Abnormal approaching object detection method and device for unmanned aerial vehicle garage, and medium

Abstract

The invention relates to the technical field of safety protection and monitoring, and discloses an abnormally approaching object detection method and device for an unmanned aerial vehicle garage and a medium, and the method comprises the steps: obtaining the space information and image information of a to-be-detected object in a preset monitoring range around the unmanned aerial vehicle garage; performing feature fusion on the image information and the spatial information to obtain fusion data; and determining whether the to-be-detected object is an abnormal approaching object based on the fusion data. According to the method, feature fusion is carried out on the space information and the image information, and judgment is carried out by using the fused data, so that the detection of the abnormal approaching object around the unmanned aerial vehicle library can be realized, the accuracy of determining the abnormal approaching object is improved, and the problem of insufficient accuracy of a detection result caused by dependence on a single information source in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of security protection and monitoring technology, specifically to a method, device and medium for detecting abnormally approaching objects in an unmanned aerial vehicle (UAV) hangar. Background Technology

[0002] As the core location for drone parking and management, the perimeter security of drone hangars is of paramount importance. Current technologies for detecting objects around drone hangars typically rely on a single type of information (images or spatial information only), leading to insufficient accuracy and a high risk of false positives or false negatives. For example, some solutions only acquire and analyze image information of objects, but image information alone is insufficient to accurately determine the spatial location of objects, thus affecting the accuracy of judging whether an object constitutes proximity. Other technologies may rely solely on spatial information for detection, but the lack of image information makes it difficult to effectively identify the specific details and features of objects, limiting the judgment of object attributes and potentially leading to inaccurate results.

[0003] Therefore, there is an urgent need for a new method for detecting abnormally approaching objects that can combine spatial and image information to detect objects around the drone hangar, in order to solve the problem that existing technologies are not accurate enough in identifying abnormally approaching objects due to their reliance on a single information source. Summary of the Invention

[0004] In view of this, this application provides a method, apparatus and medium for detecting abnormally approaching objects in an unmanned aerial vehicle (UAV) hangar, to solve the problem that the existing technology is not accurate enough in identifying abnormally approaching objects due to its reliance on a single information source.

[0005] In a first aspect, this application provides a method for detecting abnormally approaching objects in an unmanned aerial vehicle (UAV) library, the method comprising:

[0006] Acquire spatial and image information of the objects to be detected within the detection area; the detection area is a preset monitoring range around the drone hangar.

[0007] Feature fusion is performed on image information and spatial information to obtain fused data;

[0008] Based on the fused data, it is determined whether the object to be detected is an abnormally close object.

[0009] The abnormal approaching object detection method for unmanned aerial vehicle (UAV) hangars provided in this application first acquires the spatial and image information of the object to be detected. Compared with techniques that only acquire single information, this method obtains basic data about the object in two different dimensions: the object's spatial location attributes and visual appearance attributes, providing a more comprehensive information source for subsequent judgment. Based on this, feature fusion is performed on the acquired spatial and image information to obtain fused data. Through fusion processing, a unified data volume with richer information dimensions and stronger descriptive capabilities is formed. This overcomes the data partiality caused by a single information source in existing technologies. Finally, based on this fused data, it is determined whether the object to be detected is an abnormally approaching object. Because the judgment is based on comprehensive data that integrates the advantages of two information sources, rather than single, limited data, a more comprehensive and reliable assessment of the object's state can be made, thus significantly improving the accuracy of identifying abnormally approaching objects and reducing misjudgments or omissions that may be caused by insufficient information.

[0010] In one optional implementation, acquiring spatial information and image information of the object to be detected within the detection area includes:

[0011] Spatial information of the objects to be detected within the detection area is obtained by installing lidar at a preset location in the drone hangar.

[0012] The system acquires image information of the objects to be detected within the detection area using cameras installed at predetermined locations in the drone hangar.

[0013] The proposed method for detecting abnormally approaching objects in an unmanned aerial vehicle (UAV) hangar utilizes a lidar (LiDAR) sensor, an active sensor capable of directly and accurately measuring the three-dimensional coordinates of objects with minimal impact from ambient lighting variations. A camera, on the other hand, acquires two-dimensional images containing rich color and texture details. By specifying these two sensors, each with clear technological advantages in its respective field, as information sources, the method ensures high accuracy of the acquired raw spatial information and rich detail in the raw image information.

[0014] In one optional implementation, acquiring spatial information and image information of the object to be detected within the detection area further includes:

[0015] The spatial information is filtered to obtain the spatial information after noise removal;

[0016] Image enhancement processing is performed on the image information to obtain the enhanced image information.

[0017] The abnormal approaching object detection method for unmanned aerial vehicle (UAV) hangars provided in this application inevitably contains noise (such as invalid reflection points from lidar) or unclear information (such as images under low light conditions) caused by environmental factors in the raw data directly collected from sensors. Filtering can remove invalid noise points in the spatial data; image enhancement processing can improve the contrast and clarity of the image.

[0018] In one optional implementation, feature fusion is performed on image information and spatial information to obtain fused data, including:

[0019] Align the spatial information and image information in terms of coordinate systems to make them match in space;

[0020] Extract at least one of distance, velocity, or shape as a spatial feature from spatial information;

[0021] Extract at least one of color, texture, or contour as a visual feature from image information;

[0022] Spatial and visual features are combined to obtain fused data.

[0023] The abnormal approaching object detection method for unmanned aerial vehicle (UAV) hangars provided in this application relies on coordinate system alignment, which is a prerequisite for ensuring that data acquired from different sensors can be compared and correlated under the same spatial reference. Without alignment, spatial and image information are isolated from each other. Based on this, by extracting and combining two types of features separately, the originally separate geometric and visual attributes are truly correlated to the same target object at the data level, forming a unified and multi-dimensional feature description. This ensures the effectiveness and correctness of feature fusion, generating fused data that is accurately correlated both spatially and in terms of features, thereby significantly improving the accuracy and completeness of this fused data as a basis for subsequent judgments.

[0024] In one alternative implementation, determining whether the object to be detected is an abnormally close object based on the fused data includes:

[0025] The fused data is input into the detection model, and the detection model's output is used to determine whether the object to be detected is an abnormally close object.

[0026] The detection model is a pre-trained deep learning model, which includes convolutional neural network models or recurrent neural network models; the output includes the category information and motion law information of the object to be detected.

[0027] The abnormal approach object detection method for unmanned aerial vehicle (UAV) arsenals provided in this application, compared to judgment methods based on fixed thresholds or simple rules, utilizes a deep learning model capable of learning complex, non-linear mapping relationships between object features, categories, and behaviors from large amounts of data. It can automatically extract high-level semantic information from fused data, such as identifying whether an object is a person or a vehicle (category information), and analyzing its movement trajectory (motion pattern information), rather than simply concluding that an object is present.

[0028] In one optional implementation, determining whether the object to be detected is an abnormally close object based on the output of the detection model includes:

[0029] Based on the motion pattern information of the object to be detected, determine whether the object to be detected has entered the preset safety warning area;

[0030] If the object to be detected enters the preset safety warning area, the object to be detected is confirmed as an abnormally close object.

[0031] If the object to be detected does not enter the preset safety warning area, it is confirmed that the object to be detected is not an abnormally approaching object.

[0032] The abnormal approach object detection method for unmanned aerial vehicle (UAV) hangars provided in this application directly links the analysis results (motion patterns) of the model with the actual safety strategy of the UAV hangar (safety zone division). The judgment is no longer ambiguous, but based on whether the object's behavior violates explicit spatial boundary rules.

[0033] In one optional implementation, determining whether the object to be detected is an abnormally close object based on the output of the detection model includes:

[0034] Based on the category information of the object to be detected, determine whether the object to be detected is of an unauthorized type;

[0035] If the object to be detected is of an unauthorized type, then the object to be detected is confirmed as an abnormally close object;

[0036] If the object to be detected is of an authorized type, then it is confirmed that the object to be detected is not an abnormally close object.

[0037] The abnormal approach object detection method for unmanned aerial vehicle (UAV) hangars provided in this application allows the system to implement security policies based on the object's identity attributes (category), rather than just its location. For example, even if an object has not entered the core security zone, but is itself a prohibited type (such as a person appearing in an area where only vehicles are allowed), it should be considered abnormal.

[0038] In one alternative implementation, the method further includes:

[0039] Once the object to be detected is determined to be an abnormally close object, an audible and visual alarm installed in the drone hangar is triggered or an alarm message is sent to a preset terminal device.

[0040] The abnormal approach object detection method for unmanned aerial vehicle hangars provided in this application transforms internal judgments into external, perceptible actions (audible and visual alarms) or receivable information (alarm messages). Audible and visual alarms serve as on-site warnings, while remote alarm messages ensure that management personnel are promptly informed of the situation.

[0041] In summary, the abnormal approach object detection method for unmanned aerial vehicle (UAV) hangars provided in this application firstly ensures high precision and rich detail in the acquired raw spatial and image information by specifying LiDAR and cameras as sensors. Building upon this, filtering and image enhancement of the raw data, along with crucial coordinate system alignment, effectively improves data quality and ensures spatial consistency of multi-source information, laying the foundation for subsequent precise fusion. Next, deep fusion at the feature level is achieved by extracting and combining spatial and visual features, generating fused data with richer information dimensions and more comprehensive content. Subsequently, a pre-trained deep learning model is used to analyze this high-quality fused data, enabling intelligent and high-precision identification of object categories and motion patterns, far exceeding the judgment capabilities of traditional simple rule-based methods. Finally, the precise analysis results of the model are combined with specific security strategies. By determining whether an object has intruded into a preset security area or belongs to an unauthorized type, accurate definition of abnormal behavior is achieved, forming a complete response loop through audible and visual alarms and remote alerts. It achieves high-precision and high-reliability automated detection and alarm of abnormally approaching objects around the drone hangar, and ultimately realizes efficient, accurate and reliable automated monitoring of abnormally approaching objects in complex environments, which greatly improves the overall security protection level of the drone hangar.

[0042] Secondly, this application provides an abnormal approaching object detection device for an unmanned aerial vehicle (UAV) hangar, the device comprising:

[0043] The acquisition module is used to acquire spatial and image information of the objects to be detected within the detection area; the detection area is a preset monitoring range around the drone hangar.

[0044] The fusion module is used to perform feature fusion of image information and spatial information to obtain fused data;

[0045] The judgment module is used to determine whether the object to be detected is an abnormally close object based on the fused data.

[0046] Thirdly, this application provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the abnormal approach object detection method for an unmanned aerial vehicle (UAV) hangar described in the first aspect or any corresponding embodiment.

[0047] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to perform the abnormal approach object detection method for an unmanned aerial vehicle (UAV) hangar described in the first aspect or any corresponding embodiment thereof.

[0048] Fifthly, this application provides a computer program product, including computer instructions for causing a computer to execute the abnormal approach object detection method for an unmanned aerial vehicle (UAV) hangar described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating an abnormal approaching object detection method for an unmanned aerial vehicle hangar according to an embodiment of this application;

[0051] Figure 2 This is a flowchart illustrating the working module of an abnormal approach object detection system for an unmanned aerial vehicle hangar according to an embodiment of this application;

[0052] Figure 3 This is a schematic diagram of the workflow of an abnormal approach object detection system for an unmanned aerial vehicle hangar according to an embodiment of this application;

[0053] Figure 4 This is a structural block diagram of an abnormal approach object detection device for an unmanned aerial vehicle hangar according to an embodiment of this application;

[0054] Figure 5 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of this application. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] As a critical facility for storing, maintaining, and managing drones, the perimeter security of drone hangars is of paramount importance. Currently, traditional technologies for drone hangar perimeter security typically rely on a single sensor for monitoring, such as using only a camera for video surveillance. While these solutions can acquire image information of the scene, they have significant technical limitations in practical applications. First, camera performance is heavily dependent on ambient lighting and weather conditions. Image quality deteriorates significantly at night, in low light, or in adverse weather conditions such as rain, snow, or fog, making it difficult to effectively identify targets and easily leading to missed detections. Second, relying solely on two-dimensional image information, the system struggles to accurately acquire three-dimensional spatial information of objects, such as precise distance, speed of movement, and specific location. This greatly reduces the accuracy of determining whether an object poses an "abnormal approach" threat, easily resulting in false alarms.

[0057] To overcome the effects of lighting conditions, some solutions may employ active sensors such as LiDAR. While LiDAR can accurately acquire the spatial position and motion information of objects in all weather conditions, it also has its limitations. The point cloud data it generates lacks rich visual details such as texture and color, making it difficult for the system to effectively identify the specific category of objects. For example, it may not be able to distinguish between an authorized worker and an unauthorized intruder, or identify whether a fast-moving object is a vehicle or a large animal. Therefore, whether using a camera or LiDAR alone, this detection method, which relies on a single information source, suffers from the fundamental problem of insufficient information dimensions, failing to meet the high-precision, high-reliability, and highly intelligent security requirements of modern drone aviaries. Therefore, there is an urgent need for a detection method that can integrate the advantages of multiple information sources to solve the problems of low detection accuracy and poor environmental adaptability of existing technologies in complex environments.

[0058] This embodiment provides a method for detecting abnormally approaching objects in an unmanned aerial vehicle (UAV) hangar. Figure 1 This is a flowchart of an abnormal approaching object detection method for an unmanned aerial vehicle (UAV) hangar according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps:

[0059] S101. Obtain spatial information and image information of the object to be detected within the detection area.

[0060] Specifically, the core task of this step is to simultaneously collect information about an object located within a specific area, based on two different types of information: spatial information and image information. This step establishes that the data foundation of this method is multi-source, meaning it does not rely on a single type of data.

[0061] Spatial information reflects the spatial characteristics of an object, while image information is the information presented by the object through imaging. The object to be detected refers to any physical entity located within the detection area that becomes the target of this detection method for processing and analysis.

[0062] The detection area is a pre-defined and pre-set monitoring range surrounding the drone hangar. Any object appearing within this area can be detected. The pre-defined monitoring range can be divided into warning zones (e.g., 10-50m), restricted zones (e.g., 5-10m), and prohibited zones (e.g., 0-5m), with different alarm strategies for each zone.

[0063] S102. Perform feature fusion on image information and spatial information to obtain fused data.

[0064] Specifically, the purpose of this processing is not simply to put the two types of information together, but to organically combine the features they represent to generate a new, comprehensive set of data, namely, fused data. In this embodiment, the lidar point cloud coordinates (XL, YL, ZL) are converted into camera image coordinates (u, v) using a homogeneous transformation matrix T, as specifically expressed below:

[0065] The transformation formula is:

[0066] S103. Based on the fused data, determine whether the object to be detected is an abnormally close object.

[0067] Specifically, abnormally approaching objects are the targets that this method ultimately aims to identify. This encompasses two meanings: first, proximity, indicating that the object's spatial position and dynamics are somewhat close to the drone hangar; second, abnormality, indicating that this approach behavior or the object itself is determined by the system to be inconsistent with preset safety protocols, potentially posing a threat or requiring attention. For example, an object rapidly approaching the hangar door late at night might be identified as an abnormally approaching object.

[0068] The abnormal approach object detection method for unmanned aerial vehicle (UAV) hangars provided in this application first acquires the spatial and image information of the object to be detected. Compared with techniques that only acquire single information, this method obtains basic data about the object in two different dimensions: the object's spatial location attributes and visual appearance attributes, providing a more comprehensive information source for subsequent judgment. Based on this, feature fusion is performed on the acquired spatial and image information to obtain fused data. Through fusion processing, a unified data body with richer information dimensions and stronger descriptive capabilities is formed. This overcomes the data partiality caused by a single information source in existing technologies. Finally, based on this fused data, it is determined whether the object to be detected is an abnormally approaching object. Since the judgment is based on comprehensive data that integrates the advantages of two information sources, rather than single, limited data, a more comprehensive and reliable assessment of the object's state can be made, thereby significantly improving the accuracy of identifying abnormally approaching objects and reducing misjudgments or omissions that may be caused by insufficient information.

[0069] In one optional implementation, acquiring spatial and image information of the object to be detected within the detection area includes: acquiring spatial information of the object to be detected within the detection area using a LiDAR installed at a preset location in the drone hangar; and acquiring image information of the object to be detected within the detection area using a camera installed at a preset location in the drone hangar. LiDAR is an active sensor capable of directly and accurately measuring the three-dimensional coordinates of an object, with its measurement results being less affected by changes in ambient lighting. The camera, on the other hand, can acquire two-dimensional images containing rich color and texture details. By designating these two sensors, each with clear technological advantages in its respective field, as information sources, the acquired raw spatial information is ensured to have high accuracy, and the raw image information contains rich details.

[0070] In one optional implementation, acquiring spatial and image information of the object to be detected within the detection area further includes: filtering the spatial information to obtain noise-removed spatial information; and performing image enhancement processing on the image information to obtain enhanced image information. Raw data directly acquired from sensors inevitably contains noise (such as invalid reflection points from lidar) or unclear information (such as images under low light conditions) caused by environmental factors. Filtering can remove invalid noise points from the spatial data; image enhancement processing can improve the contrast and clarity of the image.

[0071] In one optional implementation, feature fusion of image information and spatial information is performed to obtain fused data. This includes: aligning the spatial information and image information in coordinate systems to match them spatially; extracting at least one of distance, velocity, or shape as spatial features from the spatial information; extracting at least one of color, texture, or contour as visual features from the image information; and combining the spatial features and visual features to obtain fused data. Coordinate system alignment is a prerequisite for ensuring that data acquired from different sensors can be compared and correlated under the same spatial reference. Without alignment, spatial information and image information are isolated from each other. Based on this, by extracting and combining the two types of features separately, the originally separate geometric and visual attributes are truly correlated to the same target object at the data level, forming a unified and multi-dimensional feature description. This ensures the effectiveness and correctness of feature fusion, generating fused data that is accurately correlated both spatially and in terms of features, thereby significantly improving the accuracy and completeness of the fused data as a basis for subsequent judgment.

[0072] In one optional implementation, determining whether an object to be detected is an abnormally close object based on fused data includes: inputting the fused data into a detection model, and determining whether the object to be detected is an abnormally close object based on the output of the detection model; wherein the detection model is a pre-trained deep learning model, including a convolutional neural network model or a recurrent neural network model; the output includes the category information and motion pattern information of the object to be detected. Compared with judgment methods based on fixed thresholds or simple rules, deep learning models can learn complex and non-linear mapping relationships between object features and categories and behaviors from large amounts of data. It can automatically extract high-level semantic information from fused data, such as identifying whether an object is a person or a vehicle (category information), and analyzing its movement trajectory (motion pattern information), rather than simply concluding that an object is present. This model uses 100,000 sets of labeled data (containing 20 categories of targets such as people, vehicles, and animals) and employs the cross-entropy loss function for model optimization.

[0073] In one optional implementation, determining whether the object to be detected is an abnormally approaching object based on the output of the detection model includes: judging whether the object to be detected has entered a preset safety warning zone based on the motion pattern information of the object; if the object to be detected has entered the preset safety warning zone, then the object to be detected is confirmed as an abnormally approaching object; if the object to be detected has not entered the preset safety warning zone, then the object to be detected is confirmed as not being an abnormally approaching object. The analysis results of the model (motion patterns) are directly linked to the actual safety strategy of the drone hangar (the division of safety zones). The judgment is no longer ambiguous, but based on whether the object's behavior violates clearly defined spatial boundary rules.

[0074] In one optional implementation, determining whether the object to be detected is an abnormally approaching object based on the output of the detection model includes: judging whether the object to be detected is of an unauthorized type based on the object's category information; if the object is of an unauthorized type, then confirming that the object is an abnormally approaching object; if the object is of an authorized type, then confirming that the object is not an abnormally approaching object. This allows the system to enforce security policies based on the object's identity attributes (category), not just its location. For example, even if an object has not entered the core security zone, but is itself a prohibited type (such as a person appearing in an area where only vehicles are allowed), it should be considered abnormal. Unauthorized types include: animals, unmarked vehicles, armed personnel, etc., which can be identified by comparison using a preset blacklist database.

[0075] In an optional implementation, the method further includes: after determining that the object to be detected is an abnormally approaching object, triggering an audible and visual alarm installed in the drone hangar or sending an alarm message to a preset terminal device. This transforms the internal judgment into an external, perceptible action (audible and visual alarm) or a receivable message (alarm message). The audible and visual alarm serves as an on-site warning, while the remote alarm message ensures that management personnel are promptly informed of the situation.

[0076] In summary, the abnormal approach object detection method for unmanned aerial vehicle (UAV) hangars provided in this application first ensures high precision and rich detail in the acquired raw spatial and image information by specifying LiDAR and cameras as sensors. Based on this, filtering and image enhancement of the raw data, along with crucial coordinate system alignment, effectively improves data quality and ensures spatial consistency of multi-source information, laying the foundation for subsequent accurate fusion. Next, deep fusion at the feature level is achieved by extracting and combining spatial and visual features, generating fused data with richer information dimensions and more comprehensive content. Subsequently, a pre-trained deep learning model is used to analyze this high-quality fused data, enabling intelligent and high-precision identification of object categories and motion patterns, far exceeding the judgment capabilities of traditional simple rule-based methods. Finally, the precise analysis results of the model are combined with specific security strategies. By determining whether an object intrudes into a preset security area or belongs to an unauthorized type, accurate definition of abnormal behavior is achieved, forming a complete response loop through audible and visual alarms and remote alerts. It achieves high-precision and high-reliability automated detection and alarm of abnormally approaching objects around the drone hangar, and ultimately realizes efficient, accurate and reliable automated monitoring of abnormally approaching objects in complex environments, which greatly improves the overall security protection level of the drone hangar.

[0077] For example, a specific example will be used below to specifically illustrate the abnormal approach object detection method for unmanned aerial vehicle (UAV) hangars provided in the above embodiments.

[0078] Based on the above-described method for detecting abnormally approaching objects in a drone hangar, an abnormally approaching object detection system for a drone hangar is constructed. The system includes an equipment compartment, a drone compartment door, a lidar mounting base, a lidar, a weather pole, a camera, a weather sensor, a communication gateway, a controller, and an alarm unit.

[0079] The equipment compartment, the main structure of the drone hangar, is a protective physical space. Its primary function is to provide a safe and stable parking place for drones and to house and protect other core electronic devices (such as controllers, communication gateways, battery charging systems, etc.) from damage caused by external environmental factors such as wind, rain, and sun.

[0080] The drone access door is the passageway for drones to enter and exit the equipment compartment. When the drone is in standby or charging mode, the door is closed to protect the drone and its internal equipment; when the drone needs to take off or land for a mission, the door automatically opens. It is a key component for enabling automated drone operations.

[0081] A lidar mounting base is a structural component specifically designed to securely mount a lidar system. Its function is to provide the lidar with a stable, unobstructed mounting position and a suitable scanning field of view, ensuring that the lidar can accurately and reliably scan the surrounding environment and preventing data inaccuracies caused by vibration or improper positioning.

[0082] LiDAR (Light Detection and Ranging) is a core spatial information acquisition sensor. By emitting laser beams and receiving reflected signals, it can accurately measure the distance, orientation, shape, and speed of objects in the surrounding environment, generating high-precision 3D point cloud maps. It is a key device for achieving accurate distance perception and moving object tracking.

[0083] A weather pole is an extended pole-like structure used to mount weather sensors and cameras at appropriate heights and locations. Its main purpose is to obtain a wider monitoring field of view and enable weather sensors to more accurately measure atmospheric parameters of the surrounding environment, avoiding interference from ground objects or the hangar itself.

[0084] The camera is the core image information acquisition sensor. It is responsible for capturing high-definition video or images of the environment surrounding the drone hangar, providing rich visual details such as the color, texture, and shape of objects. It is a key device for identifying specific categories of objects (such as people, vehicles, and animals).

[0085] Weather sensors are used to monitor environmental meteorological conditions around the drone hangar in real time, such as wind speed, wind direction, temperature, humidity, and air pressure. This data can serve as an important basis for drone take-off and landing decisions, and can also be used to analyze the impact of severe weather on the performance of monitoring sensors (such as cameras).

[0086] The communication gateway serves as the internal and external information exchange hub of the drone hangar. It is responsible for transmitting data from within the drone hangar (such as sensor data, alarm information, equipment status, etc.) to a remote monitoring center or management personnel's terminal via a network (such as wired, 4G / 5G); at the same time, it also receives control commands from remote locations.

[0087] The controller is responsible for receiving data from sensors such as LiDAR and cameras, executing algorithms for data fusion, analysis, and judgment, and controlling the actions of other units (such as alarm units and hatches) based on the judgment results. All data processing and decision-making logic is completed in the controller.

[0088] The alarm unit is the terminal for executing alarm signals. When the controller determines that an abnormal situation has occurred, the alarm unit will be activated according to the instruction, emitting a strong audible (alarm sound) and visual (warning light flashing) signal. Its main function is to deter intruders on-site and alert nearby friendly personnel.

[0089] The specific workflow of the abnormal approach object detection system for the drone hangar in this embodiment is as follows: Figure 2 and Figure 3 It mainly executes its workflow through lidar, cameras, communication gateways, controllers, and alarm units, such as... Figure 2 As shown, the structure integrates cameras and LiDAR sensors, which are distributed at different locations within the drone hangar to achieve comprehensive, blind-spot-free monitoring. An internal controller receives and processes sensor data, and executes data fusion and analysis algorithms. The drone hangar is also equipped with alarm devices, including audible and visual alarms, and a communication module that can remotely transmit alarm information to the management personnel's terminal devices.

[0090] The specific workflow of the abnormal approach object detection system for the drone hangar in this embodiment is as follows: Figure 3 As shown.

[0091] The core technology lies in the deep fusion of data from both cameras and LiDAR sensors, enabling accurate detection and timely alerts for abnormally approaching objects around the drone hangar. The specific process includes the following:

[0092] The first stage is data acquisition, during which both the LiDAR and camera operate simultaneously. The LiDAR, with its high-speed and precise scanning capabilities, performs real-time detection of the space surrounding the drone hangar. It can quickly and accurately acquire key spatial information such as the distance and orientation of surrounding objects. By continuously emitting laser beams and receiving reflected signals, the LiDAR can construct a three-dimensional spatial model. In this model, each object can be marked with its specific location and approximate direction of movement, providing fundamental spatial data support for subsequent determinations of whether objects are approaching the drone hangar.

[0093] At the same time, the camera is also simultaneously acquiring image information of the corresponding area. The camera has rich visual perception capabilities, able to capture detailed features such as the color, texture, and shape of objects. Different types of objects, such as humans, animals, and vehicles, will exhibit unique visual characteristics in the images captured by the camera; these characteristics are crucial for accurately identifying the object's category.

[0094] The next stage is data preprocessing. Since the data formats and characteristics collected by LiDAR and cameras differ, preprocessing is necessary to facilitate subsequent fusion processing. For LiDAR data, filtering is primarily performed to remove invalid data points caused by noise, clutter, and other factors, improving data accuracy and reliability. Simultaneously, coordinate transformation is performed to unify the data into a standard coordinate system for matching with camera data.

[0095] For image data captured by the camera, image enhancement processing is first performed, such as adjusting brightness, contrast, and color balance, to improve image clarity and quality, especially under low light or complex lighting conditions, to enhance image recognizability. Then, image segmentation is performed to separate different object regions in the image, preparing for subsequent feature extraction and recognition.

[0096] Next comes the data fusion stage, a crucial step in the entire detection method. Through specific data fusion algorithms, preprocessed LiDAR spatial data and camera image data are deeply fused. A common fusion strategy is feature-layer-based fusion, which involves extracting features from both the LiDAR and camera image data separately and then combining these features. For example, features such as distance, speed, and shape of an object are extracted from the LiDAR data, while features such as color, texture, and contour are extracted from the camera image data. These features are then fused together to form a more comprehensive and richer feature vector.

[0097] Another fusion strategy is decision-level fusion, which involves using LiDAR data and camera image data independently for object detection and identification, and then making a comprehensive decision based on the individual detection results. For example, if the LiDAR detects an approaching object, and the camera identifies the object as belonging to a category that may pose a threat to the drone hangar, then the object can be determined to be an abnormally approaching object.

[0098] The next stage is data analysis and judgment, where deep learning algorithms are used to conduct in-depth analysis of the fused data. Deep learning algorithms have powerful feature learning and pattern recognition capabilities. Through training with a large amount of training data, they can learn the characteristic patterns and motion laws of different types of objects. In practical applications, the fused data is input into a trained deep learning model. The model outputs information such as the object's category and features, and judges whether the object is an abnormally approaching object according to preset rules. For example, if unauthorized personnel or vehicles are detected rapidly approaching the drone hangar and entering a preset security warning area, then it can be judged as an abnormal approach.

[0099] Finally, in the alarm triggering phase, once an abnormally approaching object is detected, the system will immediately trigger the alarm mechanism. Alarm methods include several forms. First, there is an audible and visual alarm; audible and visual alarms placed around the drone hangar will emit a loud alarm and flashing lights to attract the attention of nearby personnel. Simultaneously, the system will remotely push alarm information to the terminal devices of relevant management personnel, such as mobile phones and computers, via the communication module. The alarm information includes detailed information such as the type, location, and speed of the abnormally approaching object, enabling management personnel to take appropriate measures in a timely manner.

[0100] The abnormal approach object detection system for drone hangars provided in this embodiment overcomes the shortcomings of single sensors by fusing data from cameras and LiDAR. LiDAR provides accurate spatial location information, while the camera supplements the detailed features of the object, enabling the system to more accurately identify abnormally approaching objects and reduce false alarms and missed alarms. The two sensors complement each other under different lighting and weather conditions. For example, LiDAR can still operate normally at night or in low-light environments; in adverse weather conditions such as rain, snow, and fog, the combination of camera and LiDAR ensures continuous and effective monitoring. A timely and accurate alarm mechanism enables managers to quickly take measures to address potential threats, effectively protecting the safety of drones and related equipment in the drone hangar and reducing the risk of safety accidents.

[0101] This embodiment also provides an abnormal approach object detection device for an unmanned aerial vehicle (UAV) hangar. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0102] This embodiment provides an abnormal approach object detection device for a drone hangar, such as... Figure 4 As shown, it includes:

[0103] The acquisition module 401 is used to acquire spatial information and image information of the object to be detected within the detection area; the detection area is a preset monitoring range around the drone hangar.

[0104] The fusion module 402 is used to perform feature fusion of image information and spatial information to obtain fused data;

[0105] The judgment module 403 is used to determine whether the object to be detected is an abnormally close object based on the fused data.

[0106] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0107] In this embodiment, the abnormal approach object detection device for the drone hangar is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above-mentioned functions.

[0108] This application also provides a computer device having the above-described features. Figure 4 The image shows an abnormal approach object detection device for a drone hangar.

[0109] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of this application, such as... Figure 5As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 5 Take a processor 10 as an example.

[0110] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GPA), or any combination thereof.

[0111] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0112] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0113] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0114] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0115] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0116] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0117] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0118] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for detecting abnormally approaching objects in an unmanned aerial vehicle (UAV) hangar, characterized in that, The method includes: Acquire spatial and image information of the object to be detected within the detection area; the detection area is a preset monitoring range around the drone hangar; The image information and the spatial information are fused to obtain fused data; Based on the fused data, it is determined whether the object to be detected is an abnormally close object.

2. The method according to claim 1, characterized in that, The acquisition of spatial and image information of the object to be detected within the detection area includes: Spatial information of the objects to be detected within the detection area is obtained by using lidar installed at preset locations in the drone hangar; The camera, installed at a preset location in the drone hangar, acquires image information of the object to be detected within the detection area.

3. The method according to claim 2, characterized in that, The acquisition of spatial and image information of the object to be detected within the detection area also includes: The spatial information is filtered to obtain spatial information after noise removal; The image information is then subjected to image enhancement processing to obtain enhanced image information.

4. The method according to claim 3, characterized in that, The step of fusing features from the image information and the spatial information to obtain fused data includes: The spatial information and the image information are aligned in coordinate systems to make them spatially matched. Extract at least one of distance, speed, or shape as spatial features from the spatial information; Extract at least one of color, texture, or contour as a visual feature from the image information; The spatial features and the visual features are combined to obtain fused data.

5. The method according to any one of claims 1 to 4, characterized in that, The step of determining whether the object to be detected is an abnormally close object based on the fused data includes... The fused data is input into the detection model, and the detection model's output is used to determine whether the object to be detected is an abnormally close object. The detection model is a pre-trained deep learning model, which includes a convolutional neural network model or a recurrent neural network model; the output includes the category information and motion law information of the object to be detected.

6. The method according to claim 5, characterized in that, The step of determining whether the object to be detected is an abnormally close object based on the output of the detection model includes: Based on the motion pattern information of the object to be detected, it is determined whether the object to be detected has entered the preset safety warning area; If the object to be detected enters the preset safety warning area, then the object to be detected is confirmed to be an abnormally close object. If the object to be detected does not enter the preset safety warning area, then it is confirmed that the object to be detected is not an abnormally close object.

7. The method according to claim 5, characterized in that, The step of determining whether the object to be detected is an abnormally close object based on the output of the detection model includes: Based on the category information of the object to be detected, determine whether the object to be detected is of an unauthorized type; If the object to be detected is of an unauthorized type, then the object to be detected is confirmed as an abnormally close object; If the object to be detected is of an authorized type, then it is confirmed that the object to be detected is not an abnormally close object.

8. The method according to claim 5, characterized in that, The method further includes: Once the object to be detected is determined to be an abnormally close object, an audible and visual alarm installed in the drone hangar is triggered or an alarm message is sent to a preset terminal device.

9. An abnormal approach object detection device for an unmanned aerial vehicle (UAV) hangar, characterized in that, The device includes: The acquisition module is used to acquire spatial information and image information of the object to be detected within the detection area; the detection area is a preset monitoring range around the drone hangar. The fusion module is used to perform feature fusion on the image information and the spatial information to obtain fused data; The judgment module is used to determine, based on the fused data, whether the object to be detected is an abnormally close object.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the abnormal approach object detection method for an unmanned aerial vehicle (UAV) hangar as described in any one of claims 1 to 8.

Images