Unmanned aerial vehicle identification and tracking system

Abstract

The invention discloses an unmanned aerial vehicle identification and tracking system, which can integrate multi-source visual information by calling to-be-identified images shot by a plurality of camera units, thereby determining accurate position information of a target unmanned aerial vehicle. The plurality of camera units can be monitoring camera networks which are arranged all over, the camera networks are arranged and execute tasks such as public security monitoring at ordinary times, and unmanned aerial vehicle identification and tracking can be executed, so that the hardware use cost is reduced, and the camera units can be quickly spread and implemented under the existing hardware condition; the visual field range of the camera shooting unit is matched with the low-altitude flight height of the target unmanned aerial vehicle, the blind area of the flyer detection radar can be effectively covered, and the target unmanned aerial vehicle can still be effectively recognized and tracked under the condition that the target unmanned aerial vehicle flies at a low altitude and the flyer detection radar cannot effectively detect. Supervision of the unmanned aerial vehicle is improved, and use safety of the unmanned aerial vehicle is guaranteed. The method is widely applied to the technical field of unmanned aerial vehicle management.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle management, and in particular to an unmanned aerial vehicle identification and tracking system. Background Art

[0002] Unmanned aerial vehicles can be used as carriers for tools such as cameras and sensors, and can also be used to transport goods or even people. They can be applied in fields such as geodetic surveying, news reporting, film shooting, equipment maintenance, commodity distribution, transportation, and agricultural sowing and irrigation, generating huge social and economic benefits.

[0003] Due to the wide range of activities of unmanned aerial vehicles, the use of unmanned aerial vehicles poses potential risks to national defense, security, flight safety, and social order, and it is necessary to supervise unmanned aerial vehicles. When an unmanned aerial vehicle is flying at a high altitude, a flying object detection radar can be used for detection. However, when an unmanned aerial vehicle is flying at a low altitude (for example, the flight altitude is lower than 100 meters), it is usually in the blind area of the flying object detection radar, and the flying object detection radar cannot maintain real-time tracking of the unmanned aerial vehicle. The technology relying on the flying object detection radar cannot maintain effective monitoring of the unmanned aerial vehicle, and loopholes occur in the security system. For example, in an urban environment, facilities such as buildings, utility poles, billboards, and viaducts generally exceed the low-altitude flight height range of unmanned aerial vehicles. When an unmanned aerial vehicle takes off or cruises, it is prone to flying at a low altitude for a long time. If the unmanned aerial vehicle cannot be tracked in real time, accidents such as collisions between the unmanned aerial vehicle and buildings or even vehicles are very likely to occur, posing great use risks. Summary of the Invention

[0004] Aiming at the technical problems that the current related technologies cannot maintain real-time tracking and effective monitoring of unmanned aerial vehicles and cannot effectively prevent the use risks of unmanned aerial vehicles, the purpose of the present invention is to provide an unmanned aerial vehicle identification and tracking system.

[0005] On the one hand, an embodiment of the present invention includes an unmanned aerial vehicle identification and tracking system, and the unmanned aerial vehicle identification and tracking system includes the following steps:

[0006] A processing module; the processing module is used to obtain the to-be-identified images respectively captured by a plurality of camera units, detect each of the to-be-identified images with the target unmanned aerial vehicle as the detection target, and perform field-of-view intersection according to the detection results of each of the to-be-identified images to determine the position information of the target unmanned aerial vehicle.

[0007] Further, the detecting each of the to-be-identified images with the target unmanned aerial vehicle as the detection target includes:

[0008] Using a target detection algorithm to process each of the to-be-identified images;

[0009] Obtain the anchor boxes in the image to be recognized output by the target detection algorithm; the anchor boxes are used as the detection results of the image to be recognized, and the anchor boxes represent the pixel coordinates of the target UAV in the image to be recognized.

[0010] Further, the target detection algorithm is the YOLO algorithm.

[0011] Further, the determining the position information of the target UAV according to the field-of-view intersection of the detection results of each image to be recognized includes:

[0012] For the anchor box of any image to be recognized, determine the world coordinates of the unknown depth information of the target UAV according to the anchor box;

[0013] Solve the world coordinates of the multiple unknown depth information of the target UAV to obtain the world coordinates of the known depth information; the world coordinates of the known depth information are used as the position information of the target UAV.

[0014] Further, the processing module is further configured to determine the historical flight trajectory of the target UAV according to the position information, perform trajectory prediction according to the historical flight trajectory to obtain a predicted flight trajectory, and call the corresponding camera unit of the predicted flight trajectory to perform shooting field-of-view adjustment.

[0015] Further, the processing module is further configured to identify malicious flight behaviors of the target UAV according to the historical flight trajectory, and generate a warning message when the malicious flight behaviors are identified.

[0016] Further, the identifying the malicious flight behaviors of the target UAV according to the historical flight trajectory includes:

[0017] Detect the curved segments in the historical flight trajectory;

[0018] Fit to obtain the curvature circle corresponding to the curved segment;

[0019] Obtain the center position of the curvature circle;

[0020] When the distance between the center position of the curvature circle and the installation position of any camera unit is less than a distance threshold, determine the curvature circle as the target curvature circle;

[0021] When the total number of the target curvature circles is greater than a number threshold, determine that the malicious flight behaviors are identified.

[0022] Further, the UAV identification and tracking system further includes:

[0023] A camera module; the camera module includes a plurality of camera units.

[0024] Furthermore, the UAV identification and tracking system further includes:

[0025] A camera module; the camera module is used to call a plurality of camera units.

[0026] Furthermore, the camera unit is a municipal surveillance camera.

[0027] On the other hand, an embodiment of the present invention further includes a computer device, including a memory and a processor. The memory is used to store at least one program, and the processor is used to load at least one program to execute the UAV identification and tracking system in the embodiment.

[0028] On the other hand, an embodiment of the present invention further includes a computer-readable storage medium, in which a program executable by a processor is stored. The program executable by the processor is used to execute the UAV identification and tracking system in the embodiment when executed by the processor.

[0029] The beneficial effects of the present invention are as follows: In the UAV identification and tracking system in the embodiment, by calling the to-be-identified images respectively captured by a plurality of camera units, multi-source visual information can be integrated, so as to determine the accurate position information of the target UAV; the plurality of camera units can be a network of surveillance cameras arranged throughout. These camera networks have already been deployed and perform tasks such as public security surveillance in normal times, and can be called by the UAV identification and tracking system in this embodiment to perform UAV identification and tracking. Therefore, it is beneficial to reduce the hardware usage cost and is easy to be quickly implemented under the existing hardware conditions; the camera units are generally installed on the road surface or at a low altitude, and the field of view range of the camera units matches the low-altitude flight height of the target UAV, which can effectively cover the blind area of the flying object detection radar. When the target UAV is flying at a low altitude and the flying object detection radar cannot effectively detect it, it can still maintain effective identification and tracking of the target UAV, which is beneficial to improving the supervision of UAVs and ensuring the use safety of UAVs. Description of the Drawings

[0030] Figure 1 and Figure 2 is a schematic structural diagram of the UAV identification and tracking system in the embodiment;

[0031] Figure 3 is a schematic diagram of the camera unit in the embodiment;

[0032] Figure 4 is a schematic diagram of the working principle of the UAV identification and tracking system in the embodiment;

[0033] Figure 5 is a schematic diagram of field-of-view intersection in the embodiment;

[0034] Figure 6Schematic diagram of the principle of the YOLO algorithm in the embodiment;

[0035] Figure 7 Schematic diagram of detecting the coordinates of the target UAV in the pixel coordinate system from the image to be recognized in the embodiment;

[0036] Figure 8 Schematic diagram of the principle of coordinate transformation in the embodiment;

[0037] Figure 9 Schematic diagram of the historical flight trajectory and the predicted flight trajectory in the embodiment;

[0038] Figure 10 Schematic diagram of the curvature circle in the embodiment. Detailed implementation manners

[0039] In this embodiment, the structure of the UAV recognition and tracking system is as shown in Figure 1 or Figure 2 shown. Referring to Figure 1 and Figure 2 , the UAV recognition and tracking system includes a processing module and a camera module, where the processing module can be a device such as a server. One structure of the camera module is as shown in Figure 1 shown, Figure 1 In, the camera module itself is composed of multiple camera units such as camera unit 1, camera unit 2... camera unit n, and these camera units are respectively connected to the processing module or connected to the processing module through a bus. Another structure of the camera module is as shown in Figure 2 shown, Figure 2 In, the camera module can be a device such as a server, and the camera module can call multiple external camera units such as camera unit 1, camera unit 2... camera unit n for shooting.

[0040] In this embodiment, the camera units included or called by the camera module can be municipal cameras widely arranged beside roads, buildings or utility poles, etc. In this embodiment, the style of the municipal camera is as shown in Figure 3 shown. The municipal camera can continuously shoot with a monocular camera to obtain a real-time monitoring image, so as to monitor the area within its field of view.

[0041] In this embodiment, the working principle of the UAV recognition and tracking system is as shown in Figure 4 shown.

[0042] Referring to Figure 4 , the UAV recognition and tracking system can work at the data layer, recognition layer, positioning layer and application layer.

[0043] Referring to Figure 4, at the data layer, multiple camera units perform shooting respectively. In this embodiment, the image obtained by the camera unit shooting is called the image to be recognized. Each camera unit sends the image to be recognized obtained by its shooting to the processing module.

[0044] Refer to Figure 4 , at the recognition layer, the processing module uses the target UAV as the detection target to detect each image to be recognized. Among them, the target UAV can be any UAV, or a specific UAV that needs to be continuously tracked.

[0045] Refer to Figure 4 , at the positioning layer, for any image to be recognized, its detection result is specifically that the target UAV is detected from the image to be recognized, or the target UAV is not detected from the image to be recognized. A single image to be recognized is a planar image. The processing module can determine the two-dimensional position information of the target UAV from the image to be recognized, but lacks the depth information of the target UAV, such as the distance between the target UAV and the camera unit that shoots the target UAV. If the processing module detects the target UAV from multiple images to be recognized, then since the installation positions of the camera units are fixed and known, as Figure 5 shown, the processing module can perform field-of-view intersection on the two-dimensional position information of the target UAV detected from multiple images to be recognized respectively, so as to obtain the depth information of the target UAV and finally determine the position information of the target UAV. According to the position information of the target UAV, the three-dimensional position coordinates of the target UAV can be determined, such as the distance between the target UAV and any camera unit, the flight altitude of the target UAV, etc., so that the relative position relationship between the target UAV and environmental objects such as any building can be further calculated.

[0046] Refer to Figure 4 , at the application layer, the UAV recognition and tracking system can perform trajectory tracking and behavior management on the target UAV. Specifically, since the camera unit continuously shoots the latest image to be recognized, the processing module can also continuously obtain the latest position information of the target UAV, so as to realize the trajectory tracking of the target UAV. The processing module can import the position information of the target UAV into the visual map, set areas such as no-fly zones in the visual map, and if it is detected that the position information of the target UAV is within the no-fly zone, management measures such as warning or flight restriction are taken.

[0047] In this embodiment, the function of visual map resolution is mainly to extract key information from image or video data using computer vision technology, and then generate and analyze the map. By combining means such as image processing, pattern recognition, machine learning, and optimization algorithms, the automatic extraction and analysis of spatial information are realized. The main process is as follows:

[0048] Feature extraction and descriptor calculation: Extract significant feature points from each frame of the image, such as corner points, edge points, or regions with unique textures. Calculate descriptors for each feature point to describe its local appearance and features.

[0049] Feature matching and pose estimation: Use feature descriptors for inter-frame matching to find similar feature points between different frames.

[0050] Map initialization and update: Initialize the map for the image sequence. As new image frames are acquired, continuously update and optimize the map, including adding new feature points, deleting unreliable feature points, and updating the topological structure and geometric information of the map.

[0051] Global optimization: Perform global optimization based on the constraints of all known observations, adjusting the positions of feature points in the map and the poses of the cameras to minimize the reprojection error and improve the consistency of the map.

[0052] Map representation and storage: The map is usually represented in a sparse or dense manner. In a sparse map, the positions and descriptors of feature points are stored, and a dense map may contain more geometric information such as point clouds or meshes.

[0053] In the UAV identification and tracking system of this embodiment, by calling the to-be-identified images captured by multiple camera units respectively, it is possible to integrate multi-source visual information, thereby determining the precise position information of the target UAV; the multiple camera units can be a network of surveillance cameras arranged throughout, and these camera networks are already set up and perform tasks such as public security monitoring on a daily basis, and can be called by the UAV identification and tracking system of this embodiment to perform UAV identification and tracking. Therefore, it is beneficial to reduce the hardware usage cost and is easy to quickly spread and implement under the existing hardware conditions; the camera units are generally installed on the road surface or at low altitudes, and the field of view of the camera units matches the low-altitude flight height of the target UAV, and can effectively cover the blind area of the flying object detection radar. When the target UAV is flying at low altitude and the flying object detection radar cannot effectively detect it, it is still possible to maintain effective identification and tracking of the target UAV, which is beneficial to improving the supervision of UAVs and ensuring the safe use of UAVs.

[0054] In this embodiment, each camera unit can send the to-be-identified image to the processing module in the form of a video stream. For example, camera unit 1 captures the to-be-identified image at time t1 Captures the to-be-identified image at time t2 Captures the to-be-identified image at time t3 Sends it to the processing module in the form of a video stream, then the processing module continuously receives the to-be-identified image The processing module is synchronized with each camera unit. In this way, camera unit 2 also captures the to-be-identified image at time t1 The to-be-recognized image is captured at time t2 The to-be-recognized image is captured at time t3 It is sent to the processing module in the form of a video stream, and the processing module continuously receives the to-be-recognized images The processing module can determine which camera unit captures the to-be-recognized image at which time according to the source and frame number of the received to-be-recognized image

[0055] In this embodiment, when the processing module executes the step of detecting each to-be-recognized image with the target drone as the detection target, the following steps can be specifically executed

[0056] S101. Use the target detection algorithm to process each to-be-recognized image respectively

[0057] S102. Obtain the anchor boxes in the to-be-recognized image output by the target detection algorithm

[0058] In step S101, the processing module can run the YOLO algorithm to process each to-be-recognized image respectively. YOLO is a target detection system based on a single neural network. For example Figure 6As shown in the figure, the YOLO algorithm uses a CNN model to implement object detection. First, the size of the input image is resized to 448x448, then the image is fed into the CNN network, and the network prediction results are processed to obtain the detected objects. Compared with the R-CNN algorithm, the algorithm framework of YOLO is unified and overall. YOLOv7 is a relatively new stable version in the YOLO series. It is a one-stage object detection algorithm and has obvious advantages in detection speed compared with two-stage algorithms. As an object detection algorithm in the YOLO series, YOLOv7 not only inherits the advantages of the YOLO series but also greatly improves the detection speed, which stems from its unique network architecture and optimization strategy. YOLOv7 regards object detection as a regression problem and conducts end-to-end training in a single network. The core idea of YOLO is to use the entire image as the input of the network and directly regress the position of the bounding box and the category to which the bounding box belongs at the output layer. YOLOv7 adopts an optimized network structure, including deeper network layers and more effective feature extraction methods. Usually, YOLOv7 uses Darknet as the base network and makes improvements to enhance performance. In addition, YOLOv7 uses a feature pyramid network to extract multi-scale feature maps and improves the design of anchor boxes. The optimal anchor box size is determined through more refined clustering analysis to adapt to objects of different sizes. YOLOv7 can make predictions at multiple scales, which enables it to more effectively detect various objects from small to large. While maintaining high accuracy, YOLOv7 still has real-time detection performance and is suitable for running on resource-constrained devices. It maintains the end-to-end training feature, and the entire process from image preprocessing to object detection is completed in a unified network. When designing YOLOv7, model compression and acceleration are considered. Techniques such as structured pruning and quantization are used to reduce the model size and improve the inference speed.

[0059] In step S101, if the target drone is included in the image to be recognized input to the YOLO algorithm, then the area where the target drone is located in the image to be recognized will be recognized by the YOLO algorithm. In step S102, the YOLO algorithm generates an anchor box to enclose the area where the target drone is located as the detection result of the image to be recognized. The anchor box represents the pixel coordinates of the target drone in the image to be recognized.

[0060] For example, as Figure 7 shown, for an image to be recognized, if it contains a target drone, then the coordinates (u, v) of the target drone in the pixel coordinate system can be determined according to the anchor box generated by the YOLO algorithm for recognition.

[0061] In this embodiment, the pixel coordinate system is a two-dimensional coordinate system on the image, used to define the position of each pixel on the image. For each camera unit, there is also a camera coordinate system, which is a coordinate system defined relative to the position of the camera unit. In the camera coordinate system, the origin is usually located at the optical center of the camera unit, the Z-axis is usually consistent with the shooting direction of the camera unit, and the X-axis and Y-axis can be parallel to the ground. There is also a world coordinate system, which is a global coordinate system used to define the positions of objects in the scene. In this coordinate system, the position of each object is described relative to a global origin. For example, each camera unit has its corresponding coordinates in the world coordinate system, and objects such as the target UAV and buildings also have their corresponding coordinates in the world coordinate system.

[0062] In this embodiment, the coordinates (u, v) of the target UAV directly detected from the image to be recognized in the pixel coordinate system represent the relative position of the target UAV in the image to be recognized at the shooting moment of the image to be recognized, and can be converted into the coordinates of the target UAV in the real environment where objects such as the camera unit, buildings, roads, vehicles, and the target UAV are located, that is, the coordinates in the world coordinate system, as the position information of the target UAV.

[0063] In this embodiment, when the processing module executes the step of determining the position information of the target UAV by performing field-of-view intersection based on the detection results of each image to be recognized, the following steps can be specifically executed:

[0064] S201. For the anchor box of any image to be recognized, determine the world coordinates of the unknown depth information of the target UAV according to the anchor box;

[0065] S202. Solve the world coordinates of the multiple unknown depth information of the target UAV to obtain the world coordinates of the known depth information; the world coordinates of the known depth information are used as the position information of the target UAV.

[0066] The principles of steps S201 - S202 are as Figure 8 shown. Referring to Figure 8 , the coordinates of the target UAV in the pixel coordinate system, in the camera coordinate system, and in the world coordinate system can be mutually converted. For example, if there are multiple images to be recognized taken at time t1 that all contain the target UAV, then the coordinates of the target UAV in the pixel coordinate system of each image to be recognized at time t1 can be obtained by executing steps S101 - S102. By Figure 8Through the coordinate system transformation shown above, the coordinates of the target UAV in the world coordinate system at time t1 can be obtained. Similarly, if there are multiple images captured at time t2 that contain the target UAV, then by performing steps S101 - S102, the coordinates of the target UAV in the pixel coordinate system of each image to be recognized at time t2 can be obtained. Through Figure 8 the coordinate system transformation shown above, the coordinates of the target UAV in the world coordinate system at time t2 can be obtained. Based on this principle, as long as each imaging unit continuously captures the images to be recognized at each shooting moment, the coordinates of the target UAV in the world coordinate system at any moment can be obtained.

[0067] In this embodiment, it is assumed that the internal parameter matrix of any imaging unit (for example, imaging unit 1) is K and the external parameter matrix is M ext , where the internal parameter matrix K and the external parameter matrix M ext can be obtained by calibrating the imaging unit after it is installed. If the coordinates of the target UAV in the pixel coordinate system in the image to be recognized captured by this imaging unit at any moment (for example, time t1) are (u, v), then the coordinates P of the target UAV in the world coordinate system can be obtained through the following formula

[0068]

[0069] transformation. Here, the homogeneous coordinate form of P w can be expressed as w In the above formula, R is the rotation matrix and t is the translation vector, which can be obtained according to the external parameter matrix M

[0070]

[0071] Specifically, it is ext In the above formula, the coordinates (u, v) of the target UAV in the pixel coordinate system, the rotation matrix R, the translation vector t, and the internal parameter matrix K are known quantities, and the coordinates P of the target UAV in the world coordinate system

[0072]

[0073] and the depth information Z w are unknown quantities. When at least two imaging units (for example, imaging unit 1 and imaging unit 2) capture the images to be recognized (including the target UAV) at the same shooting moment (for example, time t1), for each imaging unit, the coordinates P of the target UAV in the world coordinate system C in their respective formulas and the depth information Z w and Care the same, so we can combine their formulas to calculate the coordinates P of the target drone in the world coordinate system. w , and obtain the location information of the target UAV at the shooting time (for example, time t1).

[0074] Since the shooting time (for example, time t1) in the above embodiment is arbitrary, the drone identification and tracking system can detect the position information of the target drone at any time, that is, the coordinates of the target drone in the world coordinate system at any time, thereby tracking the target drone.

[0075] In this embodiment, since the processing module can obtain the coordinates of the target drone in the world coordinate system at any time, such as the coordinates of the target drone in the world coordinate system at times t1, t2, t3, etc., the processing module can run a curve fitting algorithm to fit these coordinates into a directional curve according to the time sequence corresponding to these coordinates, and obtain the historical flight trajectory of the target drone. The historical flight trajectory represents the actual flight position of the target drone identified by the drone identification and tracking system.

[0076] In this embodiment, the historical flight trajectory is formed by the coordinate points corresponding to each time t1, t2, t3, etc., which is equivalent to a time series. Therefore, the processing module can run a time series prediction algorithm such as a long short-term memory network (LSTM) to perform trajectory prediction and obtain a predicted flight trajectory. The predicted flight trajectory means predicting the location where the target drone may fly over in the future based on the law of the historical flight trajectory.

[0077] In this embodiment, the processing module calls the camera unit corresponding to the predicted flight trajectory to adjust the shooting field of view. Figure 9 As shown, the processing module stores the installation positions of each camera unit, such as the coordinates of the camera unit in the world coordinate system, and the predicted flight trajectory can also be represented by a combination of coordinates in the world coordinate system, so the processing module can calculate the distance between each camera unit and the predicted flight trajectory. If the distance between a camera unit and the predicted flight trajectory is less than a distance threshold (this distance threshold can be equal to the farthest shooting distance of this camera unit), then the processing module can determine that this camera unit is a camera unit corresponding to the predicted flight trajectory. For example Figure 9Among them, both the imaging unit 4 and the imaging unit 6 belong to the imaging units corresponding to the predicted flight trajectory. The imaging unit 4 and the imaging unit 6 can be cameras capable of adjusting the shooting direction, and the processing module can control the field of view ranges of the imaging unit 4 and the imaging unit 6 so that the field of view ranges of the imaging unit 4 and the imaging unit 6 include the predicted flight trajectory. In this way, the imaging unit 4 and the imaging unit 6 can adjust the field of view range in advance, so that when the target UAV actually passes through the predicted flight trajectory, the target UAV can be photographed more accurately, realizing real-time identification and tracking of the target UAV.

[0078] In this embodiment, the processing module also identifies the historical flight trajectory. If it detects the characteristics of the malicious flight behavior of the target UAV in the historical flight trajectory, the processing module generates a warning message.

[0079] In this embodiment, when the processing module executes the step of identifying the malicious flight behavior of the target UAV according to the historical flight trajectory, it can specifically execute the following steps:

[0080] S301. Detect the curve segments in the historical flight trajectory;

[0081] S302. Fit to obtain the curvature circle corresponding to the curve segment;

[0082] S303. Obtain the center position of the curvature circle;

[0083] S304. When the distance between the center position of the curvature circle and the installation position of any imaging unit is less than the distance threshold, determine the curvature circle as the target curvature circle;

[0084] S305. When the total number of target curvature circles is greater than the number threshold, determine that a malicious flight behavior is identified.

[0085] In step S301, the processing module can start from the starting point of the historical flight trajectory and sequentially intercept line segments of equal length as curve segments.

[0086] In step S301, the processing module can detect the curvature at each point in the historical flight trajectory, determine the points with curvature greater than the curvature threshold as curve points. If the length of the line segment formed by multiple consecutive curve points is greater than the length threshold, then determine the line segment formed by these multiple consecutive curve points as a curve segment. That is to say, for a curve segment, the adjacent part in the historical flight trajectory is a line segment with a smaller curvature (straighter), that is, a straight line segment, or a curved line segment that is not straight enough but relatively short.

[0087] If multiple curve segments are identified in step S301, then for each curve segment, perform the identification operations including steps S302 - S304.

[0088] For any curve segment, in step S302, as Figure 10 shown, a curvature circle corresponding to the curve segment is obtained by fitting. The curvature circle is the circle that is closest to the shape of the curve segment and can determine the position of the center of the curvature circle in step S303. In step S304, the distances between the position of the center of the curvature circle and the installation positions of each camera unit are calculated. If there is at least one camera unit and the distance between the installation position of the camera unit and the position of the center of the curvature circle is less than the distance threshold, then this curvature circle is determined as the target curvature circle.

[0089] In step S305, the number of curvature circles identified as belonging to the target curvature circle by steps S302 - S304 is counted. If the total number of target curvature circles is greater than the number threshold (for example, 3), then it can be determined that the target UAV has malicious flight behavior.

[0090] In this embodiment, the principle of executing steps S301 - S305 is as follows: The curve segment in the historical flight trajectory indicates that the target UAV has made a turning flight behavior during flight. The curvature circle corresponding to the curve segment is the circle with the shape closest to the curve segment, and the position of the center of the curvature circle represents the position around which the target UAV turns in the curve segment; if the distance between the position of the center of the curvature circle and the installation position of a camera unit is less than the distance threshold, it can be determined that the position of the center of the curvature circle coincides with the installation position of a camera unit, thereby determining that the target UAV turns and avoids around this camera unit; the number of target curvature circles indicates the number of times the target UAV turns and avoids around the camera unit. When the total number of target curvature circles is greater than the number threshold, it can be determined that the target UAV has actively turned and avoided the camera unit multiple times, thereby indicating that the target UAV is very likely to have made or will make malicious flight behavior, thus realizing the identification and prevention of the malicious flight behavior of the target UAV.

[0091] The processing module can call the communication module to send a warning message to the target UAV, thereby warning the target UAV to immediately take operations such as landing; the processing module can send the warning message to other UAVs or the receiving party, thereby reminding other UAVs or the receiving party to timely take operations such as avoiding the target UAV; the processing module can send the warning message to the aircraft management party, thereby reminding the aircraft management party to timely stop or restrict the flight of the target UAV.

[0092] By generating and sending warning messages, the processing module can effectively prevent and stop the malicious flight behavior of the target UAV, thereby contributing to ensuring the safety of the air environment.

[0093] The UAV identification and tracking system in this embodiment has the following advantages:

[0094] (1) Low cost and high economic benefits:

[0095] The monocular camera has a relatively low cost. Combining with the high efficiency of the YOLO algorithm, the entire system has low requirements for hardware and computing resources, enabling large-scale deployment of the system and achieving economic benefits.

[0096] (2) Strong real-time performance and fast response speed:

[0097] The YOLO algorithm is famous for its efficient single-step detection method and is suitable for processing real-time video stream data. This means that the drone can be identified and located in real time, providing quick feedback for drone flight monitoring.

[0098] (3) Strong scalability and flexibility:

[0099] By training different YOLO models, the system can identify and track specific types of objects, such as vehicles, people, or fixed markers. This flexibility enables the system to be applied to multiple tasks, such as cargo delivery, target tracking, security patrol, etc., and has strong scalability.

[0100] It should be noted that, unless otherwise specified, when a certain feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to another feature, or indirectly fixed or connected to another feature. In addition, the up, down, left, right, etc. descriptions used in this disclosure are only relative to the mutual positional relationship of the components of this disclosure in the drawings. The singular forms of "a", "an", and "the" used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as commonly understood by those skilled in the art of this technical field. The terms used in the specification of this embodiment are only for describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in this embodiment includes any and all combinations of one or more of the related listed items.

[0101] It should be understood that although terms such as first, second, and third may be used in this disclosure to describe various elements, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, without departing from the scope of this disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as", etc.) provided in this embodiment is only intended to better illustrate the embodiments of the present invention and will not impose a limitation on the scope of the present invention unless otherwise required.

[0102] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The methods can be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with the computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - in accordance with the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, for this purpose the program is capable of running on a programmed application-specific integrated circuit.

[0103] In addition, the operations of the processes described in this embodiment can be performed in any suitable order, unless this embodiment otherwise indicates or is otherwise clearly contradicted by the context. The processes described in this embodiment (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) executed jointly on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions executable by one or more processors.

[0104] Furthermore, the method can be implemented in any type of computing platform operatively connected, including but not limited to personal computers, minicomputers, mainframes, workstations, network or distributed computing environments, separate or integrated computer platforms, or in communication with charged particle tools or other imaging devices, etc. Aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into the computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer and, when the storage medium or device is read by the computer, can be used to configure and operate the computer to perform the processes described herein. In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network. When such media include instructions or programs that implement the above steps in conjunction with a microprocessor or other data processor, the invention of this embodiment includes these and other different types of non-transitory computer-readable storage media. When programmed in accordance with the methods and techniques of the present invention, the present invention also includes the computer itself.

[0105] A computer program can be applied to input data to perform the functions of this embodiment, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents physical and tangible objects, including a specific visual depiction of the physical and tangible objects generated on the display.

[0106] The above is only a preferred embodiment of the present invention. The present invention is not limited to the above embodiments. As long as it achieves the technical effects of the present invention by the same means, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention. Within the scope of protection of the present invention, its technical solutions and / or implementation manners can have various different modifications and changes.

Claims

1. An unmanned aerial vehicle recognition and tracking system, characterized in that, The drone identification and tracking system includes: A processing module; the processing module is used to obtain the to-be-identified images respectively captured by multiple camera units, detect each of the to-be-identified images with the target drone as the detection target, and perform field-of-view intersection according to the detection results of each of the to-be-identified images to determine the position information of the target drone.

2. The drone identification and tracking system according to claim 1, characterized in that, The detecting each of the to-be-identified images with the target drone as the detection target includes: Using a target detection algorithm to process each of the to-be-identified images respectively; Obtaining the anchor boxes in the to-be-identified images output by the target detection algorithm; the anchor boxes are used as the detection results of the to-be-identified images, and the anchor boxes represent the pixel coordinates of the target drone in the to-be-identified images.

3. The drone identification and tracking system according to claim 2, wherein The target detection algorithm is the YOLO algorithm.

4. The drone identification and tracking system according to claim 2, characterized in that, The performing field-of-view intersection according to the detection results of each of the to-be-identified images to determine the position information of the target drone includes: For the anchor box of any one of the to-be-identified images, determining the world coordinates of the unknown depth information of the target drone according to the anchor box; Solving the world coordinates of the multiple unknown depth information of the target drone to obtain the world coordinates of the known depth information; the world coordinates of the known depth information are used as the position information of the target drone.

5. The drone identification and tracking system according to claim 1, wherein: The processing module is further used to determine the historical flight trajectory of the target drone according to the position information, perform trajectory prediction according to the historical flight trajectory to obtain a predicted flight trajectory, and call the corresponding camera unit of the predicted flight trajectory to adjust the shooting field of view.

6. The drone identification and tracking system according to claim 5, wherein: The processing module is further used to identify the malicious flight behavior of the target drone according to the historical flight trajectory, and generate a warning message when the malicious flight behavior is identified.

7. The drone identification and tracking system according to claim 6, characterized in that, The identifying the malicious flight behavior of the target drone according to the historical flight trajectory includes: Detecting the curve segments in the historical flight trajectory; Fitting to obtain the curvature circle corresponding to the curve segment; Obtaining the center position of the curvature circle; When the distance between the center position of the curvature circle and the installation position of any one of the camera units is less than a distance threshold, determining the curvature circle as the target curvature circle; When the total number of the target curvature circles is greater than a number threshold, determining that the malicious flight behavior is identified.

8. The UAV identification and tracking system according to any one of claims 1-7, characterized in that The drone identification and tracking system further includes: A camera module; the camera module includes multiple camera units.

9. The UAV identification and tracking system according to any one of claims 1-7, characterized in that, The drone identification and tracking system further includes: A camera module; the camera module is used to call multiple camera units.

10. The UAV identification and tracking system according to any one of claims 1-7, characterized in that, The camera unit is a municipal surveillance camera.

Images