Unmanned aerial vehicle countering method and system based on deep learning

Abstract

The invention discloses an unmanned aerial vehicle countering method and system based on deep learning, and relates to the technical field of unmanned aerial vehicle countering, and the method comprises the steps: collecting unmanned aerial vehicle data, and carrying out the data preprocessing; constructing a first model, and optimizing the model; and analyzing the illegal unmanned aerial vehicle through a first algorithm, and making a countering strategy. According to the unmanned aerial vehicle countering method based on deep learning, the generalization ability and robustness of the model are improved through unmanned aerial vehicle image and video data acquisition, the training efficiency and accuracy of the deep learning model are improved through preprocessing the acquired data, and the accuracy of the deep learning model is improved through rule setting of the flight height, speed, course and other features. According to the unmanned aerial vehicle countering method, the capability of discovering illegal unmanned aerial vehicles is enhanced, the countering strategy is formulated by combining the classification result, the effectiveness and flexibility of unmanned aerial vehicle countering are improved, and better effects are achieved in the aspects of robustness, accuracy and effectiveness.

Description

Technical Field

[0001] The present invention relates to the technical field of drone countermeasures, and in particular to a drone countermeasure method based on deep learning. Background Art

[0002] With the rapid development of drone technology, its applications in military, civil and commercial fields are becoming increasingly extensive, such as logistics distribution, aerial photography, agricultural monitoring and infrastructure inspection. However, the use of illegal drones is becoming more and more frequent, posing a significant threat to social security, including privacy leakage, invasion of sensitive areas, etc. In response to these problems, drone countermeasure technology has emerged. Currently common methods include signal interference, physical capture, hard kill technology, etc., as well as detection and positioning technologies based on radar detection, photoelectric detection and radio signal analysis. In addition, with the rapid development of artificial intelligence, deep learning and machine learning technologies have begun to be gradually introduced into the field of drone countermeasures, especially in target detection and pattern recognition, showing great application potential.

[0003] Although the existing drone countermeasure technology has a certain detection and interception capability, it still has many shortcomings and cannot meet the increasingly complex drone countermeasure needs. First, in terms of drone detection and identification, traditional methods based on radar or photoelectric detection are easily affected by environmental factors (such as bad weather, occlusion, etc.), resulting in short detection distance and low recognition accuracy. Secondly, in terms of abnormal behavior identification, traditional methods usually rely on preset rules or simple statistical methods, lack deep feature extraction capabilities, and are difficult to deal with the complex and changeable flight trajectories of illegal drones. In terms of classification and precise countermeasures of illegal behaviors, the existing technology has too simple a classification strategy for abnormal behaviors, usually only distinguishing between "normal" and "abnormal", and lacks the ability to refine the classification of complex behavior patterns, resulting in high false alarm and missed alarm rates. In addition, in the field of real-time positioning and trajectory analysis of drones, traditional methods find it difficult to balance computational complexity and real-time performance, especially when facing multiple targets or highly dynamic environments, and often cannot provide smooth and accurate trajectory information. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: the existing drone countermeasure methods have low recognition accuracy, insufficient abnormal behavior analysis capabilities, poor real-time positioning effect, and the optimization problem of how to accurately classify abnormal behaviors and reconstruct real-time trajectories.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a drone countermeasure method based on deep learning, comprising collecting drone data and performing data preprocessing; constructing a first model and optimizing the model; analyzing illegal drones through a first algorithm and formulating a countermeasure strategy.

[0007] As a preferred solution of the deep learning-based drone countermeasure method described in the present invention, the collecting of drone data includes collecting drone image videos and flight data.

[0008] As a preferred solution of the deep learning-based drone countermeasure method described in the present invention, the data preprocessing includes preprocessing the collected data.

[0009] As a preferred solution of the deep learning-based drone countermeasure method described in the present invention, wherein: the construction of the first model includes detecting and identifying the drone.

[0010] As a preferred solution of the deep learning-based drone countermeasure method described in the present invention, the optimization of the model includes expanding the training data set.

[0011] As a preferred solution of the deep learning-based drone countermeasure method described in the present invention, the analyzing illegal drones by the first algorithm includes identifying illegal drone behaviors.

[0012] As a preferred solution of the deep learning-based drone countermeasure method described in the present invention, wherein: the formulation of the countermeasure strategy includes formulating the countermeasure strategy according to the drone behavior analysis results.

[0013] Another object of the present invention is to provide a drone countermeasure system based on deep learning, which can build a model through a model optimization module and improve the recognition accuracy and efficiency of the model through an optimization algorithm, thereby solving the problem that the current model is prone to overfitting and the recognition effect in real scenes is reduced.

[0014] As a preferred solution of the deep learning-based drone countermeasure system described in the present invention, it includes: an acquisition preprocessing module, a model optimization module, and an analysis and countermeasure module; the acquisition preprocessing module is used to collect drone data and perform data preprocessing; the model optimization module is used to build a first model and optimize the model; the analysis and countermeasure module is used to analyze illegal drones through a first algorithm and formulate a countermeasure strategy.

[0015] A computer device includes a memory and a processor, wherein the memory stores a computer program, and wherein the processor executes the computer program to implement a step of a drone countermeasure method based on deep learning.

[0016] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a drone countermeasure method based on deep learning are implemented.

[0017] Beneficial effects of the present invention: The UAV countermeasure method based on deep learning provided by the present invention improves the generalization ability and robustness of the model through the collection of UAV images and video data, realizes efficient target detection and classification functions by selecting convolutional neural network as the deep learning algorithm, improves the training efficiency and accuracy of the deep learning model by preprocessing the collected data, enhances the ability to detect illegal UAVs by setting rules for flight altitude, speed, heading and other features, and improves the effectiveness and flexibility of UAV countermeasures by formulating countermeasure strategies based on classification results. The present invention achieves better results in terms of robustness, accuracy and effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0019] Figure 1 An overall flow chart of a drone countermeasure method based on deep learning provided for the first embodiment of the present invention.

[0020] Figure 2 An overall module diagram of a UAV countermeasure system based on deep learning provided for the fourth embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0022] Example 1

[0023] Reference Figure 1 , is an embodiment of the present invention, and provides a drone countermeasure method based on deep learning, comprising:

[0024] S1: Collect drone data and perform data preprocessing.

[0025] Furthermore, collecting drone data includes collecting drone images, videos and flight data.

[0026] It should be noted that drone image videos include static image acquisition and dynamic video recording. Drone images are taken in different scenes through professional equipment (such as high-resolution cameras) to ensure that the images cover different shooting angles (front, side, bottom, etc.), lighting conditions (day, night) and environmental backgrounds (city, rural, forest). Dynamic flight scenes are recorded through drone flight videos, including the behavioral characteristics of drones at low speed, high speed, and fixed-point hovering. In order to increase the diversity of data, drone-related data are collected from public databases (such as ImageNet, DroneAV) and Internet resources (such as social media, news reports). The collected data are classified and labeled into different types (such as quadcopters, fixed-wing drones) and uses (civilian, commercial, and illegal use).

[0027] It should also be noted that the purpose of flight data collection is to obtain the behavioral characteristics of drones in dynamic environments, including the collection of flight status data, flight trajectory data and drone equipment information. Flight status data is mainly collected through the drone's built-in sensors, including real-time data output by devices such as altitude sensors, speedometers and gyroscopes. Flight trajectory data is obtained through the GPS positioning system, and the actual flight path of the drone can be reconstructed in combination with the timestamp. The drone equipment information covers its basic parameters such as model, specification, load capacity, and communication frequency band. These data are transmitted to the ground station equipment through wireless communication technology and recorded in a unified data storage platform. Through multi-sensor fusion technology, the inertial navigation system (INS) and GPS are used to jointly correct data drift and errors. By recording the flight data of drones in severe weather and complex terrain conditions in real time, the ability to identify abnormal behavior is enhanced.

[0028] Furthermore, performing data preprocessing includes preprocessing the collected data.

[0029] It should be noted that the original data is cleaned and blurred, incomplete or irrelevant images and videos are removed to ensure the high quality of the input data. To address the noise problem in the image data, median filtering or Gaussian filtering methods are used for denoising to improve the clarity and contrast of the image. In order to enhance the diversity of the data, data enhancement techniques such as rotation, cropping, scaling, and mirror flipping are used to generate more variant data, simulate the drone morphology in different scenarios, and perform frame segmentation and key frame extraction on the video data to extract representative images to reduce redundant data. For flight data, the lost GPS signal points are processed by interpolation, and the outlier detection algorithm is used to identify and eliminate unreasonable flight parameters. All preprocessed data will be converted into a unified format and stored according to the classification and labeling standards.

[0030] Example 2

[0031] Reference Figure 1 , is an embodiment of the present invention, and provides a drone countermeasure method based on deep learning, comprising:

[0032] S2: Build the first model and optimize the model.

[0033] Furthermore, building the first model includes detecting and identifying the drone.

[0034] In an embodiment of the present application, the first model is a deep learning model, a multi-layer neural network structure designed based on a convolutional neural network (CNN), the input layer of the model receives preprocessed drone images or video frame data, the input size is unified to a fixed size to be compatible with the network structure, and low-level and high-level features of the drone are extracted through multiple convolutional layers, where the low-level features include edge and texture information, and the high-level features include shape and category features. An activation function (such as ReLU) is introduced to enhance the learning ability of the network and increase the nonlinear representation ability of the model. Each convolutional layer is followed by a pooling layer (such as maximum pooling), which reduces the amount of calculation by dimensionality reduction and alleviates overfitting to a certain extent;

[0035] The CNN model combines the characteristics of the residual network (ResNet) and adds skip connections between layers to prevent the gradient vanishing problem caused by excessive network depth. In the classification stage, the fully connected layer maps the features into multiple output categories to classify and identify drones (such as quadcopters, fixed-wing drones, etc.). In order to improve the versatility of the model, a pre-training strategy is adopted by loading the model weights trained on large-scale datasets such as ImageNet as initialization.

[0036] In an optional embodiment, the first model can also be implemented in other ways, such as a recursive neural network model, using RNN to process the UAV flight trajectory data (such as speed, acceleration, GPS position, heading, etc.). Since the flight trajectory is essentially time series data, RNN can store the association information between time steps through its hidden state, thereby capturing the dynamic characteristics of the UAV flight behavior. The input layer receives the multi-dimensional time series data of the flight trajectory, and each time step contains the characteristics of the GPS position, altitude, speed and direction of the UAV. The hidden layer uses the recurrent unit in the RNN to gradually extract the sequence features and capture the time series pattern of the UAV flight behavior. The output layer maps the final output of the hidden state to a classification label (such as normal flight, suspicious flight, illegal flight);

[0037] Traditional RNN may have the problem of gradient vanishing or gradient exploding, while LSTM and GRU introduce gating mechanisms in their structures, which can more efficiently retain long-term dependencies, which is particularly important for drone flight trajectory analysis. LSTM dynamically adjusts the information flow through input gates, forget gates, and output gates to ensure that the model can effectively capture the long-term trends in drone flight behavior. GRU adopts a simplified gating mechanism to achieve accurate modeling of time series data while maintaining computational efficiency.

[0038] Compared with CNN-based methods, RNN-based models focus more on the temporal dynamic patterns of UAV behavior rather than identifying targets only from static features. This approach is particularly suitable for analyzing complex behavior patterns, such as nonlinear anomalies hidden in flight paths. In addition, the use of LSTM and GRU enhances the model's ability to capture long-term associations, improves the recognition rate of illegal behaviors, and improves the ability to parse abnormal trajectories.

[0039] Furthermore, optimizing the model involves expanding the training dataset.

[0040] It should be noted that in terms of network structure optimization, a lightweight network framework (such as MobileNet or YOLO) is introduced to replace the traditional deep CNN. The deep separable convolution is used to significantly reduce the number of parameters and improve real-time performance. At the same time, during the network training process, data enhancement techniques (such as rotation, scaling, and color adjustment) are used to expand the size of the data set to improve the generalization and robustness of the model.

[0041] It should also be noted that the present invention adopts a gradual learning rate adjustment strategy, dynamically adjusts the learning rate according to the performance of the validation set during the training process, accelerates model convergence and avoids oscillation, and introduces batch normalization technology to stabilize the model training process and prevent gradient disappearance or gradient explosion problems. In order to avoid overfitting of the model, the Dropout technology is used to randomly close some neuron connections to make the model more robust.

[0042] It should also be noted that in response to the needs of real-time drone detection, the network is further lightweighted through model pruning and quantization methods to reduce redundant parameters and computational complexity. Weight pruning is used to remove weights that have little impact on the output, reduce model complexity, and use 8-bit quantization technology to compress model weights from floating point numbers to low-order integer representation, significantly reducing inference time. In order to adapt to edge computing devices, a model deployment solution adapted to low-power hardware is designed so that it can run efficiently in embedded systems.

[0043] It should also be noted that when the flight trajectory of a drone does not conform to the rules, it is judged as abnormal behavior. Based on the clustering method, the flight trajectory of the drone is clustered, the clustering results are analyzed, abnormal behavior is identified, and the drone behavior is divided into three categories: normal, suspicious, and illegal. The flight trajectory of a drone with normal behavior conforms to the set rules and has no abnormal behavior. The flight trajectory of a drone with suspicious behavior partially conforms to the rules or has certain abnormal behaviors. The flight trajectory of a drone with illegal behavior seriously violates the rules or has obvious abnormal behaviors.

[0044] S3: Analyze illegal drones through the first algorithm and formulate countermeasure strategies.

[0045] Furthermore, analyzing illegal drones through the first algorithm includes identifying illegal drone behavior.

[0046] In the embodiment of the present application, the first algorithm is a combination of a support vector machine and a Kalman filter algorithm, which classifies the behavior of the drone in real time based on flight characteristics, and classifies its behavior into three categories: normal, suspicious, and illegal. If the classification result is suspicious or illegal, the drone will be marked as a key monitoring object;

[0047] The Kalman filter accurately reconstructs and continuously monitors the trajectory of the drone, and further compares the reconstructed trajectory with standard flight rules to identify whether there are obvious violations, such as speeding, entering a no-fly zone, etc.

[0048] Finally, the SVM classification results are combined with the Kalman filter trajectory analysis results to make a comprehensive judgment on the illegal behavior and trigger the corresponding countermeasure strategy.

[0049] In an optional embodiment, the first algorithm can also be implemented in other ways, such as based on Bayesian networks and particle filtering. Bayesian networks are a method based on probabilistic graphical models. By constructing a causal relationship graph between variables, the classification of drone behaviors is achieved. The flight status data of the drone is defined as a node of the network. A conditional probability table is used to describe the dependency relationship between these variables. For example, the joint probability of trajectory curvature and speed can reflect whether the drone has abnormal maneuvering behavior. The probability distribution of the nodes is updated according to real-time observation data, and the posterior probability that the drone behavior belongs to a normal, suspicious or illegal category is calculated. The reasoning process of the Bayesian network combines prior knowledge and observation data, and can maintain a high classification accuracy when the data is incomplete or noisy.

[0050] Particle filtering is a recursive state estimation technology based on the Monte Carlo method. It is particularly suitable for processing nonlinear and non-Gaussian distributed system states. Particle filtering generates a large number of random particles to represent the possible states of the drone, and predicts the state distribution at the next moment based on the dynamic model of the drone. Combined with real-time GPS data, the particles are weighted and updated to make the trajectory estimation closer to the actual position. It performs well in high-dynamic or multi-target scenarios, especially in high-dynamic or multi-target scenarios. According to the changing trend of particle distribution, the trajectory smoothness and continuity are analyzed to identify abnormal behaviors that obviously deviate from the normal trajectory.

[0051] The Bayesian network infers the joint distribution of flight state variables and provides preliminary behavior classification results. The particle filter further dynamically estimates the flight trajectory to make up for the shortcomings of the Bayesian network in processing dynamic characteristics. By weighted fusion of the results of the two algorithms (such as based on confidence), a comprehensive judgment of illegal UAV behavior is achieved.

[0052] Furthermore, developing countermeasure strategies includes developing countermeasure strategies based on the results of drone behavior analysis.

[0053] It should be noted that for drones with normal behavior, the monitoring status will be maintained and no countermeasures will be taken. For drones with suspicious behavior, the early warning system will be activated to alert the monitoring personnel to pay attention, and an early warning sound will be played through the loudspeaker to alert the surrounding people. The early warning information will be displayed on the display screen to alert the monitoring personnel to pay attention. For drones with illegal behavior, countermeasures will be taken to emit electromagnetic signals of specific frequencies to interfere with the drone's communication and control signals, forcing it to lose control or land.

[0054] Example 3

[0055] One embodiment of the present invention provides a drone countermeasure method based on deep learning. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0056] A simulated scenario experiment was designed to compare the performance differences between the invention and the prior art in drone detection, abnormal behavior recognition, trajectory reconstruction and illegal behavior classification. Six common types of drones on the market were selected for the experiment, and simulated flight tests were carried out in environments such as urban buildings, open farmlands and forests. Images and videos from multiple angles and under different lighting conditions were collected for each drone, and combined with the drone flight status data (such as altitude, speed, heading and GPS data) for comprehensive analysis. The experimental data was preprocessed, including denoising, outlier removal and data enhancement (rotation, scaling, etc.). The experimental system used convolutional neural networks (CNN) for target detection and behavior feature extraction, support vector machines (SVM) for abnormal behavior classification, and the Kalman filter algorithm to process trajectory data for accurate reconstruction.

[0057] The accuracy and response speed of the detection model were tested in different scenarios. The flight status data was used to identify illegal flight behaviors, including excessively high flights and abnormal trajectories. The trajectories were plotted using GPS data, and abnormal behaviors were detected and classified. The experiment compared traditional countermeasure technologies based on radar detection and rule matching to evaluate the performance improvement of the invention.

[0058] As shown in Table 1, the detection accuracy of the present invention reaches an average of 92.3%, and the response time is maintained at about 120ms, while the detection accuracy of the traditional method is usually less than 85%, and the response time is mostly above 200ms. The deep learning model improves the efficiency and accuracy of target detection by optimizing feature extraction. The anomaly recognition accuracy of the present invention reaches up to 88.9%, which is higher than the 70%-75% of the traditional rule matching method. This is due to the accurate classification of high-dimensional behavior features by SVM, which reduces misjudgments and missed judgments. Through the dynamic trajectory correction of Kalman filtering, the trajectory reconstruction error is 1.65m on average, which is about 45% lower than the 3m error of the traditional method, providing a reliable basis for subsequent precise countermeasures. In the experiment, the recognition rate of illegal behavior reached 97.1%, reflecting the fusion advantage of deep learning and behavior classification technology, and can quickly distinguish between normal and illegal operations.

[0059] Table 1 Experimental data table

[0060]

[0061] Example 4

[0062] Reference Figure 2 , is an embodiment of the present invention, and provides a drone countermeasure system based on deep learning, including: an acquisition preprocessing module, a model optimization module, and an analysis countermeasure module.

[0063] The collection and preprocessing module is used to collect drone data and perform data preprocessing; the model optimization module is used to build a first model and optimize the model; the analysis and countermeasure module is used to analyze illegal drones through a first algorithm and formulate a countermeasure strategy.

[0064] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0065] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0066] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0067] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A drone countermeasure method based on deep learning, characterized in that: include: Collect drone data and perform data preprocessing; Build the first model and optimize the model; Analyze illegal drones through the first algorithm and formulate countermeasure strategies.

2. The deep learning-based drone countermeasure method according to claim 1, characterized in that: The collecting of drone data includes collecting drone images, videos and flight data.

3. The UAV countermeasure method based on deep learning as claimed in claim 2, characterized in that: The data preprocessing includes preprocessing the collected data.

4. The method for countering drones based on deep learning as claimed in claim 3, characterized in that: The constructing of the first model includes detecting and identifying the drone.

5. The deep learning-based drone countermeasure method according to claim 4, characterized in that: The optimizing the model includes expanding the training data set.

6. The method for countering drones based on deep learning as claimed in claim 5, characterized in that: Analyzing illegal drones using a first algorithm includes identifying illegal drone behavior.

7. The method for countering drones based on deep learning as claimed in claim 6, characterized in that: The formulating of a countermeasure strategy includes formulating a countermeasure strategy based on the results of the drone behavior analysis.

8. A system using the deep learning-based drone countermeasure method as claimed in any one of claims 1 to 7, characterized in that: It includes acquisition preprocessing module, model optimization module, and analysis and countermeasure module; The acquisition and preprocessing module is used to collect drone data and perform data preprocessing; The model optimization module is used to construct a first model and optimize the model; The analysis and countermeasure module is used to analyze illegal drones through a first algorithm and formulate a countermeasure strategy.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the deep learning-based drone countermeasure method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the deep learning-based drone countermeasure method described in any one of claims 1 to 7 are implemented.

Images