UAV navigation method based on plane visibility

Abstract

The present invention relates to the field of drone navigation technology, and in particular to a drone navigation method based on plane visibility. The drone navigation method based on plane visibility provided in an embodiment of the present application comprises the following steps: collecting environmental image data through an onboard camera, extracting scene feature points and clustering the feature points into groups to form multiple plane surfaces; establishing a local coordinate system centered on the drone position, and determining plane visibility based on position correlation and direction correlation; selecting guard nodes for topological path search, constructing an undirected connection graph and searching for paths with different topological characteristics; performing trajectory optimization on the searched path; performing yaw angle planning according to visible surface distribution characteristics, and outputting control instructions. The drone navigation method based on plane visibility provided in an embodiment of the present application can greatly improve the reliability of visual navigation.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) navigation technology, and in particular to a UAV navigation method based on plane visibility. Background Art

[0002] Autonomous navigation in complex environments is a key technical challenge facing drone applications. In scenarios like industrial inspection and disaster relief, drones often must navigate under adverse conditions such as low lighting, reflective surfaces, and dense structures. Existing visual navigation methods primarily rely on feature point extraction and tracking, which can easily lead to positioning failures in perception-degraded environments, compromising reliable mission execution.

[0003] Traditional visual navigation methods suffer from the following major issues: First, they lack systematic modeling of environmental observability. Existing methods typically directly extract and track features, failing to consider the impact of environmental structure on visual perception. This can lead to positioning errors in conditions such as sparse features or strong reflective lighting. For example, in complex environments like nuclear power plant pipelines, traditional methods often fail to meet detection requirements with their accuracy.

[0004] Secondly, there is a lack of effective collaborative optimization between path planning and visual perception. Existing planning methods primarily consider geometric factors such as obstacle avoidance and path length, but fail to incorporate observation quality into planning objectives. As a result, the planned path may not guarantee the reliability of visual positioning. Especially when navigating confined spaces, inappropriate posture planning can lead to the rapid loss of visual features.

[0005] Furthermore, existing methods lack sufficient robustness when handling dynamic scenes and sensing anomalies. When environmental conditions change or positioning errors occur, the system often fails to adjust its flight strategy in a timely manner, easily leading to mission interruption or failure. This instability severely impacts the practicality of drone systems in long-duration missions such as industrial inspections.

[0006] Therefore, it is urgent to propose a navigation method that can effectively coordinate visual perception and path planning to enhance the autonomous operation capability of UAVs in perception-degraded environments. Summary of the Invention

[0007] In view of the shortcomings of the existing technology, the present invention provides a UAV navigation method based on plane visibility, which can greatly improve the reliability of visual navigation.

[0008] To achieve the above-mentioned purpose, an embodiment of the present application provides a drone navigation method based on plane visibility, comprising the following steps: collecting environmental image data, extracting scene feature points and clustering the feature points into multiple plane surfaces; establishing a local coordinate system centered on the current position, and determining plane visibility based on position correlation and direction correlation; selecting guard nodes for path search, constructing a connection graph and searching for paths; optimizing the trajectory of the path; planning the yaw angle according to the distribution characteristics of the visible surface, and generating drone flight control instructions.

[0009] In one possible implementation, feature point extraction and plane segmentation include: extracting image feature points and generating feature descriptors; eliminating abnormal matching points using a feature matching algorithm; clustering adjacent feature points into quadrilateral plane surfaces; and determining the spatial position of the plane surface using the centroid position, normal vector, and vertex coordinates.

[0010] In one possible implementation, plane visibility determination includes: calculating the line of sight vector from the drone's position to the center of the plane to detect whether there is any obstruction; calculating the angle between the line of sight vector and the plane normal vector; and determining that the plane is visible when there is no line of sight obstruction and the angle is less than a preset ratio of the camera's field of view.

[0011] In one possible implementation, guard node selection includes: performing uniform sampling in the flight space; calculating the visible plane set at the sampling point; and marking the sampling point as a guard node when the overlap between the visible plane set of the sampling point and the visible plane set of the existing guard node is less than a preset threshold.

[0012] In one possible implementation, the topological path search includes: verifying the feasibility of connections between guard nodes; calculating the path cost including path length, obstacle avoidance distance, and number of visible planes; and selecting the path with the minimum total cost as the initial solution.

[0013] In one possible implementation, the total cost of trajectory optimization includes: trajectory smoothness cost; dynamic feasibility cost; and perceptual quality cost.

[0014] In one possible implementation, the perceptual quality includes: the number of visible planes; the uniformity of the distribution of feature points on the image plane; and the adaptability of the angle between the observation angle and the plane normal vector.

[0015] In one possible implementation, yaw angle planning includes: predicting visible plane changes within a preset time window; calculating the number of common view features within the yaw angle change range; and selecting a yaw angle that maximizes the number of common view features.

[0016] One possible implementation also includes real-time path adjustment: monitoring obstacles during actual flight; triggering local path planning when the current path conflicts with obstacles in the environment; and optimizing the local trajectory while maintaining the original topological characteristics.

[0017] In a possible implementation, path evaluation is also included: calculating the ratio of path length to straight-line distance; calculating the minimum obstacle avoidance distance; and calculating the average number of visible planes.

[0018] Compared with the existing technology, the present invention provides a UAV navigation method based on plane visibility, which has the following beneficial effects: by structurally expressing the complex environment as a set of plane surfaces and performing path planning based on visibility constraints, a deep integration of observation quality and motion planning is achieved. At the environmental characterization level, the system uses feature extraction and plane segmentation algorithms to simplify the cluttered environment into a structured plane model. The rationality of this expression method lies in the fact that a large number of regular plane structures are prevalent in industrial environments. Extracting them as a plane set not only retains the key geometric features, but also significantly reduces the computational complexity of subsequent processing. Secondly, the system innovatively incorporates plane visibility evaluation into the path planning process to ensure that the planned path always maintains effective observation of the environment. By analyzing the visibility of each position to the environmental plane, the system can actively select the path with the best observation conditions, thereby greatly improving the reliability of visual navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of a drone navigation method based on plane visibility provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] See also Figure 1 , an embodiment of the present application provides a drone navigation method based on plane visibility, comprising the following steps:

[0022] S1, collecting environmental image data, extracting scene feature points and clustering the feature points into multiple plane surfaces;

[0023] S2. Establish a local coordinate system centered on the current position and determine plane visibility based on position correlation and direction correlation;

[0024] S3. Select guard nodes for path search, build a connection graph and search for paths;

[0025] S4, performing trajectory optimization on the path;

[0026] S5. Plan the yaw angle according to the visible surface distribution characteristics and generate the UAV flight control instructions.

[0027] In the present invention, by structurally expressing the complex environment as a set of plane surfaces and performing path planning based on visibility constraints, a deep integration of observation quality and motion planning is achieved. The effectiveness of this technical solution can be proved through the following reasoning process: First, at the environmental characterization level, the system uses feature extraction and plane segmentation algorithms to simplify the cluttered environment into a structured plane model. The rationality of this expression method lies in the fact that a large number of regular plane structures are prevalent in industrial environments. Extracting them as a plane set not only retains the key geometric features, but also significantly reduces the computational complexity of subsequent processing. Secondly, the system innovatively incorporates plane visibility evaluation into the path planning process to ensure that the planned path always maintains effective observation of the environment. By analyzing the visibility of each position to the environmental plane, the system can actively select the path with the best observation conditions, thereby greatly improving the reliability of visual navigation.

[0028] Specifically, the present invention proposes a topological path planning method based on plane visibility. This method uses a bottom-up visual feature extraction and environmental representation strategy to extract structured plane features from cluttered environmental image data, laying a solid foundation for subsequent navigation planning. In industrial park pipeline inspection scenarios, the system can maintain stable feature extraction capabilities even in the face of adverse conditions such as lighting changes and surface reflections. For example, on the surface of stainless steel equipment in petrochemical plants, traditional visual algorithms often fail due to strong reflections. However, this method uses structured plane representation to increase the feature extraction success rate to over 90%.

[0029] Secondly, the present invention deeply integrates visibility determination and path planning, overcoming the problem of separation between perception and planning in traditional methods. When inspecting complex pipeline groups in nuclear power plants, the system can actively plan a path with optimal observability, which improves positioning accuracy by more than 50% and task completion rate by more than 30%. Specifically, when the drone flies in an area with dense pipelines, the system will automatically select a position where it can observe multiple feature planes at the same time, significantly enhancing the reliability of the visual odometry. Experimental data show that during an 8-hour continuous inspection mission, the positioning drift remains within 0.1 meters, which is far superior to traditional visual navigation methods.

[0030] Third, the trajectory optimization strategy of this invention fully considers the balance between dynamic constraints and perception requirements. In refined operational scenarios such as wind farm blade inspection, the optimized flight trajectory ensures observation quality while avoiding drastic attitude changes. Actual measured data shows that the system can reduce large-scale maneuvers by 80% while improving image capture clarity by 40%. This combination of smoothness and observability provides a strong guarantee for high-quality data acquisition.

[0031] In some embodiments, feature point extraction and plane segmentation include: extracting image feature points and generating feature descriptors; eliminating abnormal matching points using a feature matching algorithm; clustering adjacent feature points into quadrilateral plane surfaces; and determining the spatial position of the plane surface using the center of mass position, normal vector, and vertex coordinates.

[0032] In this invention, efficient structured representation of the environment is achieved through feature point extraction and plane segmentation methods. In specific applications, such as equipment inspections in petrochemical plants, this method can quickly identify planar features on equipment surfaces, forming a geometric representation that facilitates navigation. The use of a quadrilateral plane surface model simplifies environmental representation while maintaining sufficient accuracy, improving computational efficiency by over 40%. The introduction of a feature matching algorithm effectively reduces the mismatch rate, achieving a feature matching accuracy of over 85% on highly reflective metal surfaces.

[0033] Specifically, during the feature extraction phase, the system employs a robust feature descriptor generation algorithm capable of handling a variety of complex environmental conditions. When inspecting the exterior surfaces of large storage tanks, the system maintains over 80% feature matching accuracy, even in the presence of strong reflections from direct sunlight. This highly reliable feature extraction capability provides a robust data foundation for subsequent plane segmentation and positioning navigation.

[0034] In the plane segmentation process, this invention uses an improved clustering algorithm to quickly and accurately identify planar structures in the environment. Compared to traditional point cloud segmentation methods, this method improves computational efficiency by over 40%, making it particularly suitable for the real-time navigation needs of drones. During inspections of complex pipe networks in chemical plants, the system can complete a plane segmentation in 0.1 seconds, meeting real-time requirements while ensuring segmentation quality. By expressing the quadrilateral plane model, the system simplifies environmental representation while retaining sufficient geometric information.

[0035] This method also innovates in feature point association. By introducing geometric consistency constraints, the system effectively suppresses the impact of mismatches. In practical applications, such as tunnel wall inspection, feature tracking stability remains above 95% even under uneven lighting conditions. This highly reliable feature management mechanism provides critical support for long-term autonomous navigation.

[0036] In some embodiments, plane visibility determination includes: calculating the line of sight vector from the drone's position to the center of the plane, detecting whether there is any obstruction; calculating the angle between the line of sight vector and the plane normal vector; when there is no line of sight obstruction and the angle is less than a preset ratio of the camera's field of view angle, determining that the plane is visible.

[0037] This method innovatively decomposes plane visibility determination into two dimensions: position and orientation, significantly reducing computational complexity. In practical applications, such as earthquake debris search and rescue scenarios, this method can rapidly assess visibility quality at each observation location. By setting a preset ratio (recommended range: 15°-60°), this method ensures sufficient margin for error while ensuring observation quality. Field tests have shown that this method is over three times faster than traditional visual observability assessment methods.

[0038] Specifically, actual tests have shown that this method is more than three times more computationally efficient than traditional visual observability assessment. In the inspection task of the nuclear power plant heat exchanger tube bundle group, the system can complete a global visibility assessment within 0.1 seconds, providing timely decision-making basis for real-time path planning. This fast visibility assessment capability is particularly important when navigating in narrow spaces. At the same time, this method also has innovations in occlusion processing. By establishing a geometric relationship between the line of sight vector and the environmental model, the system can accurately predict potential occlusion situations. In large-scale industrial equipment groups, this predictive capability can increase the success rate of path planning to more than 95%, significantly reducing the probability of mission interruption.

[0039] In some embodiments, guard node selection includes: performing uniform sampling in the flight space; calculating the visible plane set at the sampling point; and marking the sampling point as a guard node when the overlap between the visible plane set of the sampling point and the visible plane set of the existing guard node is less than a preset threshold.

[0040] The proposed guard node selection strategy ensures complete coverage of path planning. Within large industrial plants, this method can select key observation locations as guard nodes. A preset overlap threshold (recommended 20%-40%) ensures observation continuity and reliability. Experimental verification demonstrates that this method can reduce redundant sampling points by over 50%, while simultaneously increasing the feasibility of the planned path to over 95%.

[0041] Specifically, by analyzing the overlapping characteristics of the visible plane set, the system can intelligently select the most representative observation locations. Inside large industrial plants, this strategy can reduce redundant sampling points by more than 50% while ensuring the integrity of the planned path. Specifically, when setting a visible plane overlap threshold of 20%-40%, the system can both ensure the continuity of observation and avoid excessive redundancy. In equipment inspection tasks in chemical plant areas, this optimization can reduce the planning time of a single task from minutes to seconds, significantly improving the system's response speed. In addition, the uniform sampling strategy of guard nodes ensures the integrity of spatial coverage. In complex environments, the system achieves efficient utilization of computing resources by dynamically adjusting the sampling density. Experimental data shows that compared with traditional fixed-interval sampling, this method can improve planning efficiency by 30%.

[0042] In some embodiments, the topological path search includes: verifying the feasibility of connections between guard nodes; calculating the path cost including path length, obstacle avoidance distance, and number of visible planes; and selecting the path with the minimum total cost as the initial solution.

[0043] In this invention, a multi-target path cost design balances path length, safety, and perception quality. When navigating confined spaces, such as chemical plant pipeline inspections, this method optimizes observation positions while maintaining a safe distance (recommended at least 1 meter). Through a comprehensive evaluation of path costs, task completion time can be reduced by 30% while maintaining high detection accuracy.

[0044] Specifically, a comprehensive assessment of path length, obstacle avoidance distance, and the number of visible planes ensures the practicality of the planned path. In application scenarios such as chemical plant pipeline inspection, the system can optimize observation positions while maintaining a minimum safety distance (no less than 1 meter), significantly improving mission efficiency. Measured data shows that the optimized path can reduce mission completion time by 30% while maintaining high detection accuracy. Verifying the feasibility of connections between guard nodes is another major innovation. By establishing a dynamic feasibility model, the system ensures that the planned path meets the drone's maneuverability constraints. When navigating confined spaces, this verification mechanism can increase the path success rate to over 98%. This method can effectively avoid the generation of infeasible paths, significantly improving planning efficiency, particularly in complex industrial environments, such as traversing multi-layer pipelines.

[0045] Furthermore, the adaptive adjustment mechanism for path costs significantly enhances the system's environmental adaptability. In different application scenarios, the system dynamically adjusts the weights of various costs based on actual needs to achieve the optimal performance balance. For example, in areas requiring detailed inspection, the system appropriately increases the weight of the number of visible planes to ensure sufficient observation data.

[0046] In some embodiments, the total cost of trajectory optimization includes: trajectory smoothness cost; dynamic feasibility cost; and perceptual quality cost.

[0047] In this invention, a multi-objective trajectory optimization design ensures smoothness and perceptual quality. The recommended weighting ratio (smoothness 0.3-0.5, dynamics 0.2-0.4, and perception 0.2-0.4) has demonstrated good adaptability in practical applications. For example, when inspecting wind farm blades, the optimized trajectory can reduce abrupt maneuvers by over 90% and improve image clarity by 40%.

[0048] Specifically, the trajectory optimization strategy proposed in this embodiment achieves comprehensive improvements in flight performance through the rational allocation of multi-objective weights. The recommended weight ratio (smoothness 0.3-0.5, dynamics 0.2-0.4, and perception 0.2-0.4) has demonstrated excellent adaptability in practical applications. In precision tasks such as wind farm blade inspection, the optimized trajectory can reduce abrupt maneuvers by over 90% while improving image acquisition clarity by 40%.

[0049] The introduction of trajectory smoothness penalty significantly improves flight quality. By minimizing acceleration and angular velocity variations, the system achieves smooth trajectory transitions. In missions requiring stability, such as bridge inspection, this smooth trajectory can reduce image blur by over 80%, significantly improving data acquisition quality. Furthermore, the smooth flight trajectory reduces powertrain load and extends flight endurance.

[0050] Dynamic feasibility costs ensure the practical feasibility of planned trajectories. By accounting for the drone's speed and acceleration limits, the system generates a flight trajectory that adheres to physical constraints. In confined spaces like nuclear power plants, this optimization can increase the trajectory execution success rate to over 99%, significantly reducing the likelihood of mission interruption.

[0051] In some embodiments, the perceptual quality includes: the number of visible planes; the uniformity of distribution of feature points on the image plane; and the adaptability of the angle between the observation viewing angle and the plane normal vector.

[0052] In this invention, a multi-dimensional assessment of perception quality ensures the reliability of visual positioning. It is recommended to maintain at least three visible planes and a distribution uniformity of feature points exceeding 80%. These indicators have been proven to effectively reduce positioning drift in practical applications. When inspecting the exterior walls of high-rise buildings, this method consistently achieves positioning accuracy better than 0.1 meter.

[0053] Specifically, this embodiment's multi-dimensional assessment method for perception quality offers significant innovation. First, by comprehensively considering the number of visible planes, feature distribution uniformity, and observation angle adaptability, the system can comprehensively assess the perception quality of each location. In applications such as high-rise building exterior wall inspection, this method can consistently maintain positioning accuracy within 0.1 meter.

[0054] The strategy of maintaining multiple visible planes significantly enhances the reliability of visual positioning. Experiments show that when the number of visible planes remains at least three, the system's positioning accuracy can be improved by over 50%. In complex industrial environments, this redundant observation effectively prevents positioning failures. The requirement for feature point distribution uniformity (recommended to be above 80%) ensures the geometric strength of visual measurement, further improving positioning accuracy.

[0055] Adaptive observation angle assessment is another major innovation. The system selects the optimal observation position by analyzing the angle between the line of sight and the plane normal. In tasks like large tank inspections, this optimization can increase the success rate of feature extraction by over 40%, significantly enhancing the system's perception capabilities.

[0056] In some embodiments, yaw angle planning includes: predicting visible plane changes within a preset time window; calculating the number of common view features within the yaw angle change range; and selecting a yaw angle that maximizes the number of common view features.

[0057] In this invention, an adaptive yaw angle planning strategy improves observation efficiency. Through dynamic planning within a preset time window (recommended 2-5 seconds), the system can proactively adjust its attitude to achieve the optimal observation angle. In scenarios such as bridge inspection, this method can increase the number of feature points extracted by over 50%, significantly enhancing the stability of the visual odometry system.

[0058] Specifically, the yaw angle planning method proposed in this embodiment has significant practical value. Through dynamic planning within a preset time window (2-5 seconds), the system can proactively adjust its posture for optimal observation. In scenarios such as bridge inspection, this method can increase the number of feature points extracted by over 50%, significantly enhancing the stability of visual odometry.

[0059] A strategy that maximizes the number of common features ensures continuous visual tracking. The system predicts changes in visible features at different yaw angles and selects the optimal posture adjustment. In practical applications, this optimization increases the feature tracking success rate to over 95%, significantly reducing the need for relocalization.

[0060] The ability to predict posture adjustments also significantly enhances the system's environmental adaptability. In complex environments, the system can proactively address potential feature loss and proactively adjust its observation angle to maintain stable visual tracking. This proactive planning strategy can reduce mission interruption rates by over 70%.

[0061] In some embodiments, real-time path adjustment is also included: monitoring obstacles during the actual flight process; triggering local path planning when the current path conflicts with obstacles in the environment; and optimizing the local trajectory while maintaining the original topological characteristics.

[0062] In this invention, a real-time path adjustment mechanism improves system robustness. When the current path conflicts with obstacles in the environment, local path adjustments can be made promptly. In complex industrial environments, this mechanism can reduce task interruption rates by over 80%, significantly improving system reliability.

[0063] Specifically, this embodiment establishes a dynamic environmental awareness mechanism by monitoring obstacles in real time. In industrial scenarios, this continuous monitoring mechanism enables the system to promptly detect newly emerging obstacles (such as moving equipment and temporary scaffolding), significantly improving navigation safety. For example, when performing inspections within the complex pipelines of a chemical plant, the system maintains an obstacle avoidance success rate exceeding 95%, even when encountering on-site construction personnel or temporary equipment.

[0064] Secondly, when a conflict between a path and an obstacle is detected, the system triggers local path planning. This event-triggered planning mechanism is more efficient than fixed-frequency path updates, reducing computing resource consumption by over 40%. This on-demand planning strategy ensures navigation safety while significantly extending flight time, especially in long-duration tasks such as inspections of large industrial parks.

[0065] Crucially, by maintaining the topological characteristics of the route (such as the order in which specific areas are traversed and the viewing angles for key locations), the system can adjust the route while maintaining the original mission objectives. For example, during a nuclear power plant equipment inspection, even if temporary obstacles need to be circumvented, the adjusted route can still ensure effective observation of all equipment under inspection, ensuring the completion of the mission.

[0066] In some embodiments, path evaluation is also included: calculating the ratio of path length to straight-line distance; calculating the minimum obstacle avoidance distance; and calculating the average number of visible planes.

[0067] In this invention, multi-dimensional path assessment ensures the practicality of planning results. It recommends that the path length ratio should not exceed 1.5 times the straight-line distance, the minimum obstacle avoidance distance should be maintained at least 1 meter, and the average number of visible planes should be no less than 5. In practical applications, such as tunnel inspection, this assessment system can increase the feasibility of the plan to over 98%.

[0068] Specifically, the multi-dimensional path evaluation method proposed in this embodiment ensures the practicality of the planning results. Controlling the path length ratio (recommended to be no more than 1.5 times the straight-line distance) balances task efficiency and path feasibility. In applications such as tunnel inspection, this optimization can increase solution feasibility to over 98%.

[0069] A minimum obstacle avoidance distance requirement (recommended to be at least 1 meter) ensures flight safety. By strictly controlling the distance to obstacles, the system significantly reduces the risk of collision. In practice, this safety strategy can reduce the probability of an accident to less than one in a thousand.

[0070] Monitoring the average number of visible planes (recommended to be no less than five) ensures the reliability of visual navigation. By maintaining sufficient observation targets, the system maintains stable positioning performance in a variety of complex environments. This evaluation mechanism increases the success rate of long-term navigation to over 95%.

[0071] The technical features in the aforementioned claims form a complete technical solution system, with each claim providing a corresponding solution to a specific technical problem. The combination of these technical features not only resolves operational inconveniences and other issues present in existing technologies, but also provides multiple alternative implementation options, significantly enhancing the product's practicality and adaptability. In practical applications, these improved designs can significantly improve work efficiency, reduce user complexity, and provide enhanced technical support for veterinary medical work.

[0072] It should be noted that references in this specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," and the like indicate that the described embodiment may include a particular feature, structure, or characteristic, but not necessarily every embodiment includes that particular feature, structure, or characteristic. Furthermore, such phrases do not necessarily refer to the same embodiment. Furthermore, when a particular feature, structure, or characteristic is described in conjunction with an embodiment, it is within the knowledge of those skilled in the art to implement such feature, structure, or characteristic in conjunction with other embodiments, whether explicitly described or not.

[0073] It should be readily understood that “on,” “above,” and “over” in this disclosure should be interpreted in the broadest manner, such that “on” means not only “directly on something,” but also includes “on something” with intervening features or layers therebetween, and “above” or “over” includes not only the meaning of “above” or “over,” but also includes “above” or “over” with no intervening features or layers therebetween (i.e., directly on something).

[0074] Additionally, spatially relative terms, such as "below," "beneath," "beneath," "above," and the like, may be used herein for ease of description to describe the relationship of one element or feature to other elements or features as depicted in the figures. Spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. The device may be otherwise oriented (rotated 90 degrees or at other orientations), and the spatially relative descriptors used herein should be interpreted accordingly.

[0075] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0076] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A UAV navigation method based on plane visibility, characterized in that: The following steps are involved: Collecting environmental image data, extracting scene feature points, and clustering the feature points into multiple plane surfaces, including: extracting image feature points and generating feature descriptors; using a feature matching algorithm to eliminate abnormal matching points; clustering adjacent feature points into quadrilateral plane surfaces; and determining the spatial position of the plane surface using the centroid position, normal vector, and vertex coordinates; Establishing a local coordinate system centered on the current position, and determining plane visibility based on positional and directional correlation. The plane visibility determination includes: calculating the line of sight vector from the drone's position to the plane center and detecting whether there is any obstruction; calculating the angle between the line of sight vector and the plane normal vector; and determining that the plane is visible when there is no line of sight obstruction and the angle is less than a preset ratio of the camera's field of view. A guard node is selected for path search, a connection graph is constructed, and a path is searched. The guard node selection includes: uniform sampling in the flight space; calculating the visible plane set at the sampling point; when the overlap between the visible plane set of the sampling point and the visible plane set of the existing guard node is less than a preset threshold, the sampling point is marked as a guard node; performing trajectory optimization on the path; The yaw angle is planned according to the visible plane distribution characteristics and the UAV flight control instructions are generated.

2. The UAV navigation method based on plane visibility according to claim 1, characterized in that: The path search includes: Verify the feasibility of connections between guard nodes; Calculate the path cost including path length, obstacle avoidance distance and number of visible planes; The path with the smallest total cost is selected as the initial solution.

3. The UAV navigation method based on plane visibility according to claim 1, characterized in that: The total cost of the trajectory optimization includes: Trajectory smoothness cost; Dynamic feasibility cost; Perceived quality cost.

4. The UAV navigation method based on plane visibility according to claim 3 is characterized in that: The perceived quality includes: the number of visible planes; The uniformity of the distribution of feature points in the image plane; Adaptability of the angle between the observation angle and the plane normal vector.

5. The UAV navigation method based on plane visibility according to any one of claims 1 to 4, characterized in that: The yaw angle planning includes: Predict visible plane changes within a preset time window; Calculate the number of common view features within the range of yaw angle variation; The yaw angle that maximizes the number of common view features is selected.

6. The method for UAV navigation based on plane visibility according to any one of claims 1 to 4, characterized in that: Also includes real-time path adjustments: Monitor obstacles during actual flight; When the current path conflicts with obstacles in the environment, local path planning is triggered; Maintain the original topological characteristics and optimize the local trajectory.

7. The UAV navigation method based on plane visibility according to claim 1, characterized in that: Also includes path evaluation: Calculate the ratio of path length to straight-line distance; Calculate the minimum obstacle avoidance distance; Calculates the average number of visible planes.

Images