection is calculated. A multi-objective optimization function is constructed with the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, wherein the monitoring coverage is determined based on the comprehensive detection probability; A hybrid evolutionary algorithm is used to solve the multi-objective optimization function and output a multi-sensor deployment scheme. In the iterative optimization process, the continuous state variables and the discrete state variables are decoupled, and a dual-track cooperative evolution update is performed using an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator.

[0007] In some embodiments, solving the multi-objective optimization function using a hybrid evolutionary algorithm includes: Initialize a population containing multiple particles and an external file, wherein each particle is decoupled and configured as a continuous state vector representing the spatial coordinates and the line-of-sight angle, and a discrete state vector representing the sensor switching state; the external file is configured to store the non-dominated solution set generated during the optimization iteration process. In each iteration update, the solutions in the external archive are gridded and the current global optimal solution is selected based on the density distribution; The current individual optimal solution and the global optimal solution are decomposed into the continuous state vector and discrete state vector of the particle, respectively, and the continuous state vector and discrete state vector of the current particle are decoupled and collaboratively updated by combining the dual-track evolution mechanism. Calculate the updated objective function value of the particles, update the individual optimal solution of each particle based on the objective function value, add the newly generated non-dominated solutions to the external archive, and clean up the dominated solutions in the external archive. If the number of iterations reaches the preset maximum number of iterations, the solution that meets the preset cost constraints and has the highest monitoring coverage is selected from the external archives and used as the final multi-sensor deployment scheme.

[0008] In some embodiments, gridding the solutions in the external archive and selecting the current global optimum based on the density distribution includes: After the objective function of each solution in the external file is meshed on a two-dimensional plane, the number of particles in each mesh is counted to obtain the mesh density. The actual selection probability of each solution is calculated based on the grid density, wherein the actual selection probability is inversely proportional to the grid density; Based on the distribution of the actual selection probabilities, a solution is randomly selected from the external archives as the current global optimal solution.

[0009] In some embodiments, an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator are used for dual-track cooperative evolution updates, wherein the update method of the continuous state vector is as follows: The inertial flight mechanism of the particle swarm optimization algorithm is retained to utilize gradient information in continuous space for fine-grained search; By combining the current particle's velocity vector, the continuous state vector corresponding to the individual optimal solution, and the continuous state vector corresponding to the current global optimal solution, the continuous state vector of the current particle is calculated and updated.

[0010] In some embodiments, an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator are used for dual-track cooperative evolution updates, wherein the update method of the discrete state vector is as follows: Generate a random mask vector with the same dimension as the discrete state vector, and in the dimension where the random mask vector is activated, make the discrete state vector directly inherit the sensor usage state corresponding to the current global optimal solution; Set a mutation probability and generate a mutation mask vector with the same dimension as the discrete state vector based on the mutation probability. In the dimension where the mutation mask vector is activated, perform state flipping on the discrete state variables of the current particle. After the state transition is completed, the state variable representing the minimum number of sensors to be deployed is configured to be enabled to ensure that the generated deployment scheme meets the hardware size lower limit constraint.

[0011] In some embodiments, a multi-objective optimization function is constructed with the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, including: The comprehensive detection probability of each three-dimensional monitoring grid cell in the area to be monitored for cooperative and non-cooperative targets is calculated respectively, and the comprehensive detection probability of all grid cells in the area is arithmetically averaged to obtain the corresponding average coverage. The average coverage rate of the cooperative targets and the average coverage rate of the non-cooperative targets are weighted and fused to obtain the monitoring coverage rate that reflects the overall coverage effect; The actual number of sensor devices used is counted based on the discrete state variables, and the total deployment cost of the system is calculated based on the number of sensor devices and the unit deployment cost of each type of sensor. The multi-objective optimization function is constructed by maximizing the monitoring coverage and minimizing the total system deployment cost.

[0012] In some embodiments, a digital elevation model is used to determine the line-of-sight unobstructed nature of each 3D monitoring grid cell, and based on the results of the line-of-sight unobstructed nature determination, the comprehensive detection probability of each 3D monitoring grid cell for cooperative and non-cooperative targets under multi-sensor cooperative detection is calculated, including: Multiple equally spaced sampling points are selected on the line-of-sight vector between the sensor location and the center of each three-dimensional monitoring grid unit, and the actual ground elevation of each sampling point is obtained in the digital elevation model. If the line of sight at a certain sampling point is lower than the actual ground elevation, it is determined that the line of sight is blocked by the terrain, and its unobstructed line of sight is recorded as 0; otherwise, it is determined to be a visible target point, and its unobstructed line of sight is recorded as 1. When the monitoring point is located within the sensor's field of view and the unobstructed line-of-sight state is 1, the single-station detection probability is calculated using the distance decay function; Based on the joint probability distribution model, the single-station detection probabilities of multiple sensors covering the same three-dimensional monitoring grid cell are fused to obtain the comprehensive detection probability.

[0013] Secondly, embodiments of this application provide a heterogeneous multi-sensor joint deployment system for low-altitude target detection, the system comprising: a construction module, a judgment module, and an output module, wherein: The construction module is used to construct a three-dimensional monitoring grid cell covering the ground and low-altitude airspace of the area to be monitored, and to determine the set of multiple sensors to be deployed and their continuous state variables and discrete state variables, wherein the continuous state variables include spatial coordinates and line-of-sight angles, and the discrete state variables are the sensor on / off states. The judgment module is used to use a digital elevation model to make a line-of-sight unobstructed judgment for each three-dimensional monitoring grid unit, and to calculate the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative and non-cooperative targets under multi-sensor collaborative detection based on the result of the line-of-sight unobstructed judgment. The construction module is also used to construct a multi-objective optimization function with the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, wherein the monitoring coverage is determined based on the comprehensive detection probability; The output module is used to solve the multi-objective optimization function using a hybrid evolutionary algorithm and output a multi-sensor deployment scheme. In the iterative optimization process, the continuous state variables and the discrete state variables are decoupled, and a dual-track cooperative evolution update is performed using an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator.

[0014] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect above.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect above.

[0016] Compared to related technologies, the heterogeneous multi-sensor joint deployment method for low-altitude target detection provided in this application introduces a digital elevation model for matrix-based field-of-view analysis, which accurately reflects the line-of-sight occlusion of complex terrain and avoids the false coverage problem of two-dimensional planes. Furthermore, by establishing a weighted fusion coverage calculation model for cooperative and non-cooperative targets, low-cost wide-area surveillance and differentiated sensor deployment for different target types are achieved. In addition, through a decoupling evolution mechanism of continuous and discrete variables, an improved multi-target particle swarm optimization algorithm is used to optimize continuous parameters such as sensor position and attitude, and genetic operators are combined to dynamically adjust the sensor switching state to solve the problem of local convergence limitation. This application can improve the global search capability and convergence efficiency of the algorithm and obtain a precise deployment scheme that balances network construction cost and three-dimensional coverage. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a heterogeneous multi-sensor joint deployment method for low-altitude target detection according to an embodiment of this application; Figure 2 This is a schematic diagram of the iterative process of a hybrid evolutionary algorithm for the joint deployment of heterogeneous multi-sensor systems according to an embodiment of this application; Figure 3 This is a structural block diagram of a heterogeneous multi-sensor joint deployment system for low-altitude target detection according to an embodiment of this application; Figure 4 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0019] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0020] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0021] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0022] This embodiment provides a heterogeneous multi-sensor joint deployment method for low-altitude target detection. Figure 1 This is a flowchart of a heterogeneous multi-sensor joint deployment method for low-altitude target detection according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps: Step S101: Construct a three-dimensional monitoring grid cell covering the ground and low-altitude airspace of the area to be monitored, and determine the set of multiple sensors to be deployed and their continuous and discrete state variables. The continuous state variables include spatial coordinates and line-of-sight angles, and the discrete state variables are the sensor on / off states.

[0023] Specifically, assuming a rectangular planar ground area The system deploys M types of multi-source heterogeneous sensors. The absolute altitude of the sensors is determined by the Digital Elevation Model (DEM) of the area. Optimization variables include the sensor's horizontal coordinates (x, y) in a Cartesian coordinate system and its relative altitude h relative to the terrain elevation. Among the M types of heterogeneous sensors, the number of sensors of type i must meet preset upper and lower limits, with a minimum deployment number of [value missing]. The maximum number of deployments is The maximum total number of sensors that can be deployed in the area is then capped at 1. For the i-th sensor in the m-th class Define its deployment attribute to be optimized as

[0024] in, This represents the horizontal coordinates and relative height coordinates of the i-th sensor of class m in space. These are the azimuth and pitch angles of the sensor's line of sight, respectively. This is a binary decision variable, indicating whether the sensor is enabled (1 for enabled, 0 for disabled).

[0025] Furthermore, the cuboid monitoring area covering the ground and low-altitude airspace no higher than H is discretized into three-dimensional grid cells of equal length, width, and height, forming a set containing N monitoring points. , where the j-th monitoring point It is represented by the center coordinates of its grid cell, i.e.

[0026] Step S101 performs three-dimensional meshing discretization on the cuboid monitoring area and defines the multi-dimensional deployment attributes of heterogeneous sensors. It unifies the continuous position of the physical space with the discrete start and stop states of the equipment through mathematical modeling, providing a structured data foundation for subsequent large-scale optimization and probability calculation of complex environment detection.

[0027] Step S102: Use the digital elevation model to determine the line-of-sight unobstructedness of each three-dimensional monitoring grid unit, and calculate the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative and non-cooperative targets under multi-sensor collaborative detection based on the results of the line-of-sight unobstructedness determination.

[0028] The detection probability between the monitoring point and the sensor is calculated using distance, elevation angle, azimuth angle, and line-of-sight (LOS) parameters, where: The j-th monitoring point and the i-th sensor of the m-th class. The relative position vector between them is

[0029] The j-th monitoring point and the i-th sensor of the m-th class. The relative distance between them is

[0030] The j-th monitoring point and the i-th sensor of the m-th class. The relative pitch and azimuth angles can be calculated using local coordinate system projection. First, the sensor is constructed. Local coordinate system for

[0031] Furthermore, calculate the relationship between the j-th monitoring point and... The relative coordinates of the local coordinate system are

[0032] in, This represents the transpose, where the j-th monitoring point is transposed to the i-th sensor of the m-th class. The relative pitch and relative azimuth angles between them are

[0033] In this application, a fast discrimination algorithm based on matrix sampling is used for LOS determination, specifically including: For sensors Location relative to the target point location line-of-sight vectors In vector Select N equally spaced sampling points Based on the planar coordinates of the sampling points, an index lookup or interpolation is performed in the digital elevation model (DEM) to obtain the actual ground elevation of the corresponding sampling point location. Construct an occlusion discrimination inequality. If there exists a sampling point that satisfies... If the line of sight is below the ground elevation, then the line of sight is considered to be blocked by the terrain, and the target point... This is an invisible blind spot, i.e. Otherwise, target point For the visible target point, i.e. .

[0034] For the j-th monitoring point, it can be monitored by the sensor. The detection criteria are as follows:

[0035] in, Represents the characteristic function, Indicates intersection, The range of azimuth angles that the sensor can detect. The range of pitch angles that the sensor can detect.

[0036] In this embodiment, by introducing a digital elevation model and performing matrix-based equidistant sampling on the line-of-sight vector, accurate reconstruction of real complex terrain undulations is achieved, eliminating false coverage blind spots in traditional view analysis, and improving the physical realism and computational efficiency of three-dimensional spatial view distance occlusion determination.

[0037] In addition, when monitoring points When the sensor meets the detectability conditions, namely, when the target is simultaneously within its field of view (FOV), within its detection range, and meets the line-of-sight (LOS) unobstructed condition, the detection probability decreases as the distance between the target and the sensor increases. This process is modeled using the Sigmoid function.

[0038] Specifically, monitoring points For sensors The detection probability can be expressed as

[0039] in, For sensors (Its maximum detection range is) The distance decay function is expressed as follows:

[0040] in, The attenuation coefficient is given by the above formula, which indicates that when the target distance is... The detection probability drops to 0.5.

[0041] Step S102 combines real 3D geographic environment information with multi-sensor collaborative detection mechanisms to construct a high-fidelity target coverage assessment method. It addresses the problem of misjudgment of blind spots in complex mountainous or urban building environments through matrix-based elevation occlusion discrimination. Simultaneously, it performs differentiated detection probability fusion for cooperative and non-cooperative targets, reflecting the actual monitoring needs of targets with different threat levels, thus giving the coverage indicators of the deployment scheme fine-grained practical guiding value.

[0042] Step S103: With the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, a multi-objective optimization function is constructed, wherein the monitoring coverage is determined based on the comprehensive detection probability.

[0043] In this embodiment, to maximize the coverage of low-altitude airspace perception range under a limited construction budget, an objective function is constructed with deployment cost and regional coverage as its core components. Specifically: For monitoring points Depending on the nature of the target (cooperative / non-cooperative), the probability of it being detected by multiple sensors in a coordinated manner is also differentiated accordingly.

[0044] Assuming all sensors are fused, the monitoring point The overall detection probability for cooperative targets is The overall detection probability for non-cooperative targets is Their expressions are as follows:

[0045] in, For device type indicator variables, if If it is a cooperative target detection device (such as ADS-B, RemoteID, etc.), then ;like For non-cooperative target detection equipment (such as radar, electro-optical, radio, etc.), then .

[0046] For the entire area to be monitored, i.e., the set of monitoring points The average coverage of its cooperative objectives Average coverage of non-cooperative targets for:

[0047] Furthermore, by weighting and fusing the average coverage rates of cooperative and non-cooperative targets, a unified evaluation index reflecting the overall coverage effect of the entire monitored area can be obtained. for:

[0048] in, and These are the weight coefficients for cooperative and non-cooperative objectives, respectively, and they satisfy the following conditions:

[0049] On the other hand, network construction costs Based on discrete state variables The actual number of sensor devices deployed is counted, and the unit deployment cost is calculated accordingly.

[0050] in, The unit deployment cost for the m-th type of sensor.

[0051] After determining the coverage and cost, the following multi-objective optimization function is constructed to achieve the optimal trade-off between deployment cost and overall regional coverage:

[0052] in, .

[0053] The multi-source heterogeneous sensor optimization deployment model established in this embodiment for low-altitude security scenarios aims to minimize deployment costs and maximize the comprehensive coverage of the monitored area. Its mathematical expression is as follows:

[0054] in, This represents the constraints of the sensor site in space, i.e.

[0055] in, This represents the constraints that each variable to be optimized in the sensor needs to satisfy, i.e.

[0056] Finally, by configuring constraints Ensure that at least the m-th type of equipment is deployed indivual.

[0057] Step S103 unifies the coverage requirements of low-altitude security and the economic budget constraints of engineering construction within the same framework by constructing a multi-objective optimization function that maximizes coverage and minimizes deployment costs. In this embodiment, this dual-objective trade-off mechanism not only avoids extreme equipment redundancy caused by unilaterally pursuing extremely high coverage, but also prevents the hidden dangers of monitoring blind spots caused by excessive cost compression, providing an optimization direction for the automated search of Pareto-optimal site deployment schemes that combine high performance and cost-effectiveness.

[0058] Step S104: The multi-objective optimization function is solved using a hybrid evolutionary algorithm to output a multi-sensor deployment scheme. In the iterative optimization process, continuous state variables and discrete state variables are decoupled, and a dual-track cooperative evolution update is performed using an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator.

[0059] It should be noted that in the multi-objective optimization problem of coverage and cost in multi-sensor deployment, traditional algorithms, due to their reliance on a single velocity vector update mechanism, struggle to effectively search the mixed variable space of continuous variables (coordinates, angles) and discrete variables (sensor usage status), and are prone to getting trapped in local optima in the discrete subspace. To address this, this invention proposes a hybrid evolutionary algorithm. By decoupling the optimization of continuous deployment coordinates / line-of-sight angles and discrete sensor usage status, it employs an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator for collaborative evolution, thereby achieving a better Pareto front in the complex solution space. The specific implementation steps are as follows: S1, Population Hyperparameter Settings S1.1, based on the M-type sensing devices and their maximum deployment quantity Determine the dimension D of each particle, i.e. .

[0060] S1.2, determine the maximum number of iterations for population updates (MaxIter), population size (PopNum), file size (RepNum), velocity weight (w), learning rate (c1), learning rate (c2), and discrete mutation rate (P). MUT Table 1 is a schematic table of population initialization parameters and update parameters: Table 1: Illustration of Population Initialization and Update Parameters

[0061] S1.3, determine the constraint boundary values ​​for each variable to be optimized in the particle, i.e.

[0062] S2, Particle Swarm and External Archive Initialization S2.1, define PopNum particles It is initialized as a vector of all zeros, in the following form:

[0063] For particles In this form, it is decomposed into continuous state vectors representing the spatial coordinate system and the view angle. and discrete state variables representing the sensor's switching state ,in

[0064] S2.2, using a uniform random distribution, generates initial values ​​for the variables of each particle in the population within the feasible region of their respective variables, i.e.

[0065] in, Indicates the upper limit is The lower limit is Random numbers. For the variable representing the switching state of each sensor in the particle. First of all A value is randomly sampled uniformly within the interval. If the value is greater than 0.5, the initial state is set to 1 (representing on); otherwise, it is set to 0 (representing off).

[0066] Furthermore, to ensure that at least the m-th type of sensor is deployed Each one, its corresponding state variable is forcibly set to 1, that is

[0067] S2.3 Calculate the objective function of each particle , and each particle is used as an individual optimal solution .

[0068] It should be noted that, since the present application only uses the subgroup optimization algorithm for continuous vectors formed by deployment coordinates / line-of-sight angles, each particle is initialized with dimensional all-zero velocity vector .

[0069] S2.4 Compare the objective function value of the candidate particle with all non-dominated solutions in the external archive . If there exists an archived solution that can dominate the candidate solution, that is, the archived solution is not inferior to the candidate solution in all objective function values and is strictly better than the candidate solution in at least one objective value, then it is determined that the candidate solution is a dominated solution and will not be added to the external archive. If the candidate solution is not dominated by any existing solution in the external archive, proceed to the next step to check whether it will dominate the old solutions in the archive. If it dominates an old solution , it indicates that the old solution is no longer a Pareto optimal solution in the current set of known solutions, and it is removed from the archive. After cleaning all dominated solutions, add the candidate solution to the external archive.

[0070] S2.5 When the size of the external archive exceeds the preset upper limit RepNum, in order to prevent computational redundancy caused by excessive density of non-dominated solutions, redundant individuals in the archive are eliminated by means of retaining endpoints and random sparsing. The specific steps are as follows: first sort all solutions in the archive in ascending order according to coverage rate, retain the two extreme solutions at the first position (lowest cost) and the last position (highest coverage rate) in the sorting result, and for non-extreme solutions located between the two endpoints, remove redundant individuals by means of random uniform sampling until the size upper limit RepNum is satisfied.

[0071] S3 Hybrid evolutionary update of particle swarm Assume that the current iteration is the t-th iteration, and this step is repeated as long as t < MaxIter is satisfied. For the first iteration, both the particles and particle velocities initialized in step 2 are used as inputs.

[0072] S3.1 First, the global optimal solution of the current particle swarm needs to be selected , the steps are as follows: plot the objective function of the solutions in the external archive on a two-dimensional plane ​Perform gridding, uniformly dividing the range of each objective function along its dimension. Each interval forms There are three grids of equal size. The number of particles in each grid is counted, and the density of the i-th grid is... Furthermore, the probability of selection is inversely proportional to the density, i.e. After normalization, the actual selection probability of each solution within the grid is obtained. Based on the above probability distribution, a solution is randomly selected from the external archives using the roulette wheel method as the current global optimal solution. The cumulative probability distribution of all solutions is calculated, and a... Uniform random numbers within the interval Select the case where the cumulative probability is greater than or equal to the first time. The solution as .

[0073] Furthermore, the globally optimal particle and the current individual optimal solution Decomposed into continuous variables and discrete state variables, among which

[0074] S3.2, for the inertial flight mechanism of the particle swarm optimization algorithm that preserves the continuous state vector, gradient information in the continuous space is utilized for fine-grained search. The update formula is as follows:

[0075] Update the continuous variables in the current particle during the t-th iteration. for

[0076] S3.3 For discrete state vectors composed of sensor state variables The algorithm updates via a genetic probability perturbation strategy. First, it simulates the process of particles learning towards the global optimum. The learning probability is then set. , generation and A random mask vector of the same dimension. In the dimension where the random mask vector is activated, Directly inherit the global optimal solution The corresponding device usage status. Next, set the mutation probability. Similarly generated with A random mask vector of the same dimension. The number of bits in the activated dimension of the random mask vector is adjusted. State flip. After the flip is complete, The middle represents the minimum number of sensors to be deployed for each type (before each type of sensor). (Each state variable) is forced to be set to 1 to ensure that at least these numbers of sensors are deployed.

[0077] S3.4 After all particles have been updated, update the individual optimal solutions and external archives according to steps S2.3-S2.5. The conditions for updating the individual historical optimal solution are as follows: if the current particle dominates the historical optimal solution, then a forced update is performed; if the two do not dominate each other, then an update is performed with a 50% probability; if the current particle dominates the historical optimal solution, then no update is performed.

[0078] Finally, when the number of iterations exceeds a preset threshold, the results of the current particle swarm and the external archive are output. Based on the set maximum cost constraint, all feasible particles with costs not exceeding the threshold are selected from the external archive, and then the coverage index of these particles is calculated and compared. Finally, the particle with the highest coverage is selected as the optimal solution, thus obtaining the sensor deployment scheme that achieves the maximum coverage under the given cost constraint, including its spatial coordinates, line-of-sight angle, and on / off status information.

[0079] Step S104 employs a dual-track collaborative evolution mechanism. Continuous parameters undergo refined gradient search through particle swarm inertial flight, while discrete states rely on genetic operators for probability flipping and global solution inheritance. This can solve the optimization problem of mixed integer programming problems that are prone to local deadlock and improve the convergence efficiency of the multidimensional solution space.

[0080] This step designs and implements a highly customized hybrid evolutionary dual-track solution algorithm to address the challenge of high-dimensional mixed variables in multi-sensor joint deployment. By deeply decoupling continuous spatial attitude parameter optimization from discrete device start-up and shutdown state decisions, the algorithm retains the particle swarm's powerful local mining capabilities in continuous space while leveraging a genetic mutation mechanism to give the population the ability to jump between different device combinations. By decoupling the handling of mixed variables, it overcomes the blind search of traditional algorithms with a single velocity update mechanism when facing non-convex, discontinuous solution spaces. By introducing a discrete-dimensional probabilistic jump mechanism, the algorithm has the ability to cross cost faults, thereby avoiding the structural local optimum trap caused by the inability to dynamically remove redundant nodes.

[0081] also, Figure 2 This is a schematic diagram illustrating the iterative process of a hybrid evolutionary algorithm for the joint deployment of heterogeneous multi-sensor systems, according to an embodiment of this application. Figure 2 As shown, firstly, the deployment constraints and population hyperparameters of the sensing devices are determined, and the particle population and external archives are initialized. Each particle is decoupled into a continuous state vector representing spatial parameters and a discrete state vector representing switching states. The population size PopNum is set to 100, the external archive capacity RepNum to 60, and the inertia weight w is configured to 0.5, along with learning factors c1 and c2.

[0082] After entering the iterative optimization phase, the total deployment cost and monitoring coverage of each particle are calculated. Non-dominant solutions are stored in an external archive. When the external archive exceeds the limit, a gridding strategy is used to retain extreme solutions and sparse out dense regions to ensure the diversity of the Pareto front.

[0083] Next, the global optimal solution is selected based on the grid density distribution, and a dual-track collaborative update is performed: for continuous state vectors, the particle swarm inertial flight mechanism is used for three-dimensional spatial fine-tuning; for discrete state vectors, state inheritance and probability flip mutation are performed through random masking, and the lower limit of the deployment quantity is forcibly corrected to meet the hard constraints.

[0084] Finally, the individual optimal solution is updated based on the objective function value. The above iterative update steps are repeated until the maximum number of iterations is reached. Finally, a heterogeneous sensor deployment scheme that achieves the best trade-off between cost and coverage is output from the external archive.

[0085] On the other hand, this embodiment also provides a heterogeneous multi-sensor joint deployment system for low-altitude target detection. Figure 3 This is a structural block diagram of a heterogeneous multi-sensor joint deployment system for low-altitude target detection according to an embodiment of this application, as shown below. Figure 3 As shown, the system includes: a construction module 30, a judgment module 31, and an output module 32, wherein: The construction module 30 is used to construct a three-dimensional monitoring grid cell covering the ground and low-altitude airspace of the area to be monitored, and to determine the heterogeneous multi-sensor set to be deployed and its continuous state variables and discrete state variables. The continuous state variables include spatial coordinates and line-of-sight angles, and the discrete state variables are the sensor on / off states. The judgment module 31 is used to use the digital elevation model to make a line-of-sight unobstructed judgment on each three-dimensional monitoring grid unit, and to calculate the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative and non-cooperative targets under multi-sensor collaborative detection based on the line-of-sight unobstructed judgment result. The building module 30 is also used to construct a multi-objective optimization function with the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, wherein the monitoring coverage is determined based on the comprehensive detection probability; The output module 32 is used to solve the multi-objective optimization function using a hybrid evolutionary algorithm and output a multi-sensor deployment scheme. In the iterative optimization process, continuous state variables and discrete state variables are decoupled, and dual-track cooperative evolution updates are performed using an improved multi-objective particle swarm algorithm and a genetic mutation operator, respectively.

[0086] Through the aforementioned system, a digital elevation model is introduced for matrix-based view analysis, accurately reflecting the line-of-sight occlusion in complex terrain and avoiding the false coverage problem in two-dimensional planes. Furthermore, by establishing a weighted fusion coverage calculation model for cooperative and non-cooperative targets, low-cost wide-area surveillance and differentiated sensor deployment for different target types are achieved. In addition, by proposing a decoupling evolution mechanism for continuous and discrete variables, an improved multi-objective particle swarm optimization algorithm is used to optimize continuous parameters such as sensor position and attitude. Combined with genetic operators to dynamically adjust the sensor switching state, the local convergence limitation in mixed integer programming problems is solved, improving the algorithm's global search capability and convergence efficiency, and obtaining a precise station deployment scheme that balances network construction cost and three-dimensional coverage.

[0087] In one embodiment, Figure 4 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application. For example... Figure 4 As shown, a computer device is provided, which can be an in-vehicle computing unit or a cloud server. The computer device includes a processor, a network interface, internal memory, and non-volatile memory connected via an internal bus. The non-volatile memory stores parameters related to the operating system, computer programs, and a large language model. The processor provides computing power and executes deep neural network inference and control algorithm operations. When the processor executes the stored computer program, it implements the heterogeneous multi-sensor joint deployment method for low-altitude target detection described in any of the above embodiments.

[0088] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0089] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by hardware related to computer program instructions. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc. The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A heterogeneous multi-sensor joint deployment method for low-altitude target detection, characterized in that, The method includes: A three-dimensional monitoring grid cell covering the ground and low-altitude airspace of the area to be monitored is constructed, and the set of multiple sensors to be deployed and their continuous and discrete state variables are determined. The continuous state variables include spatial coordinates and line-of-sight angles, and the discrete state variables are the sensor on / off states. The line-of-sight unobstructed judgment is performed on each three-dimensional monitoring grid unit using a digital elevation model, and based on the result of the line-of-sight unobstructed judgment, the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative and non-cooperative targets under multi-sensor collaborative detection is calculated. A multi-objective optimization function is constructed with the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, wherein the monitoring coverage is determined based on the comprehensive detection probability; A hybrid evolutionary algorithm is used to solve the multi-objective optimization function and output a multi-sensor deployment scheme. In the iterative optimization process, the continuous state variables and the discrete state variables are decoupled, and a dual-track cooperative evolution update is performed using an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator. The solution to the multi-objective optimization function using a hybrid evolutionary algorithm includes: Initialize a population containing multiple particles and an external file, wherein each particle is decoupled and configured as a continuous state vector representing the spatial coordinates and the line-of-sight angle, and a discrete state vector representing the sensor switching state; the external file is configured to store the non-dominated solution set generated during the optimization iteration process. In each iteration update, the solutions in the external archive are gridded and the current global optimal solution is selected based on the density distribution; The current individual optimal solution and the global optimal solution are decomposed into the continuous state vector and discrete state vector of the particle, respectively, and the continuous state vector and discrete state vector of the current particle are decoupled and collaboratively updated by combining the dual-track evolution mechanism. Calculate the updated objective function value of the particles, update the individual optimal solution of each particle based on the objective function value, add the newly generated non-dominated solutions to the external archive, and clean up the dominated solutions in the external archive. The update method for the continuous state vector is as follows: The inertial flight mechanism of the particle swarm optimization algorithm is retained to utilize gradient information in continuous space for fine-grained search; By combining the current particle's velocity vector, the continuous state vector corresponding to the individual optimal solution, and the continuous state vector corresponding to the current global optimal solution, the continuous state vector of the current particle is calculated and updated. If the number of iterations reaches the preset maximum number of iterations, the solution that meets the preset cost constraints and has the highest monitoring coverage is selected from the external archives and used as the final multi-sensor deployment scheme. The discrete state vector is updated in the following way: Generate a random mask vector with the same dimension as the discrete state vector, and in the dimension where the random mask vector is activated, make the discrete state vector directly inherit the sensor usage state corresponding to the current global optimal solution; Set a mutation probability and generate a mutation mask vector with the same dimension as the discrete state vector based on the mutation probability. In the dimension where the mutation mask vector is activated, perform state flipping on the discrete state variables of the current particle. After the state transition is completed, the state variable representing the minimum number of sensors to be deployed is configured to be enabled to ensure that the generated deployment scheme meets the hardware size lower limit constraint.

2. The method according to claim 1, characterized in that, The solution in the external archive is gridded and the current global optimum is selected based on the density distribution, including: After the objective function of each solution in the external file is meshed on a two-dimensional plane, the number of particles in each mesh is counted to obtain the mesh density. The actual selection probability of each solution is calculated based on the grid density, wherein the actual selection probability is inversely proportional to the grid density; Based on the distribution of the actual selection probabilities, a solution is randomly selected from the external archives as the current global optimal solution.

3. The method according to claim 1, characterized in that, With the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, a multi-objective optimization function is constructed, including: The comprehensive detection probability of each three-dimensional monitoring grid cell in the area to be monitored for cooperative and non-cooperative targets is calculated respectively, and the comprehensive detection probability of all grid cells in the area is arithmetically averaged to obtain the corresponding average coverage. The average coverage rate of the cooperative targets and the average coverage rate of the non-cooperative targets are weighted and fused to obtain the monitoring coverage rate that reflects the overall coverage effect; The actual number of sensor devices used is counted based on the discrete state variables, and the total deployment cost of the system is calculated based on the number of sensor devices and the unit deployment cost of each type of sensor. The multi-objective optimization function is constructed by maximizing the monitoring coverage and minimizing the total system deployment cost.

4. The method according to claim 1, characterized in that, A digital elevation model is used to determine the line-of-sight unobstructed nature of each 3D monitoring grid cell. Based on the results of this determination, the comprehensive detection probability of each 3D monitoring grid cell for cooperative and non-cooperative targets under multi-sensor collaborative detection is calculated, including: Multiple equally spaced sampling points are selected on the line-of-sight vector between the sensor location and the center of each three-dimensional monitoring grid unit, and the actual ground elevation of each sampling point is obtained in the digital elevation model. If the line of sight at a certain sampling point is lower than the actual ground elevation, it is determined that the line of sight is blocked by the terrain, and its unobstructed line of sight is recorded as 0; otherwise, it is determined to be a visible target point, and its unobstructed line of sight is recorded as 1. When the monitoring point is located within the sensor's field of view and the unobstructed line-of-sight state is 1, the single-station detection probability is calculated using the distance decay function; Based on the joint probability distribution model, the single-station detection probabilities of multiple sensors covering the same three-dimensional monitoring grid cell are fused to obtain the comprehensive detection probability.

5. A heterogeneous multi-sensor joint deployment system for low-altitude target detection, characterized in that, The system includes: a construction module, a judgment module, and an output module, wherein: The construction module is used to construct a three-dimensional monitoring grid cell covering the ground and low-altitude airspace of the area to be monitored, and to determine the set of multiple sensors to be deployed and their continuous state variables and discrete state variables, wherein the continuous state variables include spatial coordinates and line-of-sight angles, and the discrete state variables are the sensor on / off states. The judgment module is used to use a digital elevation model to make a line-of-sight unobstructed judgment for each three-dimensional monitoring grid unit, and to calculate the comprehensive detection probability of each three-dimensional monitoring grid unit for cooperative and non-cooperative targets under multi-sensor collaborative detection based on the result of the line-of-sight unobstructed judgment. The construction module is also used to construct a multi-objective optimization function with the objectives of minimizing the total system deployment cost and maximizing the monitoring coverage of the area to be monitored, wherein the monitoring coverage is determined based on the comprehensive detection probability; The output module is used to solve the multi-objective optimization function using a hybrid evolutionary algorithm and output a multi-sensor deployment scheme. During the iterative optimization process, the continuous state variables and the discrete state variables are decoupled, and a dual-track cooperative evolution update is performed using an improved multi-objective particle swarm optimization algorithm and a genetic mutation operator, respectively. Solving the multi-objective optimization function using the hybrid evolutionary algorithm includes: Initialize a population containing multiple particles and an external file, wherein each particle is decoupled and configured as a continuous state vector representing the spatial coordinates and the line-of-sight angle, and a discrete state vector representing the sensor switching state; the external file is configured to store the non-dominated solution set generated during the optimization iteration process. In each iteration update, the solutions in the external archive are gridded and the current global optimal solution is selected based on the density distribution; The current individual optimal solution and the global optimal solution are decomposed into the continuous state vector and discrete state vector of the particle, respectively, and the continuous state vector and discrete state vector of the current particle are decoupled and collaboratively updated by combining the dual-track evolution mechanism. Calculate the updated objective function value of the particles, update the individual optimal solution of each particle based on the objective function value, add the newly generated non-dominated solutions to the external archive, and clean up the dominated solutions in the external archive. The update method for the continuous state vector is as follows: The inertial flight mechanism of the particle swarm optimization algorithm is retained to utilize gradient information in continuous space for fine-grained search; By combining the current particle's velocity vector, the continuous state vector corresponding to the individual optimal solution, and the continuous state vector corresponding to the current global optimal solution, the continuous state vector of the current particle is calculated and updated. If the number of iterations reaches the preset maximum number of iterations, the solution that meets the preset cost constraints and has the highest monitoring coverage is selected from the external archives and used as the final multi-sensor deployment scheme. The discrete state vector is updated in the following way: Generate a random mask vector with the same dimension as the discrete state vector, and in the dimension where the random mask vector is activated, make the discrete state vector directly inherit the sensor usage state corresponding to the current global optimal solution; Set a mutation probability and generate a mutation mask vector with the same dimension as the discrete state vector based on the mutation probability. In the dimension where the mutation mask vector is activated, perform state flipping on the discrete state variables of the current particle. After the state transition is completed, the state variable representing the minimum number of sensors to be deployed is configured to be enabled to ensure that the generated deployment scheme meets the hardware size lower limit constraint.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 4.

Images