UAV multi-dimensional perception intelligent inspection warning system and method based on edge computing

Abstract

The present application relates to a multi-dimensional perception intelligent inspection and early warning system and method for unmanned aerial vehicles based on edge computing. The system includes: a multi-dimensional perception data acquisition device, a mobile edge computing server, and an intelligent computing monitoring and decision support platform. The collected multi-dimensional perception data is sent to the mobile edge computing server through the multi-dimensional perception data acquisition device. The mobile edge computing server determines the status data of the inspection range based on the multi-dimensional perception data, generates corresponding early warning information, and sends this information to the intelligent computing monitoring and decision support platform; the intelligent computing monitoring and decision support platform feeds back the early warning information to the operator. The system transfers the calculation process to the mobile edge computing server, reduces dependence on the cloud, reduces data latency, improves security and response speed, and avoids data leakage and the risk of interference that may be suffered during the transmission process, realizing the full-domain perception of unmanned aerial vehicle inspections.

Description

Technical Field

[0001] The present application relates to the field of drone inspection and early warning technology, and in particular to a drone multi-dimensional perception intelligent inspection and early warning system and method based on edge computing. Background Art

[0002] The drone platform, with its excellent maneuverability and flexibility, combined with high-resolution imaging technology, perfectly achieves seamless integration of ground and aerial observation scales. This not only greatly broadens the accessible space and coverage of inspections, enabling it to reach areas that are difficult to reach with traditional means, such as complex terrain, high-rise buildings or remote facilities, but also ensures all-round, real-time monitoring of various types of infrastructure through automated and periodic visual inspections. As a result, the drone inspection platform has become a leader in modern inspection technology. Whether it is power lines, oil and gas pipelines, bridges and tunnels, or crop health monitoring, it can perform tasks efficiently and accurately, significantly improving inspection efficiency and safety assurance.

[0003] Current inspection equipment mainly relies on single-dimensional information for monitoring and evaluation. This method often fails to provide a comprehensive and three-dimensional on-site situation when faced with complex inspection tasks. Although the inspection data will be transmitted to the cloud for calculation and storage, this process may also bring the risk of data privacy leakage, especially when it comes to sensitive areas. In addition, drones also face the bottleneck problem of network transmission in terms of data transmission, which limits their ability to respond quickly and provide information feedback, making it impossible to achieve optimal inspection efficiency. When network conditions are poor, data delays or loss will affect the timeliness of decision-making, further restricting the emergency response capabilities of drones in emergency situations. Therefore, it is urgent to develop a more intelligent and secure inspection system to integrate multi-dimensional information analysis, improve data security, and optimize data transmission and processing processes. Summary of the invention

[0004] Based on this, it is necessary to provide a drone multi-dimensional perception intelligent inspection and warning system and method based on edge computing to address the above technical problems.

[0005] A multi-dimensional perception intelligent inspection and early warning system for unmanned aerial vehicles based on edge computing, the system comprising: a multi-dimensional perception data acquisition device, a mobile edge computing server, an intelligent computing monitoring and decision support platform, and a data center.

[0006] The multi-dimensional perception data acquisition device is mounted on the UAV platform, and is used to collect multi-dimensional perception data and send the multi-dimensional perception data to the mobile edge computing server; the multi-dimensional perception data includes one or more of scene images, scene audio and radar data.

[0007] The mobile edge computing server is mounted on the drone platform and is used to process the received multi-dimensional perception data using the trained multi-dimensional perception intelligent inspection and early warning model, generate early warning information for the inspection range, and send the early warning information to the intelligent computing monitoring and decision support platform; the early warning information includes illegal activities, fires and explosions, and leaks of hazardous materials.

[0008] The intelligent computing monitoring and decision support platform is deployed on the ground to feed back the received warning information to the operator and send the received warning information to the data center.

[0009] The data center is used to receive and store the warning information, the multi-dimensional perception data and the historical records of system operation; it is also used to fuse and analyze the multi-dimensional perception data; it is also used to realize cross-platform data sharing, so that data can be shared and exchanged between the cloud, edge and end.

[0010] In one embodiment, the multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module and a radar acquisition module.

[0011] The visible light / infrared image acquisition module is equipped with a visible light / infrared high-definition camera mounted on a rotating gimbal, and captures high-definition scene images within the inspection range through a drone platform.

[0012] The audio acquisition module is installed on the UAV platform and is used to collect scene audio of the surrounding environment when the UAV is performing inspection tasks.

[0013] The radar acquisition module includes one or more of X-band radar, millimeter wave radar and phased array radar, which are used to obtain the precise spatial position and movement data of the target object.

[0014] In one of the embodiments, the mobile edge computing server adopts a RISC-V edge computing chip, and the RISC-V edge computing chip and the multi-dimensional perception data acquisition device adopt an integrated structure.

[0015] In one embodiment, the system also includes an intelligent algorithm platform, which is connected to a mobile edge computing server.

[0016] An intelligent algorithm platform is used to obtain multi-dimensional perception data and corresponding warning information from a mobile edge computing server, and use the multi-dimensional perception data and corresponding warning information as a training data group; the training data group is used to perform online fine-tuning on the multi-dimensional perception intelligent inspection and warning model to obtain a fine-tuned multi-dimensional perception intelligent inspection and warning model.

[0017] In one embodiment, the intelligent computing monitoring and decision support platform is an intelligent platform that integrates multimodal data analysis, real-time monitoring, anomaly detection, predictive analysis and automatic decision support.

[0018] The intelligent computing monitoring and decision support platform includes: data acquisition layer, data processing and computing layer, decision support layer and user interaction and feedback layer.

[0019] The data acquisition layer is used to integrate image, audio, and radar data, and synchronize time through the PTP or NTP protocol to achieve precise alignment of data in time and space.

[0020] The data processing and computing layer is used to uniformly convert the data format of the original collected data, and to denoise and remove interference from the original collected data. It is also used to integrate deep learning large models, identify cracks and obstacles in images, identify special audio and interference audio, and predict changing trends of the environment and equipment.

[0021] The decision support layer is used to analyze the results and provide users with optimization suggestions, real-time warnings and mission planning support. It is also used to combine the results of multimodal data analysis to adjust the mission path of the drone or equipment in real time, and automatically instruct the system to slow down or stop running when the equipment is abnormal.

[0022] The user interaction and feedback layer is used to obtain information and operate equipment in real time, and to feedback task requirements to the system. It dynamically displays multimodal data by establishing a visual interface.

[0023] A multi-dimensional perception intelligent inspection and early warning method for unmanned aerial vehicles based on edge computing, the method comprising:

[0024] The UAV platform collects multi-dimensional perception data of the inspected area through the multi-dimensional perception data acquisition device carried by it; the multi-dimensional perception data includes one or more of scene images, scene audio and radar data.

[0025] The multi-dimensional sensing data acquisition device transmits the multi-dimensional sensing data to the mobile edge computing server through a wireless network.

[0026] The multi-dimensional perception data is fused in the mobile edge computing server, and the fusion results are input into the trained multi-dimensional perception intelligent inspection and early warning model for processing to generate early warning information for the inspection range. The early warning information is transmitted to the intelligent computing monitoring and decision support platform through the wireless network; the early warning information includes illegal activities, fire and explosion, and leakage of dangerous goods.

[0027] The intelligent computing monitoring and decision support platform will feed back the warning information to the operator. After completing the inspection task, the mobile edge computing server will save the perception data and warning information, and continuously learn and optimize the multi-dimensional perception intelligent inspection and warning model through the feedback mechanism.

[0028] In one of the embodiments, the method further includes: sending the warning information to the data center via the intelligent computing monitoring and decision support platform.

[0029] The above-mentioned edge computing-based drone multi-dimensional perception intelligent inspection and early warning system and method, the system includes: a multi-dimensional perception data acquisition device, a mobile edge computing server, and an intelligent computing monitoring and decision support platform. This system sends the collected multi-dimensional perception data to the mobile edge computing server through the multi-dimensional perception data acquisition device. The mobile edge computing server determines the status data of the inspection range based on the multi-dimensional perception data, generates corresponding early warning information, and sends this information to the intelligent computing monitoring and decision support platform; the intelligent computing monitoring and decision support platform feeds back the early warning information to the operator. The system transfers the calculation process to the mobile edge computing server, reduces the dependence on cloud data, reduces data latency, enables the system to respond to emergencies more promptly, and avoids data leakage and the risk of interference that may be suffered during the transmission process, realizes the full-domain perception of drone inspections, and improves safety and response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A block diagram of a multi-dimensional sensing intelligent inspection and early warning system for drones based on edge computing in one embodiment;

[0031] Figure 2 The figure is a flow chart of a multi-dimensional perception intelligent inspection and early warning method of a drone based on edge computing in one embodiment. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0033] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0034] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0035] In one embodiment, Figure 1 As shown, a multi-dimensional perception intelligent inspection and warning system for unmanned aerial vehicles based on edge computing is provided, which includes: a multi-dimensional perception data acquisition device 100, a mobile edge computing server 200, an intelligent computing monitoring and decision support platform 300 and a data center 500.

[0036] Specifically, the multi-dimensional perception data acquisition device 100 is connected to the mobile edge computing server 200 and mounted on the UAV platform. The mobile edge computing server 200 is connected to the intelligent computing monitoring and decision support platform 300 deployed on the ground.

[0037] The multi-dimensional perception data acquisition device 100 is mounted on a drone platform, and is used to collect multi-dimensional perception data and send the multi-dimensional perception data to a mobile edge computing server 200; the multi-dimensional perception data includes one or more of scene images, scene audio, and radar data.

[0038] Specifically, the multi-dimensional perception data may include one or more of scene images, scene audio, and radar information, and the types of information included in the multi-dimensional perception data correspond to the types of modules included in the multi-dimensional perception data acquisition device, because the multi-dimensional perception data is acquired by the multi-dimensional perception data acquisition device. For example, the multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module, and a radar acquisition module, and the multi-dimensional perception data may include scene images, scene audio, and radar information; for another example, the multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module, and the multi-dimensional perception data may include scene images and scene audio.

[0039] By integrating multi-dimensional perception technology, drones can obtain rich field data from multiple angles and levels in real time during the inspection process, including visual, infrared, lidar, acoustic, meteorological and vibration perception dimensions. The advantage of this multi-dimensional perception is that it can fully and accurately perceive the status of equipment and environment, avoiding the blind spots and limitations of a single perception method, and significantly improving the accuracy and comprehensiveness of inspections. Through the fusion of multi-source data, the system can more accurately identify equipment failures, environmental changes and potential risks, thereby improving response speed, reducing data transmission delays, and quickly issuing warnings when abnormalities occur. The comprehensive application of this multi-dimensional perception not only enhances the safety of inspection tasks, but also improves the intelligence level and flexibility of the entire system, providing strong technical support for drone inspections in complex environments.

[0040] Compared with a single inspection and early warning system, the edge computing-based drone multi-dimensional perception intelligent inspection and early warning system with a multi-dimensional perception data acquisition device can comprehensively and three-dimensionally monitor the inspection target by integrating multi-dimensional perception modules such as visible light / infrared images, audio and radar, and provide more accurate and reliable on-site condition assessment. A single system can often only rely on a single dimension (such as image or sound) for monitoring, and it is easy to ignore potential risks in complex environments. The multi-dimensional perception system effectively improves the accuracy of fault detection and the ability to respond to environmental changes through the fusion of multiple data. In addition, the system realizes data localization processing through edge computing, reduces the impact of transmission delays and network bottlenecks, and ensures faster decision-making responses, especially in emergency situations, and can quickly generate warnings and trigger emergency responses. In summary, the system has significant advantages over a single system in terms of accuracy, response speed, safety and emergency handling.

[0041] The mobile edge computing server 200 is mounted on the UAV platform, and is used to process the received multi-dimensional perception data using the trained multi-dimensional perception intelligent inspection and early warning model, generate early warning information for the inspection range, and send the early warning information to the intelligent computing monitoring and decision support platform 300; the early warning information includes illegal activities, fires and explosions, and leaks of hazardous materials.

[0042] Specifically, the mobile edge computing server 200 performs analysis based on the received multi-dimensional perception data to generate corresponding warning information, and sends the information to the intelligent computing monitoring and decision support platform 300.

[0043] The inspection and early warning model based on multi-dimensional perception data can extract effective information from complex environments and provide timely early warnings. It has the following significant features:

[0044] (1) Multi-dimensional perception data fusion can complement other modes when a certain perception mode is interfered with, thereby enhancing the anti-interference ability to the environment and emergencies. Multi-dimensional data fusion calibrates and aligns different data to improve the overall accuracy.

[0045] (2) Multi-dimensional data fusion enables all-round perception of the environment and targets, significantly improving the accuracy of early warning. The model is based on the fusion and cross-validation of multi-modal data, effectively reducing the probability of false alarms or missed alarms in a single mode.

[0046] (3) The adaptive fusion algorithm of multi-modal multi-dimensional data perception can dynamically adjust the weights of different modalities according to the real-time environment, and the system can maintain performance in complex environmental scenarios.

[0047] (4) The multidimensional perception model can achieve rapid response and efficient reasoning through advanced processing architecture and use edge computing devices to complete the real-time collection and fusion of multimodal data, ensuring that the delay of perception and warning is at the millisecond level.

[0048] (5) With the support of deep learning large models, the fusion of multi-bit perception data not only has the ability to detect anomalies, but can also dynamically adjust strategies based on data. That is, the model collects new data during operation, continuously updates its knowledge base through learning and migration, and adapts to new abnormal patterns.

[0049] (6) Multi-dimensional perception highlights the openness of design. Different sensors can be added to the device through different interfaces to meet different needs. Different modules are easy to maintain and replace.

[0050] The real-time analysis and processing of multi-dimensional perception data is transferred to the mobile edge computing server, and the multi-dimensional perception data related to the detected object does not need to be uploaded to the cloud. While achieving full-domain perception, it also reduces data latency and can respond to emergencies in a timely manner. At the same time, since data does not need to be uploaded to the cloud for processing, edge computing can not only effectively avoid the risk of data leakage, but also reduce the possibility of interference during information transmission.

[0051] The multi-dimensional perception intelligent inspection and early warning model is trained based on a preset training sample set, wherein the preset training sample set includes several training data groups, and each of the several training data groups includes multi-dimensional perception data and target early warning information. During the training process, the multi-dimensional perception data is input into a plurality of preset multi-dimensional perception intelligent inspection and early warning models (the model is a neural network model), and the prediction early warning information is output through the preset multi-dimensional perception intelligent inspection and early warning model, and then the loss function value is determined based on the prediction early warning information and the target early warning information, and finally the preset multi-dimensional perception intelligent inspection and early warning model is reversely learned based on the loss function value to obtain the multi-dimensional perception intelligent inspection and early warning model.

[0052] The multidimensional perception intelligent inspection and early warning model can be a deep learning neural network model. The input item of the multidimensional perception intelligent inspection and early warning model can be multidimensional perception data, and the output data can be early warning information. After obtaining the multidimensional perception data, the multidimensional perception data can be directly input into the multidimensional perception intelligent inspection and early warning model, and the early warning information can be output through the multidimensional perception intelligent inspection and early warning model.

[0053] The mobile edge computing server 200 is arranged on the UAV platform. The mobile edge computing server 200 is respectively connected to the multidimensional perception data acquisition device 100 and the intelligent computing monitoring and decision support platform 300. After the multidimensional perception data acquisition device 100 collects the multidimensional perception data, the warning information corresponding to the multidimensional perception data can be determined through the mobile edge computing server 200. In this way, the warning information can be determined by the mobile edge computing server arranged on the UAV platform, so that the multidimensional perception data involving the inspection range does not need to be sent to the cloud, thereby improving the security of the multidimensional perception data involving the inspection range.

[0054] The intelligent computing monitoring and decision support platform 300 is deployed on the ground, and is used to feed back the received warning information to the operator so that the operator can obtain the information of the inspected range, and send the received warning information to the data center 500.

[0055] Specifically, the intelligent computing monitoring and decision support platform is an intelligent system that integrates multimodal data analysis, real-time monitoring, anomaly detection, predictive analysis and automatic decision support, providing accurate anomaly detection, optimized task planning and scientific decision support.

[0056] The multi-dimensional perception data acquisition device 100 sends the collected multi-dimensional perception data to the mobile edge computing server 200. The mobile edge computing server 200 determines the warning information of the inspection range based on the multi-dimensional perception data, and sends the warning information to the intelligent computing monitoring and decision support platform 300. There is no need to send the multi-dimensional perception data related to the inspection range to the cloud for calculation, thereby reducing the dependence on cloud data. This not only reduces data latency, allowing the system to respond to emergencies more promptly, but also avoids data leakage and the risk of interference that may be suffered during transmission. In this way, the system not only achieves full-domain perception of drone inspections, but also significantly improves safety and response speed.

[0057] The intelligent computing monitoring and decision support platform is equivalent to the command center of ground equipment, and can manage equipment, users, and workstation information display.

[0058] The data center 500 is used to receive and store the warning information, the multi-dimensional perception data and the historical records of system operation; it is also used to fuse and analyze the multi-dimensional perception data; it is also used to realize cross-platform data sharing, so that data can be shared and exchanged between the cloud, edge and end. It receives and stores the warning information sent by the intelligent computing monitoring and decision support platform 300.

[0059] Specifically, (1) Data storage and management: Data storage includes multi-dimensional perception data, warning information, historical records, etc. Data storage is usually classified and stored according to data types. Multi-dimensional perception data includes original perception data collected by the multi-dimensional perception data acquisition device 100 and processed perception data. Warning information is a warning signal or status report generated during the data analysis process; for example, when the system identifies equipment failure, environmental abnormality or potential threat, it generates corresponding warning information. This information needs to be stored for subsequent response, tracing and analysis. In addition, it also includes the historical records of system operation, which include the operating status of the drone platform, the working status of the sensor, the performance record of the edge computing server, previous inspection data, historical records of warning events, logs of completed tasks, etc.

[0060] (2) Data fusion and analysis: This function is to fuse and analyze data. Integrate data from different types of sensors to achieve multi-dimensional perception and analysis. For example, the image data collected by the drone's high-definition camera is fused with the thermal imaging data collected by the infrared camera to achieve panoramic recognition of the target area. At the same time, the data from different sensors may have different positions, directions, and times in space. Data fusion needs to consider spatial positioning information, align coordinate systems, and synchronize time so that data from different sensors can be analyzed in the same coordinate system. The purpose of data analysis is to process and model the collected multi-dimensional perception data.

[0061] (3) Cross-platform data sharing: enables data to be shared and exchanged between different computing platforms, including cloud platforms, edge platforms, local platforms, smart devices, etc.

[0062] Specifically, the multidimensional perception data acquisition device 100 can be connected to the intelligent computing monitoring and decision support platform 300, and send the collected multidimensional perception data to the intelligent computing monitoring and decision support platform 300, or the intelligent computing monitoring and decision support platform 300 retrieves the multidimensional perception data collected by the multidimensional perception data acquisition device 100 according to the received retrieval instruction. When it is necessary to review or trace the data, the dispatched multidimensional perception data can be used to perform event restoration or event review, thereby increasing the function of the early warning system.

[0063] The process of using the edge computing-based drone multi-dimensional perception intelligent inspection and warning system for inspection and warning is as follows: obtaining multi-dimensional perception data collected by the multi-dimensional perception data acquisition device, and determining the warning information of the inspection range based on the multi-dimensional perception data through a mobile edge computing server; sending the warning information to the intelligent computing monitoring and decision support platform, so as to feed back the warning information to the operator through the intelligent computing monitoring and decision support platform.

[0064] In the above-mentioned edge computing-based drone multi-dimensional perception intelligent inspection and early warning system, the system includes: a multi-dimensional perception data acquisition device, a mobile edge computing server, and an intelligent computing monitoring and decision support platform. This system sends the collected multi-dimensional perception data to the mobile edge computing server through the multi-dimensional perception data acquisition device. The mobile edge computing server determines the status data of the inspection range based on the multi-dimensional perception data, generates corresponding early warning information, and sends this information to the intelligent computing monitoring and decision support platform; the intelligent computing monitoring and decision support platform feeds back the early warning information to the operator. The system transfers the calculation process to the mobile edge computing server, reduces the dependence on cloud data, reduces data latency, enables the system to respond to emergencies more promptly, and avoids data leakage and the risk of interference that may be suffered during the transmission process, realizes the full-domain perception of drone inspections, and improves safety and response speed.

[0065] In one embodiment, the multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module and a radar acquisition module.

[0066] The visible light / infrared image acquisition module is equipped with a visible light / infrared high-definition camera mounted on a rotating gimbal, and captures high-definition scene images within the inspection range through a drone platform.

[0067] Specifically, the visible light / infrared image acquisition module can be a shooting device such as a visible light / infrared high-definition camera. The high-definition camera is arranged on the UAV platform. The high-definition camera is used to collect scene images of the inspection range, so that based on the scene image, the status data of the inspection range can be analyzed and extracted, such as the location and activity status of the target object.

[0068] The visible light / infrared image acquisition module captures high-definition images under visible light conditions, and identifies targets through heat distribution under non-visible light conditions. In addition, the visible light / infrared image acquisition module supports 4K, 60FPS, dynamic scene capture, and supports automatic exposure, focus and white balance adjustment. The thermal sensitivity is below 50MK, and the dimensional measurement is between -20℃ and 1000℃.

[0069] The visible light / infrared image acquisition module also supports three-axis rotation of 360° horizontally and ±90° vertically with an accuracy of ±0.1°, and the thermal operating range is within 50°C.

[0070] The operation flow of the image processing module is as follows:

[0071] Step 1: Based on the UAV flight route and mission planning, the gimbal adjusts the angle and locates the target area.

[0072] Step 2: Capture high-resolution images or perform thermal imaging.

[0073] Step 3: Complete noise reduction, cropping or formatting through the pre-processing module.

[0074] Step 4: Send the image data to the edge computing server via a high-speed wireless network (such as 5G or Wi-Fi 6).

[0075] In order to cope with a single type of image acquisition, the focus of the visible light / infrared image acquisition module in this application is to support the simultaneous acquisition of visible light high-definition images and infrared thermal imaging dual modes, and to provide a multi-dimensional description of the target, not limited to appearance features and heat sources. Practice has shown that it can significantly improve the perception ability of complex scenes.

[0076] What’s more, the visible light / infrared image acquisition module has built-in processing algorithms that can actively identify important areas and focus on acquisition, such as embedded deep learning (YOLO).

[0077] The data preprocessing involved in the image processing module includes Gaussian filter image processing, and adjusting the image size and converting it into the corresponding format based on the requirements of the deep learning model.

[0078] Compared with traditional decentralized image acquisition devices, this module significantly improves the acquisition and processing efficiency through the combination of software and hardware, and adapts to the high-dynamic flight missions of UAVs.

[0079] The UAV platform flies within the inspection area and obtains status data within the inspection range based on the scene images captured by the high-definition camera.

[0080] For example: The visible light / infrared image acquisition module uses a high-definition camera. The mobile edge computing server performs real-time detection of the inspection range based on the scene images collected by the high-definition camera. Once a suspicious target is found, it immediately sends an early warning signal and on-site images to the intelligent computing monitoring and decision support platform. After receiving the early warning information, the platform will actively call the camera to collect real-time images for manual monitoring and analysis. In addition, the platform can also continuously track and inspect suspicious targets by controlling drones, providing multi-faceted monitoring guarantees.

[0081] The audio acquisition module is installed on the UAV platform and is used to collect scene audio of the surrounding environment when the UAV is performing inspection tasks.

[0082] Specifically, the audio acquisition module carried by the drone may include an audio pickup and other devices with recording functions. In a specific implementation, the audio acquisition module uses an audio pickup, which is configured on the drone platform to collect scene audio of the surrounding environment, so that the status data of the inspected range can be analyzed and extracted based on the scene audio, such as vehicle driving sound, explosion sound, etc.

[0083] The audio acquisition module is equipped with multiple high-sensitivity microphones to form multiple channels, which can cover the human ear and ultrasonic frequency bands. At the same time, a high-quality audio amplifier circuit is integrated internally to ensure that the collected weak sound signals can be amplified to the usable range. Built-in filters (low-pass, high-pass, band-pass filtering) are used to remove background noise or interference frequency bands.

[0084] The multi-channel audio acquisition module supports sound source positioning and cooperates with the visible light / infrared image acquisition module for tracking and positioning. Based on the beamforming algorithm, it enhances the sound in the specified direction and suppresses interference in other directions. At the same time, combined with the flight path of the drone, it provides the spatial location information of the sound source.

[0085] Furthermore, the tracking and positioning adopts the TDOA algorithm combined with image acquisition to form a three-dimensional space modeling at the data processing end to achieve accurate positioning of the sound source.

[0086] The audio acquisition module is equipped with a real-time audio processing unit that supports signal enhancement, noise reduction and preprocessing, and runs lightweight deep learning models, convolutional neural networks and RNNs for audio classification, and automatically separates different types of sounds.

[0087] In addition, the module integrates a real-time anomaly detection algorithm, analyzes the sound spectrum characteristics through short-time Fourier transform (STFT), and provides a fast sound source localization function, and cooperates with the drone's visible light / infrared image acquisition module to perform multimodal analysis of suspicious areas.

[0088] What's more, the audio acquisition module has a built-in noise processing unit. Different from the noise processing of major mainstream headphones, this noise processing uses a deep learning noise reduction algorithm DenoisingAutoencoder to achieve an adaptive noise reduction mode, actively adjusting the noise reduction according to different flight altitudes.

[0089] In order to highlight the difference of audio collection, this audio collection module has the ability to process preliminary audio data. Specifically, it uses the deep learning model: CNN+RNN to identify complex audio patterns, and uses the WaveNet model to analyze long-term audio data, so as to achieve efficient audio analysis and reduce dependence on central servers.

[0090] In order to cope with the interference of airflow generated during high-speed flight on audio, an anti-wind noise mesh cover is installed on the outside of the audio acquisition module, and a shock-absorbing device is used to reduce the impact of drone vibration on audio equipment.

[0091] The specific operation process of the audio acquisition module is as follows:

[0092] Step 1: Collect audio using a multi-channel, high-sensitivity microphone array.

[0093] Step 2: Filter out background noise or interference frequency bands through low-pass, high-pass and band-pass filters, retain only the target frequency band signal, and retain the audio, and use the amplifier circuit to amplify the weak audio signal.

[0094] Step 3: Use the beamforming algorithm to enhance the sound in the target direction and suppress interference signals in other directions. Calculate the sound source position through the TDOA algorithm and obtain the sound source positioning data in three-dimensional space in combination with the UAV flight path.

[0095] Step 4: Complete multimodal analysis of the target area through multimodal collaboration.

[0096] In addition to the above-mentioned operation process, this module also includes a real-time noise reduction system, an audio enhancement system, and a continuously optimized CNN+RNN learning model.

[0097] The radar acquisition module includes one or more of X-band radar, millimeter wave radar and phased array radar, which are used to obtain the precise spatial position and movement data of the target object.

[0098] Specifically, the radar acquisition module may include an X-band radar, a millimeter-wave radar, and a phased array radar. In a specific implementation, the radar acquisition module uses a millimeter-wave radar configured on a UAV platform to collect status information of the monitored target, such as precise spatial position and movement data.

[0099] According to different mission requirements, the radar acquisition module can only select the appropriate radar system to work. Different types of radars can be deeply integrated with other operating modules to realize a comprehensive multi-dimensional perception system, which uses a multi-dimensional deep fusion algorithm, namely multi-modal fusion (InverseAug and LearnableAlign). Multi-modal fusion can optimize target recognition and tracking in complex environments.

[0100] What's more, the radar detection system integrates an adaptive detection mode, and the system can automatically adjust the radar's operating frequency band, scanning mode, gain setting, etc. according to the flight environment.

[0101] In addition to integrating image processing and sound collection systems to form multi-dimensional perception, the radar detection system can also combine with the UAV flight path to achieve high-precision dynamic target tracking and intelligent obstacle avoidance, allowing the UAV to efficiently avoid collisions in dynamic environments.

[0102] Furthermore, based on multi-dimensional fusion data perception and combined with extended Kalman filtering, the motion trajectory of dynamic and complex targets can be processed.

[0103] At the same time, based on the convolutional neural network, the radar detection system will identify and track the target and its appearance features, and classify, detect and match different targets. The tracking method adopted by the present invention is mainly based on the Mean-Shift algorithm, which is suitable for scenes with small background changes or small targets.

[0104] Furthermore, based on the data information from the radar detection system, the drone can be controlled to achieve the purpose of intelligent obstacle avoidance. For example, the multi-dimensional fusion data mentioned above is used to establish a three-dimensional space model. The drone can identify the surrounding obstacles based on the radar detection system and deep learning algorithm, and perform intelligent obstacle avoidance.

[0105] What's more, after fusing multi-dimensional perception data, an intelligent decision-making system with high-precision dynamic target tracking and intelligent obstacle avoidance is formed.

[0106] Based on the above three modules, the fusion processing steps of the multi-dimensional perception data are as follows:

[0107] Step 1: Multi-dimensional data collection and preprocessing, that is, collecting image, audio, radar wave and other data information through different modules. Each module is embedded with different processing modules and deep learning algorithms.

[0108] Step 2: Fusion and synchronization of multi-dimensional perception data to ensure that all perception data are synchronized in time and avoid data alignment errors caused by time and space differences.

[0109] Step 3: Identify and classify the target.

[0110] Step 4: Use a multi-dimensional and multi-modal data fusion algorithm to fuse data from different perception modules, integrate image, audio and radar information, and improve target recognition and tracking accuracy.

[0111] Step 5: Based on the books and data, a three-dimensional space model is established on the data terminal, and combined with the intelligent decision-making system of multi-dimensional data fusion, dynamic tracking of the target and obstacle avoidance are achieved.

[0112] Step 6: Further control the drone through a visual platform.

[0113] In one of the embodiments, the mobile edge computing server adopts a RISC-V edge computing chip, and the RISC-V edge computing chip and the multi-dimensional perception data acquisition device adopt an integrated structure.

[0114] Specifically, the mobile edge computing server 200 can adopt a RISC-V edge computing chip, and the RISC-V edge computing chip can adopt an integrated structure with a multi-dimensional perception data acquisition device. For example, the RISC-V edge computing chip, visible light / infrared image acquisition module, audio acquisition module and radar acquisition module are all installed on the same electronic device.

[0115] In one embodiment, the system also includes an intelligent algorithm platform, which is connected to a mobile edge computing server.

[0116] An intelligent algorithm platform is used to obtain multi-dimensional perception data and corresponding warning information from a mobile edge computing server, and use the multi-dimensional perception data and corresponding warning information as a training data group; the training data group is used to perform online fine-tuning on the multi-dimensional perception intelligent inspection and warning model to obtain a fine-tuned multi-dimensional perception intelligent inspection and warning model.

[0117] Specifically, in actual applications, the inspection and early warning system may also include an intelligent algorithm platform 400, and the mobile edge computing server 200 is connected to the intelligent algorithm platform 400. The intelligent algorithm platform 400 is used to train the multi-dimensional perception intelligent inspection and early warning model. The mobile edge computing server 200 obtains the multi-dimensional perception intelligent inspection and early warning model through the intelligent algorithm platform, and feeds back the multi-dimensional perception data obtained during use and the early warning information corresponding to the multi-dimensional perception data to the intelligent algorithm platform 400, so that the intelligent algorithm platform 400 can perform online fine-tuning on the multi-dimensional perception intelligent inspection and early warning model based on the multi-dimensional perception data fed back by the mobile edge computing server 200 and the early warning information corresponding to the multi-dimensional perception data, so as to improve the accuracy and matching degree of the multi-dimensional perception intelligent inspection and early warning model.

[0118] In addition to the trained models, the intelligent algorithm platform is connected to the mobile edge computing server to enable online fine-tuning of the models. The specific steps of online fine-tuning of the models include:

[0119] (1) Data reception: The intelligent algorithm platform receives data from various sensors in real time by connecting to the mobile edge computing server. The data are all processed data.

[0120] (2) Deep learning models such as convolutional neural networks (CNN) or recurrent neural networks (RNN) can be used to process multidimensional perceptual data, or to perform classification and regression analysis on data based on traditional machine learning algorithms.

[0121] (3) Online fine-tuning: Whenever the edge computing server uploads new multi-dimensional perception data and corresponding warning information, the intelligent algorithm platform uses this data as a "new batch" of training data and inputs it into the model for learning. The model calculates the loss function based on the error between the output and the expected warning result. The platform then uses the back propagation algorithm to adjust the parameters in the model to minimize the error and optimize the model. Finally, the incremental learning method is used so that the model can be fine-tuned each time new data is received, rather than retraining the entire model.

[0122] (4) Online model update: After each training, the intelligent algorithm platform will apply the adjusted model parameters to the existing system. This approach ensures that the system can adapt to new environmental changes in real time and continuously improve its inspection and early warning capabilities.

[0123] (5) Collaborative training: The intelligent algorithm platform can also be collaboratively trained with other similar devices or systems. For example, through Federated Learning, each terminal can be trained locally and share updated parameters without transmitting the original data to the server.

[0124] In one embodiment, the intelligent computing monitoring and decision support platform is an intelligent platform that integrates multimodal data analysis, real-time monitoring, anomaly detection, predictive analysis and automatic decision support.

[0125] The intelligent computing monitoring and decision support platform includes: data acquisition layer, data processing and computing layer, decision support layer and user interaction and feedback layer.

[0126] The data acquisition layer is used to integrate image, audio, and radar data, and synchronize time through the PTP or NTP protocol to achieve precise alignment of data in time and space.

[0127] The data processing and computing layer is used to uniformly convert the data format of the original collected data, and to denoise and remove interference from the original collected data. It is also used to integrate deep learning large models, identify cracks and obstacles in images, identify special audio and interference audio, and predict changing trends of the environment and equipment.

[0128] Based on the above, data can be uploaded to the cloud platform for in-depth analysis in complex scenarios, and historical data mining and model optimization can be supported.

[0129] The decision support layer is used to analyze the results and provide users with optimization suggestions, real-time warnings, and mission planning support. By combining with deep learning, abnormal data can be marked in real time and trigger warnings, and the warning level can be adaptively evaluated. It is also used to combine the results of multimodal data analysis to adjust the mission path of drones or equipment in real time, and automatically instruct the system to slow down or stop running when the equipment is abnormal.

[0130] The user interaction and feedback layer is used to obtain information and operate equipment in real time, and to feedback task requirements to the system. By establishing a visual interface, multimodal data can be dynamically displayed, such as images, sound bands and waveforms, three-dimensional modeling, etc.

[0131] Users can remotely control and provide feedback to adjust the drone's flight altitude, path, data collection mode, etc.

[0132] In one embodiment, Figure 2 As shown, a multi-dimensional perception intelligent inspection and early warning method for drones based on edge computing is provided, and the method includes the following steps:

[0133] Step 200: The UAV platform collects multi-dimensional perception data of the inspected area through the multi-dimensional perception data acquisition device carried by the platform; the multi-dimensional perception data includes one or more of scene images, scene audio and radar data.

[0134] Specifically, the multi-dimensional perception data is obtained by a multi-dimensional perception data acquisition device, and the multi-dimensional perception data may include one or more of scene images, scene audio, and radar data. For example, the multi-dimensional perception data includes scene images and scene audio, and for another example, the multi-dimensional perception data includes radar data, etc. The warning information may include illegal activities, fires and explosions, and leaks of dangerous goods, etc.

[0135] The drone platform is equipped with high-definition visible light and infrared cameras to capture image data of the inspection area in real time; the audio module captures sound changes on the scene through an audio pickup; the radar module is used to obtain spatial data of the inspection area and monitor the equipment's positioning, movement trajectory and obstacles in the environment in real time.

[0136] Current inspection equipment mainly relies on single-dimensional information for monitoring and evaluation. This method often fails to provide a comprehensive and three-dimensional on-site situation when faced with complex inspection tasks. Although inspection data will be transmitted to the cloud for calculation and storage, this process may also bring the risk of data privacy leakage, especially when it comes to sensitive areas. In addition, drones also face the bottleneck problem of network transmission in terms of data transmission, which limits their ability to respond quickly and feedback information, making it impossible to achieve optimal inspection efficiency. When network conditions are poor, data delays or loss will affect the timeliness of decision-making, further restricting the emergency response capabilities of drones in emergency situations.

[0137] Step 202: The multi-dimensional sensing data acquisition device transmits the multi-dimensional sensing data to the mobile edge computing server via a wireless network.

[0138] Step 204: Fuse the multi-dimensional perception data in the mobile edge computing server, input the fusion result into the trained multi-dimensional perception intelligent inspection and early warning model for processing, generate early warning information for the inspection range, and transmit the early warning information to the intelligent computing monitoring and decision support platform via the wireless network; the early warning information includes illegal activities, fire and explosion, and leakage of dangerous goods.

[0139] Specifically, data from multiple sensors (images, audio, radar) are integrated to obtain comprehensive on-site information; the preset multi-dimensional perception intelligent inspection and early warning model is used to conduct in-depth analysis of the collected multi-dimensional perception data; based on the multi-dimensional analysis of the data, the data is analyzed in real time. By processing the multi-dimensional perception data, the system automatically generates early warning information and analyzes the received early warning information on the intelligent platform.

[0140] The multi-dimensional perception intelligent inspection and early warning model can pre-configure several default early warning information categories. After obtaining the multi-dimensional perception data, the multi-dimensional perception data is input into the multi-dimensional perception intelligent inspection and early warning model. The multi-dimensional perception intelligent inspection and early warning model determines the default early warning information category corresponding to the multi-dimensional perception data, and uses the early warning information corresponding to the default early warning information category as the early warning information corresponding to the multi-dimensional perception data, and uses the early warning information as the early warning information of the inspection range.

[0141] Step 206: The intelligent computing monitoring and decision support platform feeds back the warning information to the operator.

[0142] Step 208: After completing the inspection task, the mobile edge computing server saves the perception data and warning information, and continuously learns and optimizes the multi-dimensional perception intelligent inspection and warning model through the feedback mechanism.

[0143] Specifically, after completing the inspection task, the perception data and warning information are saved, and the intelligent warning model is continuously learned and optimized through the feedback mechanism to improve the efficiency and accuracy of future inspection tasks.

[0144] In one of the embodiments, the method further includes: sending the warning information to the data center via the intelligent computing monitoring and decision support platform.

[0145] In one embodiment, the multi-dimensional perception data acquisition device in step 200 includes a visible light / infrared image acquisition module, an audio acquisition module and a radar acquisition module; the visible light / infrared image acquisition module is configured with a visible light / infrared high-definition camera mounted on a rotating gimbal, and captures high-definition scene images within the inspection range through an unmanned aerial vehicle platform; the audio acquisition module is mounted on the unmanned aerial vehicle platform, and is used to collect scene audio of the surrounding environment when the unmanned aerial vehicle performs an inspection task; the radar acquisition module includes one or more of an X-band radar, a millimeter-wave radar and a phased array radar, and is used to obtain the precise spatial position and movement data of the target object.

[0146] Specifically, the multi-dimensional perception data acquisition device includes one or more of a visible light / infrared image acquisition module, an audio acquisition module, and a radar acquisition module, wherein the visible light / infrared image acquisition module is responsible for capturing high-definition image data; the audio acquisition module focuses on collecting audio information of the surrounding environment; and the radar acquisition module is used to obtain the precise spatial position and movement data of the target object. Thus, the multi-dimensional perception data may include one or more of scene images, scene audio, and radar information, and the types of information included in the multi-dimensional perception data correspond to the types of modules included in the multi-dimensional perception data acquisition device, because the multi-dimensional perception data is acquired by the multi-dimensional perception data acquisition device. For example, the multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module, and a radar acquisition module, and the multi-dimensional perception data may include scene images, scene audio, and radar information; for another example, the multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module and an audio acquisition module, and the multi-dimensional perception data may include scene images and scene audio.

[0147] In one of the embodiments, the mobile edge computing server in step 202 adopts a RISC-V edge computing chip, and the RISC-V edge computing chip and the multi-dimensional perception data acquisition device adopt an integrated structure.

[0148] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0149] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0150] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A multi-dimensional perception intelligent inspection and early warning system for unmanned aerial vehicles based on edge computing, characterized in that: The system includes: a multi-dimensional sensing data acquisition device, a mobile edge computing server, an intelligent computing monitoring and decision support platform, and a data center; The multi-dimensional perception data acquisition device is mounted on the UAV platform, and is used to collect multi-dimensional perception data, and send the multi-dimensional perception data to the mobile edge computing server and the data center; the multi-dimensional perception data includes scene images, scene audio and radar data; The mobile edge computing server is mounted on the drone platform and is used to process the received multi-dimensional perception data using the trained multi-dimensional perception intelligent inspection and early warning model to generate early warning information of the inspection range, and send the early warning information to the intelligent computing monitoring and decision support platform; the early warning information includes illegal activities, fires and explosions, and leakage of dangerous goods; The intelligent computing monitoring and decision support platform is deployed on the ground and is used to feed back the received warning information to the operator and send the received warning information to the data center; The data center is used to receive and store the warning information, the multi-dimensional perception data, and the historical records of system operation; it is also used to fuse and analyze the multi-dimensional perception data; it is also used to realize cross-platform data sharing, so that data can be shared and exchanged between the cloud, edge, and end; The mobile edge computing server adopts a RISC-V edge computing chip, and the RISC-V edge computing chip and the multi-dimensional perception data acquisition device adopt an integrated structure; The system also includes an intelligent algorithm platform, which is connected to the mobile edge computing server; The intelligent algorithm platform is used to obtain multi-dimensional perception data and corresponding warning information from a mobile edge computing server, and use the multi-dimensional perception data and corresponding warning information as a training data group; the training data group is used to perform online fine-tuning on the multi-dimensional perception intelligent inspection and warning model to obtain a fine-tuned multi-dimensional perception intelligent inspection and warning model; wherein the fine-tuning process adopts an incremental learning method, and fine-tuning is performed each time new data is received, rather than retraining the entire model; The intelligent algorithm platform is also used for collaborative training with other similar devices or systems through Federated Learning. Each terminal is trained locally and shares updated parameters.

2. According to the edge computing-based UAV multi-dimensional perception intelligent inspection and early warning system according to claim 1, it is characterized in that: The multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module and a radar acquisition module; The visible light / infrared image acquisition module is equipped with a visible light / infrared high-definition camera mounted on a rotating pan-tilt platform, and captures high-definition scene images within the inspection range through a drone platform; The audio collection module is mounted on the drone platform and is used to collect scene audio of the surrounding environment when the drone performs inspection tasks; The radar acquisition module includes one or more of an X-band radar, a millimeter-wave radar, and a phased array radar, and is used to obtain the precise spatial position and movement data of the target object.

3. The UAV multi-dimensional perception intelligent inspection and early warning system based on edge computing according to claim 1 is characterized in that: The intelligent computing monitoring and decision support platform is an intelligent platform that integrates multimodal data analysis, real-time monitoring, anomaly detection, predictive analysis and automatic decision support; The intelligent computing monitoring and decision support platform includes: a data acquisition layer, a data processing and computing layer, a decision support layer, and a user interaction and feedback layer; The data acquisition layer is used to integrate image, audio, and radar data, and synchronize time through the PTP or NTP protocol to achieve precise alignment of data in time and space; The data processing and computing layer is used to uniformly convert the data format of the original collected data, and to perform denoising and interference removal on the original collected data; it is also used to integrate the deep learning large model, identify the cracks and obstacles in the image, identify special audio and interference audio, and predict the changing trend of the environment and equipment; The decision support layer is used to analyze the results and provide users with optimization suggestions, real-time warnings and mission planning support. It is also used to adjust the mission path of the drone or equipment in real time based on the multimodal data analysis results, and automatically instruct the system to slow down or stop running when the equipment is abnormal. The user interaction and feedback layer is used to obtain information and operate equipment in real time, and to feedback task requirements to the system, and to dynamically display multimodal data by establishing a visual interface.

4. A multi-dimensional perception intelligent inspection and early warning method for drones based on edge computing, characterized in that: The method comprises: The drone platform collects multi-dimensional perception data of the inspected area through the multi-dimensional perception data collection device carried by it; the multi-dimensional perception data includes scene images, scene audio and radar data; The multi-dimensional perception data acquisition device transmits the multi-dimensional perception data to a mobile edge computing server through a wireless network; the mobile edge computing server adopts a RISC-V edge computing chip, and the RISC-V edge computing chip and the multi-dimensional perception data acquisition device adopt an integrated structure; The multi-dimensional sensing data is integrated in the mobile edge computing server, and the integrated result is input into the trained multi-dimensional sensing intelligent inspection and early warning model for processing, so as to generate early warning information of the inspection range, and transmit the early warning information to the intelligent computing monitoring and decision support platform through the wireless network; the early warning information includes illegal activities, fire and explosion, and leakage of dangerous goods; Deploy the intelligent computing monitoring and decision support platform on the ground, receive warning information, and feed back the warning information to the operator and send it to the data center; The data center receives and stores the warning information, the multi-dimensional perception data, and the historical records of system operation, and integrates and analyzes the multi-dimensional perception data in the data center; data sharing across platforms is achieved through the data center, so that data can be shared and exchanged between the cloud, edge, and end; After completing the inspection task, the mobile edge computing server saves the perception data and warning information, and continuously learns and optimizes the multi-dimensional perception intelligent inspection and warning model through the feedback mechanism; The method further includes: the intelligent algorithm platform obtains multi-dimensional perception data and corresponding warning information from the mobile edge computing server, and uses the multi-dimensional perception data and corresponding warning information as a training data group; the multi-dimensional perception intelligent inspection and warning model is fine-tuned online using the training data group to obtain a fine-tuned multi-dimensional perception intelligent inspection and warning model; wherein the fine-tuning process uses an incremental learning method to perform fine-tuning each time new data is received, rather than retraining the entire model; The intelligent algorithm platform conducts collaborative training with other similar devices or systems through Federated Learning. Each terminal is trained locally and shares updated parameters.

5. The UAV multi-dimensional perception intelligent inspection and early warning method based on edge computing according to claim 4 is characterized in that: The method further includes: sending the warning information to a data center via the intelligent computing monitoring and decision support platform.

6. The UAV multi-dimensional perception intelligent inspection and early warning method based on edge computing according to claim 4 is characterized in that: The multi-dimensional perception data acquisition device includes a visible light / infrared image acquisition module, an audio acquisition module and a radar acquisition module; The visible light / infrared image acquisition module is equipped with a visible light / infrared high-definition camera mounted on a rotating pan-tilt platform, and captures high-definition scene images within the inspection range through a drone platform; The audio collection module is mounted on the drone platform and is used to collect scene audio of the surrounding environment when the drone performs inspection tasks; The radar acquisition module includes one or more of an X-band radar, a millimeter-wave radar, and a phased array radar, and is used to obtain the precise spatial position and movement data of the target object.

Images