A method and system for managing inspection data of a drone

Abstract

The application relates to an unmanned aerial vehicle (UAV) inspection data management method and system, and relates to the field of data management.The method comprises the following steps: acquiring first on-site image and first point cloud data collected by the UAV at an inspection starting moment, and acquiring second on-site image and second point cloud data collected after a preset time interval; extracting a first object and a second object, and obtaining first object attributes of the first object and second object attributes of the second object; obtaining an inspection scheme of the UAV according to the first object attributes and the second object attributes; the UAV executes the inspection scheme to collect inspection data; and the inspection data is encapsulated according to object categories and time stamps, and is uploaded to a data management terminal.Through the above method, the utilization efficiency of the UAV and the data collection quality are improved, and accurate management of the inspection data is realized.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and in particular to a method and system for managing inspection data from unmanned aerial vehicles (UAVs). Background Technology

[0002] Traditional drone inspection systems typically use a uniform perception model to process all scene objects, failing to effectively distinguish between moving and stationary objects. In complex inspection environments, moving people or vehicles may cause motion blur, occlusion, or false triggering of recognition algorithms, interfering with the accurate identification of key stationary facilities (such as equipment nameplates, warning signs, and building structures). This results in a significant amount of manpower being required for data alignment and organization during the data management phase, leading to low data management efficiency and making it difficult to support intelligent analysis and long-term trend tracking. Summary of the Invention

[0003] This invention addresses the technical problems existing in the prior art by providing a method and system for managing inspection data of unmanned aerial vehicles (UAVs).

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0005] This invention provides a method for managing inspection data from unmanned aerial vehicles (UAVs), the method comprising:

[0006] Acquire the first on-site image and first point cloud data collected by the drone at the start of the inspection, and the second on-site image and second point cloud data collected after a preset time interval;

[0007] Extract the first object and the second object, and obtain the first object attribute of the first object and the second object attribute of the second object;

[0008] Based on the first object attribute and the second object attribute, an inspection plan for the UAV is obtained;

[0009] The drone executes the inspection plan and collects inspection data.

[0010] The inspection data is packaged according to object category and timestamp, and uploaded to the data management terminal. Multi-angle images of the same small object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0011] Optionally, the first object includes a moving object, and the second object includes a stationary object.

[0012] Optionally, the extraction of the first object and the second object includes:

[0013] Based on the first scene image, the first point cloud data, the second scene image, and the second point cloud data, a motion saliency map is generated, wherein regions with significant displacement or deformation are marked as potential first object regions, and the remaining regions are marked as potential second object regions.

[0014] The potential first object region and the potential second object region are input into a two-stream convolutional neural network, wherein the first stream channel of the two-stream convolutional neural network is used to process the potential first object region to extract the first object, and the second stream channel is used to process the potential second object region to extract the second object.

[0015] Optionally, generating a motion saliency map based on the first scene image, the first point cloud data, the second scene image, and the second point cloud data includes:

[0016] Dense optical flow field estimation is performed on the first and second scene images to generate a first motion confidence map;

[0017] The first point cloud data and the second point cloud data are respectively voxelized into three-dimensional voxel meshes to generate a second motion confidence map;

[0018] The first motion confidence map and the second motion confidence map are aligned by cross-modal projection to generate a fused motion saliency score map;

[0019] Connectivity analysis and morphological closing operations are performed on the fused motion saliency score map to segment out independent potential first object regions and output motion saliency maps.

[0020] Optionally, the step of performing dense optical flow field estimation on the first scene image and the second scene image to generate a first motion confidence map includes:

[0021] Dense optical flow field estimation is performed on the first and second scene images to obtain pixel-level displacement vector fields;

[0022] By verifying the consistency of forward and backward optical flow, reliable optical flow regions of the pixel-level displacement vector field are selected, and a first motion confidence map is generated.

[0023] Optionally, the step of voxelizing the first point cloud data and the second point cloud data into a three-dimensional voxel mesh to generate a second motion confidence map includes:

[0024] The first point cloud data and the second point cloud data are voxelized into three-dimensional voxel meshes respectively, and registered based on the ICP algorithm to obtain the registration matrix;

[0025] The second point cloud data is spatially transformed using the registration matrix to obtain an aligned voxel grid;

[0026] Calculate the Euclidean distance difference between the three-dimensional voxel grid of the first point cloud data and the aligned voxel grid;

[0027] Based on the Euclidean distance difference, a voxel-level motion confidence score is calculated to generate a second motion confidence map.

[0028] Optionally, obtaining the first object attribute of the first object and the second object attribute of the second object includes:

[0029] The first object is subjected to feature extraction to obtain a first feature; the first feature is input into a neural network model to output the first object attribute;

[0030] The second object is subjected to feature extraction to obtain the second feature; the second feature is input into the neural network model to output the second object attribute.

[0031] Optionally, obtaining the UAV inspection plan based on the first object attribute and the second object attribute includes:

[0032] Calculate the imaging priority weight of each object based on the first object attribute and the second object attribute;

[0033] Based on the imaging priority weights, a multi-priority imaging task queue is generated, and imaging resources are allocated to the multi-camera system of each UAV.

[0034] Based on the allocated imaging resources, a drone inspection scheme is constructed.

[0035] Optionally, the step of generating a multi-priority imaging task queue based on the imaging priority weights and allocating imaging resources for the multi-camera system of each UAV includes:

[0036] Objects whose imaging priority weight is higher than the first threshold are identified as high-priority objects, and high-magnification zoom cameras are scheduled to perform multi-angle zoom imaging; wherein, multi-angle imaging is achieved by N azimuth angles around the center of the object, and zoom imaging is achieved by dynamically adjusting the focal length according to the distance between the object and the drone and the size of the characters.

[0037] Objects whose imaging priority weight is lower than the first threshold and higher than the second threshold are identified as medium priority objects, and medium-focus cameras or high-magnification zoom cameras are scheduled to perform single or double imaging at a fixed angle; wherein, medium focal length and standard exposure parameters are set for single or double fixed-focus imaging.

[0038] Objects whose imaging priority weight is lower than the second threshold are identified as low-priority objects, and wide-angle cameras are scheduled to perform single-shot fixed-focus imaging; wherein, a fixed focal length and exposure parameters are set to achieve single-shot fixed-focus imaging.

[0039] This invention also provides an inspection data management system for unmanned aerial vehicles (UAVs), the system comprising:

[0040] The data acquisition module is used to acquire the first on-site image and the first point cloud data collected by the UAV at the start of the inspection, as well as the second on-site image and the second point cloud data collected after a preset time interval.

[0041] The attribute analysis module is used to extract the first object and the second object, and to obtain the first object attribute of the first object and the second object attribute of the second object;

[0042] The scheme generation module is used to obtain the inspection scheme of the UAV based on the attributes of the first object and the attributes of the second object;

[0043] The inspection module is used by the UAV to collect inspection data when executing the inspection plan;

[0044] The data management module is used to encapsulate the inspection data according to object category and timestamp, and upload it to the data management terminal. Among them, multi-angle images of the same small object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0045] In addition, to achieve the above objectives, the present invention also proposes an electronic device, comprising: a memory for storing computer software programs; and a processor for reading and executing the computer software programs, thereby realizing the UAV inspection data management method described above.

[0046] In addition, to achieve the above objectives, the present invention also proposes a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements the UAV inspection data management method described above.

[0047] The beneficial effects of this invention are:

[0048] (1) This invention acquires a first scene image and first point cloud data collected by a UAV at the start of an inspection, and a second scene image and second point cloud data collected after a preset time interval; extracts a first object and a second object, and obtains a first object attribute of the first object and a second object attribute of the second object; obtains an inspection plan for the UAV based on the first object attribute and the second object attribute; the UAV executes the inspection plan to collect inspection data; the inspection data is packaged according to object category and timestamp, and uploaded to a data management terminal, wherein multi-angle images of the same small-sized object are packaged into a data unit to support subsequent 3D reconstruction or character recognition. This improves the utilization efficiency and data acquisition quality of the UAV, thereby achieving precise management of inspection data.

[0049] (2) By combining motion saliency maps and dual-stream convolutional neural networks, the first moving object (such as a person or vehicle) and the second stationary object (such as a building) can be effectively separated, avoiding dynamic interference and improving the accuracy of static small target recognition.

[0050] (3) Calculate imaging priority weights based on object attributes and generate differentiated inspection schemes to realize intelligent scheduling of UAV camera modules and improve energy utilization efficiency and data acquisition quality. Attached Figure Description

[0051] Figure 1 A scenario diagram illustrating a drone inspection data management method provided by this invention;

[0052] Figure 2 A flowchart of a drone inspection data management method provided by the present invention;

[0053] Figure 3 A schematic diagram of the structure of an inspection data management system for unmanned aerial vehicles provided by the present invention;

[0054] Figure 4 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0055] Figure 5 This is a schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0058] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0059] Please see Figure 1 , Figure 1 This invention provides a scenario diagram for a drone inspection data management method. (Example) Figure 1 As shown, the terminal and server are connected via a network, such as a wired or wireless network. The terminal can include, but is not limited to, portable devices such as smartwatches, mobile phones, and tablets with network platform applications installed, as well as fixed terminals such as computers, kiosks, and advertising machines. The server provides users with various business services, including service push servers and user recommendation servers.

[0060] It should be noted that, Figure 1 The scenario diagram of the UAV inspection data management method shown is merely an example. The terminals, servers, and application scenarios described in the embodiments of the present invention are for the purpose of more clearly illustrating the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. As those skilled in the art will know, with the evolution of the system and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0061] The terminal has an application installed that can be used to perform the following steps:

[0062] Acquire the first on-site image and first point cloud data collected by the drone at the start of the inspection, and the second on-site image and second point cloud data collected after a preset time interval;

[0063] Extract the first object and the second object, and obtain the first object attribute of the first object and the second object attribute of the second object;

[0064] Based on the first object attribute and the second object attribute, an inspection plan for the UAV is obtained;

[0065] The drone executes the inspection plan and collects inspection data.

[0066] The inspection data is packaged according to object category and timestamp, and uploaded to the data management terminal. Multi-angle images of the same small object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0067] Please see Figure 2 The present invention provides a flowchart of a method for managing inspection data of a drone.

[0068] like Figure 2 As shown in the embodiment of the present invention, the method for managing inspection data of unmanned aerial vehicles includes:

[0069] 201. Acquire the first on-site image and first point cloud data collected by the UAV at the start of the inspection, and the second on-site image and second point cloud data collected after a preset time interval.

[0070] Specifically, at the start time of the drone's inspection mission The multi-sensor synchronous acquisition module is activated to acquire the first scene image and the first point cloud data. After completing the initial data acquisition, the UAV continues to fly along the preset route or hovers in standby. After a preset time interval... In the second moment This triggers multi-sensor synchronous data acquisition again. The preset time interval... It can be dynamically configured according to the application scenario: for power facility inspection, Δt=60s is used to detect whether personnel or vehicles have entered the safe area; for geological disaster monitoring, Δt=300s is used to capture slow displacement; in sudden emergency scenarios, Δt can be shortened to less than 10 seconds to achieve near real-time change perception.

[0071] 202. Extract the first object and the second object, and obtain the first object attribute of the first object and the second object attribute of the second object.

[0072] The first object can be a moving object, such as a pedestrian or a vehicle, and the second object can be a stationary object, such as a building or flowers. The attributes of the first object can be type, size, movement speed, and whether it contains characters, etc.; the attributes of the second object can be type, size, shape, and whether it contains characters, etc.

[0073] Optionally, feature extraction is performed on the first object to obtain a first feature; the first feature is input into a neural network model to output a first object attribute; feature extraction is performed on the second object to obtain a second feature; the second feature is input into a neural network model to output a second object attribute. The neural network model can be an existing machine learning model, and no specific limitations are imposed here.

[0074] In one implementation, step 202, extracting the first object and the second object, may include:

[0075] Dense optical flow field estimation is performed on the first and second scene images to obtain pixel-level displacement vector fields;

[0076] By verifying the consistency of forward and backward optical flow, reliable optical flow regions of the pixel-level displacement vector field are selected, and a first motion confidence map is generated.

[0077] The first point cloud data and the second point cloud data are voxelized into three-dimensional voxel meshes respectively, and registered based on the ICP algorithm to obtain the registration matrix;

[0078] The second point cloud data is spatially transformed using the registration matrix to obtain an aligned voxel grid;

[0079] Calculate the Euclidean distance difference between the three-dimensional voxel grid of the first point cloud data and the aligned voxel grid;

[0080] Based on the Euclidean distance difference, calculate the voxel-level motion confidence score and generate the second motion confidence map;

[0081] The first motion confidence map and the second motion confidence map are aligned by cross-modal projection to generate a fused motion saliency score map;

[0082] Connectivity analysis and morphological closing operations are performed on the fused motion saliency score map to segment independent potential first object regions and output motion saliency maps; among them, regions with significant displacement or deformation are marked as potential first object regions, and the remaining regions are marked as potential second object regions;

[0083] The potential first object region and the potential second object region are input into a two-stream convolutional neural network (Two-Stream CNN), which outputs the first object and the second object.

[0084] The dual-stream convolutional neural network consists of two independent convolutional neural networks: a temporal stream and a spatial stream. The temporal stream is used to process dynamic features, i.e., to extract the first object based on the potential first object region. The spatial stream is used to process static features, i.e., to extract the second object based on the potential second object region.

[0085] 203. Based on the first object attribute and the second object attribute, obtain the inspection plan for the UAV.

[0086] In one implementation, step 203 may include:

[0087] Calculate the imaging priority weight of each object based on the first object attribute and the second object attribute;

[0088] Optionally, to achieve intelligent scheduling of inspection resources, this invention calculates the imaging priority weight of each identified object based on its attribute information, including character presence factors (such as building numbers, equipment nameplates, vehicle license plates, warning signs, etc.), size factors, and motion state factors, for use in generating subsequent differentiated imaging strategies. Specifically, the system uses a weighted summation method to calculate the imaging priority weight of each object. .

[0089] Based on the imaging priority weights, a multi-priority imaging task queue is generated, and imaging resources are allocated to the multi-camera system of each UAV.

[0090] Optionally, objects with an imaging priority weight higher than a first threshold are identified as high-priority objects, and high-magnification zoom cameras are scheduled to perform multi-angle zoom imaging; wherein, multi-angle imaging is achieved by using N azimuth angles around the center of the object, and the focal length is dynamically adjusted according to the distance between the object and the drone and the size of the characters to achieve zoom imaging; objects with an imaging priority weight lower than the first threshold but higher than a second threshold are identified as medium-priority objects, and medium-focal-length cameras or high-magnification zoom cameras are scheduled to perform single or double imaging at a fixed angle; wherein, medium focal length and standard exposure parameters are set for single or double fixed-focus imaging; objects with an imaging priority weight lower than the second threshold are identified as low-priority objects, and wide-angle cameras are scheduled to perform single fixed-focus imaging; wherein, fixed focal length and exposure parameters are set for single fixed-focus imaging.

[0091] Based on the allocated imaging resources, construct a drone inspection scheme;

[0092] Specifically, after allocating imaging resources for each target object, the system further integrates the dispersed resource allocation results into a structured and executable UAV inspection scheme. All tasks are sorted from high to low execution priority, forming a structured inspection scheme. This scheme is serialized in a lightweight data format (such as JSON or Protocol Buffers) and sent to the flight control systems of the corresponding UAVs via wireless communication links (such as 4G / 5G or dedicated data transmission radios). In multi-UAV collaborative scenarios, the system also performs conflict detection and collaborative optimization: if multiple tasks involve spatially adjacent targets and need to be imaged simultaneously, the scheduling module will adjust the flight altitude or imaging sequence of each UAV to avoid mutual occlusion or collision risks, ensuring imaging integrity. In addition, this inspection scheme supports a dynamic update mechanism. During task execution, if the system continuously tracks and detects a deviation in the trajectory of the first object (such as a moving vehicle), it will recalculate its optimal imaging angle and timing in real time and generate incremental update instructions, only re-issuing the changed parts without rebuilding and transmitting the entire inspection scheme, thereby improving response efficiency.

[0093] 204. The UAV executes the inspection plan and collects inspection data.

[0094] Specifically, after receiving the structured inspection plan issued by the ground control terminal or the airborne scheduling module, each UAV enters the task execution phase; the flight control system of each UAV parses the task list in the inspection plan, extracts the imaging tasks related to itself, and executes them one by one according to the task priority.

[0095] Furthermore, after each imaging task is completed, the UAV generates a task execution log, recording information such as the actual shooting angle, focal length, image quality score (e.g., sharpness, exposure), and whether the focus was successful. This log is then transmitted back to the ground control terminal via a wireless link for subsequent scheme optimization and quality assessment.

[0096] 205. The inspection data is packaged according to object category and timestamp, and uploaded to the data management terminal. Multi-angle images of the same small object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0097] Specifically, the data encapsulation module first categorizes and organizes the raw data according to the object category associated with each data entry (such as "moving vehicle," "equipment nameplate," "warning sign," etc.) and the collection timestamp (accurate to milliseconds). All encapsulated data units are indexed according to timestamp order or object category and then uploaded in batches to the data management terminal via wireless communication links (such as 4G / 5G, satellite communication, or Wi-Fi backhaul). The upload process supports breakpoint resumption and encrypted transmission to ensure data integrity and security.

[0098] Furthermore, for data that is identified as a moving object and belongs to a small, high-value target (such as equipment number, text label, instrument reading, etc.), the system executes a special encapsulation strategy: multiple images of the same object collected from different angles and at different times, along with their associated point cloud data, are integrated into an independent data unit.

[0099] This improves the utilization efficiency and data collection quality of drones, thereby enabling precise management of inspection data.

[0100] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an inspection data management system for unmanned aerial vehicles (UAVs) provided by the present invention.

[0101] like Figure 3 As shown in the embodiment of the present invention, the UAV inspection data management system includes:

[0102] The data acquisition module 301 is used to acquire the first on-site image and the first point cloud data collected by the UAV at the start of the inspection, and the second on-site image and the second point cloud data collected after a preset time interval.

[0103] The attribute analysis module 302 is used to extract the first object and the second object, and obtain the first object attribute of the first object and the second object attribute of the second object;

[0104] The scheme generation module 303 is used to obtain the inspection scheme of the UAV based on the first object attribute and the second object attribute;

[0105] Inspection module 304 is used by the UAV to collect inspection data when executing the inspection plan;

[0106] The data management module 305 is used to encapsulate the inspection data according to object category and timestamp, and upload it to the data management terminal. Among them, multi-angle images of the same small-sized object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0107] Please see Figure 4 , Figure 4 This is a schematic diagram illustrating an embodiment of the electronic device provided in this invention. For example... Figure 4 As shown, an embodiment of the present invention provides an electronic device 400, including a memory 410, a processor 420, and a computer program 411 stored in the memory 410 and executable on the processor 420. When the processor 420 executes the computer program 411, it performs the following steps:

[0108] An application is installed on the electronic device 400. The application is used to acquire a first on-site image and a first point cloud data collected by the UAV at the start of the inspection, and a second on-site image and a second point cloud data collected after a preset time interval; extract a first object and a second object, and obtain the first object attribute of the first object and the second object attribute of the second object; obtain the UAV's inspection plan based on the first object attribute and the second object attribute; the UAV executes the inspection plan to collect inspection data; the inspection data is packaged according to object category and timestamp, and uploaded to the data management terminal, wherein multi-angle images of the same small-sized object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0109] Please see Figure 5 , Figure 5 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. For example... Figure 5 As shown, this embodiment provides a computer-readable storage medium 500, on which a computer program 411 is stored. When the computer program 411 is executed by a processor, it performs the following steps: The computer-readable storage medium 500 includes an application program, which is used to acquire a first scene image and a first point cloud data collected by the UAV at the start of the inspection, and a second scene image and a second point cloud data collected after a preset time interval; extract a first object and a second object, and obtain a first object attribute of the first object and a second object attribute of the second object; obtain an inspection plan for the UAV based on the first object attribute and the second object attribute; the UAV executes the inspection plan to collect inspection data; the inspection data is packaged according to object category and timestamp, and uploaded to a data management terminal, wherein multi-angle images of the same small-sized object are packaged into a data unit to support subsequent 3D reconstruction or character recognition.

[0110] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0111] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A system that specifies functions in one or more boxes.

[0113] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including an instruction set implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0114] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0115] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0116] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for managing inspection data of a UAV, characterized in that, The method comprises: acquiring first field image and first point cloud data collected by the unmanned aerial vehicle at a starting time of inspection, and second field image and second point cloud data collected after a preset time interval; extracting first objects and second objects, and obtaining first object attributes of the first objects and second object attributes of the second objects; the first objects include moving objects, and the second objects include stationary objects; the first object attributes include moving speed and whether to contain characters; the second object attributes include size, shape, and whether to contain characters; the extraction of the first objects and the second objects comprises: generating a motion saliency map according to the first field image, the first point cloud data, the second field image, and the second point cloud data, wherein a region with significant displacement or deformation is marked as a potential first object region, and the remaining region is marked as a potential second object region; inputting the potential first object region and the potential second object region into a double-flow convolutional neural network, wherein a first flow channel of the double-flow convolutional neural network is used to process the potential first object region to extract the first objects, and a second flow channel is used to process the potential second object region to extract the second objects; obtaining an inspection scheme of the unmanned aerial vehicle according to the first object attributes and the second object attributes; the unmanned aerial vehicle collects inspection data according to the inspection scheme; the inspection data is encapsulated according to object categories and time stamps, and is uploaded to a data management terminal, wherein multi-angle images of the same small-size object are packaged into one data unit to support subsequent three-dimensional reconstruction or character recognition. 2.The method of claim 1, wherein, the generation of the motion saliency map according to the first field image, the first point cloud data, the second field image, and the second point cloud data comprises: performing dense optical flow field estimation on the first field image and the second field image to generate a first motion confidence map; voxelizing the first point cloud data and the second point cloud data into three-dimensional voxel grids respectively to generate a second motion confidence map; performing cross-modal projection alignment on the first motion confidence map and the second motion confidence map to generate a fusion motion saliency score map; performing connected component analysis and morphological closing operation on the fusion motion saliency score map to segment out independent potential first object regions, and outputting a motion saliency map. 3.The method of claim 2, wherein, the dense optical flow field estimation on the first field image and the second field image to generate the first motion confidence map comprises: performing dense optical flow field estimation on the first field image and the second field image to obtain a pixel-level displacement vector field; filtering out reliable optical flow regions of the pixel-level displacement vector field through forward-backward optical flow consistency inspection to generate a first motion confidence map. 4.The method of claim 2, wherein, the voxelization of the first point cloud data and the second point cloud data into three-dimensional voxel grids to generate the second motion confidence map comprises: voxelizing the first point cloud data and the second point cloud data into three-dimensional voxel grids respectively, and performing registration based on an ICP algorithm to obtain a registration matrix; performing spatial transformation on the second point cloud data through the registration matrix to obtain an aligned voxel grid; calculating a Euclidean distance difference value between a three-dimensional voxel grid of the first point cloud data and the aligned voxel grid; calculating a voxel-level motion confidence score according to the Euclidean distance difference value, and generating a second motion confidence map. 5.The method of claim 1, wherein, The first object attribute of the first object and the second object attribute of the second object are obtained by: extracting features of the first object to obtain first features, and inputting the first features into a neural network model to output the first object attribute; extracting features of the second object to obtain second features, and inputting the second features into a neural network model to output the second object attribute. 6.The method of claim 1, wherein, The inspection scheme of the unmanned aerial vehicle is obtained according to the first object attribute and the second object attribute, comprising: calculating the imaging priority weight of each object according to the first object attribute and the second object attribute; generating a multi-priority imaging task queue based on the imaging priority weight, and allocating imaging resources for the multi-camera system of each unmanned aerial vehicle; constructing an unmanned aerial vehicle inspection scheme according to the allocated imaging resources. 7.The method of claim 6, wherein, The multi-priority imaging task queue is generated based on the imaging priority weight, and the imaging resources are allocated for the multi-camera system of each unmanned aerial vehicle, comprising: determining the object with an imaging priority weight higher than a first threshold as a high-priority object, and scheduling a high-magnification zoom camera for multi-angle zoom imaging; wherein N azimuth angles around the object center realize multi-angle imaging, and the focal length is dynamically adjusted according to the distance between the object and the unmanned aerial vehicle and the character size to realize zoom imaging; determining the object with an imaging priority weight lower than the first threshold and higher than a second threshold as a medium-priority object, and scheduling a medium-focus camera or a high-magnification zoom camera for single or double imaging at a fixed angle; wherein a medium focal length and standard exposure parameters are set to realize single or double fixed focus imaging; determining the object with an imaging priority weight lower than the second threshold as a low-priority object, and scheduling a wide-angle camera for single fixed focus imaging; wherein a fixed focal length and exposure parameters are set to realize single fixed focus imaging.

8. An inspection data management system of a UAV, characterized in that, The system comprises: a data acquisition module for acquiring first scene images and first point cloud data collected by the unmanned aerial vehicle at the start of the inspection, and second scene images and second point cloud data collected after a preset time interval; an attribute analysis module for extracting first objects and second objects, and obtaining first object attributes of the first objects and second object attributes of the second objects; the first objects include moving objects, and the second objects include stationary objects; the first object attributes include moving speed and whether to contain characters; and the second object attributes include size, shape, and whether to contain characters; The first objects and the second objects are extracted by: generating a motion saliency map according to the first scene images, the first point cloud data, the second scene images, and the second point cloud data, wherein regions with significant displacement or deformation are marked as potential first object regions, and the remaining regions are marked as potential second object regions; inputting the potential first object region and the potential second object region into a double-stream convolutional neural network, wherein a first stream channel of the double-stream convolutional neural network is configured to process the potential first object region to extract a first object, and a second stream channel of the double-stream convolutional neural network is configured to process the potential second object region to extract a second object; a scheme generation module configured to obtain an inspection scheme of a UAV according to the first object attribute and the second object attribute; an inspection module configured to cause the UAV to execute the inspection scheme to collect inspection data; a data management module configured to encapsulate the inspection data according to object categories and time stamps, and upload the inspection data to a data management terminal, wherein multi-angle images of a same small-size object are packed into one data unit to support subsequent three-dimensional reconstruction or character recognition.

Images