Low-altitude unmanned aerial vehicle anti-anti method based on radar positioning and image recognition technology

Abstract

The application provides a low-altitude unmanned aerial vehicle anti-countermeasure method based on radar positioning and image recognition technology, and relates to the technical field of unmanned aerial vehicle anti-countermeasure. The method comprises the following steps: establishing a coordinate system, setting an unmanned aerial vehicle no-fly area and a risk area; acquiring first coordinate data of the unmanned aerial vehicle in the coordinate system, and determining whether there is a risk of radar losing a target; if there is a risk of radar losing a target, determining a first camera and its orientation according to the first coordinate data and camera coordinate data. Determining a risk coefficient of the unmanned aerial vehicle; when the risk coefficient is higher than a preset risk threshold, determining an alarm level, recording an image and determining a countermeasure. According to the application, when there is a risk of radar losing a target, a suitable camera can be selected in time to shoot the unmanned aerial vehicle, so as to reduce the risk of losing the target. In addition, after the unmanned aerial vehicle enters the risk area, the risk of the unmanned aerial vehicle can be judged in real time, the interference on the unmanned aerial vehicle passing by without risk is reduced, and the unmanned aerial vehicle with risk is effectively counteracted.

Description

Technical Field

[0001] This invention relates to the field of anti-drone technology, and in particular to a low-altitude anti-drone method based on radar positioning and image recognition technology. Background Technology

[0002] In related technologies, radar is typically used to detect the location of drones and counter them through signal jamming and other means, forcing them away from no-fly zones. However, when drones fly at low altitudes, radar risks losing targets, and not all drones approaching no-fly zones pose a threat; related technologies struggle to determine the risk posed by approaching drones. Summary of the Invention

[0003] This invention provides a low-altitude UAV countermeasures method based on radar positioning and image recognition technology, which can solve the technical problems in related technologies such as the risk of radar losing targets and the difficulty in identifying approaching UAVs.

[0004] According to a first aspect of the present invention, a method for countering and defending against low-altitude unmanned aerial vehicles (UAVs) based on radar positioning and image recognition technology is provided, comprising:

[0005] A coordinate system is established with the radar as the origin, and no-fly zones and risk zones for drones are set in the coordinate system. The no-fly zones for drones are located within the risk zones, and the second range of the risk zones is larger than the first range of the no-fly zones for drones.

[0006] When the radar detects that the drone has entered the risk area, the first coordinate data of the drone in the coordinate system is obtained;

[0007] Based on the first coordinate data, determine whether there is a risk of radar losing the target;

[0008] If there is a risk of radar losing the target, the first camera to be activated and its orientation are determined based on the first coordinate data and the camera coordinate data of multiple cameras. The multiple cameras are set at multiple locations outside the no-fly zone for drones but within the risk zone.

[0009] Based on at least one of the first coordinate data and video frames captured by the first camera, the duration of the drone's entry into the risk area, and the first range of the drone no-fly zone, the risk coefficient of the drone is determined.

[0010] If the risk coefficient is higher than a preset risk threshold, an alarm level is determined.

[0011] Based on the first coordinate data, select the second camera to capture the recorded images from the drone;

[0012] Determine countermeasures based on the alert level.

[0013] According to the present invention, determining whether there is a risk of radar target loss based on the first coordinate data includes:

[0014] Based on the first coordinate data, determine the planar coordinates (x, y) of the UAV at the current moment. t ,y t ) and the plane coordinates (x) of the previous time step t-1 ,y t-1 ), where the current time is the t-th time after the drone enters the risk area;

[0015] if Then obtain the drone's altitude coordinate z at the current moment. t And the height coordinate z of the previous time step t-1 ;

[0016] According to the formula

[0017]

[0018] Determine judgment conditions C1, C2, and C3, where h ll δ1 is the lower limit of radar detection altitude, δ1 is a coefficient greater than 1, Δt is the time difference between the current moment and the previous moment, and n is the lower limit of radar detection altitude. d The number of moments corresponding to the preset camera calibration duration;

[0019] If all three conditions C1, C2, and C3 are met simultaneously, it is determined that there is a risk of radar losing the target.

[0020] According to the present invention, if there is a risk of radar losing the target, the first camera to be activated and its orientation are determined based on the first coordinate data and the camera coordinate data of multiple cameras, including:

[0021] Input the first coordinate data of the drone at multiple moments after it enters the risk area into the trained position prediction model to obtain the predicted coordinate data of the first preset number of moments in the future.

[0022] Determine the first line connecting the plane coordinates corresponding to the camera coordinate data and the plane coordinates corresponding to the first coordinate data, and determine the second line connecting the plane coordinates corresponding to the camera coordinate data and the plane coordinates corresponding to the predicted coordinate data;

[0023] A third extent is defined within the risk area, wherein the third extent includes the first extent;

[0024] Among multiple cameras, identify the undetermined camera whose first and second lines do not intersect with the third range;

[0025] The optimal selection coefficient of the camera to be determined is determined based on the camera coordinate data of the camera to be determined, the first coordinate data, the predicted coordinate data, and the third range.

[0026] Based on the optimization coefficients of each candidate camera, the first camera is determined;

[0027] The orientation of the first camera is determined based on the camera coordinate data of the first camera and the first coordinate data.

[0028] According to the present invention, determining the preferred coefficient of the camera to be determined based on the camera coordinate data of the camera to be determined, the first coordinate data, the predicted coordinate data, and the third range includes:

[0029] According to the formula

[0030]

[0031] Determine the optimization coefficient C for the i-th undetermined camera. S,i , where L 1,i Let l be the distance between the first coordinate data and the camera coordinate data of the i-th camera to be determined. 3,i L represents the minimum planar distance between the planar coordinates corresponding to the first coordinate data and the third range. 1,i,p,j Let l be the distance between the predicted coordinates of the j-th future time and the coordinates of the i-th undetermined camera. 3,i,p,j Let γ be the minimum planar distance between the predicted coordinate data at the j-th time in the future and the third range, where γ is a coefficient less than 1 and greater than 0, N1 is the first preset quantity, j≤N1, and both j and N1 are positive integers.

[0032] According to the present invention, the risk coefficient of a drone is determined based on at least one of the first coordinate data and video frames captured by the first camera, the duration of the drone's entry into the risk area, and the first range of the drone no-fly zone, including:

[0033] If the radar does not lose the target, the first coordinate data will be used as the UAV's positioning data;

[0034] If the radar loses the target, the drone's positioning data is determined based on the image coordinates of the drone in the video frames captured by multiple first cameras and the calibration parameters of the multiple first cameras;

[0035] The risk factor of the drone is determined based on the location data of the drone at multiple times after entering the risk area, the duration of the drone's stay in the risk area, and the first range of the drone no-fly zone.

[0036] According to the present invention, the risk coefficient of a drone is determined based on positioning data of the drone at multiple times after entering the risk area, the duration of the drone's entry into the risk area, and a first range of the drone no-fly zone, including:

[0037] Determine the coordinates of the centroid plane of the first region;

[0038] Obtain the third line connecting the centroid plane coordinates and the plane coordinates corresponding to the positioning data at each time point;

[0039] Determine the first intersection point between the third line and the projected edge of the first range;

[0040] Obtain the first planar distance between the first intersection point and the planar coordinates corresponding to the positioning data at each time point;

[0041] The planar distance function is obtained by fitting the first planar distance at each time point to the distance at each time point.

[0042] Obtain the first distance between the first intersection point at each time and the first intersection point at the time when the radar detects the drone entering the risk area, on the first projection edge of the first range;

[0043] Determine a first ratio between the first distance and the total length of the projected edge of the first range;

[0044] The first proportion corresponding to each time point is fitted to the proportion of each time point to obtain the observation proportion function;

[0045] The risk coefficient of the drone is determined based on the observation scale function, the planar distance function, and the duration of the drone's entry into the risk area.

[0046] According to the present invention, the risk coefficient of a UAV is determined based on an observation scale function, a planar distance function, and the duration of the UAV's entry into the risk area, including:

[0047] According to the formula

[0048]

[0049] Determine the risk factor R of the drone, where f o (T) is the observation scale function, f l (T) is the planar distance function, and T is the time variable. t The duration of the drone's stay in the risk area is given by r, where r is the radius of the risk area. Let T be the average planar velocity of the drone at multiple moments after it enters the risk area, 0 ≤ T ≤ T t w1 and w2 are preset weights.

[0050] According to the present invention, the method further includes:

[0051] The first camera is repositioned when at least one first camera loses the drone's field of view, or when a first preset number of time intervals have elapsed.

[0052] According to the present invention, selecting a second camera to capture recorded images of the drone based on the first coordinate data includes:

[0053] Immediately re-identify the first camera and designate the re-identified first camera as the second camera to capture and record images.

[0054] By adopting the above technical solution, the present invention can achieve the following technical effects:

[0055] According to the present invention, after a drone enters a risk area and approaches a no-fly zone, its coordinate data can be detected by radar. If there is a risk of radar loss of the drone, a suitable camera can be selected to capture the drone in a timely manner, thereby reducing the risk of target loss. Furthermore, the risk of a drone entering a risk area can be assessed in real time, reducing interference with passing drones that are not at risk. When assessing whether the radar is at risk of losing the target, it can determine whether the drone entering the risk area is leaving the risk area. If the drone has not left the risk area, it can be determined whether the drone is at a low altitude and continuing to decrease in altitude, and whether the time required for the drone to reach the radar's detection altitude limit is short, thus assessing whether the radar is at risk of losing the target. Additionally, sufficient calibration time is reserved for the camera to track and capture the drone, thereby enabling precise drone tracking, improving tracking accuracy, and reducing the probability of target loss. When selecting the first camera, the probability that the camera can continuously, clearly, and stably track and film the drone at each time step can be determined by the ratio of the minimum planar distance of the first coordinate data to the third range, the distance between the first coordinate data and the camera coordinate data of the i-th camera to be selected, and the ratio of the distance between the planar coordinate data corresponding to the predicted coordinate data at the j-th time step to the minimum planar distance of the third range, the distance between the predicted coordinate data at the j-th time step and the camera coordinate data of the i-th camera to be selected. Furthermore, when solving for the probability that the camera can continuously, clearly, and stably track and film the drone at future times, a discount rate is set for future possibilities, taking into account the accuracy of the position prediction model, thereby improving the accuracy of the selection coefficient. This provides an accurate and objective standard for selecting the first camera, improves the accuracy and stability of tracking and filming the drone, and reduces the probability of losing the target. When determining the risk coefficient of a drone, an integral method can be used to process the risks of the clarity and comprehensiveness of the drone's observation of the no-fly zone. This yields the risk of the drone observing the no-fly zone. The ratio of the time the drone spends in the risk zone to the longest time it takes to traverse the risk zone at its average speed represents the likelihood of the drone deliberately lingering to observe the no-fly zone. This provides the drone's risk coefficient, which accurately indicates the risk of the drone deliberately lingering in the risk zone to observe the no-fly zone and obtaining more information. It also accurately represents the risk of information prohibited from being photographed in the no-fly zone being leaked. Attached Figure Description

[0056] Figure 1 An exemplary flowchart illustrates a low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to an embodiment of the present invention;

[0057] Figure 2A schematic diagram illustrating no-fly zones and risk zones for drones according to embodiments of the present invention is provided.

[0058] Figure 3 A schematic diagram of first coordinate data and predicted coordinate data according to an embodiment of the present invention is shown exemplarily;

[0059] Figure 4 A schematic diagram of the third connection according to an embodiment of the present invention is shown as an example. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0062] Figure 1 An exemplary flowchart illustrates a low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to an embodiment of the present invention, the method comprising:

[0063] Step S1: Establish a coordinate system with the radar as the origin, and set up no-fly zone and risk zone for drones in the coordinate system. The no-fly zone for drones is located within the risk zone, and the second range of the risk zone is larger than the first range of the no-fly zone for drones.

[0064] Step S2: When the radar detects that the UAV has entered the risk area, acquire the first coordinate data of the UAV in the coordinate system;

[0065] Step S3: Based on the first coordinate data, determine whether there is a risk of radar losing the target;

[0066] Step S4: If there is a risk of radar losing the target, then based on the first coordinate data and the camera coordinate data of multiple cameras, determine the first camera that needs to be activated and the orientation of the first camera. The multiple cameras are set at multiple locations outside the no-fly zone of the drone and within the risk zone.

[0067] Step S5: Determine the risk coefficient of the drone based on at least one of the first coordinate data and video frames captured by the first camera, the duration of the drone entering the risk area, and the first range of the drone no-fly zone.

[0068] Step S6: If the risk coefficient is higher than a preset risk threshold, determine the alarm level;

[0069] Step S7: Based on the first coordinate data, select the second camera to capture the recorded image of the drone;

[0070] Step S8: Determine countermeasures based on the alarm level.

[0071] According to embodiments of the present invention, the low-altitude UAV countermeasures method based on radar positioning and image recognition technology can detect the coordinate data of a UAV by radar after the UAV enters a risk area and approaches a no-fly zone. When there is a risk of losing the target, a suitable camera can be selected in time to photograph the UAV, thereby reducing the risk of losing the target. It can also judge the risk of the UAV in real time after it enters the risk area, reducing interference to passing UAVs that are not at risk.

[0072] According to an embodiment of the present invention, in step S1, the radar may be located within a risk area or within a no-fly zone for drones. The location of the radar can be used as the origin of a coordinate system. For example, the ground plane can be set as the xoy plane and the vertical direction can be determined as the z-axis direction.

[0073] Figure 2 A schematic diagram of no-fly zones and risk zones for drones according to embodiments of the present invention is shown as an example. Figure 2 This is a planar schematic diagram of a no-fly zone and a risk zone for drones. The no-fly zone is located within the risk zone. The second range of the risk zone (which, from a top-down view, is a circular area centered on the radar's location, indicated by the dotted circular line) is larger than the first range of the no-fly zone (which, from a top-down view, is indicated by the dotted rectangular line). The heights of both ranges can be set to infinity. In other words, if the drone's planar coordinates (x-axis and y-axis coordinates) are within the circular range, the drone is considered to have entered the risk zone; if its planar coordinates are within the rectangular range, it is considered to have entered the no-fly zone. A radar is installed within the no-fly zone (located at the origin, i.e., its coordinates are (0,0,0)). Outside the no-fly zone, within the risk zone, multiple cameras are installed at multiple locations, arranged around the no-fly zone, for example, evenly distributed around it.

[0074] According to one embodiment of the present invention, in step S2, the radar can scan the surroundings once every preset time interval (e.g., 2 seconds). If the drone is detected to have entered the second range of the risk area, the first coordinate data of the drone can be determined. For example, the distance between the drone and the radar can be determined by the time difference between the transmitted signal and the echo. The direction of the strongest echo can be determined by the rotation of the radar antenna or electronic scanning, and the azimuth angle of the drone can be determined by the direction. The altitude of the drone can be determined by the elevation angle of the echo, thereby determining the relative positional relationship between the drone and the radar. Then, based on the coordinates of the radar in the coordinate system, the first coordinate data of the drone in the coordinate system can be determined.

[0075] According to one embodiment of the present invention, in step S3, due to factors such as the radar having a certain elevation angle, the detection difficulty of aircraft flying at low altitudes increases, and the target may be lost. For example, if the UAV's flight altitude is below 50 meters, the risk of the radar not being able to detect the UAV (i.e., losing the target) increases. Therefore, the risk of radar losing the target can be determined based on whether the UAV's flight altitude is low or whether the UAV has a trend of decreasing flight altitude.

[0076] According to an embodiment of the present invention, in step S3, determining whether there is a risk of radar target loss based on the first coordinate data includes:

[0077] Step S31: Determine the planar coordinates (x, y) of the UAV at the current moment based on the first coordinate data. t ,y t ) and the plane coordinates (x) of the previous time step t-1 ,y t-1 ), where the current time is the t-th time after the drone enters the risk area;

[0078] Step S32, if Then obtain the drone's altitude coordinate z at the current moment. t And the height coordinate z of the previous time step t-1 ;

[0079] Step S33: Determine judgment condition C1, judgment condition C2, and judgment condition C3 according to formula (1).

[0080]

[0081] Among them, h ll δ1 is the lower limit of radar detection altitude, δ1 is a coefficient greater than 1, Δt is the time difference between the current moment and the previous moment, and n is the lower limit of radar detection altitude. d The number of moments corresponding to the preset camera calibration duration;

[0082] Step S34: If judgment conditions C1, C2 and C3 are satisfied simultaneously, it is determined that there is a risk of radar losing the target.

[0083] According to an embodiment of the present invention, in step S31, the first coordinate data of the UAV can be acquired at each moment based on the above-described method of acquiring the first coordinate data. Based on the first coordinate data, the planar coordinates of the UAV at the current moment and the planar coordinates at the previous moment are determined. That is, the coordinate data of the first two dimensions (i.e., the x-axis coordinate and the y-axis coordinate) in the first coordinate data (three-dimensional coordinate data) are determined as planar coordinates to obtain the planar coordinates of the UAV at the current moment and the planar coordinates at the previous moment. The interval between the two moments is the aforementioned preset duration (e.g., 2 seconds).

[0084] According to one embodiment of the present invention, in step S32, it can be determined whether the drone is leaving the risk area; if That is, the distance between the drone and the coordinate origin is decreasing; in other words, the drone is approaching the center of the risk area (i.e., the drone no-fly zone), and it is possible to continue to determine whether there is a risk of radar losing the target. Otherwise, if If this indicates that the drone is leaving the risk area, then there is no need to further assess the risk of the radar losing the target.

[0085] According to an embodiment of the present invention, in step S33, formula (1) describes three judgment conditions for determining whether there is a risk of radar losing the target. In judgment condition C1, the lower limit of the radar detection altitude can be set to 50 meters. When the UAV's flight altitude is below 50 meters, the probability that the radar cannot detect the UAV (i.e., the target is lost) increases, that is, there is a risk of losing the target. The judgment altitude can be set to δ1h. ll δ1 can be set to 1.5 or 2 if the drone's altitude coordinates are less than or equal to δ1h. ll Then, the lower limit of the detection altitude of the approaching drone is determined.

[0086] According to one embodiment of the present invention, in the judgment condition C2, The rate of change of altitude between the current moment and the previous moment. This indicates that the direction of the altitude change rate is downward, meaning that the drone's altitude is decreasing.

[0087] According to one embodiment of the present invention, in the judgment condition C3, the altitude change z of the UAV t -z t-1 The required time is Δt, which is a preset time for the drone to descend from its current altitude to the lower limit of the radar's detection altitude at this descent rate. When a drone approaches the lower limit of radar detection altitude, a camera can be used to photograph it. This allows the camera to continue tracking the drone even when it is difficult for the radar to detect it. However, the process of the camera pointing at the drone and taking a picture takes a certain amount of time. For example, the camera needs to adjust its orientation angle and focus, which takes time. This is collectively referred to as calibration time, n. d Δt is the average duration of camera calibration, which can be used as the preset camera calibration duration (e.g., n). d =1), if the time required for the drone to descend from its current altitude to the radar's detection altitude limit is less than the preset camera calibration time, the camera will also have difficulty capturing the drone, potentially resulting in a loss of target. Therefore, in Right now, In this way, a judgment can be made in advance, thus giving the camera time to calibrate, which makes it easier for the camera to track and film drones at low altitudes.

[0088] According to an embodiment of the present invention, in step S34, if judgment condition C1, judgment condition C2 and judgment condition C3 are satisfied at the same time, it indicates that the drone is at a low altitude and the altitude is still decreasing. Furthermore, the time required for the drone to reach the lower limit of the radar detection altitude is short, close to the preset camera calibration time. In this case, it can be determined that there is a risk of the radar losing the target, and the camera is given calibration time so that it can be aimed at the drone for timely tracking and shooting. Thus, when the radar has difficulty detecting the drone, the camera can continue to track the drone.

[0089] In this way, it can be determined whether a drone entering a risk area is leaving the risk area. If the drone has not left the risk area, it can be determined whether the drone is at a low altitude and is still decreasing in altitude. In addition, the time required for the drone to reach the radar's detection altitude limit is short, thereby determining whether the radar is at risk of losing the target. It also allows time for the camera to calibrate and track the drone, thus enabling precise tracking of the drone, improving tracking accuracy, and reducing the probability of losing the target.

[0090] According to one embodiment of the present invention, in step S4, as shown above, if there is a risk of radar losing the target, the drone can be tracked and photographed by a camera in order to accurately track the drone, thereby reducing the risk of completely losing the target. A suitable first camera can be selected from multiple cameras to photograph the drone. The first camera can be a camera that is close to the drone and there are no obstacles in the space between the camera and the drone, so that the drone's image can be successfully captured.

[0091] According to an embodiment of the present invention, in step S4, if there is a risk of radar losing the target, the first camera to be activated and its orientation are determined based on the first coordinate data and the camera coordinate data of the multiple cameras, including:

[0092] Step S41: Input the first coordinate data of the drone at multiple times after it enters the risk area into the trained position prediction model to obtain the predicted coordinate data of the first preset number of times in the future.

[0093] Step S42: Determine the first line connecting the plane coordinates corresponding to the camera coordinate data and the plane coordinates corresponding to the first coordinate data, and determine the second line connecting the plane coordinates corresponding to the camera coordinate data and the plane coordinates corresponding to the predicted coordinate data.

[0094] Step S43: Determine the third range of the obstacle area within the risk area, wherein the third range includes the first range;

[0095] Step S44: Among the multiple cameras, identify the camera to be determined that has no intersection between the first and second connecting lines and the third range.

[0096] Step S45: Determine the preferred coefficient of the camera to be determined based on the camera coordinate data of the camera to be determined, the first coordinate data, the predicted coordinate data, and the third range;

[0097] Step S46: Determine a second preset number of first cameras based on the optimization coefficients of each candidate camera;

[0098] Step S47: Determine the orientation of the first camera based on the camera coordinate data of the first camera and the first coordinate data.

[0099] Figure 3 A schematic diagram of first coordinate data and predicted coordinate data according to an embodiment of the present invention is shown as an example.

[0100] According to an embodiment of the present invention, in step S41, the position prediction model can be a BP neural network model, an LSTM model, etc. The present invention does not limit the type of position prediction model. The position prediction model can be trained using historical coordinate data of the UAV collected over a past time period. For example, the position at the 1st to nth historical time is input into the position prediction model, and the model can output the predicted position at the (n+1)th to (n+m)th time. The historical position data at the (n+1)th to (n+m)th historical time can then be compared with the predicted position to determine the error in the predicted position, thereby determining the loss function of the position prediction model. The position prediction model can be trained by backpropagating the loss function. After multiple training iterations and verification of the accuracy of the position prediction model, a trained position prediction model is obtained, where n and m are both positive integers. The trained position prediction model can be used to analyze the first coordinate data (e.g., ...) at multiple times after the UAV enters the risk area. Figure 3 The data (shown as a small rectangle with a solid line in the middle) is processed to obtain the predicted coordinate data for the first preset number of future time moments (e.g., ...). Figure 3 (As shown by the small rectangle with the dashed line in the middle).

[0101] According to an embodiment of the present invention, in step S42, the camera coordinate data is the three-dimensional coordinate data of the camera in the coordinate system, and the planar coordinates corresponding to the camera coordinate data are the coordinate data of the first two dimensions of the three-dimensional coordinate data, namely, the x-axis coordinate data and the y-axis coordinate data. Both the first coordinate data and the predicted coordinate data are three-dimensional coordinate data, and their corresponding planar coordinate data are also the coordinate data of the first two dimensions of the three-dimensional coordinate data, namely, the x-axis coordinate data and the y-axis coordinate data. The planar coordinates corresponding to the camera coordinate data can be connected with the planar coordinates corresponding to the first coordinate data to obtain a first connecting line, and the planar coordinates corresponding to the camera coordinate data can be connected with the planar coordinates corresponding to the predicted coordinate data to obtain a second connecting line. The first and second connecting lines can be used to determine whether there are obstacles obstructing the view between the camera and the drone. If the first and second connecting lines do not cross obstacles on the two-dimensional plane, then there are no obstacles obstructing the view between the drone and the camera in three-dimensional space. Furthermore, using two-dimensional coordinates simplifies calculations and saves computational resources.

[0102] According to one embodiment of the present invention, in step S43, the obstacle area is the range of the area where the obstacle is located, such as the area where obstacles such as trees are located in the risk area. The drone no-fly zone may also block the camera's field of vision. Therefore, the drone no-fly zone can also be used as an obstacle area. That is, the third range may include the first range.

[0103] According to an embodiment of the present invention, in step S44, the first connection ( Figure 3The solid line connecting the small solid rectangle corresponding to the first coordinate data (the middle camera) and the third range have no intersection, indicating that there are no obstacles obstructing the current position of the drone and the camera. The second line ( Figure 3 The fact that the dashed line connecting the small rectangle corresponding to the camera and the predicted coordinate data has no intersection with the third range indicates that there are no obstacles obstructing the predicted coordinate data of the drone and the camera. The fact that the first and second lines have no intersection with the third range means that the camera can continuously capture images of the drone at multiple moments without being affected by obstacles, and can be used as a camera to be determined.

[0104] According to one embodiment of the present invention, based on Figure 2 Regarding the distribution of cameras, when a drone is in a risk area, there may be multiple cameras waiting to be positioned, meaning multiple cameras could continuously capture images of the drone at multiple moments. However, only 2-3 cameras are needed to locate the drone's coordinates. Therefore, the most suitable camera can be selected from these multiple waiting cameras; for example, the camera closest to the drone and capturing the clearest image can be designated as the primary camera. The camera distribution can also be... Figure 2 While the results may be inconsistent, multiple potential cameras may still exist, requiring selection. If the number of potential cameras is small (e.g., due to numerous obstacles, fewer cameras can continuously capture images of the drone), such as only two or three, all potential cameras can be directly identified as the first camera. If there is only one potential camera, or even none, the amount of predicted coordinate data for the drone can be reduced (e.g., the first preset number is reduced by one each time, thus reducing the number of predicted coordinate data points). This reduces the number of second connections, thereby relaxing the selection criteria for potential cameras. In other words, more cameras can continuously capture images of the drone in fewer consecutive moments, unaffected by obstacles. If there are still no potential cameras or only one, the amount of predicted coordinate data can be further reduced until the first preset number is reduced to zero, or until two or more potential cameras are found.

[0105] According to an embodiment of the present invention, in step S45, as described above, only 2-3 cameras are needed to locate the coordinates of the drone. Therefore, the most suitable first camera for shooting can be selected from multiple undetermined cameras. The preferred coefficient of the undetermined camera is determined based on the camera coordinate data of the undetermined camera, the first coordinate data, the predicted coordinate data, and the third range, including: determining the preferred coefficient C of the i-th undetermined camera according to formula (2). S,i ,

[0106]

[0107] Among them, L 1,iLet l be the distance between the first coordinate data and the camera coordinate data of the i-th camera to be determined. 3,i L represents the minimum planar distance between the planar coordinates corresponding to the first coordinate data and the third range. 1,i,p,j Let l be the distance between the predicted coordinates of the j-th future time and the coordinates of the i-th undetermined camera. 3,i,p,j Let γ be the minimum planar distance between the predicted coordinate data at the j-th time in the future and the third range, where γ is a coefficient less than 1 and greater than 0, N1 is the first preset quantity, j≤N1, and both j and N1 are positive integers.

[0108] According to an embodiment of the present invention, in formula (2), This is the ratio of the minimum planar distance between the current coordinate data and the first coordinate data at the current moment, and the minimum planar distance within the third range, to the distance between the first coordinate data and the camera coordinate data of the i-th undetermined camera. The larger the distance between the current coordinate data and the minimum planar distance within the third range, the lower the probability that the drone will be obstructed by obstacles in the next moment; in other words, the higher the probability of capturing the drone continuously for multiple moments. The smaller the distance between the first coordinate data and the camera coordinate data of the i-th undetermined camera, the clearer the image captured by the camera of the drone, and the less susceptible to interference from other flying objects. For example, if birds fly near the drone, and the camera is too far from the drone, the drone will appear small in the camera's image and be difficult to distinguish, potentially leading to confusion between the bird and the drone, and subsequent tracking and filming of the bird. Therefore, the smaller the distance between the first coordinate data and the camera coordinate data of the i-th undetermined camera, the higher the probability that the camera can capture the drone clearly and stably. This can comprehensively describe the probability that the i-th camera can clearly and stably track and film the drone at multiple consecutive moments. The higher this ratio, the better the tracking effect on the drone.

[0109] According to one embodiment of the present invention, Meaning and Similarly, this can represent the probability that, at time j in the future, the camera to be determined will still be able to clearly and stably track and film the drone for multiple consecutive moments. The higher this ratio, the better the tracking performance of the camera to be determined on the drone at that time. However, the predicted coordinates of the drone in the future are obtained through a position prediction model, which may contain errors. There may be errors. Therefore, when analyzing the probability that the camera will still be able to clearly and stably track and film the drone for multiple consecutive moments in the future at the j-th moment, this probability will be discounted. The larger the time difference between the j-th future moment and the current moment, the smaller the discount. In other words, it is necessary to... Set a smaller coefficient. γ can be used as the discount rate coefficient, which can be the accuracy of the trained location prediction model. For example, when validating the trained location prediction model on a validation set, the difference between the historical coordinate data in the validation set and the coordinate data predicted by the trained location prediction model can be calculated to obtain a difference vector. The magnitude of the difference vector is then calculated, and the ratio of this magnitude to the magnitude of the historical coordinate data can be used as the prediction accuracy of the trained location prediction model for that coordinate. The prediction accuracy of the trained location prediction model for multiple coordinates can be averaged to obtain the accuracy of the trained location prediction model, i.e., the aforementioned discount rate coefficient (e.g., γ = 0.95). This means analyzing the probability that the camera will still be able to clearly and stably track and film the drone for multiple consecutive moments in the future at the current moment. In other words, the probability that the camera will still be able to clearly and stably track and film the drone for multiple consecutive moments in the future at the current moment is discounted j times. This represents the total probability that the camera will be able to clearly and stably track and film the drone for a predetermined number of future moments, based on the current moment. This indicates the probability that the camera to be determined can continuously, clearly and stably track and film the drone at each moment, and can be used as the optimization coefficient of the camera to be determined. Similarly, the optimization coefficient of each camera can be obtained. The higher the optimization coefficient, the more stable the shooting and tracking quality of the camera, and the more likely it is to be selected as the first camera.

[0110] According to one embodiment of the present invention, in step S46, two or three cameras with the highest preferred coefficient (i.e., a second preset number) can be selected as the first camera. In step S47, the orientation of the first camera is determined. For example, if adjusting the orientation of the first camera requires a preset time, the predicted coordinate data of the UAV at the first future moment can be determined, and the direction of the vector between the camera coordinate data and the predicted coordinate data can be determined, which is the orientation of the first camera. For example, the orientation angle of the projection of the vector onto the xoy plane can be used as the yaw angle of the first camera, and the angle between the vector and the xoy plane can be used as the pitch angle of the first camera.

[0111] In this way, the probability that a camera can continuously, clearly, and stably track and film the drone at each time step can be determined by the ratio of the minimum planar distance between the first coordinate data and the third range, and the distance between the first coordinate data and the camera coordinate data of the i-th camera to be determined, as well as the ratio of the ratio of the planar coordinate data corresponding to the predicted coordinate data at the j-th time step to the minimum planar distance of the third range, and the distance between the predicted coordinate data at the j-th time step and the camera coordinate data of the i-th camera to be determined. Furthermore, when solving for the probability that a camera can continuously, clearly, and stably track and film the drone at future times, a discount rate is set for future possibilities, taking into account the accuracy of the position prediction model, thereby improving the accuracy of the selection coefficient. This provides an accurate and objective standard for selecting the first camera, improves the accuracy and stability of tracking and filming the drone, and reduces the probability of losing the target.

[0112] According to one embodiment of the present invention, the method further includes: re-determining the first camera when at least one first camera loses the drone's field of view, or when a first preset number of time intervals have ended. If the first camera loses the drone's field of view, it needs to be replaced to film the drone; or, if the drone's position is expected to have changed significantly by the end of the first preset number of time intervals, the first camera can be re-determined to film the drone. The method for re-determining the first camera is similar to that described above and will not be repeated here.

[0113] According to one embodiment of the present invention, in step S5, if the radar does not lose the target, the first coordinate data determined by the radar is used to determine the flight trajectory of the drone, thereby determining the risk of the drone. If the radar loses the target, the coordinates of the drone in the video frame captured by the first camera can be used to calculate the drone's coordinates in the coordinate system, thereby determining the drone's flight trajectory and thus determining the risk of the drone. The no-fly zone can be an area where drones are prohibited from staying or filming. The longer the drone stays in the risk zone, the more images of the no-fly zone it may capture, and the greater the risk of the drone. Furthermore, if the drone's flight trajectory can surround the boundary of the no-fly zone, the images of the no-fly zone captured by the drone are more comprehensive, and the risk of the drone is greater.

[0114] According to an embodiment of the present invention, in step S5, the risk coefficient of the drone is determined based on at least one of the first coordinate data and video frames captured by the first camera, the duration of the drone's entry into the risk area, and the first range of the drone no-fly zone, including:

[0115] Step S51: If the radar has not lost the target, the first coordinate data is used as the UAV's positioning data.

[0116] Step S52: If the radar loses the target, determine the UAV's positioning data based on the UAV's image coordinates in the video frames captured by the multiple first cameras and the calibration parameters of the multiple first cameras.

[0117] Step S53: Determine the risk coefficient of the drone based on the positioning data of the drone at multiple times after entering the risk area, the duration of the drone's entry into the risk area, and the first range of the drone no-fly zone.

[0118] According to one embodiment of the present invention, in step S51, as described above, if the radar has not lost the target, more accurate radar positioning data (i.e., first coordinate data) is used as the UAV's positioning data. In step S52, if the radar has lost the target, the UAV's positioning data is determined based on the UAV's image coordinates in video frames captured by multiple first cameras, the calibration parameters of the multiple first cameras, and the camera coordinate data of the multiple first cameras. For example, each camera can be calibrated before being put into use, for example, using a checkerboard calibration board. Triangulation is performed based on the calibration parameters (intrinsic and extrinsic parameters) of at least two first cameras and the UAV's image coordinates to obtain the UAV's coordinates in the coordinate system, i.e., the positioning data.

[0119] According to an embodiment of the present invention, in step S53, the risk coefficient of the UAV is determined based on the positioning data of the UAV at multiple times after entering the risk area, the duration of the UAV's entry into the risk area, and the first range of the UAV no-fly zone, including:

[0120] Step S531: Determine the coordinates of the centroid plane of the first range;

[0121] Step S532: Obtain the third line connecting the centroid plane coordinates and the plane coordinates corresponding to the positioning data at each time point;

[0122] Step S533: Determine the first intersection point between the third line and the projection edge of the first range;

[0123] Step S534: Obtain the first planar distance between the first intersection point and the planar coordinates corresponding to the positioning data at each time point;

[0124] Step S535: Fit the first plane distance corresponding to each time step with the plane distance function at each time step.

[0125] Step S536: Obtain the first distance between the first intersection point corresponding to each time moment and the first intersection point corresponding to the time when the radar detects the drone entering the risk area on the projection edge of the first range;

[0126] Step S537: Determine a first ratio between the first distance and the total length of the projected edge of the first range;

[0127] Step S538: Fit the first proportion corresponding to each time point to each time point to obtain the observation proportion function;

[0128] Step S539: Determine the risk coefficient of the UAV based on the observation scale function, the planar distance function, and the duration of the UAV's entry into the risk area.

[0129] According to an embodiment of the present invention, in step S531, the centroid plane coordinates of the first range are the projections of the first range (spatial range) onto the xoy plane. Figure 2 The centroid of the area (shown by the dashed rectangle) is determined. In step S532, after the UAV enters the risk area, multiple moments have passed. The planar coordinates (i.e., x-axis and y-axis coordinates) corresponding to the positioning data at each moment can be connected to the centroid planar coordinates to obtain a third connecting line corresponding to each moment. In step S533, the third connecting line intersects with the projection edge of the first range, and the coordinates of the first intersection point can be determined. In step S534, the first planar distance can be used as the planar projection distance between the UAV and the UAV no-fly zone. In step S535, the first planar distance corresponding to each moment can be fitted to each moment using interpolation or polynomial fitting to obtain a planar distance function, which is used to describe the change law of the planar projection distance between the UAV and the UAV no-fly zone over time.

[0130] Figure 4 A schematic diagram of the third connection according to an embodiment of the present invention is shown as an example.

[0131] According to one embodiment of the present invention, the line connecting the planar coordinates corresponding to the positioning data and the centroidal plane coordinates is a third line. This third line intersects the projection edge of the first range (the edge of the first range projected onto the xoy plane, i.e., Figure 4 The intersection of the dashed rectangular line and the first intersection point is defined as the first intersection point. In step S536, the first distance between the first intersection point at each moment and the first intersection point at the initial moment when the drone enters the risk area, on the projected edge of the first range, can be determined. Figure 4 As shown, the solid-line rectangle labeled 1 represents the planar coordinates corresponding to the initial positioning data of the UAV entering the risk area, and the solid-line rectangle labeled 2 represents the planar coordinates corresponding to the second positioning data of the UAV entering the risk area. Based on the above method, the first intersection points corresponding to the two can be obtained respectively. The first distance between the two first intersection points on the projection edge of the first range is the distance traveled by the first intersection point corresponding to the initial moment along the projection edge to the first intersection point corresponding to the second moment, that is, Figure 4 The total length of the portion within the dashed rectangle that has been changed to a solid line. Using a similar approach, the first distance at each time step can be determined.

[0132] According to one embodiment of the present invention, in step S537, a first ratio of the total length of the projected edge of the first distance to the first range can be determined. For example, it is the ratio of the total length of the portion of the rectangular dashed line that has been modified to a solid line to the perimeter of the rectangular frame. Similarly, the first ratio corresponding to each time moment can be determined. In step S538, the first ratio corresponding to each time moment is fitted to each time moment to obtain an observation ratio function, which is used to describe the change law of the first ratio over time, and can also be used to describe the change law of the ratio of the drone observing the drone no-fly zone over time.

[0133] According to an embodiment of the present invention, in step S539, the risk coefficient of the UAV is determined based on the observation scale function, the planar distance function, and the duration of the UAV's entry into the risk area, including: determining the risk coefficient R of the UAV according to formula (3).

[0134]

[0135] Among them, f o (T) is the observation scale function, f l (T) is the planar distance function, and T is the time variable. t The duration of the drone's stay in the risk area is given by r, where r is the radius of the risk area. Let T be the average planar velocity of the drone at multiple moments after it enters the risk area, 0 ≤ T ≤ T t w1 and w2 are preset weights.

[0136] According to one embodiment of the present invention, in formula (3), at time T after the UAV enters the risk area, the planar projection distance between the UAV and the UAV no-fly zone can be expressed as f. l (T) represents the proportion of the drone's flight path that can surround the no-fly zone, which is the first proportion. It can be used to represent the proportion of no-fly zones observed by a drone, and can be expressed as f. o (T) indicates that f o The higher the (T) value, the more comprehensive the drone's observation of the no-fly zone, and the greater the risk posed by the drone. It can be used to indicate the risk associated with the comprehensiveness of a drone's observation of the no-fly zone. l The smaller (T) is, the closer the drone is to the no-fly zone, the clearer the drone can observe the no-fly zone, and the greater the risk to the drone. Therefore, it can be determined by... This indicates the risk to the clarity of drone observations of no-fly zones.

[0137] According to one embodiment of the present invention, both of the above risks can accumulate over time. For example, the risk of incomplete observation of no-fly zones by drones can occur even if the drone remains stationary.o (T) remains constant, but the drone can still acquire more information about the no-fly zone over time from the same location. For example, it can take more pictures of the no-fly zone and learn more details from the pictures. Therefore, the risk of the drone's comprehensive observation of the no-fly zone can accumulate over time. On the other hand, at the first moment, the risk of the drone's comprehensive observation of the no-fly zone is 15%. At the second moment, the risk of the drone's comprehensive observation of the no-fly zone is 20%. At the third moment, the drone moves in the opposite direction, causing the risk of the drone's comprehensive observation of the no-fly zone to drop to 13%. This does not mean that the total risk of the drone's comprehensive observation of the no-fly zone decreases. On the contrary, it means that the drone repeatedly observes the same part of the no-fly zone. That is, the total risk of the drone's comprehensive observation of the no-fly zone increases. Therefore, the risk of the drone's comprehensive observation of the no-fly zone can accumulate over time.

[0138] According to one embodiment of the present invention, on the other hand, the risk of the drone's observation clarity in the no-fly zone also accumulates over time. For example, if the drone's flight process involves first approaching the no-fly zone and then moving away from it, the risk value of the drone's observation clarity in the no-fly zone will first increase and then decrease. However, this does not mean that the total risk of the drone's observation clarity in the no-fly zone decreases. On the contrary, it means that the drone repeatedly observes the no-fly zone from far to near and then from near to far, obtaining more images of the no-fly zone and understanding more details in the images. That is, the total risk of the drone's observation clarity in the no-fly zone increases. Therefore, the risk of the drone's observation clarity in the no-fly zone accumulates over time.

[0139] According to one embodiment of the present invention, the instantaneous risk values ​​of the two risks mentioned above at the same moment can be weighted and summed. The two preset weights can be set to equal values, for example, both equal to 0.5, or they can be set to unequal values; the present invention does not limit this. After weighted summation, the instantaneous value of the observation risk of the UAV over the UAV no-fly zone can be obtained. Integrating this instantaneous value yields the risk of observing the UAV in the no-fly zone since the UAV entered the risk area. Furthermore, the longest time it takes for a drone to traverse a risk area at its average speed is In other words, the time it takes for a drone to traverse the risk zone along its diameter; the longer the drone spends inside the risk zone, the longer it can observe the no-fly zone. This is the ratio of the time a drone spends entering a risk area to the longest time it takes for the drone to traverse the risk area at its average speed. The larger this ratio, the greater the likelihood that the drone is not simply passing through the risk area, but rather observing the no-fly zone. When this ratio exceeds 1, it can be considered that the drone is deliberately lingering in the risk area to observe the no-fly zone. This can be used to describe the likelihood that a drone is specifically observing a no-fly zone rather than simply passing through it; that is, it describes the risk of a drone deliberately loitering. Multiplying the risk of a drone deliberately loitering by the risk of a drone observing a no-fly zone yields a risk coefficient. A higher risk coefficient indicates that the drone is more likely to deliberately loiter within the risk area to observe the no-fly zone, and that it observes more information; in other words, the higher the risk of the drone and the higher the risk of information leaked that is prohibited from being captured by the drone. Specifically, the ratio of displacement to time between multiple adjacent moments after the drone enters the risk area can be calculated, i.e., the velocity over multiple time intervals, and the average of these velocities is obtained.

[0140] In this way, the risks of the clarity and comprehensiveness of drone observations of no-fly zones can be processed through integration to obtain the risk of drones observing no-fly zones. The probability of a drone deliberately lingering to observe no-fly zones is represented by the ratio of the time a drone spends entering the risk zone to the longest time it takes for a drone to cross the risk zone at its average speed. This yields a risk coefficient for the drone, which can accurately represent the risk of a drone deliberately lingering in the risk zone to observe no-fly zones and observing more information, and also accurately represent the risk of information that is prohibited from being photographed in no-fly zones being leaked.

[0141] According to one embodiment of the present invention, in step S6, an alarm level is determined when the risk coefficient is higher than a preset risk threshold. As the time the drone spends in the risk area increases and the location the drone flies over increases, its risk coefficient will become higher and higher until it exceeds the preset risk threshold (e.g., 3). At this point, the drone can be determined to be of high risk, and its risk coefficient can be updated in real time to determine the alarm level. For example, a risk coefficient higher than 3 but less than or equal to 5 is a low alarm level, a risk coefficient higher than 5 but less than or equal to 10 is a medium alarm level, and a risk coefficient higher than 10 is a high alarm level.

[0142] According to one embodiment of the present invention, in step S7, when the risk coefficient is higher than a preset risk threshold, a second camera can be selected to capture recorded images of the drone based on the first coordinate data. Step S7 may include: immediately re-determining the first camera and designating the re-determined first camera as the second camera to capture recorded images. That is, the method of selecting the second camera is the same as that of selecting the first camera; a camera that is close to the drone, unobstructed, and expected to continuously capture images of the drone at multiple future moments can be used as the second camera, thereby obtaining more recorded images of the drone and thus obtaining more information about the drone for further investigation and evidence collection.

[0143] According to one embodiment of the present invention, in step S8, countermeasures can be determined based on the alarm level. For example, if the drone leaves the risk area after reaching the lowest alarm level, recorded images can be saved for later observation. If the drone leaves the risk area after reaching the highest alarm level, recorded images can be saved, and the relevant drone management agency can be notified for investigation, while the drone operator is also alerted. If the drone remains in the risk area after reaching the highest alarm level, measures such as radio interference can be used to prevent it from continuing to fly or to drive the drone away. Recorded images can be saved, the relevant drone management agency can be notified for investigation, and the drone operator is also alerted. This improves the safety of drone no-fly zones.

[0144] The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to embodiments of the present invention can detect the coordinate data of a UAV by radar after it enters a risk area and approaches a no-fly zone. When there is a risk of radar loss of the target, a suitable camera can be selected to photograph the UAV in a timely manner, thereby reducing the risk of target loss. Furthermore, the method can assess the risk of a UAV entering a risk area in real time, reducing interference with passing, harmless UAVs. When assessing the risk of radar loss of the target, it can determine whether the UAV entering the risk area is leaving the risk area. If the UAV has not left the risk area, it can determine whether the UAV is at a low altitude and continuing to decrease in altitude, and whether the time required for the UAV to reach the radar's detection altitude limit is short, thus assessing the risk of radar loss of the target. It also allows sufficient calibration time for the camera to track and photograph the UAV, thereby enabling precise tracking, improving tracking accuracy, and reducing the probability of target loss. When selecting the first camera, the probability that the camera can continuously, clearly, and stably track and film the drone at each time step can be determined by the ratio of the minimum planar distance of the first coordinate data to the third range, the distance between the first coordinate data and the camera coordinate data of the i-th camera to be selected, and the ratio of the distance between the planar coordinate data corresponding to the predicted coordinate data at the j-th time step to the minimum planar distance of the third range, the distance between the predicted coordinate data at the j-th time step and the camera coordinate data of the i-th camera to be selected. Furthermore, when solving for the probability that the camera can continuously, clearly, and stably track and film the drone at future times, a discount rate is set for future possibilities, taking into account the accuracy of the position prediction model, thereby improving the accuracy of the selection coefficient. This provides an accurate and objective standard for selecting the first camera, improves the accuracy and stability of tracking and filming the drone, and reduces the probability of losing the target. When determining the risk coefficient of a drone, an integral method can be used to process the risks of the clarity and comprehensiveness of the drone's observation of the no-fly zone. This yields the risk of the drone observing the no-fly zone. The ratio of the time the drone spends in the risk zone to the longest time it takes to traverse the risk zone at its average speed represents the likelihood of the drone deliberately lingering to observe the no-fly zone. This provides the drone's risk coefficient, which accurately indicates the risk of the drone deliberately lingering in the risk zone to observe the no-fly zone and obtaining more information. It also accurately represents the risk of information prohibited from being photographed in the no-fly zone being leaked.

[0145] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0146] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments, and any modifications or variations of the embodiments of the present invention may be made without departing from the stated principles.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for countering and defending against low-altitude unmanned aerial vehicles (UAVs) based on radar positioning and image recognition technology, characterized in that, include: Establish a coordinate system with the radar as the origin, and set up no-fly zones and risk zones for drones in the coordinate system. The no-fly zones for drones are located within the risk zones, and the second range of the risk zones is larger than the first range of the no-fly zones for drones. When the radar detects that the drone has entered the risk area, the first coordinate data of the drone in the coordinate system is obtained; Based on the first coordinate data, determine whether there is a risk of radar losing the target; If there is a risk of radar losing the target, the first camera to be activated and its orientation are determined based on the first coordinate data and the camera coordinate data of multiple cameras. The multiple cameras are set at multiple locations outside the no-fly zone for drones but within the risk zone. Based on at least one of the first coordinate data and video frames captured by the first camera, the duration of the drone's entry into the risk area, and the first range of the drone no-fly zone, the risk coefficient of the drone is determined. If the risk coefficient is higher than a preset risk threshold, an alarm level is determined. Based on the first coordinate data, select the second camera to capture the recorded images from the drone; Determine countermeasures based on the alert level; Based on the first coordinate data, determine whether there is a risk of radar losing the target, including: Based on the first coordinate data, determine the planar coordinates (x, y) of the UAV at the current moment. t ,y t ) and the plane coordinates (x) of the previous time step t-1 ,y t-1 ), where the current time is the t-th time after the drone enters the risk area; if Then obtain the drone's altitude coordinate z at the current moment. t And the height coordinate z of the previous time step t-1 ; According to the formula Determine judgment conditions C1, C2, and C3, where h ll The lower limit of radar detection altitude, δ1 is a coefficient greater than 1, Δt is the time difference between the current moment and the previous moment, and n d The number of moments corresponding to the preset camera calibration duration; If all three conditions C1, C2, and C3 are met simultaneously, it is determined that there is a risk of radar losing the target.

2. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 1, characterized in that, If there is a risk of radar losing the target, then based on the first coordinate data and the camera coordinate data of multiple cameras, the first camera that needs to be activated, and the orientation of the first camera, are determined, including: Input the first coordinate data of the drone at multiple moments after it enters the risk area into the trained position prediction model to obtain the predicted coordinate data of the first preset number of moments in the future. Determine the first line connecting the plane coordinates corresponding to the camera coordinate data and the plane coordinates corresponding to the first coordinate data, and determine the second line connecting the plane coordinates corresponding to the camera coordinate data and the plane coordinates corresponding to the predicted coordinate data; A third extent is defined within the risk area, wherein the third extent includes the first extent; Among multiple cameras, identify the undetermined camera whose first and second lines do not intersect with the third range; The optimal selection coefficient of the camera to be determined is determined based on the camera coordinate data of the camera to be determined, the first coordinate data, the predicted coordinate data, and the third range. Based on the optimization coefficients of each candidate camera, the first camera is determined; The orientation of the first camera is determined based on the camera coordinate data of the first camera and the first coordinate data.

3. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 2, characterized in that, Based on the camera coordinate data of the camera to be determined, the first coordinate data, the predicted coordinate data, and the third range, the optimization coefficient of the camera to be determined is determined, including: According to the formula Determine the optimization coefficient C for the i-th undetermined camera. S,i , where L 1,i Let l be the distance between the first coordinate data and the camera coordinate data of the i-th camera to be determined. 3,i L represents the minimum planar distance between the planar coordinates corresponding to the first coordinate data and the third range. 1,i,p,j Let l be the distance between the predicted coordinates of the j-th future time and the coordinates of the i-th undetermined camera. 3,i,p,j Let γ be the minimum planar distance between the predicted coordinate data at the j-th time in the future and the third range, where γ is a coefficient less than 1 and greater than 0, N1 is the first preset quantity, j≤N1, and both j and N1 are positive integers.

4. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 1, characterized in that, Based on at least one of the first coordinate data and video frames captured by the first camera, the duration of the drone's entry into the risk area, and the first range of the drone no-fly zone, the risk factor of the drone is determined, including: If the radar does not lose the target, the first coordinate data will be used as the UAV's positioning data; If the radar loses the target, the drone's positioning data is determined based on the image coordinates of the drone in the video frames captured by multiple first cameras and the calibration parameters of the multiple first cameras; The risk factor of the drone is determined based on the location data of the drone at multiple times after entering the risk area, the duration of the drone's stay in the risk area, and the first range of the drone no-fly zone.

5. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 4, characterized in that, Based on the drone's location data at multiple points after entering the risk area, the duration of the drone's stay in the risk area, and the first boundary of the drone no-fly zone, the risk factor of the drone is determined, including: Determine the coordinates of the centroid plane of the first region; Obtain the third line connecting the centroid plane coordinates and the plane coordinates corresponding to the positioning data at each time point; Determine the first intersection point between the third line and the projected edge of the first range; Obtain the first planar distance between the first intersection point and the planar coordinates corresponding to the positioning data at each time point; The planar distance function is obtained by fitting the first planar distance at each time point to the distance at each time point. Obtain the first distance between the first intersection point at each time and the first intersection point at the time when the radar detects the drone entering the risk area, on the first projection edge of the first range; Determine a first ratio between the first distance and the total length of the projected edge of the first range; The first proportion corresponding to each time point is fitted to the proportion of each time point to obtain the observation proportion function; The risk coefficient of the drone is determined based on the observation scale function, the planar distance function, and the duration of the drone's entry into the risk area.

6. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 5, characterized in that, The risk coefficient of the UAV is determined based on the observation scale function, the planar distance function, and the duration of the UAV's entry into the risk area, including: According to the formula Determine the risk factor R of the drone, where f o (T) is the observation scale function, f l (T) is the planar distance function, and T is the time variable. t The duration of the drone's stay in the risk area is given by r, where r is the radius of the risk area. Let T be the average planar velocity of the drone at multiple moments after it enters the risk area, 0 ≤ T ≤ T t w1 and w2 are preset weights.

7. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 2, characterized in that, The method further includes: The first camera is repositioned when at least one first camera loses the drone's field of view, or when a first preset number of time intervals have elapsed.

8. The low-altitude UAV countermeasures method based on radar positioning and image recognition technology according to claim 7, characterized in that, Based on the first coordinate data, the second camera is selected to capture the recorded images from the drone, including: Immediately re-identify the first camera and designate the re-identified first camera as the second camera to capture and record images.

Images