Unmanned aerial vehicle identification method and apparatus, electronic device, and storage medium
By acquiring images from target base stations and overlaying them with labels, combining trajectory information, and utilizing panoramic databases and classification models, the problems of low accuracy and high false alarm rate in existing UAV identification methods are solved, achieving accurate identification of UAVs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA MOBILE CHENGDU INFORMATION & TELECOMM TECH CO LTD
- Filing Date
- 2026-01-30
- Publication Date
- 2026-06-02
AI Technical Summary
Existing drone identification methods have low accuracy in identifying unregistered drones, and these methods are costly, cumbersome to deploy, and have a high false alarm rate, making it difficult to effectively identify drones that do not comply with registration agreements.
By acquiring multiple images centered on the target base station, obtaining target images based on the location information of the sensing sensors, overlaying labels, and combining trajectory information for drone identification, preliminary screening and accurate identification are performed using a panoramic database and classification model.
It effectively eliminates false alarm information, reduces subsequent computation, improves the accuracy and efficiency of drone identification, and achieves precise identification of drones.
Smart Images
Figure CN121617039B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, electronic device, and storage medium for drone identification. Background Technology
[0002] In existing technologies, drone identification is generally based on forcing drones to use a remote identification identifier (remoteID), and then determining whether the target is a drone based on the captured remoteID information.
[0003] The existing drone identification process requires drones to register in accordance with the agreement, which is not applicable to drones that do not comply with the registration agreement. This can lead to missed identification of drones, resulting in a low drone identification accuracy. Summary of the Invention
[0004] This application provides a drone identification method, apparatus, electronic device, and storage medium to improve the accuracy of drone identification.
[0005] This application provides a method for identifying unmanned aerial vehicles (UAVs), including the following steps:
[0006] Multiple images of the detection range are acquired with the target base station as the center; the detection range is the sensing range of the sensing sensors in the target base station.
[0007] The location information of the target to be identified, sensed by the sensing sensor, is obtained, and based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images.
[0008] Based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target;
[0009] The trajectory information of the target to be identified is obtained, and based on the trajectory information, the target to be identified is determined to be a drone.
[0010] According to the drone identification method provided in this application, based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images, including:
[0011] Based on the multiple images and their index information, a panoramic database of the detection range is constructed.
[0012] Based on the location information and the index information of the multiple images, the target image of the location of the target to be identified is obtained from the panoramic database.
[0013] According to the UAV identification method and panoramic database construction method provided in this application, the method includes:
[0014] Centered on the target base station, acquire multiple forward views and multiple top views of the sensing range;
[0015] Based on each of the forward graphs and the index information of each of the forward graphs, multiple forward graph dictionaries are constructed;
[0016] Based on each of the top views and the index information of each of the top views, multiple top view dictionaries are constructed;
[0017] The panoramic database is constructed based on the multiple forward graph dictionaries and the multiple top-view dictionaries.
[0018] According to the UAV identification method provided in this application, taking the target base station as the center, multiple forward views and multiple top views of the sensing range are obtained, including:
[0019] Centered on the target base station, images are acquired at preset angular resolutions within the horizontal and vertical field-of-view angle ranges of the sensing range to obtain multiple forward images within the sensing range.
[0020] Based on a preset segmentation threshold, the three-dimensional grid of the sensing range is divided to obtain multiple sensing regions. A top view of each sensing region is obtained with the target base station as the center, resulting in multiple top views of the sensing range.
[0021] According to the UAV identification method provided in this application, based on the location information and the index information of the plurality of images, a target image of the location of the target to be identified is obtained from the panoramic database, including:
[0022] Based on the location information, the coordinate information of the target to be identified in the three-dimensional grid is determined, and based on the coordinate information of the target to be identified in the three-dimensional grid and the index information of the multiple images, the forward target image of the location of the target to be identified is determined from the panoramic database.
[0023] Based on the location information, the horizontal and vertical angles of the target to be identified within the detection range are determined, and based on the horizontal and vertical angles and index information of multiple images, a top-down target image of the target's location is determined from the panoramic database.
[0024] According to the UAV identification method provided in this application, based on the target overlay image, the method determines the target to be identified as a flying target, including:
[0025] The positive target image is input into the classification model to obtain the first classification result output by the classification model;
[0026] The top-view target image is input into the classification model to obtain the second classification result output by the classification model;
[0027] Based on the first classification result and the second classification result, the target to be identified is determined to be a flying target;
[0028] The classification model is trained based on sample images and the corresponding flight target labels of the sample images.
[0029] According to the drone identification method provided in this application, based on the trajectory information, the method determines that the target to be identified is a drone, including:
[0030] The trajectory information is input into the anomaly detection network model to obtain the detection result output by the anomaly detection network model;
[0031] Based on the detection results, the target to be identified is determined to be a drone;
[0032] The anomaly detection network model is trained based on historical trajectory data of UAVs.
[0033] According to the UAV identification method provided in this application, based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, including:
[0034] Based on the camera parameter information of the target image, the position information is converted into coordinate information in the pixel coordinate system;
[0035] Based on the coordinate information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image.
[0036] This application also provides a drone identification device, including the following modules:
[0037] The image acquisition module is used to acquire multiple images of the detection range centered on the target base station; the detection range is the sensing range of the sensing sensor in the target base station.
[0038] The filtering module is used to obtain the location information of the target to be identified sensed by the sensing sensor, and based on the location information, to obtain the target image of the location of the target to be identified from the plurality of images;
[0039] The flight target determination module is used to overlay the identifier of the target to be identified onto the target image based on the location information to obtain a target overlay image, and to determine the target to be identified as a flight target based on the target overlay image;
[0040] The drone identification module is used to acquire the trajectory information of the target to be identified, and to determine the target to be identified as a drone based on the trajectory information.
[0041] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement any of the above-described drone identification methods.
[0042] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the unmanned aerial vehicle (UAV) identification method as described above.
[0043] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described drone identification methods.
[0044] The UAV identification method, apparatus, electronic device, and storage medium provided in this application acquire multiple images centered on a target base station, extract the target image based on the target location information provided by the sensing sensor, and then superimpose the identifiers to form a target superimposed image to confirm the target as a flying target. This eliminates a large amount of false alarm information, reduces the computational load of the subsequent accurate judgment process, improves judgment efficiency, and realizes the initial screening process. After confirming that it is a flying target, its trajectory information is further acquired and analyzed to achieve the accurate identification process of the UAV, thus improving the accuracy of UAV identification. Attached Figure Description
[0045] To more clearly illustrate the technical solutions in this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0046] Figure 1 This is a flowchart illustrating the drone identification method provided in this application.
[0047] Figure 2 This is a schematic diagram of the judgment process provided in this application.
[0048] Figure 3 This is a schematic diagram of the testing process provided in this application.
[0049] Figure 4 This is a schematic diagram of the structure of the drone identification device provided in this application.
[0050] Figure 5 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation
[0051] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0052] Sensor-based systems play a crucial role in low-altitude airspace surveillance. First, the sensing system can detect and identify various targets within low-altitude airspace, including drones, aircraft, and birds, providing regulators with real-time target information. Second, the sensing system provides situational awareness of low-altitude airspace, monitoring parameters such as target position, speed, and altitude, helping regulators understand dynamic changes within the airspace. Furthermore, the sensing system can monitor and identify abnormal behaviors, such as overtaking or exceeding altitude limits, providing regulators with a basis for timely action. Through data collection, analysis, and prediction, the sensing system provides regulators with more comprehensive information support, helping them effectively manage and control low-altitude airspace and ensure aviation safety.
[0053] The relevant methods for identifying and regulating drones generally include:
[0054] One related method is based on forcing drones to use a remote ID, and then determining whether the target is a drone based on the captured remote ID information. This approach requires drones to register according to the agreed-upon terms, which is not suitable for drones that do not comply with the registration agreement, and may lead to missed detections of drones, resulting in a low accuracy rate for drone identification.
[0055] The second approach uses a perception sensor to detect the target, then controls a moving camera to track the target, and determines whether the target is a drone based on the tracked image information. This approach has a series of drawbacks, including high cost (requiring equipment such as cameras and gimbals), complicated deployment, and a high false alarm rate of current perception sensors (easily misidentifying cars, ground or airborne moving targets as drones).
[0056] To address the shortcomings of existing methods, this application provides a method for identifying unmanned aerial vehicles (UAVs). Figure 1 This is a flowchart illustrating the drone identification method provided in this application, as shown below. Figure 1 As shown, the method includes the following:
[0057] Step 110: With the target base station as the center, acquire multiple images of the detection range; the detection range is the sensing range of the sensing sensor in the target base station.
[0058] Step 120: Obtain the location information of the target to be identified by the sensing sensor, and based on the location information, obtain the target image of the location of the target to be identified from the plurality of images;
[0059] Step 130: Based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target;
[0060] Step 140: Obtain the trajectory information of the target to be identified, and determine the target to be identified as a drone based on the trajectory information.
[0061] The drone identification method provided in this application can be implemented by an electronic device, a component within an electronic device, an integrated circuit, or a chip. The electronic device can be a mobile electronic device or a non-mobile electronic device. For example, a mobile electronic device can be a mobile phone, tablet computer, laptop computer, PDA, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., while a non-mobile electronic device can be a server, network attached storage (NAS), or personal computer (PC), etc., and this application does not impose specific limitations.
[0062] The following example, using a computer executing the drone identification method provided in this application, illustrates the technical solution of this application in detail.
[0063] In step 110, multiple images of the detection range are acquired with the target base station as the center.
[0064] It should be noted that the target base station is a communication base station with sensing capabilities, such as a 5G-A (5G-Advanced) base station, which integrates sensing sensors. These sensors can be radar, lidar, or other radio frequency sensing systems capable of sensing the position of objects in space. The detection range specifically refers to the spatial area or coverage area where the sensing sensors can effectively operate.
[0065] Multiple images can be acquired by installing one or more cameras on the target base station to capture images of the detection range from all angles without blind spots, and then storing the captured images for later use. It should be noted that these images are static background pictures.
[0066] In step 120, the location information of the target to be identified perceived by the sensing sensor is obtained, and based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images.
[0067] Specifically, when a sensing sensor detects an object within its detection range, it will treat that object as a target to be identified and determine the three-dimensional spatial location information of the target, such as longitude, latitude, and altitude coordinates.
[0068] After determining the location information of the target to be identified, the image showing the location of the target is selected from multiple pre-stored images based on the location information of the target to be identified, and this image is used as the target image.
[0069] Understandably, each image acquired within the detection range displays a region of that range. After determining the location information of the target to be identified, the target image at its location can be located based on which region the target is situated in.
[0070] In step 130, based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target.
[0071] The identifier for the target to be identified can be a pre-defined visual symbol used to mark the target's location on an image, such as a special icon, an arrow, or a cartoon character, to facilitate subsequent image analysis model recognition. Overlaying the identifier onto the target image actually involves superimposing the identifier at the corresponding pixel location in the target image, forming an image with a clearly marked target, i.e., a target overlay image.
[0072] The superimposed image of the target is analyzed, for example, using a pre-trained image classification model. If the model determines that the label is located in the air, rather than attached to a building, tree, or the ground, then the target to be identified can be determined as a flying target. The image classification model can be trained based on acquiring a large number of sample images containing flying target labels.
[0073] It should be noted that step 130 is a preliminary assessment of the target to be identified. This preliminary assessment can effectively filter out a large number of false alarms caused by stationary ground objects and non-flying targets.
[0074] Understandably, filtering the perceived data can eliminate a large amount of false alarm information, reduce the computational load of the subsequent accurate judgment process, and improve judgment efficiency.
[0075] In step 140, the trajectory information of the target to be identified is obtained, and based on the trajectory information, the target to be identified is determined to be a drone.
[0076] For an object identified as a flight target, its position information can be continuously tracked, and the position information sequence from multiple consecutive time points can be combined to form its trajectory information. The process of acquiring trajectory information can involve continuously recording the target's position within a preset time window (e.g., 100 sensing cycles).
[0077] The trajectory information is analyzed. For example, its dynamic characteristics, such as flight speed, acceleration, turning radius, and hovering ability, can be analyzed and compared with known drone flight characteristics. If the trajectory characteristics of the flight target match the typical motion pattern of a drone, the target is ultimately identified as a drone, achieving accurate drone identification; otherwise, if its trajectory characteristics do not match the drone's operating mode, it is excluded.
[0078] Understandably, the drone identification process based on trajectory judgment can transform the classification problem into an anomaly detection problem, thereby solving the problem of the difficulty in obtaining flight data of balloons and birds.
[0079] The UAV identification method provided in this application acquires multiple images centered on a target base station, extracts the target image based on the target location information provided by the sensing sensors, and then superimposes the identifiers to form a target overlay image to confirm the target as a flying target. This method can eliminate a large number of false alarms, reduce the computational load of the subsequent accurate judgment process, improve judgment efficiency, and realize the initial screening process. After confirming that it is a flying target, its trajectory information is further acquired and analyzed to achieve the accurate identification process of the UAV, thus improving the accuracy of UAV identification.
[0080] In one embodiment, obtaining a target image of the location of the target to be identified from the plurality of images based on the location information includes:
[0081] Based on the multiple images and their index information, a panoramic database of the detection range is constructed.
[0082] Based on the location information and the index information of the multiple images, the target image of the location of the target to be identified is obtained from the panoramic database.
[0083] It should be noted that the index information is metadata attached to each image to facilitate rapid retrieval. This metadata could include the azimuth and elevation angles when the image was taken within the detection range, the corresponding spatial grid number, and the associated base station ID. The panoramic database is a structured collection that associates and stores all background images with their index information, forming a library that can be quickly queried.
[0084] After constructing a panoramic database, the target image at the location of the target can be retrieved from the database based on the target's location information and the index information of multiple images. Once the sensing sensor determines the target's location information, the corresponding target image can be quickly located and extracted based on this location information and the index information in the database. For example, the target's location coordinates can be converted into a specific index key, which can then be used for efficient searching within the panoramic database.
[0085] The UAV identification method provided in this application improves the efficiency of acquiring target images by pre-constructing a panoramic database, enabling structured management and rapid retrieval of background images.
[0086] In one embodiment, the method for constructing the panoramic database includes:
[0087] Centered on the target base station, acquire multiple forward views and multiple top views of the sensing range;
[0088] Based on each of the forward graphs and the index information of each of the forward graphs, multiple forward graph dictionaries are constructed;
[0089] Based on each of the top views and the index information of each of the top views, multiple top view dictionaries are constructed;
[0090] The panoramic database is constructed based on the multiple forward graph dictionaries and the multiple top-view dictionaries.
[0091] Specifically, taking the target base station as the center, multiple forward views and multiple top-view views are acquired within the sensing range. A forward view can be understood as an image taken horizontally or at a certain elevation angle from the base station's location. A top-view view is an image taken looking down from above the target area. Simultaneously acquiring these two different perspectives allows for a more comprehensive description of the three-dimensional environment within the detection range.
[0092] After obtaining the forward image, the index information of each forward image (e.g., a combination string of horizontal angle-vertical angle-base station number) can be used as the key of a dictionary, and the corresponding image data can be used as the value of the dictionary. This creates one or more forward image dictionaries that can be quickly queried using angle information.
[0093] The entire detection range can be divided into local units according to a preset segmentation threshold. Each unit is a three-dimensional grid or sensing area. The index information of each top view (such as a combination string of grid ID and base station number) can be used as the key of the dictionary, and the corresponding image data can be used as the value of the dictionary, thereby constructing one or more top view dictionaries that can be quickly queried through the spatial grid location.
[0094] A panoramic database is constructed based on multiple forward view dictionaries and multiple top-view dictionaries. This constructed panoramic database logically integrates all the aforementioned dictionaries, enabling fast image queries.
[0095] The UAV identification method provided in this application organizes image data from different perspectives by constructing separate forward-view and top-view dictionaries, making the structure of the panoramic database clearer and the data retrieval path more explicit. This multi-view, multi-index approach provides a foundation for more accurate target localization and analysis.
[0096] In one embodiment, taking the target base station as the center, acquiring multiple forward views and multiple top views of the sensing range includes:
[0097] Centered on the target base station, images are acquired at preset angular resolutions within the horizontal and vertical field-of-view angle ranges of the sensing range to obtain multiple forward images within the sensing range.
[0098] Based on a preset segmentation threshold, the three-dimensional grid of the sensing range is divided to obtain multiple sensing regions. A top view of each sensing region is obtained with the target base station as the center, resulting in multiple top views of the sensing range.
[0099] For the process of acquiring forward images, for example, if the horizontal field of view of a base station is [30°, 90°] and the vertical field of view is [-30°, 30°], and the set angular resolution is 5°, then the camera will rotate and take pictures in the horizontal and vertical directions in 5° increments, and a total of 12×12=144 forward images can be taken, thereby achieving full coverage of the front field of view.
[0100] For the top-down view process, the entire detection range can be divided into local units according to a preset segmentation threshold. Each unit is a 3D grid or sensing area. Then, a top-down view is obtained for each sensing area.
[0101] In one embodiment, obtaining the target image of the location of the target to be identified from the panoramic database based on the location information and the index information of the plurality of images includes:
[0102] Based on the location information, the coordinate information of the target to be identified in the three-dimensional grid is determined, and based on the coordinate information of the target to be identified in the three-dimensional grid and the index information of the multiple images, the forward target image of the location of the target to be identified is determined from the panoramic database.
[0103] Based on the location information, the horizontal and vertical angles of the target to be identified within the detection range are determined, and based on the horizontal and vertical angles and index information of multiple images, a top-down target image of the target's location is determined from the panoramic database.
[0104] The location information of the target to be identified can be converted into the ID of the 3D grid in which it is located. Then, the grid ID can be used as an index to retrieve the corresponding forward graph from the forward graph dictionary.
[0105] Specifically, based on location information, the horizontal and vertical angles of the target to be identified relative to the target base station can be calculated. For example, trigonometric function calculations can be performed using the WGS84 coordinates of the target base station and the location information of the target. Then, based on the calculated horizontal and vertical angles, and combined with the image index information (horizontal and vertical angles), the forward target image at the location of the target to be identified is determined from the forward image dictionary of the panoramic database.
[0106] Simultaneously, based on the location information, the coordinate information of the 3D grid where the target to be identified is located can be determined, i.e., the grid ID. Then, based on the grid ID and the image index information, the top-view image of the target's location is determined from the top-view dictionary of the panoramic database.
[0107] In one embodiment, determining that the target to be identified is a flying target based on the target overlay image includes:
[0108] The positive target image is input into the classification model to obtain the first classification result output by the classification model;
[0109] The top-view target image is input into the classification model to obtain the second classification result output by the classification model;
[0110] Based on the first classification result and the second classification result, the target to be identified is determined to be a flying target;
[0111] The classification model is trained based on sample images and the corresponding flight target labels of the sample images.
[0112] It should be noted that the classification model can be a pre-trained deep learning network, such as the ResNet-50 network. A schematic diagram of the judgment process implemented based on the ResNet-50 network can be shown below. Figure 2 The judgment process provided in this application is illustrated in the diagram. A forward-facing target image is input into the classification model to obtain a first classification result; a top-view target image is input into the Siamese network of the classification model to obtain a second classification result; based on the first and second classification results, the final output judgment result is determined. This model is trained to perform binary classification judgment, that is, to determine whether the marker in the image is in a flying state or a non-flying state.
[0113] The training data for the classification model consists of a large number of labeled sample images, specifically including frontal and top-down views, along with corresponding flight target labels. For example, a positive label indicates a flight target, while a negative label indicates it is not a flight target.
[0114] After obtaining the first and second classification results, a fusion strategy can be set up so that the target is finally determined to be a flying target only if both the first and second classification results are positive. Optionally, a continuous frame judgment mechanism can also be set up, for example, confirming it as a flying target only if the judgment results of three consecutive frames are positive, to increase the stability of the judgment.
[0115] The UAV identification method provided in this application classifies and judges images from two different dimensions, namely frontal and top-down views, and then fuses the two results to achieve dual verification of flying targets. This multi-view verification mechanism can greatly improve the accuracy of judgment, effectively avoid misjudgments that may be caused by a single view, and thus more reliably filter out non-flying targets.
[0116] In one embodiment, determining that the target to be identified is a drone based on the trajectory information includes:
[0117] The trajectory information is input into the anomaly detection network model to obtain the detection result output by the anomaly detection network model;
[0118] Based on the detection results, the target to be identified is determined to be a drone;
[0119] The anomaly detection network model is trained based on historical trajectory data of UAVs.
[0120] It should be noted that anomaly detection network models are machine learning models specifically designed to identify whether data patterns are abnormal. For example, they can be models based on variational autoencoders (VAEs). These models are trained using a large amount of historical drone trajectory data to learn the normal flight patterns and dynamic characteristics of drones.
[0121] When the trajectory information of a target to be identified is input into an anomaly detection network model, the model outputs a detection result, which can be an anomaly score or a reconstruction error. If the score is low, it means the input trajectory closely matches the normal trajectory pattern of a drone learned by the model, and is not considered an anomaly; therefore, the target can be identified as a drone. If the score is high, it means its movement pattern is significantly different from that of a drone, such as birds or balloons flying erratically, which is considered an anomaly, and it is judged to be a non-drone target.
[0122] Optionally, the detection process based on the anomaly detection network model can be as follows: Figure 3 The testing process diagram provided in this application is shown.
[0123] First, a panoramic database of the detection range is constructed based on multiple images acquired within the detection range and their index information. Target images of the locations of the target to be identified are then retrieved from this database. The target's identifier is overlaid onto the target images to obtain a target overlay image. Based on this overlay image, it is determined whether the target is a flying target. If the target is determined to be a flying target, an anomaly detection network model is used to determine whether the target is a drone.
[0124] The drone identification method provided in this application transforms a multi-class classification problem of distinguishing drones from other interfering targets into an anomaly detection problem. It only requires collecting and training drone trajectory data, eliminating the need for acquiring large amounts of difficult-to-collect trajectory data from other flying objects such as birds and balloons. This reduces the difficulty of model training while effectively achieving accurate drone identification.
[0125] In one embodiment, based on the location information, the identifier of the target to be identified is superimposed onto the target image to obtain a target superimposed image, including:
[0126] Based on the camera parameter information of the target image, the position information is converted into coordinate information in the pixel coordinate system;
[0127] Based on the coordinate information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image.
[0128] The camera parameter information for the target image includes the camera's internal parameters, such as focal length and principal point, as well as its external parameters, all of which are predetermined. These parameters allow for the establishment of a mapping from real-world 3D coordinates to the image's 2D pixel coordinates. Using this mapping, the 3D position information of the target to be identified can be accurately converted into its corresponding 2D pixel coordinates on the target image.
[0129] Based on the determined coordinate information, the identifier of the target to be identified is superimposed onto the target image. At the calculated pixel coordinate positions, the preset identifier image is drawn or fused onto the target image, ultimately obtaining the target superimposed image.
[0130] The drone identification device provided in this application is described below. The drone identification device described below can be referred to in correspondence with the drone identification method described above.
[0131] like Figure 4 As shown, the device includes:
[0132] The image acquisition module 410 is used to acquire multiple images of the detection range centered on the target base station; the detection range is the sensing range of the sensing sensor in the target base station.
[0133] The filtering module 420 is used to obtain the location information of the target to be identified perceived by the sensing sensor, and based on the location information, obtain the target image of the location of the target to be identified from the plurality of images;
[0134] The flight target determination module 430 is used to overlay the identifier of the target to be identified onto the target image based on the location information to obtain a target overlay image, and to determine the target to be identified as a flight target based on the target overlay image;
[0135] The drone identification module 440 is used to acquire the trajectory information of the target to be identified, and to determine the target to be identified as a drone based on the trajectory information.
[0136] The UAV identification device provided in this application acquires multiple images centered on a target base station, extracts the target image based on the target location information provided by the sensing sensors, and then superimposes the identifiers to form a target overlay image to confirm the target as a flying target. This eliminates a large amount of false alarm information, reduces the computational load of the subsequent accurate judgment process, improves judgment efficiency, and realizes the initial screening process. After confirming that it is a flying target, its trajectory information is further acquired and analyzed to achieve the accurate identification process of the UAV, thus improving the accuracy of UAV identification.
[0137] In one embodiment, the image acquisition module 410 is specifically used for:
[0138] Based on the location information, obtaining the target image of the location of the target to be identified from the plurality of images includes:
[0139] Based on the multiple images and their index information, a panoramic database of the detection range is constructed.
[0140] Based on the location information and the index information of the multiple images, the target image of the location of the target to be identified is obtained from the panoramic database.
[0141] In one embodiment, the image acquisition module 410 is further configured to:
[0142] Methods for constructing a panoramic database include:
[0143] Centered on the target base station, acquire multiple forward views and multiple top views of the sensing range;
[0144] Based on each of the forward graphs and the index information of each of the forward graphs, multiple forward graph dictionaries are constructed;
[0145] Based on each of the top views and the index information of each of the top views, multiple top view dictionaries are constructed;
[0146] The panoramic database is constructed based on the multiple forward graph dictionaries and the multiple top-view dictionaries.
[0147] In one embodiment, the image acquisition module 410 is further configured to:
[0148] Centered on the target base station, multiple forward views and multiple top views of the sensing range are obtained, including:
[0149] Centered on the target base station, images are acquired at preset angular resolutions within the horizontal and vertical field-of-view angle ranges of the sensing range to obtain multiple forward images within the sensing range.
[0150] Based on a preset segmentation threshold, the three-dimensional grid of the sensing range is divided to obtain multiple sensing regions. A top view of each sensing region is obtained with the target base station as the center, resulting in multiple top views of the sensing range.
[0151] In one embodiment, the filtering module 420 is specifically used for:
[0152] Based on the location information and the index information of the multiple images, the target image of the location of the target to be identified is obtained from the panoramic database, including:
[0153] Based on the location information, the coordinate information of the target to be identified in the three-dimensional grid is determined, and based on the coordinate information of the target to be identified in the three-dimensional grid and the index information of the multiple images, the forward target image of the location of the target to be identified is determined from the panoramic database.
[0154] Based on the location information, the horizontal and vertical angles of the target to be identified within the detection range are determined, and based on the horizontal and vertical angles and index information of multiple images, a top-down target image of the target's location is determined from the panoramic database.
[0155] In one embodiment, the filtering module 420 is further configured to:
[0156] Based on the overlay image of the target, determining that the target to be identified is a flying target includes:
[0157] The positive target image is input into the classification model to obtain the first classification result output by the classification model;
[0158] The top-view target image is input into the classification model to obtain the second classification result output by the classification model;
[0159] Based on the first classification result and the second classification result, the target to be identified is determined to be a flying target;
[0160] The classification model is trained based on sample images and the corresponding flight target labels of the sample images.
[0161] In one embodiment, the drone identification module 440 is specifically used for:
[0162] Based on the trajectory information, determining that the target to be identified is a drone includes:
[0163] The trajectory information is input into the anomaly detection network model to obtain the detection result output by the anomaly detection network model;
[0164] Based on the detection results, the target to be identified is determined to be a drone;
[0165] The anomaly detection network model is trained based on historical trajectory data of UAVs.
[0166] In one embodiment, the filtering module 420 is further configured to:
[0167] Based on the location information, the identifier of the target to be identified is superimposed onto the target image to obtain a target superimposed image, including:
[0168] Based on the camera parameter information of the target image, the position information is converted into coordinate information in the pixel coordinate system;
[0169] Based on the coordinate information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image.
[0170] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, communications interface 520, and memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a UAV identification method, which includes: acquiring multiple images of a detection range centered on a target base station; the detection range is the sensing range of the sensing sensors in the target base station.
[0171] The location information of the target to be identified, sensed by the sensing sensor, is obtained, and based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images.
[0172] Based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target;
[0173] The trajectory information of the target to be identified is obtained, and based on the trajectory information, the target to be identified is determined to be a drone.
[0174] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0175] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the UAV identification method provided by the above methods. The method includes: acquiring multiple images of a detection range centered on a target base station; the detection range is the sensing range of the sensing sensor in the target base station.
[0176] The location information of the target to be identified, sensed by the sensing sensor, is obtained, and based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images.
[0177] Based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target;
[0178] The trajectory information of the target to be identified is obtained, and based on the trajectory information, the target to be identified is determined to be a drone.
[0179] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the UAV identification method provided by the above methods, the method comprising: acquiring multiple images of a detection range centered on a target base station; the detection range being the sensing range of a sensing sensor in the target base station;
[0180] The location information of the target to be identified, sensed by the sensing sensor, is obtained, and based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images.
[0181] Based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target;
[0182] The trajectory information of the target to be identified is obtained, and based on the trajectory information, the target to be identified is determined to be a drone.
[0183] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0184] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0185] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for identifying unmanned aerial vehicles (UAVs), characterized in that, include: Multiple images of the detection range are acquired, centered on the target base station; The detection range is the sensing range of the sensing sensors in the target base station; The location information of the target to be identified, sensed by the sensing sensor, is obtained, and based on the location information, a target image of the location of the target to be identified is obtained from the plurality of images. Based on the location information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image, and based on the target superimposed image, the target to be identified is determined to be a flying target; The trajectory information of the target to be identified is obtained, and based on the trajectory information, the target to be identified is determined to be a drone; The step of obtaining the target image of the location of the target to be identified from the plurality of images based on the location information includes: Based on the multiple images and their index information, a panoramic database of the detection range is constructed. Based on the location information and the index information of the multiple images, the target image of the location of the target to be identified is obtained from the panoramic database; The method for constructing the panoramic database includes: Centered on the target base station, acquire multiple forward views and multiple top views of the sensing range; Based on each of the forward graphs and the index information of each of the forward graphs, multiple forward graph dictionaries are constructed; Based on each of the top views and the index information of each of the top views, multiple top view dictionaries are constructed; The panoramic database is constructed based on the multiple forward graph dictionaries and the multiple top-view dictionaries.
2. The UAV identification method according to claim 1, characterized in that, The step of acquiring multiple forward views and multiple top views of the sensing range, centered on the target base station, includes: Centered on the target base station, images are acquired at preset angular resolutions within the horizontal and vertical field-of-view angle ranges of the sensing range to obtain multiple forward images within the sensing range. Based on a preset segmentation threshold, the three-dimensional grid of the sensing range is divided to obtain multiple sensing regions. A top view of each sensing region is obtained with the target base station as the center, resulting in multiple top views of the sensing range.
3. The UAV identification method according to claim 2, characterized in that, The step of obtaining the target image of the location of the target to be identified from the panoramic database based on the location information and the index information of the multiple images includes: Based on the location information, the coordinate information of the target to be identified in the three-dimensional grid is determined, and based on the coordinate information of the target to be identified in the three-dimensional grid and the index information of the multiple images, the forward target image of the location of the target to be identified is determined from the panoramic database. Based on the location information, the horizontal and vertical angles of the target to be identified within the detection range are determined, and based on the horizontal and vertical angles and index information of multiple images, a top-down target image of the target's location is determined from the panoramic database.
4. The UAV identification method according to claim 3, characterized in that, The step of determining the target to be identified as a flying target based on the target overlay image includes: The positive target image is input into the classification model to obtain the first classification result output by the classification model; The top-view target image is input into the classification model to obtain the second classification result output by the classification model; Based on the first classification result and the second classification result, the target to be identified is determined to be a flying target; The classification model is trained based on sample images and the corresponding flight target labels of the sample images.
5. The UAV identification method according to claim 1, characterized in that, The step of determining the target to be identified as a drone based on the trajectory information includes: The trajectory information is input into the anomaly detection network model to obtain the detection result output by the anomaly detection network model; Based on the detection results, the target to be identified is determined to be a drone; The anomaly detection network model is trained based on historical trajectory data of UAVs.
6. The UAV identification method according to claim 1, characterized in that, The step of overlaying the identifier of the target to be identified onto the target image based on the location information to obtain a target overlay image includes: Based on the camera parameter information of the target image, the position information is converted into coordinate information in the pixel coordinate system; Based on the coordinate information, the identifier of the target to be identified is superimposed on the target image to obtain a target superimposed image.
7. A drone identification device, characterized in that, include: The image acquisition module is used to acquire multiple images within the detection range, centered on the target base station; The detection range is the sensing range of the sensing sensors in the target base station; The filtering module is used to obtain the location information of the target to be identified sensed by the sensing sensor, and based on the location information, to obtain the target image of the location of the target to be identified from the plurality of images; The flight target determination module is used to overlay the identifier of the target to be identified onto the target image based on the location information to obtain a target overlay image, and to determine the target to be identified as a flight target based on the target overlay image; The drone identification module is used to acquire the trajectory information of the target to be identified, and to determine the target to be identified as a drone based on the trajectory information; The step of obtaining the target image of the location of the target to be identified from the plurality of images based on the location information includes: Based on the multiple images and their index information, a panoramic database of the detection range is constructed. Based on the location information and the index information of the multiple images, the target image of the location of the target to be identified is obtained from the panoramic database; The method for constructing the panoramic database includes: Centered on the target base station, acquire multiple forward views and multiple top views of the sensing range; Based on each of the forward graphs and the index information of each of the forward graphs, multiple forward graph dictionaries are constructed; Based on each of the top views and the index information of each of the top views, multiple top view dictionaries are constructed; The panoramic database is constructed based on the multiple forward graph dictionaries and the multiple top-view dictionaries.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the drone identification method as described in any one of claims 1 to 6.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the drone identification method as described in any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the drone identification method as described in any one of claims 1 to 6.