An unmanned aerial vehicle swarm induction countermeasure method based on surrounding swarm repulsion

Abstract

The application discloses a kind of based on surrounding group repulsion UAV group induction countermeasure method, comprising the following steps:Step1: obtain UAV group multi-target state information;Step2: extract UAV group shape and distribution area;Step3: locate UAV group boundary leader;Step4: exert navigation decoy signal;Step5: observe UAV group induction countermeasure effect.The application obtains feature map by saliency shape detection model, then carries out second-order differential operation to obtain UAV group boundary, leader can be accurately positioned, can efficiently filter out useless information, enhance feature representation ability, to accurately locate to UAV group leader, then exerts navigation induction signal Δ i (t) to leader, and constructs decoy signal action model δ i (t), fully utilizes the structure characteristics of "leading group encirclement" UAV group, so as to efficiently countermeasure UAV group with small cost with point to surface.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of counteracting unmanned aerial vehicle swarms, in particular to a method for counteracting unmanned aerial vehicle swarms based on surrounding swarm repulsion. BACKGROUND

[0002] Unmanned aerial vehicle swarms have been increasingly widely used in civil and military fields due to their advantages such as large quantity, low cost and intelligent cooperation, but along with this, various potential dangers and security risks have been caused, especially in military confrontation. Therefore, the research on countermeasures against unmanned aerial vehicle swarms has important practical significance for economic livelihood and national security, and the many advantages of unmanned aerial vehicle swarms have thus become many difficulties in counteracting them.

[0003] At present, countermeasures against unmanned aerial vehicles can be divided into hard countermeasures and soft countermeasures. Hard countermeasures use physical means to intercept or destroy unmanned aerial vehicles, and soft countermeasures use electronic information technology to interfere with, seize control of or lure unmanned aerial vehicles. When facing large-scale unmanned aerial vehicle swarms, hard countermeasures have the problems of high cost and poor results. If the unmanned aerial vehicle swarm is regarded as a simple aggregation of numerous individuals, soft countermeasures will also be limited by factors such as device power, communication bandwidth and technical level, and cannot achieve ideal results.

[0004] Under normal circumstances, a large threat unmanned aerial vehicle swarm is not a simple aggregation, but a distributed system with special structural characteristics to fully exert its advantages and cover up its disadvantages, which becomes a starting point for countermeasures. Using soft countermeasures to attack important joints of the structure of the unmanned aerial vehicle swarm can efficiently and at low cost counteract it. In the distributed system of the unmanned aerial vehicle swarm, there is a "leader encirclement" structure in which a leader aircraft forms an encirclement formation and the "leader encirclement" structure is randomly constrained inside the formation boundary. At present, there is no countermeasure method for this structure. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art and provide a method for counteracting unmanned aerial vehicle swarms based on surrounding swarm repulsion. The method identifies the boundary leader aircraft by targeting the special structural form of the "leader encirclement" unmanned aerial vehicle swarm and applies a navigation deception signal to it to achieve an efficient counteracting effect.

[0006] To achieve the above purpose, the technical solution adopted by the present application is as follows:

[0007] A method for counteracting unmanned aerial vehicle swarms based on surrounding swarm repulsion, comprising the following steps:

[0008] Step1: Obtain multi-target state information of the UAV group: use a laser radar and a camera multi-source sensor to respectively detect the monitored airspace, and fuse the obtained point cloud data and image data to perceive the UAV group target state information, obtain the accurate position and trajectory of the UAV individual, and obtain multi-source image information of the UAV group;

[0009] Step2: Extract the shape and distribution area of the UAV group: perform significant shape detection on the multi-source image information of the UAV group obtained in Step1 to obtain a UAV group shape map S0;

[0010] Step3: Position the boundary leader of the UAV group: obtain the edge line of the UAV group according to the UAV group shape map S0 output in Step2, and position the UAV near the edge line, i.e., the boundary leader;

[0011] Step4: Apply a navigation deception signal: according to the expected UAV group deviation path, apply a navigation deception signal Δ i (t) to the boundary leader positioned in Step3, and calculate the UAV deception signal affected δ i (t);

[0012] Step5: Observe the induced countermeasure effect of the UAV group: after the navigation deception signal in Step4 is applied to the boundary leader, it will be propagated to all UAVs through the UAV group leader surrounding structure, so that the entire UAV group tends to the expected deception position. If the UAV group tends to the expected position, the countermeasure is successful, otherwise go to Step1.

[0013] Preferably, Step1 specifically comprises the following steps:

[0014] Step101: Use a laser radar and a camera multi-source sensor to respectively detect the monitored airspace to obtain point cloud data and image data;

[0015] Step102: Coordinate transform the coordinate system of the laser radar to the camera coordinate system, and register and fuse the time stamps of the point cloud data and image data obtained in Step101 to obtain calibration information;

[0016] Step103: Cluster the point cloud data obtained in Step101 to obtain a UAV point group, and project the UAV point group to the image data obtained in Step101 according to the calibration information obtained in Step102 to obtain the position state information of the UAV;

[0017] Step 104, tracking the drone-like point group obtained in Step 103, and predicting the position of the drone point group in the next frame, then projecting the predicted drone point group position into the image data obtained in Step 101, updating the drone position in real time and recording the motion trajectory, obtaining the multi-source image information of the drone group.

[0018] Preferably, Step 2 specifically comprises the following steps:

[0019] Step 201, sequentially passing the multi-source image information of the drone group obtained in Step 1 through 5 convolution layers, and the output of Level i (i=1,…,5) is compressed into a feature map m i with 16 channels by convolution.

[0020] Step 202, using a pyramid module to extract the context information of multi-level images for the Level 5 with the largest receptive field in Step 201, outputting a feature map m6 with 4 channels, and after obtaining the level outputs (m1,…,m6), starting from the highest level Level 5, stacking m i (i=6,5,…,1) in descending order, respectively as the feature map of the i-th layer of the salient shape extraction network.

[0021] Step 203, passing the feature maps of each layer obtained in Step 202 through a convolution, and after Level 1 and Level 2 correspondingly performing a upsampling process, outputting an activation score map S1' ～5 with the same size as the input image and 1 channel. ～5 After passing through an activation function or a convolution layer, respectively outputting a salient shape map S 1～5 and a drone group shape map S0.

[0022] Preferably, Step 3 specifically comprises the following steps:

[0023] Step 301, performing a second-order differential operation on the drone group shape map S0 output by Step 2 to obtain an operator Δf(x,y);

[0024] Step 302, performing a threshold operation on the second-order differential operator Δf(x,y) obtained in Step 301 to obtain the edge line of the drone group, and positioning the drones near the edge line, which are the boundary leaders.

[0025] Preferably, in Step 301, the calculation formula of the second-order differential operation is as follows:

[0026] Δf(x,y)=f(x+1,y)+f(x-1,y)+f(x,y+1)+f(x,y-1)-4f(x,y)

[0027] Wherein, f(x, y) is the image information function at pixel position (x, y).

[0028] Preferably, in Step 4, the UAV is affected by the decoy signal δ i (t) is calculated by the following decoy signal action model:

[0029]

[0030] Wherein, M and N represent the boundary leader and the follower set respectively, δ i (t) is the navigation decoy signal input to the UAV i, The decoy formation after the boundary leader receives the navigation decoy signal, Δ i (t) is the navigation decoy signal, f i (t) is the task expected formation, The decoy state of the UAV i after being affected by the navigation decoy signal, N i The set composed of the neighbors of the UAV i, w ij (t) represents the local interaction relationship weight between the UAV j and i, Indicates the mutual avoidance action between the follower in the cluster.

[0031] Compared with the prior art, the present application has the beneficial effects that:

[0032] The present application obtains a feature map through a saliency shape detection model, and then performs a second-order differential operation to obtain the boundary of the UAV group, and the leader can be accurately positioned. This boundary leader recognition method based on shape edge segmentation can efficiently filter out useless information, enhance feature representation ability, and accurately position the leader of the UAV group. Then, the navigation decoy signal Δ i (t) is applied to the leader, and the decoy signal action model δ i (t) is constructed, and the "lead encirclement" UAV group structure characteristics are fully utilized, so that the UAV group can be efficiently counteracted with small cost. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The flowchart of the UAV group decoy countermeasure method based on surrounding group driving provided by the embodiment of the present application;

[0034] Figure 2 The saliency shape detection model based on double attention pyramid convolution;

[0035] Figure 3 The navigation decoy strategy principle diagram of the boundary leader in the leader encirclement structure;

[0036] Figure 4 Fig. 1 is a schematic diagram of a navigation deception expected deviation trajectory of a surrounding structure unmanned aerial vehicle group. DETAILED DESCRIPTION

[0037] Exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments explicitly described. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0038] As Figure 1 shown, the embodiment of the present application provides a method for unmanned aerial vehicle group decoy countermeasure based on surrounding group repulsion, comprising the following steps:

[0039] Step 1: Obtain multi-target state information of the unmanned aerial vehicle group: use laser radar and camera multi-source sensors to detect the monitored airspace respectively, and fuse the obtained point cloud data and image data to perceive the unmanned aerial vehicle group target state information, obtain the accurate position and trajectory of the unmanned aerial vehicle individual, and obtain the multi-source image information of the unmanned aerial vehicle group;

[0040] Step 1 specifically comprises the following steps:

[0041] Step 101, use laser radar and camera multi-source sensors to detect the monitored airspace respectively to obtain point cloud data and image data;

[0042] Step 102, coordinate transformation of the coordinate system of the laser radar to the camera coordinate system, and registration and fusion of the time stamp of the point cloud data and the image data obtained in Step 101 to obtain calibration information;

[0043] Step 103, clustering the point cloud data obtained in Step 101 to obtain a UAV point group, and projecting the UAV point group into the image data obtained in Step 101 according to the calibration information obtained in Step 102 to obtain the position state information of the UAV; specifically, the point cloud data obtained in Step 101 is clustered by DBSCAN clustering algorithm, the clustering parameters Eps and MinPts are determined in advance, the core points are determined according to the parameters, the corresponding point group is found by expanding point by point, the process is repeatedly to find all point groups, and they are divided into different categories, including the UAV point group, and the UAV point group is projected into the image data obtained in Step 101 according to the calibration information obtained in Step 102, that is, the position state information of the UAV is obtained.

[0044] Step 104: Track the drone cluster obtained in Step 103, predict the position of the drone cluster in the next frame, and then project the predicted drone cluster position onto the image data obtained in Step 101. Update the drone position in real time and record the motion trajectory to obtain multi-source image information of the drone cluster. Use the particle filter tracking algorithm to track the drone cluster obtained in Step 103. The process is as follows: select a neighborhood centered on the drone cluster, scatter particles in it, use particle filtering to predict the position of the drone cluster in the next frame, and then project the predicted drone cluster position onto the image obtained in Step 101. Update its position in real time and record its motion trajectory.

[0045] Step 2: Extract the shape and distribution area of ​​the drone swarm: Perform salient shape detection, input the multi-source image information of the drone swarm obtained in Step 1, and output the drone swarm shape map S0;

[0046] Step 2 specifically includes the following steps:

[0047] For a swarm of drones with a lead-and-encircle structure, the boundary leader drone is located at the edge of the formation, forming a specific shape, while simultaneously constraining the other drones to remain inside the formation. To accurately detect the boundary leader drone, a method integrating formation shape extraction and drone swarm edge segmentation is proposed. The formation structure of a drone swarm is time-varying during mission execution, but it maintains a relatively stable formation at each stage of the mission, and the boundary of the lead-and-encircle drone swarm exhibits relatively regular shape characteristics. The boundary leader drone identification method based on shape edge segmentation first searches for drone swarm shapes and distribution areas with regular attributes, and then extracts the boundary leader drone at the boundary curve position.

[0048] Step 201: Pass the multi-source image information of the drone swarm obtained in Step 1 through 5 convolutional layers. Level i The input for (i = 2, ..., 5) is Level i-1 The output of the feature map is reduced by half the number of channels after one convolution, while Level i The output of (i = 1, ..., 5) is convolved and compressed into a feature map m with 16 channels. i This is used for subsequent feature map fusion;

[0049] Step 202, the pyramid module is used to extract the context information of the multi-level image for the Level 5 with the largest receptive field in Step 201, and a feature map m6 with a channel number of 4 is output, the PPM includes four branches, each branch is inputted to be reduced in dimension according to a corresponding pooling kernel, then a convolution kernel with a size of 1x1 is used to compress the channel number to 1, finally, the feature maps of the four branches are converted into the resolution through an upsampling layer, and the feature maps with the same features are inputted, and the stacked feature maps are outputted as m6 of the PPM module, after obtaining the output (m1,..., m5) of each level and the output m6 of the PPM module, the feature maps of the levels are stacked in descending order from the highest level Level 5, and the feature maps are outputted as the feature maps of the i-th layer of the salient shape extraction network i (i = 6, 5,..., 1), respectively.

[0050] Step 203, the feature maps of each layer obtained in Step 202 are convoluted once, the feature maps corresponding to Level 1 and Level 2 are increased by one upsampling process to ensure that the resolution of each layer has the same spatial size, after the feature maps corresponding to Level 1 and Level 2 are increased by one upsampling process, an activation score map S1' with the same size as the input image and a channel number of 1 is outputted. ～5 , the activation score map S1' ～5 After the activation function or a convolution layer, the salient shape map S 1～5 and the UAV group shape map S0 are outputted, respectively.

[0051] To prevent the problem of introducing a large number of useless features caused by the way of cross-layer connection to fuse the features of the encoding end and the decoding end, before the feature fusion, the model first calibrates the response weight between the channels of the double attention module inputted by the single feature of the encoding, highlights the response intensity of the foreground region pixels, enhances the task-related feature representation ability, and weakens the background and noise influence. To avoid the weakening and loss of feature information caused by the series connection of multi-level attention modules, the DAM fuses the two types of attention in parallel.

[0052] Step 3: positioning the boundary leader of the UAV group: the UAV group edge line is obtained according to the UAV group shape map S0 outputted in Step 2, and the UAV near the edge line is positioned as the boundary leader;

[0053] Step 3 specifically includes the following steps:

[0054] Step 301, a second-order differential operation is performed on the UAV group shape map S0 outputted in Step 2 to obtain an operator Δf(x, y);

[0055] Step 302, threshold operation is performed on the second-order differential operator Δf(x, y) obtained in Step 301, and if greater than a threshold, an edge exists, an edge line of the UAV group is obtained, and the UAVs near the edge line are located as the boundary leader.

[0056] In Step 301, the calculation formula of the second-order differential operation is as follows:

[0057] Δf(x, y) = f(x+1, y) + f(x-1, y) + f(x, y+1) + f(x, y-1) - 4f(x, y)

[0058] Wherein, f(x, y) is the image information function at pixel position (x, y).

[0059] The local edge of the image is the transition between two regions with obviously different intensities, and the gradient function of the image, that is, the rate of change of the image gray scale, has a maximum value at the edge position. Therefore, the gradient direction of the image gray scale change is estimated based on the gradient operator, and then threshold operation is performed on the gradient or edge extraction operator. If greater than a given threshold, an edge exists. However, in a wide area with equal slope, the gradient operator may extract all the areas as edges. Further calculating the rate of change of the slope can solve this problem.

[0060] Step 4: Apply the navigation deception signal: according to the expected UAV group deviation path, the boundary leader located in Step 3 is applied with a navigation deception signal Δ i (t), and the UAV affected by the deception signal δ i (t) is calculated.

[0061] In Step 4, the UAV affected by the deception signal δ i (t) is calculated by the following deception signal action model:

[0062]

[0063] Wherein, M and N represent the boundary leader set and the follower set respectively, δ i (t) is the action input of the navigation deception signal to the UAV i, represents the deception formation of the boundary leader after receiving the navigation deception signal, Δ i (t) is the navigation deception signal, f i (t) is the task expected formation, is the deception state of the UAV i after being affected by the navigation deception signal, N i is a set composed of neighbors of the UAV i, w ij (t) represents the local interaction relationship weight between the UAV j and i, This indicates the collision avoidance mechanism between followers within a cluster.

[0064] Note: Analysis of the "leader-encirclement" structure of drone swarms:

[0065] For a swarm of N drones, the dynamic characteristics of each drone are described as follows:

[0066]

[0067] Where i∈{1,2,…,N},x i (t)∈R n For the state variables of the drone, For x i The first differential of (t), y i (t)∈R q For the output variable of the drone, u i (t)∈R m For controlling input.

[0068] Suppose there are M (M < N) follower drones and NM boundary navigators in a drone swarm, where the follower drones have state x. k (t)(k=1,2,…,M), the boundary navigator state is x. i (t)(i=M+1,M+2,…,N). Define the indices of the follower and the boundary leader as ξ={1,2,…,M} and ζ={M+1,M+2,…,N}, respectively. For each follower in the UAV swarm, there exists at least one directed path from the boundary leader to that follower. When the UAV swarm achieves encirclement, for any k∈ξ, there exists a path satisfying… nonnegative constant β k,j (j∈ξ) such that:

[0069]

[0070] Define the desired formation of the boundary navigator as in Let be a piecewise, continuously differentiable vector function. Consider the following "leader-encirclement" protocol based on output feedback:

[0071]

[0072]

[0073] K1, K2, and K3 are constant gain matrices with matching dimension.

[0074] According to the above-mentioned "navigation encirclement" agreement, it can be deduced that the boundary navigation aircraft of the "navigation encirclement" structure unmanned aerial vehicle group can form a convex hull according to the set formation, and the following random has the surrounding absorption effect, that is, the action of approaching the boundary formation and entering the convex hull; the inter-machine collision avoidance mechanism, that is, the following random is scattered in the convex hull by the holding position effect to avoid collision.

[0075] When the navigation deception is implemented on the navigation encirclement structure unmanned aerial vehicle group, the following deception signal action model is constructed for unmanned aerial vehicle i:

[0076]

[0077] Wherein, ζ and ξ respectively represent the boundary navigation aircraft set and the following random set, δ i (t) is the action input of the error signal to unmanned aerial vehicle i, represents the error formation of the boundary navigation aircraft after receiving the error signal, Δ i (t) is the error signal, f i (t) is the task expected formation, is the error state of unmanned aerial vehicle i after being affected by the error signal, N i is the set composed of the neighbors of unmanned aerial vehicle i, w ij (t) represents the local interaction relationship weight between unmanned aerial vehicle j and i.

[0078] Step 5: observe the effect of the countermeasure of the unmanned aerial vehicle group: after the navigation deception signal in Step 4 is applied to the boundary navigation aircraft, it will be transmitted to all unmanned aerial vehicles through the navigation encirclement structure of the unmanned aerial vehicle group, so that the entire unmanned aerial vehicle group tends to the expected deception position, and the countermeasure is successful, otherwise go to Step 1.

[0079] As shown in Figure 3 and Figure 4 , after the constructed navigation deception signal is applied to the navigation aircraft, the navigation deception signal will be transmitted to the entire unmanned aerial vehicle group. When the boundary navigation aircraft deviates from the flight path, the following random is affected by the surrounding attraction and collision avoidance repulsion mechanism, and always in the convex hull of the navigation encirclement structure, and cannot escape from the surrounding area, so the entire unmanned aerial vehicle group will deviate from the flight path. On the one hand, since the surrounding formation formed by the boundary navigation aircraft has a higher priority, when its positioning is wrong, the remaining ordinary unmanned aerial vehicles will reach the wrong surrounding area according to the error surrounding formation through the formation attraction principle. On the other hand, the following random is scattered in the surrounding area through the inter-machine collision avoidance principle. Through the cooperation of the two principles, the entire unmanned aerial vehicle group is counteracted.

[0080] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A UAV swarm decoy countermeasure method based on enveloping swarm repulsion, characterized in that, The method comprises the following steps: Step 1: obtaining multi-target state information of the UAV group: using a laser radar and a camera multi-source sensor to detect the monitored airspace respectively, fusing the obtained point cloud data and image data, sensing the UAV group target state information, obtaining the accurate position and trajectory of the UAV individual, and obtaining the multi-source image information of the UAV group; Step 2: extracting the shape and distribution area of the UAV group: performing significant shape detection on the multi-source image information of the UAV group obtained in Step 1 to obtain a UAV group shape graph S0; Step 3: positioning the boundary leader of the UAV group: obtaining the UAV group edge line according to the UAV group shape graph S0 output in Step 2, and positioning the UAV near the edge line, that is, the boundary leader; Step 4: Apply a navigational deception signal: Based on the desired deviation of the drone swarm from the path, apply a navigational deception signal Δ to the boundary leader identified in Step 3 i (t), the calculation of the drone's deception signal impact δ i (t); The drone is affected by the decoy signal delta i (t) by a decoy signal action model calculated as follows: where M and N represent the set of border leaders and the set of followers, respectively, δ i (t) is the effect of the navigation deception signal on UAV i, is the deception formation after the border leaders receive the navigation deception signal, Δ i (t) is the navigation deception signal, f i (t) is the mission desired formation, is the deception state of UAV i after it is affected by the navigation deception signal, N i is the set of neighbors of UAV i, w ij (t) represents the local interaction relationship weight between UAV j and i, represents the mutual avoidance effect between the followers in the cluster; Step 5: observing the UAV group induction countermeasure effect: after the navigation deception signal in Step 4 is applied to the boundary leader, it is propagated to all UAVs through the UAV group leader encirclement structure, so that the entire UAV group tends to the expected deception position, the UAV group tends to the expected position, and the countermeasure is successful, otherwise, go to Step 1.

2. The swarm of drones based on the surrounding swarm repulsion-based drone swarm induction countermeasure method of claim 1, wherein, Step 1 specifically comprises the following steps: Step 101, using a laser radar and a camera multi-source sensor to detect the monitored airspace respectively, obtaining point cloud data and image data; Step 102, performing coordinate conversion on the coordinate system of the laser radar to the camera coordinate system, and registering and fusing the time stamps of the point cloud data and the image data obtained in Step 101 to obtain calibration information; Step 103, clustering the point cloud data obtained in Step 101 to obtain a UAV point group, and projecting the UAV point group to the image data obtained in Step 101 according to the calibration information obtained in Step 102, that is, obtaining the position state information of the UAV; Step 104, tracking the UAV point group obtained in Step 103, and predicting the position of the UAV point group in the next frame, then projecting the predicted UAV point group position to the image data obtained in Step 101, updating the UAV position in real time and recording the motion trajectory, and obtaining the multi-source image information of the UAV group.

3. The swarm-inducing countermeasure method based on enveloping swarm repulsion of claim 1, wherein, Step 2 specifically comprises the following steps: Step 201, the multi-source image information of the UAV group obtained in Step 1 is sequentially subjected to 5 convolution layers, and the output of Level i is compressed into a feature map m i with 16 channels through convolution, i = 1, …, 5; Step 202, for the Level 5 with the largest receptive field in Step 201, the context information of the multi-level image is extracted using the pyramid module, and a feature map m6 with a channel number of 4 is output, and after obtaining the output m1,..., m6 of each level, starting from the highest level Level 5, the m i , i = 6, 5,..., 1, are stacked in descending order as the feature map of the i-th layer of the salient shape extraction network; Step 203, the feature maps of each layer obtained in Step 202 are subjected to one convolution, and the Level 1 and Level 2 corresponding feature maps are subjected to one upsampling process, thereby outputting an activation score map S' with the same size as the input image and a channel number of 1 1～5 , the activation score map S' 1～5 After being subjected to an activation function or one convolution layer, a significant shape map S 1～5 and a UAV group shape map S0 are respectively outputted.

4. The swarm-inducing countermeasure method based on enveloping swarm repulsion of claim 1, wherein, Step 3 specifically comprises the following steps: Step 301, performing second-order differential operation on the UAV group shape graph S0 output in Step 2 to obtain an operator Δf(x, y); Step 302, performing threshold operation on the second-order differential operator Δf(x, y) obtained in Step 301 to obtain the UAV group edge line, and positioning the UAV near the edge line, that is, the boundary leader.

5. The swarm-inducing countermeasure method based on enveloping swarm repulsion of claim 4, wherein, In Step 301, the calculation formula of the second-order differential operation is as follows: Δf(x, y) = f(x+1, y) + f(x-1, y) + f(x, y+1) + f(x, y-1) - 4f(x, y) Wherein, f(x, y) is the image information function at pixel position (x, y).

Images