Unmanned aerial vehicle equipment for monitoring confined space and autonomous operation method of unmanned aerial vehicle equipment

Abstract

The invention provides unmanned aerial vehicle equipment for closed space monitoring and an autonomous operation method thereof, and particularly relates to the technical field of mechanical automation, the unmanned aerial vehicle equipment comprises a fuselage main body, a power system, a data acquisition and control module and other components; the fuselage main body adopts a combined design of an upper anti-collision frame, a lower anti-collision frame and a load-bearing structure, the power system is in a vertically overlapped and staggered six-rotor layout, the data acquisition and control module integrates a flight control computer, an airborne calculation unit and a standardized secondary development interface, and an external special sensor can be flexibly mounted; according to the autonomous operation method, positioning and map construction in a GPS-free environment are realized through fusion of a laser radar and a visual sensor, and path planning, dynamic obstacle avoidance, task execution and autonomous return are completed in combination with an autonomous navigation algorithm.

Description

Technical Field

[0001] This invention provides an unmanned aerial vehicle (UAV) device for monitoring confined spaces and its autonomous operation method, specifically relating to the field of mechanical automation technology. Background Technology

[0002] In enclosed space operations such as indoor 3D mapping, warehouse inspection, and tunnel detection, multi-rotor drones, with their vertical take-off and landing, hovering, and flexible maneuverability, have become core equipment for detection and data acquisition tasks, effectively solving problems such as the difficulty and low efficiency of manual operations. However, enclosed spaces present objective constraints such as the lack of GPS signals, limited space dimensions, and complex internal structures. The application of existing multi-rotor drones in this field still faces many insurmountable technical bottlenecks, severely restricting operational efficiency and reliability. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the present invention provides an unmanned aerial vehicle (UAV) equipment for monitoring confined spaces and its autonomous operation method, which can effectively solve the related technical problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A drone-based monitoring system for confined spaces includes: The main body of the fuselage is composed of a crash protection frame and a load-bearing structure; wherein, the crash protection frame is a double-layer frame structure formed by a bracket; and the load-bearing structure is fixed to the upper area of ​​the crash protection frame. The power system adopts a six-rotor layout with overlapping and interlacing rotors, including first to sixth motors and first to sixth propellers respectively fixed on the shafts of each motor; the first to sixth motors are arranged in two layers in the lower part of the load-bearing structure of the fuselage, wherein the first motor and the second motor are located in the upper layer, and the third to sixth motors are located in the lower layer. The data acquisition and control module integrates a flight control computer, an onboard computing unit, and a GPS module. The flight control computer receives detection data from the perception and navigation system to achieve autonomous positioning, path planning, dynamic obstacle avoidance, and control of the power system. The data acquisition and control module also has a standardized secondary development interface.

[0005] Furthermore, the drone equipment includes a fuselage, a power system, a data acquisition and control module, a camera, supplementary lighting, and a vision sensor; the load-bearing structure is fixed to the upper area of ​​the anti-collision frame; the first to sixth motors of the power system are arranged on the load-bearing structure; the first and second vision sensors of the vision sensor are symmetrically arranged on one side of the data acquisition and control module; the camera is fixed to one side of the load-bearing structure; the first and second supplementary lighting are symmetrically installed on both sides of the camera.

[0006] Furthermore, the drone equipment used for monitoring confined spaces also includes a camera, which is fixedly connected to one side of the load-bearing structure and electrically connected to the data acquisition and control module through the standardized interface.

[0007] Furthermore, the drone equipment used for monitoring confined spaces also includes supplementary lighting, consisting of a first supplementary light and a second supplementary light; the first and second supplementary lights are symmetrically installed on both sides of the camera, and the two are electrically connected to the data acquisition and control module through the standardized interface.

[0008] Furthermore, the UAV equipment used for monitoring confined spaces also includes the aforementioned visual sensors, comprising a first visual sensor and a second visual sensor; the first visual sensor and the second visual sensor are respectively symmetrically mounted on one side of the data acquisition and control module, and both are electrically connected to the data acquisition and control module through the aforementioned standardized interface.

[0009] This invention also proposes an autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces, comprising the following steps: In a confined space without satellite navigation signals, the airborne lidar and visual sensors are activated, and the supplementary lighting device is selectively activated based on the ambient lighting conditions. By fusing the perception data from the lidar and the vision sensor, the drone's pose is calculated in real time, and a three-dimensional environmental map of the enclosed space is constructed simultaneously. Based on the aforementioned 3D environment map and the preset mission objectives, a flight path is planned; When executing the flight path, dynamic obstacle detection and avoidance are performed based on the real-time perception data of the lidar and the vision sensor, and the flight trajectory is adjusted accordingly; The drone is controlled to fly along the adjusted flight path and to perform monitoring tasks through onboard mission sensors, while simultaneously transmitting monitoring data back. After the monitoring task is completed, the return path is planned based on the real-time pose, and the UAV is controlled to return to the starting point and land.

[0010] Furthermore, the step of fusing the perception data from the lidar and the vision sensor to calculate the drone's pose in real time and simultaneously construct a three-dimensional environmental map of the enclosed space includes: The raw point cloud data collected by the lidar is preprocessed to obtain lidar point cloud data, and the image data collected by the vision sensor is corrected. Three-dimensional geometric features are extracted from the preprocessed lidar point cloud data to generate point cloud feature descriptors. At the same time, two-dimensional image feature points and corresponding feature descriptors are extracted from the corrected image data. Based on the point cloud feature descriptor and the similarity measure of the feature descriptor, the cross-modal matching of the three-dimensional geometric features of the lidar and the two-dimensional geometric features of the vision is performed by the random sampling consensus algorithm to establish the correlation relationship of multi-sensor data; Using feature matching results as constraints, a vision-LiDAR fusion positioning equation is constructed. Combined with short-term attitude data from the UAV IMU, an extended Kalman filter algorithm is used to fuse multi-source data in real time. The position coordinates and attitude angles of the UAV are solved by iterative calculation to perform pose estimation in a GPS-free environment; Based on the real-time calculated UAV pose, the lidar point cloud data is transformed to the global coordinate system through coordinate transformation; A voxel grid algorithm is used to spatially partition and optimize the density of the global point cloud, removing redundant point cloud data. By combining texture information extracted from visual images, the image texture is mapped onto the corresponding 3D point cloud surface to generate a textured 3D environment map.

[0011] Furthermore, the step of planning the flight path based on the three-dimensional environment map and the preset mission objective includes: The pre-set task objectives are broken down into structures, clarifying the task type, operation coverage, key monitoring areas, data acquisition accuracy requirements, and task priority; among them, the spatial boundaries of the full-coverage scan must be marked for the 3D mapping task, the location coordinates or feature markers of the equipment to be inspected must be specified for the equipment to be inspected, and the sampling point distribution density of parameters such as gas, temperature, and humidity must be set for the environmental monitoring task. Based on the constructed 3D environment map, the flyable area, obstacle area and buffer zone are divided by voxel segmentation algorithm; Starting from the mission start point and targeting the mission end point or key sampling point, the optimal path is searched within the flyable area. During path generation, obstacle buffer zones are avoided first. At the same time, the distribution of path nodes is optimized according to the mission type. 3D mapping tasks use a gridded uniform point distribution. Collision detection is performed between the optimized path and the 3D environment map, and the distance between all nodes and obstacles on the path is verified by point cloud distance calculation to see if the distance between them meets the preset constraints. Preset path update trigger conditions include incremental updates of the 3D environment map and adjustments to the task objective; when the trigger conditions are met, map data or task information is updated.

[0012] Furthermore, the step of dynamically detecting and avoiding obstacles and adjusting the flight trajectory based on real-time perception data from the lidar and the vision sensor includes: The lidar and dual vision sensors synchronously collect environmental data at a preset frequency. The lidar outputs a three-dimensional point cloud data stream, and the vision sensors output continuous image frames. The three-dimensional contours, spatial coordinates, and distance information of obstacles are extracted to initially screen dynamic targets that deviate from the three-dimensional environment map. Moving target detection is performed on the image frames of the visual sensor to identify the contour features and movement trends of dynamic obstacles. The detection results of the LiDAR and dual visual sensors are then fused. Based on the detected obstacle parameters, a risk assessment model is constructed, and the flight trajectory is adjusted using the risk assessment model.

[0013] Furthermore, the steps of controlling the UAV to fly along the adjusted flight path, performing monitoring tasks through onboard mission sensors, and simultaneously transmitting monitoring data back include: Based on the adjusted flight trajectory, the flight control computer combines real-time attitude data from lidar and vision sensors, and uses a PID control algorithm to adjust the speed difference between the upper and lower six rotors. During the control process, it corrects trajectory deviations caused by airflow disturbances and load changes in real time.

[0014] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: Firstly, the power layout is efficient and compact: the power system adopts a six-rotor layout with overlapping upper and lower layers. Through the layered arrangement of dual motors on the upper layer and four motors on the lower layer, the power output is maximized within a limited space, while taking into account both load capacity and flight stability. Compared with traditional single-layer six-rotor systems, this layout is more compact, reduces the space occupied by the fuselage, and facilitates flexible movement in narrow and confined spaces. At the same time, the multi-rotor design ensures redundant flight capability in the event of a single point of failure, thereby improving operational reliability.

[0015] The components are arranged in a reasonable and orderly manner: the GPS module and the lidar are installed adjacent to each other above the data acquisition and control module, the visual sensors are symmetrically arranged on the side of the module, and the camera and the fill light are symmetrically combined. The components are arranged in an orderly manner around the core module, which not only avoids signal interference, but also optimizes the distribution of the body's center of gravity, ensuring that the drone remains stable during complex actions such as hovering and turning, and is suitable for high-precision operations in confined spaces.

[0016] Secondly, it provides full coverage in GPS-free environments: Through the fusion design of LiDAR and dual vision sensors, combined with the fusion positioning logic of LiDAR point cloud data and visual image data in the autonomous operation method, it can complete real-time positioning and high-precision 3D point cloud map construction in confined spaces without relying on GPS signals, completely solving the pain point of positioning failure of traditional drones in GPS-free scenarios such as indoors and tunnels.

[0017] Reliable positioning redundancy backup: The data acquisition and control module is integrated, and the GPS module and LiDAR work together to form positioning redundancy. In open environments, GPS can quickly obtain accurate positioning references, while in enclosed environments, it automatically switches to the fusion positioning mode of LiDAR and visual sensors. The two positioning methods are seamlessly connected, which improves the positioning stability and fault tolerance in different scenarios.

[0018] Thirdly, the standardized interface is widely compatible: the data acquisition and control module is equipped with a standardized serial port, CAN interface or Ethernet interface, which supports flexible mounting and data interaction of various external special sensors such as gas sensors, infrared thermal imagers, and multispectral sensors. It can quickly switch and adapt to the sensor according to different task requirements such as 3D mapping, environmental monitoring, and equipment inspection, without the need for major modifications to the body structure, thus reducing the task adaptation cost.

[0019] Autonomous operation closed-loop process: Through autonomous operation methods, drones can automatically complete equipment initialization, GPS-free positioning and map building, path planning, dynamic obstacle avoidance, task execution and autonomous return, without the need for real-time human control. This greatly reduces the dependence on operator skills for confined space operations and improves operational efficiency. It is especially suitable for confined environments that are difficult for humans to enter, such as toxic, oxygen-deficient, and narrow spaces, avoiding the safety risks of manual operation. Attached Figure Description

[0020] Figure 1 This is a three-dimensional structural diagram of an unmanned aerial vehicle (UAV) device for monitoring confined spaces and its autonomous operation method proposed in this invention. Figure 2 This is a schematic diagram of a camera mounted on a drone for monitoring confined spaces, as proposed in this invention. Figure 3 This is a schematic diagram of a supplementary lighting device for monitoring confined spaces on a drone, as proposed in this invention. Figure 4 This is a schematic diagram of a visual sensor for monitoring confined spaces on a drone, as proposed in this invention. Figure 5 This is a flowchart illustrating an autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces, as proposed in this invention.

[0021] The labels in the diagram represent: 1-Fuselage body, 11-Anti-collision frame, 12-Supporting structure, 2-Power system, 21-Motor, 211-First motor, 212-Second motor, 213-Third motor, 214-Fourth motor, 215-Fifth motor, 216-Sixth motor, 22-Propeller, 221-First propeller, 222-Second propeller, 223-Third propeller, 224-Fourth propeller, 225-Fifth propeller, 226-Sixth propeller, 3-Data acquisition and control module, 4-Camera, 5-Fill light, 51-First fill light, 52-Second fill light, 6-Vision sensor, 61-First vision sensor, 62-Second vision sensor. Detailed Implementation

[0022] The present invention will be further described below with reference to embodiments.

[0023] Achieving a balance between high payload capacity, long endurance, high reliability, and modular functionality in unmanned aerial vehicles (UAVs) while meeting the dimensional constraints of confined spaces has become a critical technical challenge in this field. Specifically: First, it is difficult to balance payload and endurance. The confined space imposes strict limitations on the external dimensions of the drone's fuselage. Traditional multi-rotor drones employ a single-planar layout, and within this limited size, the total area of ​​the rotor disk is difficult to increase, resulting in low aerodynamic efficiency. To adapt to space requirements, existing solutions often employ miniaturized designs and lightweight sensors. While this meets the size constraints, it sacrifices the load-bearing capacity required for industrial-grade sensors. Furthermore, the limited rotor output efficiency significantly shortens endurance, making it impossible to complete long-distance, long-duration confined space operations.

[0024] Secondly, the reliability of the power system is insufficient. The complex internal structure of confined spaces makes drones prone to collisions with obstacles during flight, potentially damaging power units such as motors or propellers. Traditional multi-rotor drones lack effective power redundancy design; if a single power unit fails, it will directly lead to loss of aircraft attitude control, increasing the risk of a crash. This not only damages equipment but may also affect operational safety in confined spaces, especially in high-risk, enclosed environments such as tunnels and mines, where this deficiency is even more pronounced.

[0025] Third, poor functional expandability. Most existing multi-rotor drones are enclosed flight platforms with fixed sensor types, lacking modular design to meet the diverse operational needs of confined spaces. When facing different tasks such as gas detection, multispectral scanning, and 3D mapping, rapid sensor replacement and functional adaptation are not possible, requiring customized drone platforms for specific tasks. This not only increases the user's equipment investment costs but also reduces the equipment's versatility and operational flexibility.

[0026] To overcome the aforementioned drawbacks, the present invention employs the following embodiments to address the current situation.

[0027] Example 1: Reference Appendix Figure 1 This is a schematic diagram of an unmanned aerial vehicle (UAV) device used for monitoring confined spaces, which includes: The main body 1 is composed of a crash protection frame 11 and a load-bearing structure 12; wherein, the crash protection frame 11 is a double-layer frame structure formed by a bracket; and the load-bearing structure 12 is fixed to the upper area of ​​the crash protection frame 11. The power system 2 adopts a six-rotor layout with overlapping and interleaved rotors, including first to sixth motors 211-216 and first to sixth propellers 221-226 respectively fixed on the shafts of each motor; the first to sixth motors 211-216 are arranged in two layers in the lower part of the load-bearing structure 12 of the fuselage body 1, wherein the first motor 211 and the second motor 212 are located in the upper layer, and the third to sixth motors 213-216 are located in the lower layer; The data acquisition and control module 3 integrates a flight control computer, an airborne computing unit, and a GPS module 4. The flight control computer is used to receive detection data from the perception and navigation system to achieve autonomous positioning, path planning, dynamic obstacle avoidance, and control of the power system. The data acquisition and control module is also equipped with a standardized secondary development interface for mounting external dedicated sensors.

[0028] Through the combined design of upper and lower double-layer anti-collision frames and upper load-bearing structure, on the one hand, the double-layer anti-collision frames form a full-enclosed protection to avoid damage to core components such as propellers and motors caused by collisions with obstacles during flight, adapting to the complex and narrow flight environment of confined spaces; on the other hand, the load-bearing structure enables the orderly fixation of various functional components, ensuring the rational layout of components such as the power system, battery, data acquisition and control module, taking into account the stability of the fuselage structure and the balance of the center of gravity, and providing structural support for precise hovering and flexible turning.

[0029] Employing a layered, overlapping six-rotor layout with dual motors on the upper layer and four motors on the lower layer, this design maximizes the total rotor disk area without increasing the fuselage's overall dimensions, thereby improving aerodynamic efficiency and achieving a balance between heavy payload and long endurance. This also meets the requirements for mounting industrial-grade sensors. Furthermore, leveraging the power redundancy of the six-rotor configuration, if any motor or propeller fails, the remaining power units can maintain attitude stability through flight control adjustments, preventing loss of control and significantly enhancing the reliability and safety of operations in confined spaces.

[0030] The data acquisition and control module integrates a flight control computer, an onboard computing unit, and a GPS module, enabling autonomous control in GPS-free environments: receiving perception data from LiDAR, visual sensors, etc., and completing autonomous positioning, path planning, and dynamic obstacle avoidance, adapting to scenarios where satellite signals are lacking in confined spaces; at the same time, through standardized secondary development interfaces, it supports the flexible mounting of external dedicated sensors, breaking the limitations of traditional UAVs with fixed functions, and realizing multi-task adaptation such as 3D mapping, environmental monitoring, and equipment inspection, improving equipment versatility and operational flexibility.

[0031] In another embodiment, the secondary development interface of the data acquisition and control module 3 is a standardized serial port, CAN interface or Ethernet interface, which supports the access and data interaction of one or more external devices such as gas sensors, infrared thermal imagers and multispectral sensors.

[0032] The secondary development interface uses industry-standard serial, CAN, or Ethernet interfaces, leveraging the mature communication protocols of these three types of interfaces to achieve plug-and-play adaptation with different types of external sensors. This eliminates the need to modify the drone's airframe structure or core control logic, lowering the technical barrier to sensor replacement, ensuring the efficiency and stability of the adaptation process, and solving the problems of poor compatibility and high adaptation costs associated with traditional non-standard interfaces.

[0033] In another embodiment, the drone equipment includes a fuselage 1, a power system 2, a data acquisition and control module 3, a camera 4, a fill light 5, and a vision sensor 6; the load-bearing structure 12 is fixed to the upper area of ​​the anti-collision frame 11; the first to sixth motors 211-216 of the power system 2 are arranged on the load-bearing structure 12; the first vision sensor 61 and the second vision sensor 62 of the vision sensor 6 are symmetrically arranged on one side of the data acquisition and control module 3; the camera 4 is fixed to one side of the load-bearing structure 12, and the first fill light 51 and the second fill light 52 of the fill light 5 are symmetrically installed on both sides of the camera 4.

[0034] like Figure 2 As shown, in one embodiment, a camera 4 is also included. The camera 4 is fixedly connected to one side of the load-bearing structure 12 and electrically connected to the data acquisition and control module 3 through the standardized interface.

[0035] The installation angle and field of view of Camera 4 have been optimized to cover the core area of ​​the drone's flight path and operation zone, adapting to diverse monitoring needs in confined spaces. In 3D mapping tasks, it can simultaneously acquire environmental texture images and fuse them with LiDAR point cloud data to generate a high-precision 3D model with texture. In equipment inspection tasks, it can clearly capture detailed information such as equipment appearance defects and component status. In scenarios such as tunnel inspection, it can record spatial structural features, providing intuitive image evidence for subsequent analysis, completely solving the problems of limited field of view and insufficient data targeting of traditional drone cameras.

[0036] like Figure 3 As shown, in one embodiment, it also includes a fill light 5, which consists of a first fill light 51 and a second fill light 52; the first fill light 51 and the second fill light 52 are symmetrically installed on both sides of the camera 4, and the two are electrically connected to the data acquisition and control module 3 through the standardized interface.

[0037] The supplementary lights 5 are installed on both sides of the camera, forming a symmetrical supplementary lighting layout, which can provide uniform and shadow-free lighting coverage to the camera's field of view. In dimly lit enclosed spaces such as tunnels and underground warehouses, it can effectively improve the ambient brightness and avoid problems such as increased image noise, blurred texture, and loss of detail caused by insufficient light. At the same time, the supplementary light intensity can be dynamically adjusted by the data acquisition and control module according to the ambient lighting conditions, which can meet the brightness requirements of image acquisition while avoiding overexposure caused by strong light reflection, adapting to enclosed space operation scenarios with different lighting intensities.

[0038] like Figure 4 As shown, in one embodiment, it further includes: The vision sensor 6 includes a first vision sensor 61 and a second vision sensor 62; the first vision sensor 61 and the second vision sensor 62 are respectively symmetrically mounted on one side of the data acquisition and control module 3, and the two are electrically connected to the data acquisition and control module 3 through the standardized interface.

[0039] The visual sensor 6 is symmetrically composed of a first visual sensor 61 and a second visual sensor 62, which are mounted symmetrically on one side of the data acquisition and control module, forming a complementary visual perception layout. This design can expand the field of view for environmental detection and effectively reduce the blind spots of a single sensor, making it particularly suitable for working environments with dense obstacles and complex structures in confined spaces. The dual sensors simultaneously acquire image data, capturing different angular features of the same target and achieving coordinated close-range detail detection and mid-range environmental scanning. For example, the distance to the target can be calculated using stereo vision principles, with a measurement accuracy of ≤±3cm. This provides multi-dimensional visual evidence for obstacle detection and spatial structure recognition, solving the problems of incomplete or inaccurate measurements caused by traditional single-vision sensors.

[0040] As attached Figure 5 In one embodiment, the following steps are included: S1. In a confined space without satellite navigation signals, activate the airborne lidar and visual sensors, and selectively activate the supplementary lighting device based on ambient lighting conditions; S2. By fusing the perception data from the lidar and the vision sensor, the pose of the UAV is calculated in real time, and a three-dimensional environmental map of the enclosed space is constructed simultaneously. S3. Based on the aforementioned three-dimensional environment map and the preset mission objectives, plan the flight path; S4. When executing the flight path, dynamic obstacle detection and avoidance are performed based on the real-time perception data of the lidar and the vision sensor, and the flight trajectory is adjusted. S5. Control the UAV to fly along the adjusted flight path and perform monitoring tasks through the airborne mission sensors, while transmitting monitoring data back; S6. After the monitoring task is completed, the return path is planned based on the real-time pose, and the UAV is controlled to return to the starting point and land.

[0041] In one embodiment, the step of calculating the UAV's pose in real time and simultaneously constructing a three-dimensional environmental map of the enclosed space by fusing the perception data from the lidar and the vision sensor includes: The raw point cloud data collected by the lidar is preprocessed to obtain lidar point cloud data, and the image data collected by the vision sensor is corrected. Three-dimensional geometric features are extracted from the preprocessed lidar point cloud data to generate point cloud feature descriptors. At the same time, two-dimensional image feature points and corresponding feature descriptors are extracted from the corrected image data. Based on the point cloud feature descriptors and the similarity measure of the feature descriptors, a random sampling consensus algorithm is used to perform cross-modal matching between the three-dimensional geometric features of the lidar and the two-dimensional geometric features of the vision, thereby establishing the correlation between multi-sensor data.

[0042] Using feature matching results as constraints, a vision-LiDAR fusion positioning equation is constructed. Combined with short-term attitude data from the UAV's IMU, an extended Kalman filter algorithm is used to fuse multi-source data in real time. The position coordinates and attitude angles of the UAV are solved by iterative calculation to perform pose estimation in a GPS-free environment.

[0043] Based on the real-time calculated UAV pose, the lidar point cloud data is transformed to the global coordinate system through coordinate transformation; the voxel grid algorithm is used to perform spatial partitioning and density optimization of the global point cloud to remove redundant point cloud data; By combining the texture information extracted from the visual image, the image texture is mapped onto the corresponding 3D point cloud surface to generate a textured 3D environment map. In one embodiment, the step of planning a flight path based on the three-dimensional environment map and a preset mission objective includes: The pre-set task objectives are broken down into structures, clarifying the task type, operation coverage, key monitoring areas, data acquisition accuracy requirements, and task priority. Among them, the spatial boundary of the full coverage scan needs to be marked for the 3D mapping task, the location coordinates or feature marks of the equipment to be inspected need to be specified for the equipment to be inspected, and the sampling point distribution density of parameters such as gas, temperature, and humidity needs to be set for the environmental monitoring task.

[0044] Based on the constructed 3D environment map, the flyable area, obstacle area, and buffer zone are divided using a voxel segmentation algorithm. At the same time, combined with the UAV's dynamic characteristics and the size constraints of the confined space, flight parameter thresholds are set, including maximum flight speed ≤ 5m / s, minimum turning radius ≤ 0.3m, flight altitude and distance from obstacle ≥ 0.5m, and the operating altitude range corresponding to the effective detection distance of the sensor.

[0045] An improved fast exploration random tree-star algorithm is adopted, which uses the task start point as the starting point and the task end point or key sampling point as the target point to search for the optimal path in the flightable area. During the path generation process, obstacle buffer zones are avoided first, and the distribution of path nodes is optimized according to the task type. The 3D mapping task adopts a gridded uniform point distribution.

[0046] The initial path is smoothed and optimized for energy consumption: sharp inflection points in the path are eliminated by B-spline curve fitting, making the flight trajectory continuous and differentiable, and reducing the energy consumption for attitude adjustment; based on the energy consumption model of the UAV power system, redundant turn-off segments in the path are eliminated, shortening the total flight distance and extending the endurance time; the curvature change rate of the optimized path is ≤0.2rad / m, and the total path length is shortened by ≥10% compared with the initial path.

[0047] The optimized path is subjected to collision detection with the 3D environment map. The distance between all nodes and obstacles on the path is verified by point cloud distance calculation to see if the distance meets the preset constraints. At the same time, the attitude change of the UAV flying along the path is simulated to verify whether the turning radius, flight speed and other parameters are within the range of the power system. If there is a collision risk or the constraints are not met, the process returns to step 3 to regenerate the path.

[0048] Preset path update trigger conditions include incremental updates of the 3D environment map and adjustments to task objectives. When the trigger conditions are met, based on the updated map data or task information, steps 2-5 are repeated to quickly generate a corrected path, ensuring that the path matches the operational requirements and environmental conditions in real time.

[0049] In one embodiment, the step of dynamically detecting and avoiding obstacles and adjusting the flight trajectory based on real-time perception data from the lidar and the vision sensor includes: The lidar and dual vision sensors synchronously collect environmental data at a frequency of no less than 10Hz. The lidar outputs a three-dimensional point cloud data stream, and the vision sensors output continuous image frames. Through the timestamp synchronization mechanism of the data acquisition and control module, the time deviation between the two types of data is corrected to ≤10ms to avoid misjudgment of obstacle positions due to asynchronous data.

[0050] Real-time clustering analysis is performed on the synchronized lidar point cloud data to extract the 3D contours, spatial coordinates, and distance information of obstacles, and to initially screen out dynamic targets that deviate from the 3D environment map; moving target detection is performed on the image frames of the visual sensor to identify the contour features and movement trends of dynamic obstacles; the detection results of the two types of sensors are fused by the Kalman filter algorithm to eliminate false targets and determine the true position, size, and motion parameters of dynamic obstacles, with a detection accuracy of ≥95%.

[0051] Based on the detected obstacle parameters, combined with the UAV's current flight speed ≤5m / s, maneuverability, minimum turning radius ≤0.3m, and confined space size constraints, a risk assessment model is constructed: High risk: The distance between the obstacle and the drone is less than 0.5m, or the moving speed is greater than 1m / s, and the direction of movement is towards the drone; Medium risk: Obstacles are 0.5-1.0m away from the drone, and the drone is moving at a speed of 0.3-1m / s; Low risk: The distance between the obstacle and the drone is >1.0m, the moving speed is <0.3m / s or it is stationary; the obstacle type is identified simultaneously to provide a basis for avoidance strategy.

[0052] Adaptive selection of avoidance strategy based on risk level: High-risk scenarios: Activate emergency avoidance mode, quickly adjust motor speed through the power redundancy characteristics of the six-rotor to achieve lateral movement, lifting or sharp turning, avoidance response time ≤0.5s, and ensure that the distance between the avoidance and the obstacle is ≥0.3m after avoidance; For low- to medium-risk scenarios: An improved local A* algorithm is used to generate the shortest transition trajectory under preset safety constraints, avoiding significant deviations from the original operation path; the replanned trajectory must simultaneously meet the power output limits of the UAV to ensure attitude stability during the adjustment process.

[0053] The flight control computer converts the replanned trajectory into power system control commands. By adjusting the speed difference between the upper and lower six rotors, a smooth transition of flight attitude is achieved. At the same time, it integrates real-time attitude data from the IMU and uses a PID control algorithm to correct flight deviations, avoiding fuselage shaking or instability caused by avoidance maneuvers, and ensuring that the adjusted trajectory tracking error is ≤±3cm.

[0054] After the avoidance maneuver is completed, the shortest return trajectory is planned based on the 3D environment map and the original mission path, so that the drone gradually returns to the preset operation path. During this process, the obstacle status is continuously monitored through LiDAR and visual sensors. If the obstacle's movement trajectory changes, steps 2-5 are repeated to make dynamic adjustments to ensure that there is no risk of collision throughout the process.

[0055] In one embodiment, the step of controlling the UAV to fly along an adjusted flight path, performing monitoring tasks via onboard mission sensors, and simultaneously transmitting monitoring data back includes: Based on the preset monitoring task type, the corresponding airborne task sensors are activated through the standardized interface of the data acquisition and control module. The sensor operating parameters are configured according to the task requirements. In the 3D mapping task, the lidar sampling frequency is set to 10-50Hz, the point cloud density is ≥500 points / ㎡, the camera resolution is ≥4K, and the frame rate is ≥30fps. In the environmental monitoring task, the gas sensor sampling interval is ≤1s, the detection accuracy is ≤±5% FS, the temperature and humidity sensor sampling interval is ≤0.5s, and the error is ≤±0.5℃ / ±2% RH. In the equipment inspection task, the infrared thermal imager temperature measurement range is -20℃~150℃, the temperature measurement accuracy is ≤±2%, and the multispectral sensor has ≥4 spectral bands.

[0056] Based on the adjusted flight trajectory, the flight control computer combines real-time attitude data from lidar and visual sensors, and uses a PID control algorithm to adjust the speed difference between the upper and lower six rotors to achieve precise tracking of flight attitude and trajectory. During the control process, it corrects trajectory deviations caused by airflow disturbances and load changes in real time, ensuring that the distance error between the UAV and the preset trajectory is ≤±3cm and the flight speed fluctuation is ≤±0.2m / s, providing attitude support for stable data acquisition by the sensors.

[0057] Adaptive data collection based on the enclosed space environment and mission requirements: In dimly lit scenes, the brightness level of the fill light is automatically adjusted to ensure that the images captured by the camera and vision sensor have clear textures and are free from shadow interference. In a 3D mapping scenario, LiDAR and cameras simultaneously acquire 3D point cloud data and 2D texture images, ensuring complete coverage of point cloud data and an image overlap rate of ≥60%. In environmental monitoring scenarios, sensors continuously collect target parameters and record the three-dimensional coordinates of the collection points; In equipment inspection scenarios, visual sensors and cameras work together to focus on key parts of the equipment under inspection, capturing detailed images or videos with magnification of ≥2 times and no distortion.

[0058] The airborne computing unit performs real-time preprocessing on the raw data collected by the sensors. It performs noise reduction, downsampling, and coordinate calibration on the lidar point cloud data, eliminating noise points with a proportion of ≤3%. It performs distortion correction, white balance optimization, and edge enhancement on the image data to improve feature recognition. It filters and removes jitter from the environmental sensor data. It performs temperature calibration and pseudo-color encoding on the infrared thermal image data to generate an intuitive temperature distribution map. The effectiveness of the preprocessed data is ≥98%.

[0059] The data acquisition and control module uses a timestamp synchronization mechanism to align preprocessed sensor data, UAV real-time pose data, and flight status data with a time synchronization error of ≤10ms. The synchronized data is then encapsulated according to a preset standard format, including key information such as data type identifier, acquisition timestamp, 3D coordinates of acquisition points, sensor parameters, and data verification codes, ensuring data integrity and traceability.

[0060] Monitoring data is processed using a dual-mode approach: real-time data transmission and local storage. Real-time transmission: Encapsulated data is transmitted to the ground control terminal via the airborne wireless communication module. The transmission rate is adaptively adjusted according to the signal attenuation in the confined space to ensure that the data transmission packet loss rate is ≤1%. In areas without wireless signal coverage, the temporarily stored data is automatically retransmitted after the signal is restored. The raw data and preprocessed data are stored synchronously on the onboard high-speed storage medium, and a cyclic overwrite strategy is adopted to prevent data loss due to interruption of real-time transmission.

[0061] The airborne computing unit verifies the quality of transmitted and stored data in real time. Verification indicators include whether the data resolution, integrity, and accuracy meet preset thresholds. If data loss, inaccuracy, or omissions are detected, a supplementary acquisition mechanism is automatically triggered. The flight control system adjusts the flight trajectory, causing the UAV to return to the target area and re-execute the acquisition task. During the supplementary acquisition process, the sensor operating parameters remain consistent with the original acquisition, ensuring the compatibility of the supplementary data with the original data.

[0062] The system provides real-time statistical monitoring of task completion and feeds back task progress, sensor operating status, and data transmission status to the ground control terminal at a frequency of ≥1 time / 5s, facilitating real-time monitoring of the operation process and timely intervention by operators.

[0063] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the present invention.

Claims

1. A drone equipment for monitoring confined spaces, characterized in that, include: The fuselage body (1) is composed of a crash protection frame (11) and a load-bearing structure (12); wherein, the crash protection frame (11) is a double-layer frame structure formed by a bracket; the load-bearing structure (12) is fixed in the upper area of ​​the crash protection frame (11); The power system (2) adopts a six-rotor layout with overlapping and interlacing upper and lower rotors, including the first to sixth motors (211-216) and the first to sixth propellers (221-226) respectively fixed on the shafts of each motor; the first to sixth motors (211-216) are arranged in two layers in the lower part of the load-bearing structure (12) of the fuselage body (1), wherein the first motor (211) and the second motor (212) are located in the upper layer, and the third to sixth motors (213-216) are located in the lower layer; The data acquisition and control module (3) integrates a flight control computer, an airborne computing unit and a GPS module (4). The flight control computer is used to receive detection data from the perception and navigation system.

2. The unmanned aerial vehicle (UAV) equipment for monitoring confined spaces according to claim 1, characterized in that, The UAV equipment includes a fuselage (1), a power system (2), a data acquisition and control module (3), a camera (4), a fill light (5), and a vision sensor (6); the load-bearing structure (12) is fixed to the upper area of ​​the anti-collision frame (11); the first to sixth motors (211-216) of the power system (2) are arranged on the load-bearing structure (12); the first vision sensor (61) and the second vision sensor (62) of the vision sensor (6) are symmetrically arranged on one side of the data acquisition and control module (3); the camera (4) is fixed to one side of the load-bearing structure (12), and the first fill light (51) and the second fill light (52) of the fill light (5) are symmetrically installed on both sides of the camera (4).

3. The unmanned aerial vehicle (UAV) equipment for monitoring confined spaces according to claim 1, characterized in that, It also includes a camera (4), which is fixedly connected to one side of the load-bearing structure (12) and electrically connected to the data acquisition and control module (3) through the standardized interface.

4. The unmanned aerial vehicle (UAV) equipment for monitoring confined spaces according to claim 1, characterized in that, It also includes a fill light (5), which consists of a first fill light (51) and a second fill light (52); the first fill light (51) and the second fill light (52) are symmetrically installed on both sides of the camera (4), and the two are electrically connected to the data acquisition and control module (3) through the standardized interface.

5. The unmanned aerial vehicle (UAV) equipment for monitoring confined spaces according to claim 1, characterized in that, It also includes the vision sensor (6), which includes a first vision sensor (61) and a second vision sensor (62); the first vision sensor (61) and the second vision sensor (62) are respectively symmetrically installed on one side of the data acquisition and control module (3), and the two are electrically connected to the data acquisition and control module (3) through the standardized interface.

6. An autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces, characterized in that... The method of using the unmanned aerial vehicle (UAV) equipment for monitoring confined spaces as described in any one of claims 1-5 includes the following steps: In a confined space without satellite navigation signals, the airborne lidar and visual sensors are activated, and the supplementary lighting device is selectively activated based on the ambient lighting conditions. By fusing the perception data from the lidar and the vision sensor, the drone's pose is calculated in real time, and a three-dimensional environmental map of the enclosed space is constructed simultaneously. Based on the aforementioned 3D environment map and the preset mission objectives, a flight path is planned; When executing the flight path, dynamic obstacle detection and avoidance are performed based on the real-time perception data of the lidar and the vision sensor, and the flight trajectory is adjusted accordingly; The drone is controlled to fly along the adjusted flight path and to perform monitoring tasks through onboard mission sensors, while simultaneously transmitting monitoring data back. After the monitoring task is completed, the return path is planned based on the real-time pose, and the UAV is controlled to return to the starting point and land.

7. The autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces according to claim 6, characterized in that, The step of calculating the UAV's pose in real time and simultaneously constructing a 3D environmental map of the enclosed space by fusing the perception data from the lidar and the vision sensor includes: The raw point cloud data collected by the lidar is preprocessed to obtain lidar point cloud data, and the image data collected by the vision sensor is corrected. Three-dimensional geometric features are extracted from the preprocessed lidar point cloud data to generate point cloud feature descriptors. At the same time, two-dimensional image feature points and corresponding feature descriptors are extracted from the corrected image data. Based on the point cloud feature descriptor and the similarity measure of the feature descriptor, the cross-modal matching of the three-dimensional geometric features of the lidar and the two-dimensional geometric features of the vision is performed by the random sampling consensus algorithm to establish the correlation relationship of multi-sensor data; Using feature matching results as constraints, a vision-LiDAR fusion positioning equation is constructed. Combined with short-term attitude data from the UAV IMU, an extended Kalman filter algorithm is used to fuse multi-source data in real time. The position coordinates and attitude angles of the UAV are solved by iterative calculation to perform pose estimation in a GPS-free environment; Based on the real-time calculated UAV pose, the lidar point cloud data is transformed to the global coordinate system through coordinate transformation; A voxel grid algorithm is used to spatially partition and optimize the density of the global point cloud, removing redundant point cloud data. By combining texture information extracted from visual images, the image texture is mapped onto the corresponding 3D point cloud surface to generate a textured 3D environment map.

8. The autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces according to claim 6, characterized in that, The step of planning the flight path based on the aforementioned 3D environment map and the preset mission objective includes: The pre-set task objectives are broken down into structures, clarifying the task type, operation coverage, key monitoring areas, data acquisition accuracy requirements, and task priority; among them, the spatial boundaries of the full-coverage scan must be marked for the 3D mapping task, the location coordinates or feature markers of the equipment to be inspected must be specified for the equipment to be inspected, and the sampling point distribution density of parameters such as gas, temperature, and humidity must be set for the environmental monitoring task. Based on the constructed 3D environment map, the flyable area, obstacle area and buffer zone are divided by voxel segmentation algorithm; Starting from the mission start point and targeting the mission end point or key sampling point, the optimal path is searched within the flyable area. During path generation, obstacle buffer zones are avoided first. At the same time, the distribution of path nodes is optimized according to the mission type. 3D mapping tasks use a gridded uniform point distribution. Collision detection is performed between the optimized path and the 3D environment map, and the distance between all nodes and obstacles on the path is verified by point cloud distance calculation to see if the distance between them meets the preset constraints. Preset path update trigger conditions include incremental updates of the 3D environment map and adjustments to the task objective; when the trigger conditions are met, map data or task information is updated.

9. The autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces according to claim 6, characterized in that, The step of dynamically detecting and avoiding obstacles and adjusting the flight trajectory based on real-time perception data from the lidar and the vision sensor includes: The lidar and dual vision sensors synchronously collect environmental data at a preset frequency. The lidar outputs a three-dimensional point cloud data stream, and the vision sensors output continuous image frames. The three-dimensional contours, spatial coordinates, and distance information of obstacles are extracted to initially screen dynamic targets that deviate from the three-dimensional environment map. Moving target detection is performed on the image frames of the visual sensor to identify the contour features and movement trends of dynamic obstacles. The detection results of the LiDAR and dual visual sensors are then fused. Based on the detected obstacle parameters, a risk assessment model is constructed, and the flight trajectory is adjusted using the risk assessment model.

10. The autonomous operation method for unmanned aerial vehicle (UAV) equipment used for monitoring confined spaces according to claim 6, characterized in that, The steps of controlling the UAV to fly along the adjusted flight path, performing monitoring tasks through onboard mission sensors, and transmitting monitoring data back include: Based on the adjusted flight trajectory, the flight control computer combines real-time attitude data from lidar and vision sensors, and uses a PID control algorithm to adjust the speed difference between the upper and lower six rotors. During the control process, it corrects trajectory deviations caused by airflow disturbances and load changes in real time.

Images