Intrusion drone identification method and device applied to airport area and electronic equipment

Abstract

This disclosure presents embodiments of a method, apparatus, and electronic device for identifying intruding drones in airport areas. One specific implementation of the method includes: acquiring a set of point information and a set of infrared images within the same time window; aggregating the point information to obtain a first set of detection point information groups; for each first detection point information group in the first detection point information group set, performing directional multi-target detection on the infrared images in the infrared image set corresponding to the first detection point information group, to obtain a second detection point information group; aggregating the detection points based on the first and second detection point information groups to obtain a third detection point information set; and identifying the drone based on the third detection point information set and a pre-trained drone identification model. This implementation enables effective and rapid identification of multiple drones in multi-drone intrusion scenarios.

Description

Technical Field

[0001] Embodiments of this disclosure relate to the fields of computer technology, radar technology, and imaging technology, and specifically to methods, apparatus, and electronic devices for identifying intruding drones in airport areas. Background Technology

[0002] Unmanned aerial vehicles (UAVs), also known as drones, are unmanned aircraft controlled by radio remote control equipment and their own program control devices. With the rapid development of UAV-related technologies, the number of UAVs in ownership continues to rise. At the same time, unauthorized UAV flights in controlled airspace are also common, especially in airport areas. These unauthorized and unapproved UAV flights seriously affect the operational safety of civil aviation aircraft within airports.

[0003] Currently, the main drone identification technologies are as follows:

[0004] (1) Identification of UAVs is carried out by pulse radar, but pulse radar is susceptible to clutter interference and has a weak ability to detect low-speed small targets.

[0005] (2) Identification of drones is performed by radio spectrum detection, but this method is ineffective for drones controlled by private control protocols or those that are radio silent.

[0006] (3) UAV identification is performed by photoelectric recognition through visible light images, but visible light images are easily affected by weather factors, resulting in poor recognition accuracy and robustness.

[0007] (4) UAV identification is performed by sound wave monitoring, but there are many noise sources in the airport area, and environmental noise will seriously interfere with the identification accuracy of UAV.

[0008] It is evident that the drone identification methods listed above all have certain limitations, making it difficult to guarantee identification accuracy and robustness. Summary of the Invention

[0009] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0010] Some embodiments of this disclosure propose methods, apparatuses, and electronic devices for identifying intruding drones in airport areas to address the technical problems mentioned in the background section above.

[0011] In a first aspect, some embodiments of this disclosure provide a method for identifying intruding drones in airport areas. The method includes: acquiring a set of point information and a set of infrared images within the same time window, wherein the set of point information is acquired by an active phased array radar facing low-altitude airspace within the airport area, and the set of infrared images is acquired by an infrared camera facing low-altitude airspace; performing point aggregation based on the set of point information to obtain a first detection point information group set, wherein the first detection point information group represents multiple detection points located within the same acquisition area; for each first detection point information group in the first detection point information group set, based on the first detection point information group set... The measurement point information group is obtained by performing directional multi-target detection on the infrared images in the aforementioned infrared image set that correspond to the aforementioned first detection point information group, resulting in a second detection point information group. The detection points are then aggregated based on the aforementioned first and second detection point information groups to obtain a third detection point information set. The third detection point information includes: detection point trace features, detection point infrared features, and detection point location. Finally, drone identification is performed based on the aforementioned third detection point information set and a pre-trained drone identification model to obtain a drone information set. The drone information includes: drone location information, drone behavior features, drone identification identifier, and identification timestamp.

[0012] Secondly, some embodiments of this disclosure provide an intrusion drone identification device applied to an airport area. The device includes: an acquisition unit configured to acquire a set of point information and a set of infrared images within the same time window, wherein the set of point information is acquired by an active phased array radar facing low-altitude airspace within the airport area, and the set of infrared images is acquired by an infrared camera facing low-altitude airspace; a point aggregation unit configured to aggregate points based on the set of point information to obtain a first detection point information group set, wherein the first detection point information group represents multiple detection points located within the same acquisition area; and a directional multi-target detection unit configured to, for each first detection point information group in the first detection point information group set, according to the above... The system comprises: a first detection point information group; and a second detection point information group. The first detection point information group is used to perform directional multi-target detection on the infrared images corresponding to the first detection point information group in the aforementioned infrared image set. The second detection point information group is configured to aggregate detection points based on the first and second detection point information groups to obtain a third detection point information set. The third detection point information includes: detection point trace features, detection point infrared features, and detection point positions. The drone identification unit is configured to perform drone identification based on the third detection point information set and a pre-trained drone identification model to obtain a drone information set. The drone information includes: drone position information, drone behavior features, drone identification identifier, and identification timestamp.

[0013] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0014] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0015] The above embodiments of this disclosure have the following beneficial effects: The intrusion drone identification method applied to airport areas through some embodiments of this disclosure achieves accurate drone identification. Specifically, the reason why identification accuracy is difficult to guarantee is that each single drone identification method has certain limitations, leading to difficulties in guaranteeing identification accuracy and robustness. Based on this, the intrusion drone identification method applied to airport areas through some embodiments of this disclosure first acquires a set of point information and a set of infrared images within the same time window. The point information set is acquired by an active phased array radar facing the low-altitude airspace within the airport area, and the infrared image set is acquired by an infrared camera facing the low-altitude airspace. In practice, malicious drone intrusion behavior usually adopts a multi-drone intrusion mode, where different drones are pre-set with corresponding intrusion targets. Therefore, conventional pulse radar is difficult to handle multi-target identification scenarios, while active phased array radar has a faster scanning speed than conventional pulse radar. Therefore, the active phased array radar used in this disclosure can effectively meet the above identification scenario. Considering that UAVs are low-altitude, small, and slow-moving targets, and that they can be considered stable heat sources during flight, infrared images are further combined for identification to avoid misidentification based on radar signals. Secondly, point clustering is performed on the aforementioned point information set to obtain a first detection point information group set, where each first detection point information group represents multiple detection points located within the same acquisition area. In practice, the original radar echo may contain a large amount of noise, as well as multiple echoes targeting the same detection point. Point clustering can eliminate noise, reducing noise interference and subsequent data processing volume. It can also aggregate point data for the same detection point, avoiding the existence of false detection points. Next, for each first detection point information group in the aforementioned first detection point information group set, directional multi-target detection is performed on the infrared images in the aforementioned infrared image set corresponding to the first detection point information group, resulting in a second detection point information group. In practice, considering that infrared images mainly reflect differences in heat sources, directly identifying UAVs in this single dimension involves many interference factors. Therefore, this disclosure combines the first detection point information to assist in target detection using infrared images, thereby improving identification speed and accuracy. Furthermore, based on the aforementioned first and second sets of detection point information, detection points are aggregated to obtain a third set of detection point information. This third set includes: detection point trace features, detection point infrared features, and detection point location. Detection point aggregation merges the detection points identified in the radar echo dimension and the infrared dimension. Finally, based on the third set of detection point information and a pre-trained UAV recognition model, UAV recognition is performed to obtain a UAV information set. This UAV information includes: UAV location information, UAV behavioral characteristics, UAV identification identifier, and identification timestamp.This method enables effective and rapid identification of multiple drones in multi-drone intrusion scenarios. Attached Figure Description

[0016] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0017] Figure 1 This is a flowchart of some embodiments of the intrusion drone identification method applied to airport areas according to the present disclosure;

[0018] Figure 2 This is a schematic diagram illustrating the process of acquiring dot information and infrared images;

[0019] Figure 3 This is a schematic diagram illustrating the generation process of the weight matrix;

[0020] Figure 4 This is a schematic diagram of the structure of some embodiments of an intrusion drone identification device applied to airport areas according to the present disclosure;

[0021] Figure 5 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0022] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0023] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0024] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0025] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0026] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0027] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] refer to Figure 1 The diagram illustrates a flow 100 of some embodiments of an intrusion drone identification method applied to an airport area according to the present disclosure. The intrusion drone identification method applied to an airport area includes the following steps:

[0029] Step 101: Obtain the set of dot information and the set of infrared images within the same time window.

[0030] In some embodiments, the execution subject (e.g., a computing device) of the intrusion drone identification method applied to airport areas can acquire a set of point information and a set of infrared images within the same time window via a wired or wireless connection.

[0031] The point-of-sight information set was acquired by an active phased array radar facing the low-altitude airspace within the airport area. Point-of-sight information represents the original measurement result detected by the radar for a single detection point within a single scan cycle. The aforementioned active phased array radar is a Ku-band dual-polarization active phased array radar. The aforementioned infrared image set was acquired by an infrared camera facing the aforementioned low-altitude airspace. The infrared camera is an over-the-horizon infrared camera. Specifically, the infrared camera uses multiple infrared sensors. The infrared sensors are uncooled infrared focal plane detectors using vanadium oxide as the thermosensitive material. The resolution of the infrared sensors is 1280 × 1024. The thermal sensitivity of the infrared sensors is ≤ 40 mK (millikelvin). The resolution of the infrared camera is 3840 × 1024. The infrared camera and the active phased array radar simultaneously acquire point-of-sight information and infrared images from the same viewpoint. The low-altitude airspace represents the airspace corresponding to the common flight altitudes of UAVs.

[0032] In practice, for multi-aircraft intrusion scenarios, multiple aircraft typically intrude into controlled areas at low altitudes. Therefore, the active phased array radar and infrared camera disclosed herein are mainly used to scan low-altitude airspace to collect point information and infrared images, thereby avoiding redundant data generated during the scanning of invalid areas.

[0033] As an example, see Figure 2 A schematic diagram illustrating the acquisition process of dot information and infrared images shows the data acquisition process of an active phased array radar and an infrared camera from a top-down perspective. The active phased array radar and the infrared camera simultaneously acquire dot information and infrared images from the same viewpoint. Figure 2 For example, an active phased array radar and an infrared camera continuously scan the low-altitude airspace in a clockwise rotation direction. Within time window W1, the active phased array radar and the infrared camera acquire a set of point information D1 and a set of infrared images I1. Within time window Wn, the active phased array radar and the infrared camera acquire a set of point information Dn and a set of infrared images In.

[0034] It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (Ultra Wide Band) connections, and other currently known or future wireless connection methods.

[0035] It should be noted that the aforementioned computing devices can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is software, it can be installed on the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here.

[0036] Step 102: Perform point aggregation based on the point information set to obtain the first detection point information group set.

[0037] In some embodiments, the aforementioned execution entity can aggregate the dots based on the dot information set to obtain a first detection point information group set.

[0038] The first detection point information group represents multiple detection points located within the same acquisition area. The first detection point information in the first detection point information set includes: the initial detection point trace features and the detection point location.

[0039] In practice, point information sets can be aggregated using Euclidean distance-based point aggregation to obtain the first set of detection point information. Specifically, within a single time window, an active phased array radar will perform partitioned vertical scanning of a local area in the low-altitude airspace corresponding to that time window, and control the orientation change through horizontal rotation. In this process, a large amount of point information is generated. This point information may correspond to noise or multiple echoes from the same detection point. Therefore, it is necessary to determine the detection points in each acquisition area through partitioned aggregation.

[0040] In some optional implementations of certain embodiments, the execution entity performs point aggregation based on the aforementioned point information set to obtain a first detection point information group set, including:

[0041] Step S1: Based on the spatial voxels to which the corresponding point information belongs, group the point information in the above point information set to obtain a point information group set.

[0042] Among them, the airspace voxel is a square pyramid-shaped spatial region with the location of the aforementioned active phased array radar as its vertex, and multiple points corresponding to the point information group are located within the same airspace voxel.

[0043] In practice, considering that active phased array radar mainly includes 360-degree horizontal rotation and a fixed range of elevation angle changes, this disclosure uses the installation position of the active phased array radar as the vertex and divides the detection range of the active phased array radar into multiple spatial voxels, each of which is a square pyramid-shaped spatial region. Based on this, the point information is grouped according to the positional constraints of the spatial voxels, resulting in a set of point information groups. Specifically, during the point information acquisition process, since the rotation and elevation angle changes of the active phased array radar are known, the point information can be bound to the corresponding spatial voxel during the point information acquisition stage. Alternatively, point grouping can be performed by determining the position, i.e., judging whether the corresponding points are located within the same spatial voxel.

[0044] Step S2: For each trace information group in the above trace information group set, perform the following processing steps:

[0045] Step S21: Extract the feature of each dot information in the above dot information group to obtain the dot feature.

[0046] The features of a target point consist of position features, motion features, and target features. Position features represent the position of the target point relative to the active phased array radar. Motion features represent the velocity of the target corresponding to the target point. Target features represent the characteristics of the target corresponding to the target point. Position features include slant range, azimuth, and elevation angle. Motion features include radial velocity. Target features include radar cross-section, signal-to-noise ratio (SNR), and scintillation characteristics. Slant range represents the straight-line distance between the target point and the active phased array radar. Azimuth represents the direction angle of the target point on the horizontal plane. Elevation represents the angle between the target point and the horizontal plane. Radial velocity represents the velocity component of the target point along the line-of-sight of the active phased array radar. Radial cross-section represents the target point's reflectivity to radar waves. SNR represents the ratio of the target point's echo intensity to the background noise intensity. Scintillation characteristics represent the amplitude fluctuations of the target point. In particular, for common rotary-wing drones, the rotors often have periodic rotation characteristics, and the corresponding radar echoes will show regular amplitude changes, which can be used as one of the identification dimensions for subsequent judgment of whether it is a drone.

[0047] In practice, after locating the target point corresponding to the target point information, firstly, the slant range, azimuth, and elevation angles can be determined based on the relative position of the target point to the active phased array radar. Secondly, the radial velocity corresponding to the target point can be determined through Doppler processing. Next, the radar cross-section can be determined by the radar echo power to assess the reflectivity of the target point against the radar echo. Furthermore, scintillation characteristics are generated by analyzing the periodic radar echo amplitude variations corresponding to the target point.

[0048] Step S22: Based on the motion features and target features included in the dot features, filter the dot information in the above dot information group to obtain the filtered dot information group.

[0049] In practice, intrusion drones are often in motion, exhibiting distinct motion and target characteristics compared to the traces corresponding to other static objects. Therefore, based on the motion and target characteristics included in the trace features, the trace information in the aforementioned trace information group can be filtered to obtain a filtered trace information group. This serves as a preliminary filter for traces corresponding to noise and non-moving targets. Specifically, the aforementioned execution entity can filter trace information with empty flashing features or whose motion features are not within a preset speed range to obtain a filtered trace information group.

[0050] Step S23: Perform multi-detection point recognition based on the point features corresponding to the filtered point information in the above filtered point information group to obtain the first detection point information group.

[0051] In practice, the aforementioned execution entity can use a detection point recognition model, taking the characteristics of the filtered point information as input, to determine whether the points corresponding to the filtered point information are suspected drone targets. Specifically, the detection point recognition model consists of three input nodes, three hidden layers, and one output node. The three input nodes correspond to the positional features, motion features, and target features of the points corresponding to the filtered point information, respectively. The one output layer corresponds to the classification category. Since the points have already undergone corresponding feature extraction, to avoid feature forgetting due to overly deep feature extraction and considering the data volume issue during real-time point information acquisition, the detection point recognition model uses a lightweight model to quickly determine the detection point category. When the category of the detection point corresponds to a suspected drone target, the characteristics of the filtered point information are used as the initial detection point characteristics of the corresponding first detection point information, and the positional features of the filtered point information are converted into position coordinates as the position of the corresponding first detection point information.

[0052] Step 103: For each first detection point information group in the first detection point information group set, perform directional multi-target detection on the infrared images in the infrared image set that correspond to the first detection point information group based on the first detection point information group to obtain the second detection point information group.

[0053] In some embodiments, the execution entity may perform directional multi-target detection on the infrared images in the infrared image set corresponding to the first detection point information group for each first detection point information group in the first detection point information group set, thereby obtaining a second detection point information group.

[0054] The second detection point information group represents multiple detection points located within the same acquisition area. The second detection point information in the second detection point information set includes: the infrared characteristics of the initial detection point and the location of the detection point.

[0055] In practice, since active phased array radar and infrared cameras simultaneously acquire point information and infrared images from the same viewpoint, under the premise of infrared camera calibration, three-dimensional coordinates can be converted into two-dimensional coordinates within the image coordinate system. Specifically, since the detection point positions included in the first detection point information are in three-dimensional coordinates, it is necessary to combine the rotation matrix and translation vector corresponding to the infrared camera to convert the detection point positions included in the first detection point information into two-dimensional coordinates. Then, the anchor frame region centered on the two-dimensional coordinates is taken as the first region of interest. At the same time, the aforementioned execution entity can perform directional multi-target detection on the infrared image corresponding to the first detection point information group using a model such as Tiny-YOLO to obtain the second region of interest. Next, feature one-dimensional mapping is performed on the local image features of each of the multiple first regions of interest and each of the multiple second regions of interest to obtain the initial detection point infrared features included in the second detection point information, and the region center of the first region of interest or the second region of interest is taken as the detection point position included in the second detection point information.

[0056] In some optional implementations of certain embodiments, the execution entity performs directional multi-target detection on the infrared images in the infrared image set corresponding to the first detection point information group based on the first detection point information group, to obtain a second detection point information group, including:

[0057] Step S1: Perform detection point position transformation on each first detection point information in the first detection point information group above to generate the transformed position and obtain the transformed position group.

[0058] The transformed position represents the image coordinates of the detection point corresponding to the first detection point information within the corresponding infrared image.

[0059] In practice, since active phased array radar and infrared cameras simultaneously acquire point information and infrared images from the same viewpoint, under the premise of infrared camera calibration, three-dimensional coordinates can be converted into two-dimensional coordinates within the image coordinate system. Specifically, since the detection point information includes the detection point position in three-dimensional coordinates, it is necessary to combine the rotation matrix and translation vector corresponding to the infrared camera to convert the detection point position included in the first detection point information into two-dimensional coordinates, which serve as the converted position.

[0060] Step S2: Divide the infrared image corresponding to the first detection point information group into image partitions to obtain partitioned infrared images.

[0061] The partitioned infrared image consists of K image blocks of the same size.

[0062] As an example, see Figure 3 The diagram shown illustrates the generation process of the weight matrix, where... Figure 3 The partitioned infrared image shown includes 24 (K) image blocks. Assuming the image size of the infrared image corresponding to the first detection point information group is M×N, the image block size of each image block is M / 6×N / 4.

[0063] Step S3: Based on the transformed location information group, determine the number of hits for each image block in the K image blocks included in the partitioned infrared image.

[0064] In practice, the transformed location information can be projected onto the corresponding image block, and the number of transformed location information contained in each image block can be counted as the hit count of the corresponding image block.

[0065] As an example, see further. Figure 3 Assuming the first detection point information group includes 10 first detection information points, the hit situation of the corresponding transformed position group containing 10 transformed positions in the partitioned infrared image is as follows: Figure 3 As shown, the image patch in the first row and second column has a hit count of 4. The image patch in the second row and fourth column has a hit count of 3. The image patch in the third row and second column has a hit count of 1. The image patch in the third row and fifth column has a hit count of 1. The image patch in the fourth row and first column has a hit count of 1. The remaining image patches have a hit count of 0.

[0066] Step S4: Generate a weight matrix based on the number of hits corresponding to the image patches.

[0067] In practice, the matrix dimension of the weight matrix is ​​consistent with the image size of the infrared image. The matrix dimension of the sub-weight matrix corresponding to each image patch is also consistent with the image size of the image patch. Specifically, the matrix values ​​of the sub-weight matrix corresponding to each image patch can be determined by normalizing the number of hits.

[0068] As an example, see further. Figure 3 After normalization based on the maximum and minimum hit counts, the weight matrix has a size of M×N, containing 24 sub-weight matrices of size M / 6×N / 4. The sub-weight matrix in the first row and second column is [1], meaning all M / 6×N / 4 values ​​within the matrix are 1. The sub-weight matrix in the second row and fourth column is [0.75], meaning all M / 6×N / 4 values ​​within the matrix are 0.75. The sub-weight matrix in the third row and second column is [0.25], meaning all M / 6×N / 4 values ​​within the matrix are 0.25. The sub-weight matrix in the third row and fifth column is [0.25], meaning all M / 6×N / 4 values ​​within the matrix are [0.25]. The sub-weight matrix in the fourth row and first column is [0.25], meaning all M / 6×N / 4 values ​​within the matrix are [0.25]. The remaining sub-weight matrices are all [0] matrices. The reason for not directly performing pixel-level weight mapping, but instead using image patches as the granularity, is that although the three-dimensional coordinates can be projected into the image coordinate system using the pre-calibrated rotation matrix and translation vector of the infrared camera, the echo signal cannot be completely noise-suppressed, and is also affected by the signal acquisition environment, resulting in certain accuracy errors. Directly constructing a pixel-level weight matrix in this situation might lead to errors in multi-target detection of the infrared image, especially target loss. Therefore, this disclosure uses image patch-level weight mapping to construct the weight matrix, thereby minimizing detection errors.

[0069] Step S5: Perform image block encoding on the K image blocks included in the above partitioned infrared image to obtain the image block feature matrix.

[0070] In practice, the aforementioned execution entity can use the TransformerEncoder module in the VIT (Vision Transformer) model to encode the K image blocks included in the partitioned infrared image, thereby obtaining the image block feature matrix.

[0071] Step S6: Based on the above weight matrix, perform feature enhancement on the above image patch feature matrix to obtain the enhanced image patch matrix.

[0072] The image patch feature matrix and the enhanced image patch matrix have the same matrix dimension.

[0073] In practice, firstly, the execution entity can project the weight matrix to obtain a projected weight matrix. Matrix projection ensures that the dimensions of the projected weight matrix and the image patch feature matrix are consistent. Then, the execution entity can multiply the corresponding matrix values ​​of the projected weight matrix and the image patch feature matrix to obtain the enhanced image patch matrix.

[0074] Step S7: Generate the second detection point information group based on the enhanced image block matrix and the infrared target detection model described above.

[0075] The infrared target detection model described above is a cascaded weak classifier. Specifically, the infrared target detection model consists of three cascaded weak convolutional neural network classifiers. Each weak convolutional neural network classifier includes a lightweight convolutional network, a binary classifier, and a region of interest localizer.

[0076] In practice, the enhanced image patch matrix is ​​input into a cascaded weak classifier, and the suspected drone location is located through target detection. Local image features within this region are extracted as initial detection point infrared features, and the region center is used as the location of the detection point included in the second detection point information. The reason for using a cascaded weak classifier is that the generation of the second detection point information is mainly used to extract the initial detection point infrared features included in the second detection point information. Conventional strong feature extractors have complex network structures, making it difficult to improve the task execution speed for large-scale image processing tasks in airport drone detection scenarios, while also requiring significant hardware resources. Therefore, this disclosure uses multiple cascaded weak convolutional neural network classifiers. The cascading method ensures the accuracy of extraction and classification, while the weak classifier structure is simple and suitable for the aforementioned airport drone detection scenario.

[0077] Step 104: Aggregate the detection points based on the first detection point information set and the second detection point information set to obtain the third detection point information set.

[0078] In some embodiments, the aforementioned execution entity may aggregate detection points based on the first set of detection point information groups and the second set of detection point information groups to obtain a third set of detection point information.

[0079] The information for the third detection point includes: the detection point trace characteristics, the detection point infrared characteristics, and the detection point location.

[0080] In practice, the aforementioned implementing entity can merge the information of the first and second detection points corresponding to the same detection point location as the third detection point information, thus obtaining the third detection point information set.

[0081] In some optional implementations of certain embodiments, the execution entity aggregates detection points based on the first set of detection point information groups and the second set of detection point information groups to obtain a third set of detection point information, including:

[0082] Step S1: Based on the converted position corresponding to the second detection point information and the detection point positions included in the first detection point information, perform detection point matching on the first detection point information set and the second detection point information set to obtain an initial detection point information pair set.

[0083] The initial detection point information includes the matched first detection point information and second detection point information.

[0084] In practice, since the detection point positions included in the first detection point information and the detection point positions included in the second detection point position information are in different coordinate systems, while the transformed position corresponding to the second detection point information is in the same coordinate system as the detection point positions included in the first detection point information, the detection point matching of the above-mentioned first detection point information set and the above-mentioned second detection point information set can be performed by calculating the distance between coordinates to obtain an initial detection point information pair set.

[0085] Step S2: For each initial detection point information pair in the above initial detection point information pair set, perform the following matching steps:

[0086] Step S21: Perform feature mapping on the initial detection point information, including the initial detection point trace features and the initial detection point infrared features, to obtain the mapped detection point trace features and the mapped detection point infrared features.

[0087] In this case, the trace features of the detected points after mapping and the infrared features of the detected points after mapping are located in the same feature space.

[0088] In practice, a feature mapping network with two branches can be used to perform feature mapping on the initial detection point information, including the initial detection point trace features and the initial detection point infrared features, to obtain the mapped detection point trace features and the mapped detection point infrared features. The two-branch feature mapping network employs a weight-sharing dual-path convolutional neural network. Through feature mapping, the initial detection point trace features and the initial detection point infrared features are transformed into feature representations within the same feature space.

[0089] Step S22: Perform multi-dimensional feature similarity determination on the above-described mapped detection point trace features and the above-described mapped detection point infrared features to obtain a similarity value.

[0090] In practice, the similarity between the trace features of the mapped detection points and the infrared features of the mapped detection points can be calculated by using cosine similarity, and this similarity value can be used as the similarity value.

[0091] Step S23: In response to the similarity value being greater than a preset threshold, the mapped detection point trace features are determined as the detection point trace features included in the third detection point information, the mapped detection point infrared features are determined as the detection point infrared features included in the third detection point information, and the detection point positions included in the second detection point information in the initial detection point information pair are determined as the detection point positions included in the third detection point information.

[0092] Step 105: Perform drone identification based on the third detection point information set and the pre-trained drone identification model to obtain the drone information set.

[0093] In some embodiments, the aforementioned execution entity can perform drone identification based on the third detection point information set and a pre-trained drone identification model to obtain a drone information set.

[0094] The drone information includes: drone location information, drone behavior characteristics, drone identification, and identification timestamp.

[0095] In practice, since feature extraction has already been performed on the trace information and infrared image when determining the first and second detection point information, and mapping of the initial detection point trace features and initial detection point infrared features expressed in two different feature spaces has also been performed in the same feature space when determining the third detection point information, it can be understood that corresponding feature extraction has already been performed on the trace information and infrared image. Therefore, when combining the third detection point information, a UAV identification network containing only a position discriminator and a feature mapper can be set up to generate UAV information. The position discriminator can be based on an RNN (Recurrent Neural Network) as the backbone network to determine the position of the UAV target based on the third detection point information (determining whether the detection point positions included in the third detection point information correspond to a UAV), serving as the UAV position information. The feature mapper uses multiple fully connected layers in series to map the detection point trace features and detection point infrared features included in the third detection point information into 256-dimensional UAV behavioral features. UAV identification identifiers can be automatically assigned according to the recognition order. The recognition timestamp can be the current timestamp, thus constructing the UAV identification information.

[0096] In some optional implementations of certain embodiments, the aforementioned execution entity performs drone identification based on the aforementioned third detection point information set and a pre-trained drone identification model to obtain a drone information set, including:

[0097] Step S1: Associate the information set of the third detection point and the information set of historical detection points to obtain the associated detection point information set.

[0098] Among them, the historical detection point information set represents the third detection point information set generated in the previous time window.

[0099] In practice, drone targets often correspond to continuous flight behavior. To predict the position of a drone target in the next moment, it is necessary to combine historical detection point information. Specifically, the positions can be correlated based on the detection point locations included in the third detection point information and the detection point locations included in the historical detection point information to obtain the correlated detection point information.

[0100] Step S2: For each associated detection point in the above associated detection point information set, perform the following recognition steps:

[0101] Step S21: Convert the above-mentioned associated detection point information into a spatial feature map.

[0102] In practice, since the third detection point information includes the detection point positions in three-dimensional coordinates, the associated detection point information can be projected into three-dimensional space to generate a spatial feature map. In particular, the projected spatial feature map can be understood as a sparse three-dimensional point cloud representation.

[0103] Step S22: Using the location encoding module included in the above UAV recognition model, the spatial feature map is position encoded to obtain a location feature sequence.

[0104] The location encoding module uses the PointNet++ model. The projected spatial feature map can be regarded as a sparse 3D point cloud representation, so location features can be extracted through point cloud segmentation.

[0105] Step S23: Using the detection point feature extraction module included in the above UAV recognition model, perform fusion feature extraction on the detection point trace features and detection point infrared features included in the above associated detection point information to obtain the detection point fusion feature sequence.

[0106] Among them, the location features and the detection point fusion features are in one-to-one correspondence.

[0107] In practice, since the two sets of detection point trace features and detection point infrared features included in the associated detection point information (the detection point trace features and detection point infrared features included in the third detection point information, and the detection point trace features and detection point infrared features included in the historical detection point information) only differ in the time dimension, and there is no difference in the feature dimension and feature space, feature fusion can be directly performed by feature superposition to obtain the detection point fused feature sequence.

[0108] Step S24: Using the location discriminator included in the above drone identification model and the above location feature sequence, generate drone location information including drone information.

[0109] In practice, since the location feature sequence consists of multiple location values ​​expanded over time, an RNN (Recurrent Neural Network) can be used as the backbone network of the location discriminator. A binary classifier can then be used to determine whether the location of the detection point included in the third detection point information corresponds to the UAV, thereby obtaining the UAV location information included in the UAV information.

[0110] Step S25: Generate drone behavior features, including drone information, by using the behavior feature mapper included in the above drone identification model and the above detection point fusion feature sequence.

[0111] In practice, the behavior feature mapper includes multiple serially connected fully connected layers to shape the fused feature sequence of detection points into a one-dimensional (1×256) feature representation, thereby obtaining the UAV behavior features included in the UAV information.

[0112] Step S26: Determine whether there is drone identity information in the drone identity information database that matches the drone behavior characteristics included in the drone information.

[0113] Among them, the drone identity information database is a pre-built database that stores the identity information of drones that have flown into low-altitude airspace.

[0114] In practice, a database search can be used to determine whether drone identity information exists in the drone identity information database that matches the drone behavior characteristics included in the drone information.

[0115] Step S27: In response to the existence, determine the drone identity identifier corresponding to the matched drone identity information as the drone identity identifier included in the drone identity information.

[0116] Step S28: In response to non-existence, increment the drone identity information including the drone identity identifier.

[0117] Step S29: Determine the current time as the identification timestamp included in the drone's identity information.

[0118] In some optional implementations of some embodiments, the above method further includes:

[0119] Step S1: Based on the drone behavior characteristics included in the drone information, perform behavior clustering on the drone information in the above drone information set to obtain a drone information group set.

[0120] Among them, the drone information group corresponds to multiple drones with the same drone behavior.

[0121] In practice, the K-means clustering algorithm can be used to cluster the drone information in the above drone information set according to the drone behavior characteristics included in the drone information, thus obtaining a drone information group set.

[0122] Step S2: For each drone information group in the above drone information group set, perform feature depth extraction on the drone behavior features included in the drone information in the above drone information group to obtain an initial behavior feature matrix.

[0123] In practice, a symmetrical feature pyramid network can be used to perform feature depth extraction on the drone behavior features included in the drone information group, resulting in an initial behavior feature matrix. Since the drone behavior features have already been compressed to 256 dimensions, a symmetrical feature pyramid network is used for feature depth extraction to avoid feature dimensionality reduction during downsampling. This ensures that the obtained initial behavior feature matrix is ​​consistent with the scale of the drone behavior features.

[0124] Step S2: Perform matrix concatenation on the obtained initial set of behavioral feature matrices to obtain the target behavioral feature matrix.

[0125] Step S3: Based on the above target behavior feature matrix, identify the group behavior of the UAVs to obtain group behavior information.

[0126] The aforementioned group behavior information includes: group risk level. The group risk level represents the risk level of a group of drones exhibiting the same drone behavior.

[0127] In practice, a multi-classifier with multiple serially connected convolutional layers can be used to identify drone swarm behavior using the target behavior feature matrix as input, thereby obtaining the swarm risk level. For multi-drone intrusion scenarios, different intrusion tasks are often set for different groups of drones. Single-drone identification methods are insufficient to effectively identify swarm behavior and risk. Therefore, treating drones exhibiting similar behavior as a whole for risk identification can improve the overall identification of multi-drone intrusion behavior.

[0128] In some optional implementations of some embodiments, the above method further includes:

[0129] Step S1: Simultaneously display the above-mentioned set of drone information within the airport area twin.

[0130] The airport area twin is a 3D data structure used for visualization. Data is synchronized between the airport area twin and sensors (such as infrared cameras and active phased array radars) installed within the airport area, allowing the real-time visualization of the airport area's operational status.

[0131] Step S2: Generate drone intrusion warning information based on the pre-set warning trigger and the above-mentioned group behavior information.

[0132] In practice, different response plans are often required depending on the degree of danger of the intrusion. Therefore, multiple early warning triggers with different triggering conditions can be set in advance, and automatic condition triggering can be performed based on group behavior information to automatically generate corresponding drone intrusion early warning information.

[0133] Step S3: Simultaneously display the aforementioned drone intrusion warning information on the twin of the aforementioned airport area.

[0134] The above embodiments of this disclosure have the following beneficial effects: The intrusion drone identification method applied to airport areas through some embodiments of this disclosure achieves accurate drone identification. Specifically, the reason why identification accuracy is difficult to guarantee is that each single drone identification method has certain limitations, leading to difficulties in guaranteeing identification accuracy and robustness. Based on this, the intrusion drone identification method applied to airport areas through some embodiments of this disclosure first acquires a set of point information and a set of infrared images within the same time window. The point information set is acquired by an active phased array radar facing the low-altitude airspace within the airport area, and the infrared image set is acquired by an infrared camera facing the low-altitude airspace. In practice, malicious drone intrusion behavior usually adopts a multi-drone intrusion mode, where different drones are pre-set with corresponding intrusion targets. Therefore, conventional pulse radar is difficult to handle multi-target identification scenarios, while active phased array radar has a faster scanning speed than conventional pulse radar. Therefore, the active phased array radar used in this disclosure can effectively meet the above identification scenario. Considering that UAVs are low-altitude, small, and slow-moving targets, and that they can be considered stable heat sources during flight, infrared images are further combined for identification to avoid misidentification based on radar signals. Secondly, point clustering is performed on the aforementioned point information set to obtain a first detection point information group set, where each first detection point information group represents multiple detection points located within the same acquisition area. In practice, the original radar echo may contain a large amount of noise, as well as multiple echoes targeting the same detection point. Point clustering can eliminate noise, reducing noise interference and subsequent data processing volume. It can also aggregate point data for the same detection point, avoiding the existence of false detection points. Next, for each first detection point information group in the aforementioned first detection point information group set, directional multi-target detection is performed on the infrared images in the aforementioned infrared image set corresponding to the first detection point information group, resulting in a second detection point information group. In practice, considering that infrared images mainly reflect differences in heat sources, directly identifying UAVs in this single dimension involves many interference factors. Therefore, this disclosure combines the first detection point information to assist in target detection using infrared images, thereby improving identification speed and accuracy. Furthermore, based on the aforementioned first and second sets of detection point information, detection points are aggregated to obtain a third set of detection point information. This third set includes: detection point trace features, detection point infrared features, and detection point location. Detection point aggregation merges the detection points identified in the radar echo dimension and the infrared dimension. Finally, based on the third set of detection point information and a pre-trained UAV recognition model, UAV recognition is performed to obtain a UAV information set. This UAV information includes: UAV location information, UAV behavioral characteristics, UAV identification identifier, and identification timestamp.This method enables effective and rapid identification of multiple drones in multi-drone intrusion scenarios.

[0135] Further reference Figure 4 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of an intrusion drone identification device applied to airport areas. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, this intrusion drone identification device applied to airport areas can be specifically applied to various electronic devices.

[0136] like Figure 4 As shown, an intrusion drone identification device 400 applied to an airport area in some embodiments includes: an acquisition unit 401, a point aggregation unit 402, a directional multi-target detection unit 403, a detection point aggregation unit 404, and a drone identification unit 405. The acquisition unit 401 is configured to acquire a set of point information and a set of infrared images within the same time window. The point information set is acquired by an active phased array radar facing low-altitude airspace within the airport area, and the infrared image set is acquired by an infrared camera facing low-altitude airspace. The point aggregation unit 402 is configured to aggregate points based on the point information set to obtain a first detection point information group set, wherein the first detection point information group represents multiple detection points located within the same acquisition area. The directional multi-target detection unit 403 is configured to... Each first detection point information group in a set of detection point information groups is used to perform directional multi-target detection on the infrared images in the set of infrared images corresponding to the first detection point information group, to obtain a second detection point information group; the detection point aggregation unit 404 is configured to aggregate detection points according to the first detection point information group set and the second detection point information group set to obtain a third detection point information set, wherein the third detection point information includes: detection point trace features, detection point infrared features, and detection point positions; the UAV recognition unit 405 is configured to perform UAV recognition according to the third detection point information set and a pre-trained UAV recognition model to obtain a UAV information set, wherein the UAV information includes: UAV position information, UAV behavior features, UAV identification, and recognition timestamp.

[0137] It is understandable that the units described in the intrusion drone identification device 400 applied to airport areas are similar to those in the reference device. Figure 1 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the intrusion drone identification device 400 and its constituent units applied in airport areas, and will not be repeated here.

[0138] The following is for reference. Figure 5It shows a schematic diagram of the structure of an electronic device (e.g., a computing device) 500 suitable for implementing some embodiments of the present disclosure. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0139] like Figure 5 As shown, the electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory 502 or a program loaded from a storage device 508 into a random access memory 503. The random access memory 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, the read-only memory 502, and the random access memory 503 are interconnected via a bus 504. An input / output interface 505 is also connected to the bus 504.

[0140] Typically, the following devices can be connected to the input / output interface 505: input devices 506 including, for example, a touchscreen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; output devices 507 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 508 including, for example, magnetic tape, hard disk, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 5 Each box shown can represent a device or multiple devices as needed.

[0141] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a read-only memory 502. When the computer program is executed by the processing device 501, it performs the functions defined above in the methods of some embodiments of this disclosure.

[0142] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0143] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0144] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire a set of point trace information and a set of infrared images within the same time window, wherein the set of point trace information is acquired by an active phased array radar facing low-altitude airspace within the airport area, and the set of infrared images is acquired by an infrared camera facing low-altitude airspace; perform point trace aggregation based on the set of point trace information to obtain a first set of detection point information groups, wherein the first set of detection point information groups represents multiple detection points located within the same acquisition area; for each first set of detection point information groups in the first set of detection point information groups, according to the aforementioned... The first detection point information group is used to perform directional multi-target detection on the infrared images corresponding to the first detection point information group in the aforementioned infrared image set, resulting in the second detection point information group. Detection points are aggregated based on the first and second detection point information groups to obtain the third detection point information set, where the third detection point information includes: detection point trace features, detection point infrared features, and detection point positions. UAV identification is performed based on the third detection point information set and a pre-trained UAV recognition model to obtain the UAV information set, where the UAV information includes: UAV position information, UAV behavior features, UAV identification identifier, and identification timestamp.

[0145] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0146] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0147] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0148] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for identifying intruding drones in airport areas, characterized in that, include: Acquire a set of point information and a set of infrared images within the same time window, wherein the set of point information is acquired by an active phased array radar facing the low-altitude airspace within the airport area, and the set of infrared images is acquired by an infrared camera facing the low-altitude airspace. Based on the set of point information, point clustering is performed to obtain a first set of detection point information groups, wherein the first set of detection point information groups represents multiple detection points located in the same acquisition area; For each first detection point information group in the first detection point information group set, based on the first detection point information group, perform directional multi-target detection on the infrared images in the infrared image set that correspond to the first detection point information group to obtain the second detection point information group; The detection points are aggregated based on the first set of detection point information and the second set of detection point information to obtain a third set of detection point information, wherein the third set of detection point information includes: detection point trace features, detection point infrared features, and detection point positions. Based on the third detection point information set and the pre-trained drone recognition model, drone identification is performed to obtain a drone information set. This drone information set includes: drone location information, drone behavior characteristics, drone identification identifier, and identification timestamp. The step of performing directional multi-target detection on the infrared images in the infrared image set corresponding to the first detection point information group based on the first detection point information group to obtain the second detection point information group includes: The detection point position is transformed for each first detection point information in the first detection point information group to generate a transformed position, resulting in a transformed position group. The transformed position represents the image coordinates of the detection point corresponding to the first detection point information in the corresponding infrared image. The infrared image corresponding to the first detection point information group is divided into image partitions to obtain a partitioned infrared image, wherein the partitioned infrared image consists of K image blocks of the same size. Based on the transformed position group, determine the number of hits for each of the K image blocks in the partitioned infrared image; Generate a weight matrix based on the number of hits corresponding to the image patches; The K image blocks included in the partitioned infrared image are encoded to obtain an image block feature matrix; Based on the weight matrix, feature enhancement is performed on the image patch feature matrix to obtain the enhanced image patch matrix; Based on the enhanced image patch matrix and the infrared target detection model, a second detection point information group is generated, wherein the infrared target detection model is a cascaded weak classifier.

2. The method for identifying intruding drones in airport areas according to claim 1, characterized in that, The method further includes: Based on the drone behavior characteristics included in the drone information, the drone information in the drone information set is clustered by behavior to obtain a drone information group set, wherein the drone information group corresponds to multiple drones with the same drone behavior. For each drone information group in the drone information group set, feature depth extraction is performed on the drone behavior features included in the drone information in the drone information group to obtain an initial behavior feature matrix; The initial set of behavioral feature matrices is concatenated to obtain the target behavioral feature matrix. The drone swarm behavior is identified based on the target behavior feature matrix to obtain swarm behavior information, which includes: swarm risk level.

3. The method for identifying intruding drones in airport areas according to claim 2, characterized in that, The step of clustering the dots based on the dot information set to obtain a first detection dot information group set includes: Based on the spatial voxel to which the corresponding point information belongs, the point information in the point information set is grouped to obtain a point information group set. The spatial voxel is a square pyramid-shaped spatial region with the setting position of the active phased array radar as the vertex. Multiple points corresponding to the point information group are located in the same spatial voxel. For each trace information group in the set of trace information groups, perform the following processing steps: For each piece of information in the piece of information group, the piece of information features are extracted to obtain the piece of information features, wherein the piece of information features consist of position features, motion features and target features; Based on the motion features and target features included in the dot features, the dot information in the dot information group is filtered to obtain the filtered dot information group. Based on the trace features corresponding to the filtered trace information in the filtered trace information group, multi-detection point identification is performed to obtain the first detection point information group.

4. The method for identifying intruding drones in airport areas according to claim 3, characterized in that, The first detection point information in the first detection point information set includes: initial detection point trace features and detection point positions; the second detection point information in the second detection point information set includes: initial detection point infrared features and detection point positions; wherein, the step of aggregating detection points based on the first detection point information set and the second detection point information set to obtain a third detection point information set includes: Based on the converted position corresponding to the second detection point information and the detection point position included in the first detection point information, the first detection point information set and the second detection point information set are matched to obtain an initial detection point information pair set, wherein the initial detection point information pair includes the matched first detection point information and second detection point information; For each initial detection point information pair in the set of initial detection point information pairs, the following matching steps are performed: The initial detection point information includes the initial detection point trace features and the initial detection point infrared features. Feature mapping is performed on the initial detection point information to obtain the mapped detection point trace features and the mapped detection point infrared features, wherein the mapped detection point trace features and the mapped detection point infrared features are located in the same feature space. Multidimensional feature similarity is determined for the trace features and infrared features of the mapped detection points to obtain a similarity value; In response to a similarity value greater than a preset threshold, the mapped detection point trace features are determined as the detection point trace features included in the third detection point information, the mapped detection point infrared features are determined as the detection point infrared features included in the third detection point information, and the detection point position included in the second detection point information in the initial detection point information pair is determined as the detection point position included in the third detection point information.

5. The method for identifying intruding drones in airport areas according to claim 4, characterized in that, The process of identifying drones based on the third detection point information set and a pre-trained drone recognition model, resulting in a drone information set, includes: The third detection point information set and the historical detection point information set are associated to obtain the associated detection point information set, wherein the historical detection point information set represents the third detection point information set generated in the previous time window. For each associated post-detection point in the associated post-detection point information set, perform the following identification steps: The associated detection point information is converted into a spatial feature map; The location encoding module included in the UAV identification model is used to encode the spatial feature map to obtain a location feature sequence. The detection point feature extraction module included in the UAV recognition model performs fusion feature extraction on the detection point trace features and detection point infrared features included in the associated detection point information to obtain a detection point fusion feature sequence, wherein the position features and the detection point fusion features correspond one-to-one. The drone location information, including the drone location information, is generated by using the location predictor included in the drone identification model and the location feature sequence. By using the behavior feature mapper included in the drone identification model and the fusion feature sequence of the detection points, drone information including drone behavior features is generated; Determine whether drone identity information exists in the drone identity information database that matches the drone behavior characteristics included in the drone information; In response to the existence of the drone identity information, the drone identity identifier corresponding to the matched drone identity information is determined as the drone identity identifier included in the drone identity information; In response to non-existence, incrementally generate drone identity information including the drone identity identifier; The current time is determined as the identification timestamp included in the drone's identity information.

6. The method for identifying intruding drones in airport areas according to claim 5, characterized in that, The method further includes: The drone information set is simultaneously displayed within the twin body of the airport area; Based on the pre-set warning triggers and the group behavior information, a drone intrusion warning message is generated; The drone intrusion warning information is displayed synchronously on the twin in the airport area.

7. An intrusion drone identification device applied to airport areas, characterized in that, include: The acquisition unit is configured to acquire a set of point information and a set of infrared images within the same time window, wherein the set of point information is acquired by an active phased array radar facing the low-altitude airspace within the airport area, and the set of infrared images is acquired by an infrared camera facing the low-altitude airspace. The dot clustering unit is configured to perform dot clustering based on the dot information set to obtain a first detection point information group set, wherein the first detection point information group represents multiple detection points located in the same acquisition area. The directional multi-target detection unit is configured to perform directional multi-target detection on the infrared images in the infrared image set corresponding to the first detection point information group for each first detection point information group in the first detection point information group set, based on the first detection point information group, to obtain a second detection point information group; The detection point aggregation unit is configured to aggregate detection points based on the first detection point information set and the second detection point information set to obtain a third detection point information set, wherein the third detection point information includes: detection point trace features, detection point infrared features, and detection point positions. The drone identification unit is configured to identify drones based on the third detection point information set and a pre-trained drone identification model, obtaining a drone information set. The drone information includes: drone location information, drone behavior characteristics, drone identification identifier, and identification timestamp. The step of performing directional multi-target detection on the infrared images in the infrared image set corresponding to the first detection point information group based on the first detection point information group to obtain the second detection point information group includes: The detection point position is transformed for each first detection point information in the first detection point information group to generate a transformed position, resulting in a transformed position group. The transformed position represents the image coordinates of the detection point corresponding to the first detection point information in the corresponding infrared image. The infrared image corresponding to the first detection point information group is divided into image partitions to obtain a partitioned infrared image, wherein the partitioned infrared image consists of K image blocks of the same size. Based on the transformed position group, determine the number of hits for each of the K image blocks in the partitioned infrared image; Generate a weight matrix based on the number of hits corresponding to the image patches; The K image blocks included in the partitioned infrared image are encoded to obtain an image block feature matrix; Based on the weight matrix, feature enhancement is performed on the image patch feature matrix to obtain the enhanced image patch matrix; Based on the enhanced image patch matrix and the infrared target detection model, a second detection point information group is generated, wherein the infrared target detection model is a cascaded weak classifier.

8. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 6.

9. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by a processor, implements the method as described in any one of claims 1 to 6.

Images