Unmanned aerial vehicle inspection shooting pose optimization method and system based on structure perception

Abstract

The invention provides an unmanned aerial vehicle inspection shooting pose optimization method and system based on structure perception. The method comprises the following steps: determining a key detection area according to three-dimensional structure data of a detected object; generating a detectability constraint condition for the key detection area according to the imaging system parameters; generating a security constraint condition according to the security specification information; according to the three-dimensional structure data, shielding detection is carried out on a connecting line of the candidate pose and the key detection area, and a shielding constraint condition is generated; legal poses are sampled according to constraint conditions; selecting a representative pose set from the legal poses of the key detection areas according to the coverage amount, the pose number and the track length of the key detection areas; path planning is carried out based on the representative pose set, and an unmanned aerial vehicle inspection track is generated; the imaging quality of the key defect part is improved by utilizing a detectability index; shielding analysis is carried out based on a three-dimensional structure, and information dead angles are reduced; constraint conditions are explicitly unified, and risks caused by human experience are reduced.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of unmanned aerial vehicle control, and in particular to a structure perception based unmanned aerial vehicle inspection shooting pose optimization method and system. BACKGROUND

[0002] Existing unmanned aerial vehicle inspection task planning usually adopts fixed height, equidistant waypoints or fly-around trajectories, and uses geometric coverage rate as the main evaluation index. However, the following problems exist: the minimum defect size, pixel resolution and incident angle requirements are not explicitly considered, resulting in that key components may appear in the image but cannot meet the identification requirements; the occlusion analysis of the three-dimensional structure of the measured object is not performed, and there are stable information dead angles on the back and intersection of complex structures such as towers, busbars and pipe galleries; the waypoints and poses are configured by the experience of the operator, and the visual angle difference of different batches of inspection is large, and it is difficult to quantitatively compare the historical data; the safety distance requirement and the imaging quality are not uniformly modeled, and there is a risk of too close flight or insufficient image quality. SUMMARY

[0003] The following is a summary of the subject matter described in detail herein.

[0004] The purpose of the present application is to at least partially solve one of the technical problems existing in the related art. The embodiments of the present application provide a structure perception based unmanned aerial vehicle inspection shooting pose optimization method and system.

[0005] In an embodiment of the first aspect of the present application, a structure perception based unmanned aerial vehicle inspection shooting pose optimization method comprises: obtaining three-dimensional structure data of a measured object; determining a key detection area according to the three-dimensional structure data; generating detectability constraint conditions for the key detection area according to imaging system parameters; generating safety constraint conditions according to safety specification information; performing occlusion detection on the connecting line between the candidate pose and the key detection area according to the three-dimensional structure data, and generating occlusion constraint conditions; sampling legal poses that satisfy the detectability constraint conditions, the safety constraint conditions and the occlusion constraint conditions; selecting a representative pose set from the legal poses of each key detection area according to key detection area coverage, pose quantity and flight path length; performing path planning based on the representative pose set to generate an unmanned aerial vehicle inspection flight path.

[0006] According to certain embodiments of the first aspect of the present application, the three-dimensional structure data comprises one or more of a design model, a BIM model or a point cloud model of the measured object.

[0007] According to certain embodiments of the first aspect of the application, the critical detection area records a position coordinate, an approximate normal direction, an expected defect type, and a minimum defect size.

[0008] According to certain embodiments of the first aspect of the application, the detectability constraint includes a maximum allowed shooting distance. The maximum allowed shooting distance is calculated according to the following formula: ; In the formula, is the maximum allowed shooting distance, is the critical detection area, is the candidate pose, is the distance between the candidate pose and the center point of the critical detection area, s is the pixel size of the imaging system parameter, and f is the focal length of the imaging system parameter.

[0009] According to certain embodiments of the first aspect of the application, the detectability constraint includes an angle range between the shooting optical axis and the normal vector of the detection surface, and the angle range between the shooting optical axis and the normal vector of the detection surface is set according to the type of the critical detection area.

[0010] According to certain embodiments of the first aspect of the application, the safety constraint defines the allowed flight space and the forbidden flight space of the unmanned aerial vehicle.

[0011] According to certain embodiments of the first aspect of the application, occlusion detection is performed on the line connecting the candidate pose and the critical detection area according to the three-dimensional structure data, and an occlusion constraint condition is generated, including: When the line connecting the candidate pose and the critical detection area intersects with the three-dimensional structure data, the candidate pose is determined to be a pose that meets the occlusion constraint condition; When the line connecting the candidate pose and the critical detection area does not intersect with the three-dimensional structure data, the candidate pose is determined to be a pose that does not meet the occlusion constraint condition.

[0012] According to certain embodiments of the first aspect of the application, a representative pose set is selected from the legal poses of each critical detection area according to the critical detection area coverage, the number of poses, and the length of the flight path, including: The representative pose set is selected from the legal poses of each critical detection area to maximize the critical detection area coverage, minimize the number of poses, and minimize the length of the flight path.

[0013] According to certain embodiments of the first aspect of the application, the path planning is performed based on the representative pose set to generate an unmanned aerial vehicle inspection flight path, including: The path planning is performed on the representative pose set according to the maximum speed, the maximum acceleration, the minimum turning radius, and the forbidden flight volume boundary of the unmanned aerial vehicle to generate an unmanned aerial vehicle inspection flight path.

[0014] An embodiment of the second aspect of this application, a structure-aware UAV inspection and shooting pose optimization system, includes: The structural modeling module is used to acquire the three-dimensional structural data of the object under test. The key region annotation module is used to determine key detection regions based on the three-dimensional structural data; The constraint generation module is used to generate inspectability constraints for the key detection area based on the preset minimum defect size and imaging system parameters, generate safety constraints based on safety specification information, and generate occlusion constraints based on the occlusion detection of the connection between the candidate pose and the key detection area based on the three-dimensional structural data. The pose selection module is used to sample legal poses that satisfy the detectability constraint, the safety constraint, and the occlusion constraint. The optimization module is used to select a representative set of poses from the legal poses of each key detection area based on the coverage, number of poses, and track length of the key detection area. The trajectory generation module is used to perform path planning based on the representative pose set and generate UAV inspection trajectories.

[0015] The above scheme has at least the following beneficial effects: It determines key detection areas based on 3D structural data; generates inspectability constraints for key detection areas based on imaging system parameters; generates safety constraints based on safety specifications; performs occlusion detection on the connection between candidate poses and key detection areas based on 3D structural data, generating occlusion constraints; samples legal poses that satisfy inspectability, safety, and occlusion constraints; selects representative pose sets from the legal poses of each key detection area based on the coverage, number of poses, and track length of the key detection area; performs path planning based on the representative pose sets to generate UAV inspection tracks; replaces coverage indicators with inspectability indicators to improve the imaging quality of key defect areas; performs occlusion analysis based on 3D structure to reduce blind spots; and explicitly unifies safety and imaging constraints to reduce risks associated with human experience. Attached Figure Description

[0016] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0017] Figure 1 This is a step diagram of a structure-aware UAV inspection and shooting pose optimization method; Figure 2 This is a structural diagram of a structure-aware UAV inspection and photography pose optimization system. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, or the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0020] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0021] The embodiments of this application provide a method and system for optimizing the pose of drone inspection and shooting based on structure awareness.

[0022] Reference Figure 1 The method for optimizing the pose of drone inspection and shooting includes, but is not limited to, the following steps: Step S100: Obtain the three-dimensional structural data of the object under test; Step S200: Determine the key detection area based on the three-dimensional structural data; Step S300: Generate detectability constraints for key detection areas based on imaging system parameters; Step S400: Generate safety constraints based on safety specification information; Step S500: Based on the 3D structural data, perform occlusion detection on the connection between the candidate pose and the key detection area, and generate occlusion constraints. Step S600: Sample valid poses that satisfy the inspectability constraints, safety constraints, and occlusion constraints; Step S700: Select a representative pose set from the legal poses of each key detection area based on the coverage of the key detection area, the number of poses, and the track length. Step S800: Path planning is performed based on the representative pose set to generate the UAV inspection track.

[0023] For step S100, the three-dimensional structural data of the object under test is obtained.

[0024] The 3D structural data includes one or more of the following: the design model of the object under test, the BIM model, or the point cloud model. The point cloud model is obtained through LiDAR or multi-view image reconstruction.

[0025] For step S200, the key detection area is determined based on the three-dimensional structural data.

[0026] Key inspection areas include insulator strings, hardware connection points, welds, flanges, valves, switch contacts, busbar connections, etc.

[0027] The key inspection area records the location coordinates, approximate normal direction, expected defect type, and minimum defect size.

[0028] For step S300, detectability constraints are generated for the key detection area based on the preset minimum defect size and imaging system parameters.

[0029] Detectability constraints include the maximum permissible shooting distance; The maximum permissible shooting distance is calculated using the following formula: ; In the formula, To the maximum permissible shooting distance, This is a key detection area. As candidate pose, is the distance between the candidate pose and the center point of the key detection area, s is the pixel size of the imaging system parameter, and f is the focal length of the imaging system parameter.

[0030] The detectability constraints include the range of the angle between the shooting optical axis and the normal vector of the detection surface. The range of the angle between the shooting optical axis and the normal vector of the detection surface is set according to the type of the key detection area.

[0031] For step S400, safety constraints are generated based on safety specification information. For example, based on safety specification information such as electrical safety, explosion-proof safety, equipment clearance, and management requirements, safety constraints that limit the permissible and no-fly zones for the UAV are generated.

[0032] For step S500, occlusion detection is performed on the connection between the candidate pose and the key detection area based on the 3D structural data, generating occlusion constraints. Specifically: if the connection between the candidate pose and the key detection area intersects with the 3D structural data, the candidate pose is determined to meet the occlusion constraints; if the connection between the candidate pose and the key detection area does not intersect with the 3D structural data, the candidate pose is determined to not meet the occlusion constraints. This eliminates occluded poses.

[0033] For step S600, sample valid poses that satisfy the inspectability constraint, safety constraint, and occlusion constraint.

[0034] Specifically, within the safe flight space boundary or designated search area, a set of candidate sampling points is generated for each key detection area; for each sampling point, it is determined whether it simultaneously satisfies: detectability constraint, safety constraint, and occlusion constraint; the sampling points that satisfy the conditions and their corresponding camera poses are recorded as the set of legal poses for the key detection area.

[0035] For step S700, based on the coverage of the key detection area, the number of poses, and the track length, a representative pose set is selected from the legal poses of each key detection area. Specifically, with the goal of maximizing the coverage of the key detection area, minimizing the number of poses, and minimizing the track length, a representative pose set is selected from the legal poses of each key detection area. This achieves the following requirements: (1) covering all key detection areas; (2) reducing the number of poses and the track length; and (3) increasing the amount of integrated information from multiple sensors.

[0036] For step S800, based on the UAV's maximum speed, maximum acceleration, minimum turning radius, and no-fly zone boundary, path planning is performed on the representative pose set to generate the UAV inspection track.

[0037] Specifically, based on the dynamic constraints of the UAV and the distribution of obstacles, representative shooting poses are sorted and paths are planned to generate smooth trajectories that meet the requirements of speed, acceleration and safe distance, and the output is a standard task file that can be executed directly.

[0038] The standard mission file includes a list of waypoints, camera trigger commands, speed configurations, etc.; it is sent to the ground station or the drone nesting system, and the drone can execute it with one click.

[0039] Embodiments of this application also provide a drone inspection and shooting pose optimization system.

[0040] Reference Figure 2 The UAV inspection and photography pose optimization system includes: a structural modeling module, a key area annotation module, a constraint generation module, a pose selection module, an optimization module, and a trajectory generation module.

[0041] The structural modeling module is used to acquire the three-dimensional structural data of the object under test; The key region annotation module is used to determine key detection regions based on 3D structural data; The constraint generation module is used to generate inspectability constraints for key detection areas based on preset minimum defect size and imaging system parameters, generate safety constraints based on safety specification information, and generate occlusion constraints based on occlusion detection of the connection between candidate poses and key detection areas based on 3D structural data. The pose selection module is used to sample legal poses that meet the inspectability constraints, safety constraints, and occlusion constraints. The optimization module is used to select a representative set of poses from the legal poses of each key detection area based on the coverage, number of poses, and track length of the key detection area. The trajectory generation module is used to perform path planning based on a representative pose set and generate UAV inspection trajectories.

[0042] It is understandable that the UAV inspection and shooting pose optimization system and the UAV inspection and shooting pose optimization method adopt the same technical solution. Each module of the UAV inspection and shooting pose optimization system corresponds one-to-one with each step of the UAV inspection and shooting pose optimization method. The two solve the same technical problems and have the same technical effects.

[0043] The following is an explanation through specific examples.

[0044] Example 1: Optimization of shooting pose in the scenario of power transmission tower inspection.

[0045] Example 1 illustrates the application of a structure-aware UAV inspection pose optimization method for 220kV transmission towers to generate UAV shooting poses and trajectories that meet safety constraints and defect inspectability requirements.

[0046] The operation and maintenance unit already has one of the following data: a 3D model or BIM model of the tower provided by the design department; or a point cloud model obtained by one-time LiDAR aerial scanning + multi-view image reconstruction.

[0047] The system's structural modeling module performs coordinate unification and simplification on the above data, generating three-dimensional structural data including components such as tower body, crossarm, insulator hanging point, conductor, and ground wire, and establishes a local coordinate system with the tower base center as the origin and the vertical upward Z-axis.

[0048] By combining a built-in rule base with manual verification, key inspection areas are marked on the 3D model, including but not limited to: the entire surface of each string of suspension or tension insulators; the connection points of fittings and the positions of U-rings, hanging plates, and clamps; the fixing points of ground wires and down conductors; the welds connecting the crossarms and main materials; and the vibration dampers, spacers, and corresponding clamps.

[0049] For each key detection area It records: spatial location and extent (point set or bounding box); approximate normal direction. ; Expected defect types (e.g., umbrella skirt cracks, hardware corrosion, wire clamp ablation, etc.); Preset minimum inspectable defect size (For example, a crack of 0.5 mm or an edge defect of 1 mm).

[0050] For each key inspection area, inspectability constraints are generated based on the camera parameters carried by the inspection drone.

[0051] For example, the camera resolution is 5472×3648; pixel size The focal length is f=24mm; the ground sampling interval GSD(p,i) of the target in the image is required to be ≤0.2mm / pixel.

[0052] For the distance d(p,i) between the candidate pose p and the center point of the key detection region, we have: .

[0053] Based on this, the maximum permissible shooting distance for this key detection area is: .

[0054] Based on experience and optical imaging characteristics, the imaging optical axis and the normal of the detection surface are... The included angle The range is defined as follows: .

[0055] The angle between the shooting optical axis and the normal vector of the detection surface is set according to the type of the key detection area. For example, the angle between the shooting optical axis and the normal vector of the detection surface for insulator skirts is set to 20° to 60°, while 10° to 70° is recommended for hardware surfaces.

[0056] The detectability constraint generation module will , Parameters such as range are stored in the task configuration.

[0057] Based on voltage levels and safety regulations of operation and maintenance units, a three-dimensional safety distance buffer zone is generated for live conductors, fittings, and other electrical components, forming permissible flight space and no-fly zone.

[0058] In the system structure, the safety shell is centered on conductors, hardware, busbars, etc.; the inner side of the tower body, narrow gaps and other unsuitable areas for flight are designated as no-fly zones.

[0059] Safety constraints are used to limit the permissible flight space of drones. .

[0060] The occlusion detection module, based on 3D structural data, compares candidate pose p with key detection regions. The connection between the elements is checked for occlusion; if the connection intersects with components such as tower materials, crossarms, or conductors, then the pose is determined to be obstructed. If there is occlusion, it is marked as invalid; if there is no intersection, it passes the occlusion test.

[0061] exist In this process, candidate pose sampling bands are generated centered on each key detection region. For example, from... Heading outwards to On the spherical or toroidal surface between them, samples are uniformly taken according to azimuth and elevation angles; or, each sampling point p is assigned a direction. The camera orientation is used as the initial pose.

[0062] For each candidate pose p, the following constraint conditions are applied: Detectability constraints: Safety constraints: Obstruction constraint: The connecting line and the 3D structure of the tower do not intersect; Angle constraint: .

[0063] When all constraints are met, the pose p is recorded as the valid pose of the key detection region and added to the set. .

[0064] It can generate dozens to hundreds of legal pose points for each key detection area.

[0065] The optimization module applies to all Joint optimization was carried out.

[0066] The objectives include: for each key detection area Select at least one or more legal poses to cover; minimize the total number of poses selected N; minimize the total expected track length L; and retain reasonable view redundancy to cope with disturbances such as wind drift.

[0067] Each valid pose is considered as a "candidate point" that can cover several key detection regions; a joint objective of set coverage problem and path cost is constructed; a heuristic algorithm (greedy + local search or genetic algorithm) is used to select a representative pose set. .

[0068] The path planning module is based on As a set of necessary points, a smooth and feasible three-dimensional trajectory is generated by combining the maximum speed, maximum acceleration, minimum turning radius, and no-fly zone boundary of the UAV, including the climb segment, tower-circling segment, and return segment.

[0069] Will The program employs a constrained traveling salesman problem to determine the order of visits; interpolates between adjacent poses to generate multi-segment spline curves; and ensures that the generated track does not intrude into no-fly zones at any point. The final output is a standard mission file, including a list of waypoints, camera trigger commands, and speed configurations, which is then distributed to the ground station or drone nesting system for one-click execution by the UAV.

[0070] In actual deployment, the following effects are achieved: images of key parts of the tower all meet the preset GSD and viewing angle requirements; traditional blind spots such as the back side of insulators and the shadow area of ​​hardware are effectively covered; different batches of tasks reuse the same planning logic, and the differences in track and pose are controlled within the preset range, which is conducive to change detection.

[0071] Example 2: Optimization of shooting pose in complex substation scenarios.

[0072] Example 2 addresses image inspection under conditions of densely packed equipment and complex restricted areas within outdoor substations. It illustrates the application of a structure-aware UAV inspection pose optimization method to generate UAV shooting poses and trajectories that meet safety constraints and defect inspectability requirements.

[0073] The operation and maintenance unit already has one of the following data: a 3D model or BIM model of the substation provided by the design department; or a point cloud model obtained by one-time LiDAR aerial scanning + multi-view image reconstruction.

[0074] The system's structural modeling module performs coordinate unification and simplification on the above data, generating three-dimensional structural data including components such as main transformers, circuit breakers, disconnect switches, busbars, frames, fences, and control cabinets, and establishing a local coordinate system.

[0075] By combining a built-in rule base with manual verification, key detection areas are marked on the 3D model, including but not limited to: circuit breaker contacts and connection points; disconnector switch blade closing position; bushing root, oil tank, heat sink; busbar connection point and suspension hardware.

[0076] For each key detection area The record includes: spatial location and extent; approximate normal direction. ; Expected defect type; Preset minimum inspectable defect size .

[0077] For each key inspection area, inspectability constraints are generated based on the camera parameters carried by the inspection drone.

[0078] For example, the camera resolution is 5472×3648; pixel size The focal length is f=24mm; the ground sampling interval GSD(p,i) of the target in the image is required to be ≤0.2mm / pixel.

[0079] For the distance d(p,i) between the candidate pose p and the center point of the key detection region, we have: .

[0080] Based on this, the maximum permissible shooting distance for this key detection area is: .

[0081] Based on experience and optical imaging characteristics, the imaging optical axis and the normal of the detection surface are... The included angle The range is defined as follows: .

[0082] The detectability constraint generation module will , Parameters such as range are stored in the task configuration.

[0083] According to the voltage level and the safety regulations of the operation and maintenance unit, "No flying over volume" signs are set above the main transformer and high-voltage equipment; the minimum flying height is restricted above personnel passages and roads within the station; and horizontal and vertical safety distance zones are set according to explosion-proof and electrical safety requirements.

[0084] Safety constraints are used to limit the permissible flight space of drones. .

[0085] The occlusion detection module, based on 3D structural data, compares candidate pose p with key detection regions. The connection between the elements is checked for occlusion; if the connection intersects with components such as tower materials, crossarms, or conductors, then the pose is determined to be obstructed. If there is occlusion, it is marked as invalid; if there is no intersection, it passes the occlusion test.

[0086] exist In this process, candidate pose sampling bands are generated centered on each key detection region. For example, from... Heading outwards to On the spherical or toroidal surface between them, samples are uniformly taken according to azimuth and elevation angles; or, each sampling point p is assigned a direction. The camera orientation is used as the initial pose.

[0087] For each candidate pose p, the following constraint conditions are applied: Detectability constraints: Safety constraints: Obstruction constraint: The connecting line and the 3D structure of the tower do not intersect; Angle constraint: .

[0088] When all constraints are met, the pose p is recorded as the valid pose of the key detection region and added to the set. .

[0089] It can generate dozens to hundreds of legal pose points for each key detection area.

[0090] In addition, instead of using circling flight, candidate points are generated in a safe space, distributed on both sides and above the equipment; for parts such as knife switch contacts, a combination of side and top views is preferred to ensure unobstructed views.

[0091] The optimization module applies to all Joint optimization was carried out.

[0092] The objectives include: for each key detection area Select at least one or more legal poses to cover; minimize the total number of poses selected N; minimize the total expected track length L; and retain reasonable view redundancy to cope with disturbances such as wind drift.

[0093] The legal positions of multiple devices within the station are unified into a single optimization problem; at the same time, a "zonal inspection sequence" (such as main transformer area first, then outgoing line area) is added to reduce large-scale horizontal jumps.

[0094] The path planning module is based on As a set of necessary points, a smooth and feasible three-dimensional trajectory is generated by combining the maximum speed, maximum acceleration, minimum turning radius, and no-fly zone boundary of the UAV, including the climb segment, tower-circling segment, and return segment.

[0095] Will The program employs a constrained traveling salesman problem to determine the order of visits; interpolates between adjacent poses to generate multi-segment spline curves; and ensures that the generated track does not intrude into no-fly zones at any point. The final output is a standard mission file, including a list of waypoints, camera trigger commands, and speed configurations, which is then distributed to the ground station or drone nesting system for one-click execution by the UAV.

[0096] In actual deployment, the following effects are achieved: automatically generated flight paths avoid densely populated areas and high-risk equipment, ensuring regional safety and controllability; small targets such as circuit breaker contacts and knife switch contacts achieve detectable resolution on the image; and compared to the traditional method of "manually drawing flight paths + test flight adjustments", planning time and the number of trials and errors are significantly reduced.

[0097] Example 3: Shooting pose optimization under multi-sensor joint constraints.

[0098] Example 3 illustrates the image inspection of a UAV equipped with both a visible light camera and an infrared camera, demonstrating the application of a structure-aware UAV inspection shooting pose optimization method to generate UAV shooting poses and trajectories that meet safety constraints and defect inspectability requirements.

[0099] The operation and maintenance unit has one of the following data: a 3D model or BIM model of the object under test provided by the design department; or a point cloud model obtained by one-time LiDAR aerial scanning + multi-view image reconstruction.

[0100] The system's structural modeling module performs coordinate unification and simplification on the above data, generates three-dimensional structural data, and establishes a local coordinate system.

[0101] By combining a built-in rule base with manual verification, key detection areas are marked on the 3D model.

[0102] For each key detection area The record includes: spatial location and extent; approximate normal direction. ; Expected defect type; Preset minimum inspectable defect size .

[0103] For each key detection area In addition to visible light defects (geometric cracks, corrosion), it is also equipped with infrared detection requirements (such as joint overheating).

[0104] For each key inspection area, inspectability constraints are generated based on the camera parameters carried by the inspection drone.

[0105] The detectability constraint is extended to: Visible light: GSDv(p,i)≤GSDvmax(i), the incident angle satisfies [θvmin, θvmax]; Infrared: The preferred viewing angle is close to the normal to reduce reflection error. The temperature measurement viewing angle satisfies [θirmin, θirmax], and the distance range is limited to ensure detection sensitivity.

[0106] Based on the voltage level and the safety specifications of the operation and maintenance unit, a three-dimensional safety distance buffer zone is generated, forming a permitted flight space and a no-fly space.

[0107] Safety constraints are used to limit the permissible flight space of drones. .

[0108] The occlusion detection module, based on 3D structural data, compares candidate pose p with key detection regions. Occlusion detection is performed on the lines connecting the points; if the line intersects with the 3D structure, then the pose is determined to be... If there is occlusion, it is marked as invalid; if there is no intersection, it passes the occlusion test.

[0109] exist In this process, candidate pose sampling bands are generated centered on each key detection region. For example, from... Heading outwards to On the spherical or toroidal surface between them, samples are uniformly taken according to azimuth and elevation angles; or, each sampling point p is assigned a direction. The camera orientation is used as the initial pose.

[0110] For each candidate pose p, the following constraint conditions are applied: when the candidate pose p simultaneously satisfies the safety constraint, the visible light constraint, and the infrared constraint, it is recorded as a joint legal pose; otherwise, it can be classified into poses that are "preferred for visible light" or "preferred for infrared" according to their weights.

[0111] It can generate dozens to hundreds of legal pose points for each key detection area.

[0112] The optimization module applies to all Joint optimization was carried out.

[0113] The objectives include maximizing the combined information gain of the two types of sensors while covering all key detection areas, and reducing the number of poses that need to be added for a single sensor.

[0114] Each valid pose is considered as a "candidate point" that can cover several key detection regions; a joint objective of set coverage problem and path cost is constructed; a heuristic algorithm (greedy + local search or genetic algorithm) is used to select a representative pose set. .

[0115] The path planning module is based on As a set of necessary points, a smooth and feasible three-dimensional trajectory is generated by combining the maximum speed, maximum acceleration, minimum turning radius, and no-fly zone boundary of the UAV, including the climb segment, tower-circling segment, and return segment.

[0116] Will The program employs a constrained traveling salesman problem to determine the order of visits; interpolates between adjacent poses to generate multi-segment spline curves; and ensures that the generated track does not intrude into no-fly zones at any point. The final output is a standard mission file, including a list of waypoints, camera trigger commands, and speed configurations, which is then distributed to the ground station or drone nesting system for one-click execution by the UAV.

[0117] In actual deployment, the following effects are achieved: the same flight path serves both visible light and infrared inspections simultaneously, reducing redundant flights; and it ensures stable viewing angles at infrared temperature measurement points, facilitating cross-time-series temperature trend analysis.

[0118] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A structure-aware based unmanned aerial vehicle (UAV) inspection shooting pose optimization method, characterized in that, The method comprises: acquiring three-dimensional structure data of a measured object; determining a key detection region according to the three-dimensional structure data; generating a detectability constraint condition for the key detection region according to imaging system parameters; generating a safety constraint condition according to safety specification information; performing occlusion detection on a line connecting a candidate pose and the key detection region according to the three-dimensional structure data to generate an occlusion constraint condition; sampling legal poses that satisfy the detectability constraint condition, the safety constraint condition, and the occlusion constraint condition; selecting a representative pose set from the legal poses of each key detection region according to key detection region coverage, pose quantity, and flight path length; performing path planning based on the representative pose set to generate a UAV inspection flight path. 2.The structure perception based UAV inspection shooting pose optimization method of claim 1, wherein, The three-dimensional structure data comprises one or more of a design model, a BIM model, or a point cloud model of the measured object. 3.The structure-aware based UAV inspection shooting pose optimization method of claim 1, wherein, The key detection region records position coordinates, an approximate normal direction, an expected defect type, and a minimum defect size. 4.The structure-aware based UAV inspection shooting pose optimization method of claim 1, wherein, The detectability constraint condition comprises a maximum allowed shooting distance. The maximum allowed photographing distance is calculated according to the following formula: ; wherein, is the maximum allowed shooting distance, is the key detection area, is the candidate pose, is the distance between the candidate pose and the center point of the key detection area, s is the pixel size of the imaging system parameter, and f is the focal length of the imaging system parameter.

5. The structure perception based UAV inspection shooting pose optimization method according to claim 1, characterized in that, The detectability constraint condition comprises a range of angles between a shooting optical axis and a detection surface normal vector, and the range of angles between the shooting optical axis and the detection surface normal vector is set according to the type of the key detection region. 6.The structure-aware based UAV inspection shooting pose optimization method of claim 1, wherein, The safety constraint condition defines an allowed flight space and a forbidden flight space of the UAV.

7. The structure perception based UAV inspection shooting pose optimization method according to claim 1, characterized in that, The occlusion constraint condition is generated by performing occlusion detection on a line connecting a candidate pose and the key detection region according to the three-dimensional structure data, comprising: when the line connecting the candidate pose and the key detection region intersects the three-dimensional structure data, determining that the candidate pose is a pose that satisfies the occlusion constraint condition; when the line connecting the candidate pose and the key detection region does not intersect the three-dimensional structure data, determining that the candidate pose is a pose that does not satisfy the occlusion constraint condition. 8.The structure-aware based UAV inspection shooting pose optimization method of claim 1, wherein, The representative pose set is selected from the legal poses of each key detection region according to key detection region coverage, pose quantity, and flight path length, comprising: selecting the representative pose set from the legal poses of each key detection region to maximize key detection region coverage, minimize pose quantity, and minimize flight path length. 9.The structure-aware based UAV inspection shooting pose optimization method of claim 1, wherein, The path planning is performed based on the representative pose set to generate the UAV inspection flight path, comprising: performing path planning on the representative pose set according to a maximum speed, a maximum acceleration, a minimum turning radius, and a forbidden flight volume boundary of the UAV to generate the UAV inspection flight path.

10. A UAV inspection shooting pose optimization system based on structure perception, comprising: a structure modeling module configured to acquire three-dimensional structure data of a measured object; a key region labeling module configured to determine a key detection region according to the three-dimensional structure data; a constraint generation module configured to generate a detectability constraint condition for the key detection region according to a preset minimum defect size and imaging system parameters, generate a safety constraint condition according to safety specification information, and perform occlusion detection on a line connecting a candidate pose and the key detection region according to the three-dimensional structure data to generate an occlusion constraint condition; a pose screening module configured to sample legal poses that satisfy the detectability constraint condition, the safety constraint condition, and the occlusion constraint condition; and An optimization module is configured to select a representative pose set from the legal poses of each key detection area according to the key detection area coverage, the number of poses, and the length of the flight path. A flight path generation module is configured to perform path planning based on the representative pose set and generate a UAV inspection flight path.

Images