Image Recognition-Based UAV Threat Assessment Method and Related Equipment

Abstract

This invention provides a drone threat assessment method based on image recognition. The method includes: acquiring a current environmental image during the drone's flight; predicting the trajectory of the target object when a target object is detected in the current environmental image; and assessing the threat of the drone based on the trajectory. By performing image recognition on the drone's flight process and predicting the target object's trajectory when it is detected in the current environmental image to assess the drone's threat, accurate threat assessment can be performed during drone flight, allowing the drone to take evasive action in advance and improving its safety during flight.

Description

Technical Field

[0001] This invention relates to the field of image processing, and in particular to a method and related equipment for determining drone threats based on image recognition. Background Technology

[0002] With the widespread adoption and large-scale use of civilian drones, they have brought convenience to people's work, such as for field exploration, high-speed traffic aerial photography, and scenic area aerial photography. When performing tasks over large areas, drones generally fly along planned flight routes. However, various threats may exist during flight, such as flying organisms, other drones, objects blown into the air by the wind, and natural terrain. The presence of these threats directly affects the safety of drones during flight, thus reducing their safety. Summary of the Invention

[0003] This invention provides an image recognition-based method for drone threat assessment, aiming to address the issue of reduced safety during drone flight. By performing image recognition on the drone's flight process and detecting target objects in the current environmental image, the method predicts the target object's trajectory to assess the drone threat. This allows for accurate threat assessment during drone flight, enabling the drone to take evasive action in advance and improving its safety during flight.

[0004] In a first aspect, embodiments of the present invention provide a method for determining drone threats based on image recognition, the method comprising:

[0005] Acquire images of the current environment during the drone's flight;

[0006] When a target object is detected in the current environment image, the motion trajectory of the target object is predicted;

[0007] The threat assessment of the drone is performed based on its movement trajectory.

[0008] Optionally, before predicting the motion trajectory of the target object when the presence of a target object in the current environmental image is detected, the method includes:

[0009] Obtain the current flight altitude of the drone;

[0010] Using the current flight altitude as prior knowledge, target detection is performed on the current environmental image.

[0011] Optionally, the step of performing target detection on the current environmental image based on the current flight altitude as prior knowledge includes:

[0012] Based on the current flight altitude, a corresponding target detection model is matched, and the target detection model is trained based on image sets at different altitudes;

[0013] The target detection model is used to perform target detection on the current environment image.

[0014] Optionally, when a target object is detected in the current environmental image, predicting the motion trajectory of the target object includes:

[0015] When a target object is detected in the current environment image, the target object is tracked to obtain a tracked image of the target object;

[0016] Trajectory analysis is performed on the tracking image of the target object to obtain the tracking trajectory of the target object;

[0017] Based on the tracking trajectory, the motion trajectory of the target object is predicted.

[0018] Optionally, the step of performing trajectory analysis on the tracking image of the target object to obtain the tracking trajectory of the target object includes:

[0019] Key point extraction is performed on the tracked image to obtain the feature key points of the target object;

[0020] Trajectory analysis is performed based on the key feature points to obtain the tracking trajectory of the target object.

[0021] Optionally, predicting the motion trajectory of the target object based on the tracking trajectory includes:

[0022] Obtain the fluid motion parameters at the current flight altitude;

[0023] By combining the fluid motion parameters and the tracking trajectory, the motion trajectory of the target object is predicted using a pre-trained time series model, thus obtaining the motion trajectory of the target object.

[0024] Optionally, the step of determining the threat of the drone based on the movement trajectory includes:

[0025] Obtain the flight path and contour parameters of the UAV;

[0026] Based on the flight path and the contour parameters, a three-dimensional dynamic route of the UAV is generated;

[0027] Calculate the intersection probability between the motion trajectory and the three-dimensional dynamic route;

[0028] Threat determination is performed on the drone based on the cross-probability.

[0029] Secondly, embodiments of the present invention provide a drone threat determination device based on image recognition, the device comprising:

[0030] The acquisition module is used to acquire images of the current environment during the drone's flight.

[0031] The prediction module is used to predict the motion trajectory of the target object when the presence of the target object is detected in the current environmental image;

[0032] The determination module is used to determine the threat of the drone based on the movement trajectory.

[0033] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the image recognition-based drone threat determination method provided in embodiments of the present invention.

[0034] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the image recognition-based drone threat determination method provided in the embodiments of the present invention.

[0035] In this embodiment of the invention, a current environmental image is acquired during the flight of the drone; when a target object is detected in the current environmental image, the trajectory of the target object is predicted; and a threat assessment is performed on the drone based on the trajectory. By performing image recognition on the drone's flight process and predicting the trajectory of the target object when it is detected in the current environmental image to determine the drone's threat, accurate threat assessment can be performed during drone flight, allowing the drone to take evasive action in advance and improving the drone's safety during flight. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of a drone threat determination method based on image recognition provided in an embodiment of the present invention;

[0038] Figure 2This is a schematic diagram of the structure of a drone threat determination device based on image recognition provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0040] 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.

[0041] Please see Figure 1 , Figure 1 This is a flowchart of a drone threat determination method based on image recognition provided in an embodiment of the present invention, such as... Figure 1 As shown, the image recognition-based drone threat assessment method includes the following steps:

[0042] 101. Acquire images of the current environment during the flight of the UAV.

[0043] In this embodiment of the invention, the aforementioned drone can be understood as any type of flight device performing a mission, including rotary-wing and jet-powered aircraft. The drone is equipped with image acquisition devices, which can be multiple image acquisition devices positioned at different locations to capture images from different directions; for example, there could be six image acquisition devices positioned in front, behind, to the left, right, up, and down.

[0044] Multiple image acquisition devices positioned at different locations capture images of the current environment during the drone's flight. Specifically, these multiple environmental images can include images from various directions captured by the multiple image acquisition devices.

[0045] 102. When a target object is detected in the current environment image, predict the motion trajectory of the target object.

[0046] In this embodiment of the invention, the current environment image includes images from multiple directions. Image recognition can be performed sequentially on the images from each direction to detect whether a target object exists in the current environment image. For example, image recognition can be performed sequentially in the directions of front, left, right, up, back, and down.

[0047] The aforementioned target objects are those that may affect the flight of the drone, such as flying organisms, other drones, objects blown into the air by the wind, and natural terrain.

[0048] The image recognition described above can be based on object detection algorithms. These algorithms classify objects and then identify which category an object in the image belongs to. Among these categories, those that have an impact on the drone's flight process are the corresponding target categories, and the objects in these target categories are the target objects.

[0049] When a target object is detected, it can be tracked using a corresponding image acquisition device, and the trajectory of the target object can be predicted based on the tracked image.

[0050] 103. Determine the threat level of the drone based on its movement trajectory.

[0051] In this embodiment of the invention, the motion trajectory represents the motion trend of the target object and the degree of threat it poses to the drone over a future period of time. Based on different levels of threat, the drone is assessed for threat.

[0052] Specifically, the threat level can be divided into three levels, with Level 1 being the highest, Level 2 being the most common, and Level 3 being the lowest. Level 1 threat can be understood as the target object's trajectory intersecting directly with the drone in the future, causing a collision. Level 2 threat can be understood as the target object's trajectory being within a predetermined distance from the drone in the future, with a probability of collision. Level 3 threat can be understood as the target object's trajectory being outside a predetermined distance from the drone in the future, with no probability of collision under normal circumstances.

[0053] In one possible embodiment, the drone control system can make corresponding control adjustments to the drone based on the threat level. For example, when a level one or two threat is detected, the drone's flight path can be temporarily modified. When a level three threat is detected, the threat assessment of the target object can be continuously performed until the target object disappears.

[0054] In another possible embodiment, after adjusting the flight path, the threat assessment of the target object can continue until the threat level is reduced to level three or the target object disappears.

[0055] In this embodiment of the invention, a current environmental image is acquired during the flight of the drone; when a target object is detected in the current environmental image, the trajectory of the target object is predicted; and a threat assessment is performed on the drone based on the trajectory. By performing image recognition on the drone's flight process and predicting the trajectory of the target object when it is detected in the current environmental image to determine the drone's threat, accurate threat assessment can be performed during drone flight, allowing the drone to take evasive action in advance and improving the drone's safety during flight.

[0056] Optionally, when a target object is detected in the current environmental image, before predicting the target object's trajectory, the current flight altitude of the UAV can be obtained; using the current flight altitude as prior knowledge, target detection can be performed on the current environmental image.

[0057] In this embodiment of the invention, the current position of the drone can be obtained in real time according to the positioning system on the drone, and the current flight altitude of the drone can be determined based on the current position.

[0058] In one possible embodiment, when it is determined that there is no deviation in the flight status of the drone, the flight altitude set in the flight path of the drone can be determined as the current flight altitude of the drone.

[0059] Optionally, in the step of performing target detection on the current environmental image based on the current flight altitude as prior knowledge, a corresponding target detection model can be matched according to the current flight altitude. The target detection model is trained on image sets at different altitudes. The target detection model is then used to perform target detection on the current environmental image.

[0060] In this embodiment of the invention, since different altitudes in the air may contain different objects, it can be understood that different altitudes have different object distributions. For example, the distribution of birds at different altitudes may also correspond to different objects. For instance, at lower flight altitudes, leaves may be blown by the wind, while at higher altitudes, there may be no leaves. Other objects such as balloons, kites, and sky lanterns also have a maximum reach. Therefore, different altitudes will have different object distributions, and this object distribution can be considered as prior knowledge.

[0061] By matching the target detection model with the current flight altitude, a more targeted target detection model can be obtained to detect target objects, thereby improving the accuracy of target object detection.

[0062] Specifically, image sets can be collected based on different flight altitudes, with each flight altitude corresponding to a separate image set containing images of objects that may appear at that altitude. The object detection model is then trained in a supervised manner using these different image sets, resulting in trained object detection models corresponding to each image set. The number of trained object detection models corresponds one-to-one with the number of image sets.

[0063] Optionally, in the step of predicting the motion trajectory of a target object when a target object is detected in the current environmental image, the target object can be tracked to obtain a tracking image of the target object; trajectory analysis can be performed on the tracking image of the target object to obtain the tracking trajectory of the target object; and the motion trajectory of the target object can be predicted based on the tracking trajectory.

[0064] In this embodiment of the invention, when a target object is detected in the current environmental image, it can be determined that the image contains the target object. The direction of the target object is determined based on the image containing the target object, and the image acquisition device in that direction is controlled to track the target object to obtain a tracking image of the target object.

[0065] The tracking images of the target object can be consecutive frames. Based on the position information of the target object in the consecutive frames, the tracking trajectory of the target object is obtained. The tracking trajectory is essentially a spatiotemporal sequence. Based on this spatiotemporal sequence, the spatiotemporal changes of the target object are predicted, and the motion trajectory of the target object is obtained.

[0066] Specifically, the position of the target object in a series of frames can be determined based on its detection bounding box. For example, the detection bounding box can be (x, y, w, h, γ), where (x, y) are the coordinates of the center point of the detection bounding box, w and h are the width of the detection bounding box, and γ is the confidence level of the detection bounding box. The coordinates of the center point of the detection bounding box can be used as the position information of the target object, and the tracking trajectory of the target object can be obtained based on its position information in the series of frames.

[0067] The tracking trajectory can be predicted using a trained temporal model to obtain the motion trajectory of the target object. This motion trajectory includes the tracking trajectory and the predicted trajectory, which is the motion trajectory of the target object after the current time.

[0068] Specifically, sample motion trajectories of various objects can be collected and divided into two segments: a tracking trajectory and a label trajectory, with the label trajectory occurring after the tracking trajectory. The tracking trajectory is input into a time-series model, which outputs a predicted trajectory. The error loss between the predicted and label trajectories is calculated, and the time-series model is trained to minimize this error loss, ensuring its output closely resembles the label trajectory. In use, the tracking trajectory is input into the trained time-series model, which outputs a predicted trajectory. This predicted trajectory is then concatenated with the tracking trajectory to obtain the target object's motion trajectory.

[0069] In one possible embodiment, sample motion trajectories of various objects can be collected. A segment of the trajectory from the first middle part of the sample motion trajectory is extracted as the sample tracking trajectory, and this sample motion trajectory is used as the label trajectory. The sample tracking trajectory is input into a temporal model, which outputs a predicted trajectory. The error loss between the predicted trajectory and the label trajectory is calculated. The temporal model is trained with the goal of minimizing the error loss, making the output of the temporal model similar to the label trajectory, thus completing the training and obtaining a trained temporal model. In use, the tracking trajectory is input into the trained temporal model, and the motion trajectory of the target object is obtained through the output of the trained temporal model.

[0070] Optionally, in the step of performing trajectory analysis on the tracking image of the target object to obtain the tracking trajectory of the target object, key points can be extracted from the tracking image to obtain the feature key points of the target object; trajectory analysis can be performed based on the feature key points to obtain the tracking trajectory of the target object.

[0071] In this embodiment of the invention, the aforementioned key feature points can be the geometric center point of the target object, or they can be contour points of the target object. When extracting key points from the tracking image, image segmentation can be performed on each frame to obtain a segmented image of the target object. Key point extraction is then performed based on the segmented image to obtain the key feature points of the target object.

[0072] In one possible embodiment, the aforementioned key features can be SIFT (Scale-invariant feature transform) key points. SIFT features are based on points of interest in the local appearance of an object and are independent of the image size and rotation. They also have a high tolerance for changes in lighting, noise, and micro-viewpoints. Based on these characteristics, they are highly salient and relatively easy to extract, making it easy to identify objects in a large feature database with few false positives. Using SIFT features also results in a high detection rate for partially occluded objects; therefore, only three or more SIFT features are sufficient to calculate the object's position and orientation. The target object's position information can be determined based on the SIFT key points, thereby determining the target object's tracking trajectory.

[0073] Optionally, in the step of predicting the motion trajectory of the target object based on the tracking trajectory, the fluid motion parameters at the current flight altitude can be obtained; combining the fluid motion parameters and the tracking trajectory, the motion trajectory of the target object can be predicted by a pre-trained time series model to obtain the motion trajectory of the target object.

[0074] In this embodiment of the invention, the fluid motion parameters of the current flight altitude may include parameters such as air flow speed, air flow direction, and air density. By correcting the trajectory prediction process of the target object through the fluid motion parameters, a more accurate trajectory of the target object can be obtained.

[0075] Specifically, assuming the tracking trajectory is denoted as 'an', representing the position of the target object in n frames of images, and the motion trajectory is denoted as 'cm', the prediction formula for the time-series model is cm = w × an + b, where w is the weight parameter and b is the bias parameter, both obtained during training. After incorporating fluid motion parameters, the prediction formula for the time-series model becomes cm = λ(w × an + b), where λ represents the fluid motion parameters. More specifically, λ = (ν1 × μ1 × ρ1) / (ν0 × μ0 × ρ0), where ν1 represents the current airflow velocity, μ1 represents the current airflow direction, ρ1 represents the current air density, ν0 represents the reference airflow velocity, μ0 represents the reference airflow direction, and ρ0 represents the reference air density. The reference airflow velocity, reference airflow direction, and reference air density can be set empirically.

[0076] Optionally, in the step of determining the threat of a drone based on its motion trajectory, the drone's flight path and contour parameters can be obtained; a three-dimensional dynamic route of the drone can be generated based on the flight path and contour parameters; the intersection probability between the motion trajectory and the three-dimensional dynamic route can be calculated; and the threat of the drone can be determined based on the intersection probability.

[0077] In this embodiment of the invention, the flight path of the drone can be pre-planned and can be directly retrieved from the drone's memory. The drone's contour parameters can be preset. The drone will have different contour parameters in different directions. For example, the first contour corresponding to the front and rear is the same, the second contour corresponding to the left and right is the same, and the third contour corresponding to the top and bottom is the same, but the first, second, and third contours are different from each other. Therefore, the contour parameters of the drone can be obtained according to the direction of the target object.

[0078] The aforementioned contour parameters represent the outer contour of the UAV. Using the flight path as the direction of motion, the outer contour of the UAV is moved along the flight path to obtain the UAV's three-dimensional dynamic path. The intersection probability between the aforementioned trajectory and the three-dimensional dynamic path can be understood as the probability that the trajectory passes through the three-dimensional dynamic path. Specifically, the intersection probability can be expressed by the formula K = l0 / (l1+l2), where l0 is the distance from the target object to the UAV, l1 is the length of the tracking trajectory, and l2 is the length of the trajectory.

[0079] Specifically, the threat level can be divided into three levels: Level 1 is the highest, Level 2 is the average, and Level 3 is the lowest. Specifically, when K is greater than the first threshold, it can be classified as Level 1 threat; when K is less than the first threshold but greater than the second threshold, it can be classified as Level 2 threat; and when K is less than the second threshold, it can be classified as Level 3 threat. Level 1 threat can be understood as the trajectory of the target object directly intersecting the drone within a certain period of time, causing a collision. Level 2 threat can be understood as the trajectory of the target object being within a predetermined distance from the drone within a certain period of time, with a probability of collision. Level 3 threat can be understood as the trajectory of the target object being outside the predetermined distance from the drone within a certain period of time, with no probability of collision under normal circumstances.

[0080] It should be noted that the image recognition-based drone threat determination method provided in this embodiment of the invention can be applied to drones, mobile phones, servers and other devices that can perform image recognition-based drone threat determination.

[0081] It should be noted that the image recognition-based drone threat determination method provided in this embodiment of the invention can be applied to devices such as smartphones, computers, and servers that can perform image recognition-based drone threat determination.

[0082] Optional, please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a drone threat determination device based on image recognition provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the device includes:

[0083] The acquisition module 201 is used to acquire images of the current environment during the flight of the UAV;

[0084] Prediction module 202 is used to predict the motion trajectory of the target object when the presence of the target object is detected in the current environment image;

[0085] The determination module 203 is used to determine the threat of the drone based on the movement trajectory.

[0086] Optionally, the device further includes:

[0087] An altitude acquisition module is used to acquire the current flight altitude of the UAV;

[0088] The detection module is used to perform target detection on the current environmental image based on the current flight altitude as prior knowledge.

[0089] Optionally, the detection module is further configured to match a corresponding target detection model based on the current flight altitude, wherein the target detection model is trained based on image sets at different altitudes; and to perform target detection on the current environment image using the matched target detection model.

[0090] Optionally, the prediction module 202 is further configured to, when a target object is detected in the current environment image, track the target object to obtain a tracking image of the target object; perform trajectory analysis on the tracking image of the target object to obtain a tracking trajectory of the target object; and predict the motion trajectory of the target object based on the tracking trajectory.

[0091] Optionally, the prediction module 202 is further configured to extract key points from the tracking image to obtain the feature key points of the target object; and to perform trajectory analysis based on the feature key points to obtain the tracking trajectory of the target object.

[0092] Optionally, the prediction module 202 is further configured to acquire fluid motion parameters at the current flight altitude; and combine the fluid motion parameters with the tracking trajectory to predict the motion trajectory of the target object using a pre-trained time series model, thereby obtaining the motion trajectory of the target object.

[0093] Optionally, the determination module 203 is further configured to acquire the flight path of the UAV and the contour parameters of the UAV; generate a three-dimensional dynamic route of the UAV based on the flight path and the contour parameters; calculate the intersection probability of the motion trajectory and the three-dimensional dynamic route; and determine the threat of the UAV based on the intersection probability.

[0094] It should be noted that the image recognition-based drone threat determination device provided in this embodiment of the invention can be applied to drones, mobile phones, servers and other devices that can perform image recognition-based drone threat determination.

[0095] The image recognition-based drone threat determination device provided in this embodiment of the invention can implement all the processes of the image recognition-based drone threat determination method in the above-described method embodiments, and can achieve the same beneficial effects. To avoid repetition, further details are omitted here.

[0096] See Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 3 As shown, it includes: a memory 302, a processor 301, and a computer program for a drone threat determination method based on image recognition, stored in the memory 302 and executable on the processor 301, wherein:

[0097] The processor 301 is used to call the computer program stored in the memory 302 and perform the following steps:

[0098] Acquire images of the current environment during the drone's flight;

[0099] When a target object is detected in the current environment image, the motion trajectory of the target object is predicted;

[0100] The threat assessment of the drone is performed based on its movement trajectory.

[0101] Optionally, before predicting the motion trajectory of the target object when the presence of a target object is detected in the current environmental image, the method executed by the processor 301 includes:

[0102] Obtain the current flight altitude of the drone;

[0103] Using the current flight altitude as prior knowledge, target detection is performed on the current environmental image.

[0104] Optionally, the process executed by processor 301 to perform target detection on the current environment image based on the current flight altitude as prior knowledge includes:

[0105] Based on the current flight altitude, a corresponding target detection model is matched, and the target detection model is trained based on image sets at different altitudes;

[0106] The target detection model is used to perform target detection on the current environment image.

[0107] Optionally, when the processor 301 detects the presence of a target object in the current environmental image, predicting the motion trajectory of the target object includes:

[0108] When a target object is detected in the current environment image, the target object is tracked to obtain a tracked image of the target object;

[0109] Trajectory analysis is performed on the tracking image of the target object to obtain the tracking trajectory of the target object;

[0110] Based on the tracking trajectory, the motion trajectory of the target object is predicted.

[0111] Optionally, the process of processor 301 performing trajectory analysis on the tracking image of the target object to obtain the tracking trajectory of the target object includes:

[0112] Key point extraction is performed on the tracked image to obtain the feature key points of the target object;

[0113] Trajectory analysis is performed based on the key feature points to obtain the tracking trajectory of the target object.

[0114] Optionally, the process of predicting the motion trajectory of the target object based on the tracking trajectory, performed by processor 301, includes:

[0115] Obtain the fluid motion parameters at the current flight altitude;

[0116] By combining the fluid motion parameters and the tracking trajectory, the motion trajectory of the target object is predicted using a pre-trained time series model, thus obtaining the motion trajectory of the target object.

[0117] Optionally, the process of determining the threat of the drone based on the motion trajectory, executed by the processor 301, includes:

[0118] Obtain the flight path and contour parameters of the UAV;

[0119] Based on the flight path and the contour parameters, a three-dimensional dynamic route of the UAV is generated;

[0120] Calculate the intersection probability between the motion trajectory and the three-dimensional dynamic route;

[0121] Threat determination is performed on the drone based on the cross-probability.

[0122] The electronic device provided in this embodiment of the invention can implement all the processes of the image recognition-based drone threat determination method in the above-described method embodiments, and can achieve the same beneficial effects. To avoid repetition, further details are omitted here.

[0123] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the image recognition-based drone threat determination method or the application-side image recognition-based drone threat determination method provided in this invention, and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0124] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0125] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for determining drone threats based on image recognition, characterized in that, Includes the following steps: Acquire images of the current environment during the drone's flight; Obtain the current flight altitude of the drone; Using the current flight altitude as prior knowledge, target detection is performed on the current environmental image; When a target object is detected in the current environment image, the motion trajectory of the target object is predicted; Based on the movement trajectory, a threat assessment is performed on the drone; The step of performing target detection on the current environment image based on the current flight altitude as prior knowledge includes: Based on the current flight altitude, a corresponding target detection model is matched, and the target detection model is trained based on image sets at different altitudes; The target detection model is used to perform target detection on the current environment image; The threat determination of the drone based on the movement trajectory includes: Obtain the flight path and contour parameters of the UAV; Based on the flight path and the contour parameters, a three-dimensional dynamic route of the UAV is generated; Calculate the intersection probability between the motion trajectory and the three-dimensional dynamic route; Threat determination is performed on the drone based on the cross-probability.

2. The drone threat determination method based on image recognition as described in claim 1, characterized in that, When a target object is detected in the current environment image, predicting the motion trajectory of the target object includes: When a target object is detected in the current environment image, the target object is tracked to obtain a tracked image of the target object; Trajectory analysis is performed on the tracking image of the target object to obtain the tracking trajectory of the target object; Based on the tracking trajectory, the motion trajectory of the target object is predicted.

3. The drone threat determination method based on image recognition as described in claim 2, characterized in that, The step of performing trajectory analysis on the tracking image of the target object to obtain the tracking trajectory of the target object includes: Key point extraction is performed on the tracked image to obtain the feature key points of the target object; Trajectory analysis is performed based on the key feature points to obtain the tracking trajectory of the target object.

4. The drone threat determination method based on image recognition as described in claim 3, characterized in that, The step of predicting the motion trajectory of the target object based on the tracking trajectory includes: Obtain the fluid motion parameters at the current flight altitude; By combining the fluid motion parameters and the tracking trajectory, the motion trajectory of the target object is predicted using a pre-trained time series model, thus obtaining the motion trajectory of the target object.

5. A drone threat assessment device based on image recognition, characterized in that, For the drone threat determination method based on image recognition as described in any one of claims 1 to 4, the drone threat determination device based on image recognition includes: The acquisition module is used to acquire images of the current environment during the drone's flight. The prediction module is used to predict the motion trajectory of the target object when the presence of the target object is detected in the current environmental image; The determination module is used to determine the threat of the drone based on the movement trajectory.

6. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the image recognition-based drone threat determination method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the image recognition-based drone threat determination method as described in any one of claims 1 to 4.

Images