Heavy-load unmanned aerial vehicle landing point identification processing method and system based on visual identification

Abstract

The invention provides a heavy-load unmanned aerial vehicle landing point identification processing method and system based on visual identification, and relates to the technical field of computer vision, and the method comprises the steps: obtaining the airflow speed direction and airflow change data above a landing area of a heavy-load unmanned aerial vehicle, real-time flight data, and an enhanced panoramic image; based on the initial positioning position of the heavy-load unmanned aerial vehicle, the initial coordinates of the preset landing area and the flight task parameters, generating an initial landing trajectory to determine a qualified landing point position; based on the qualified landing point position, the enhanced panoramic image, the airflow velocity direction and the airflow change data, simulating a fuselage stress simulation result of the heavy-load unmanned aerial vehicle in the landing process; based on the qualified landing point position and the fuselage stress simulation result, obtaining corrected flight data; and based on the qualified landing point position and the corrected flight data, an optimized landing trajectory is obtained, and high-precision and high-stability landing control of the heavy-load unmanned aerial vehicle in a complex environment is realized.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and in particular to a method and system for identifying and processing landing points of heavy-duty unmanned aerial vehicles based on visual recognition. Background Technology

[0002] In typical application scenarios such as power line inspection, emergency material delivery, and industrial logistics, heavy-duty drones often need to perform high-precision autonomous landing missions in areas without fixed take-off and landing platforms or in complex environments. This not only requires the landing point surface to have extremely high flatness, but also requires the drone to be able to sense environmental disturbances in real time, accurately identify safe landing areas, and dynamically adjust its flight attitude to maintain a stable descent trajectory under conditions of strong crosswinds, gusts, or turbulent airflow.

[0003] To address these requirements, current mainstream solutions employ a landing control strategy based on the fusion of single-frame high-resolution visual recognition and inertial navigation. However, existing solutions have certain limitations. For example, single-frame images are susceptible to changes in lighting, lens distortion, or brief occlusions, making it difficult to fully and robustly reproduce the microscopic smoothness characteristics of the landing surface, leading to misjudgments of qualified areas. Furthermore, the control logic does not consider the coupling effect of aerodynamic loads dynamically changing with altitude and wind field during landing, making it impossible to finely coordinate the thrust of the multi-rotor, resulting in landing accuracy and stability in turbulent airflow environments failing to meet the requirements of industrial-grade heavy-duty operations. Summary of the Invention

[0004] The purpose of this invention is to provide a visual recognition-based method and system for identifying and processing landing points of heavy-duty UAVs, in order to solve problems such as misjudging qualified areas in the prior art and the difficulty in meeting the requirements of industrial-grade heavy-duty operations in terms of landing accuracy and stability in turbulent airflow environments.

[0005] In a first aspect, the present invention provides a visual recognition-based method for identifying and processing the landing point of a heavy-duty unmanned aerial vehicle (UAV), comprising:

[0006] Data on airflow speed, direction, and changes in airflow over the landing area of ​​the heavy-duty UAV, as well as real-time flight data and landing area images, are acquired. The landing area images are then stitched together and enhanced to obtain an enhanced panoramic image.

[0007] The initial landing trajectory is generated based on the initial positioning of the heavy-duty UAV, the initial coordinates of the preset landing area, and the flight mission parameters.

[0008] The visual detection dataset of landing qualification features of heavy-duty drones in the field of industrial internet is compared with the enhanced panoramic image to determine the location of qualified landing points.

[0009] Based on the qualified landing point location, the enhanced panoramic image, the airflow speed and direction, and the airflow change data, the simulation results of the fuselage force of the heavy-load UAV during the landing process are simulated using digital twin technology.

[0010] Based on the qualified landing point location and the stress simulation results of the fuselage, the lift distribution ratio of each rotor of the heavy-load UAV is calculated to correct the real-time flight data and obtain the corrected flight data.

[0011] Based on the qualified landing point location and the corrected flight data, the initial landing trajectory is optimized to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV.

[0012] Optionally, based on the initial positioning of the heavy-load UAV, the initial coordinates of the preset landing area, and flight mission parameters, an initial landing trajectory is generated, including:

[0013] Calculate the straight-line distance and relative orientation between the initial positioning position of the heavy-load UAV and the initial coordinates of the preset landing area;

[0014] Based on the straight-line distance and the relative orientation, and combined with the maximum flight speed, maximum descent speed and allowable flight altitude range in the flight mission parameters of the heavy-load UAV, multiple trajectory nodes of the heavy-load UAV are set.

[0015] Based on the turning limit angle in the flight mission parameters, the flight direction and speed of the heavy-load UAV from the previous trajectory node to the next trajectory node are calculated, and the initial flight path is generated by combining the arrangement order of the trajectory nodes.

[0016] Based on the maximum flight speed, the distance between adjacent trajectory nodes in the initial flight path, and the height difference between nodes, the descent rate parameters of each trajectory node in the initial flight path are calculated, and the initial landing trajectory is generated by combining the initial flight path.

[0017] Optionally, the visual detection dataset of landing qualification features of heavy-duty drones in the industrial internet field is compared with the enhanced panoramic image to determine the location of qualified landing points, including:

[0018] The visual detection dataset of landing qualification features of heavy-duty UAVs in the field of industrial Internet is compared with the enhanced panoramic image to obtain the comparison results, which include shape comparison results, size comparison results and distribution continuity comparison results.

[0019] Based on the comparison results, target sub-regions are selected from the enhanced panoramic image;

[0020] Based on the boundary range and center coordinates of each target sub-region, multiple candidate landing points are determined;

[0021] The terrain flatness and coverage of each candidate landing point are calculated, and the location of the candidate landing point that meets the preset conditions in terms of both terrain flatness and coverage is taken as the qualified landing point location.

[0022] Optionally, based on the comparison results, the target sub-region is selected from the enhanced panoramic image, including:

[0023] Based on the comparison results, and combined with the shape qualification threshold, size qualification threshold, and distribution continuity qualification threshold in the visual detection dataset, the qualification dimension ratio of each sub-region to be detected in the enhanced panoramic image is calculated.

[0024] The sub-regions to be detected with a qualified dimension ratio greater than or equal to the preset qualified dimension ratio threshold are designated as target sub-regions. If there are sub-regions to be detected with overlapping regions, the sub-region with the larger region is designated as the target sub-region.

[0025] Optionally, based on the qualified landing point location, the enhanced panoramic image, the airflow velocity direction, and the airflow change data, digital twin technology is used to simulate the fuselage forces of the heavy-load UAV during landing, including:

[0026] Based on the qualified landing point location and the terrain features of the enhanced panoramic image, a digital twin simulation scene is constructed;

[0027] The airflow velocity direction and the airflow change data are converted into airflow parameters that can be recognized by the digital twin simulation scene. Combined with the airflow obstruction and disturbance law and the terrain features, the airflow action parameters of the heavy-load UAV are set.

[0028] Based on the initial landing trajectory, the qualified landing point location in the digital twin simulation scenario, and the airflow parameters, combined with the fuselage structure parameters and flight performance parameters of the heavy-load UAV, the airflow force, gravity, and air resistance experienced by the heavy-load UAV in each landing phase are calculated to form the fuselage force simulation results.

[0029] Secondly, the present invention provides a visual recognition-based landing point identification and processing system for heavy-duty unmanned aerial vehicles, comprising:

[0030] The acquisition module is used to acquire airflow speed, direction and airflow change data above the landing area of ​​the heavy-load UAV, as well as real-time flight data and landing area images. The landing area images are then stitched and enhanced to obtain an enhanced panoramic image.

[0031] The generation module is used to generate the initial landing trajectory based on the initial positioning of the heavy-load UAV, the initial coordinates of the preset landing area, and the flight mission parameters.

[0032] The comparison module is used to compare the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial Internet field with the enhanced panoramic image to determine the location of qualified landing points.

[0033] The simulation module is used to simulate the stress on the fuselage of the heavy-load UAV during the landing process using digital twin technology, based on the qualified landing point location, the enhanced panoramic image, the airflow speed and direction, and the airflow change data.

[0034] The correction module is used to calculate the lift distribution ratio of each rotor of the heavy-load UAV based on the qualified landing point location and the stress simulation results of the fuselage, so as to correct the real-time flight data and obtain the corrected flight data.

[0035] The optimization module optimizes the initial landing trajectory based on the qualified landing point location and the corrected flight data to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV.

[0036] Thirdly, the present invention provides an electronic device, comprising:

[0037] Memory, used to store computer programs;

[0038] A processor, configured to execute the computer program to implement the steps of the vision-based heavy-load UAV landing point identification processing method as described in the first aspect above.

[0039] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the visual recognition-based heavy-load UAV landing point identification processing method described in the first aspect above.

[0040] The present invention provides a visual recognition-based landing point identification and processing method for heavy-load UAVs. This method acquires airflow velocity, direction, and change data above the landing area of ​​the heavy-load UAV, as well as real-time flight data and landing area images. The landing area images are then stitched and enhanced to obtain an enhanced panoramic image. An initial landing trajectory is generated based on the initial positioning of the heavy-load UAV, the initial coordinates of the preset landing area, and flight mission parameters. A visual detection dataset of landing qualification features for heavy-load UAVs in the industrial internet field is compared with the enhanced panoramic image to determine the qualified landing point location. Based on the qualified landing point location and the enhanced panoramic image... The airflow velocity direction and airflow change data are used to simulate the fuselage stress of the heavy-load UAV during landing using digital twin technology. Based on the qualified landing point location and the fuselage stress simulation results, the lift distribution ratio of each rotor of the heavy-load UAV is calculated to correct the real-time flight data, resulting in corrected flight data. Based on the qualified landing point location and the corrected flight data, the initial landing trajectory is optimized to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV. Multi-frame image stitching and enhancement processing overcome the problem of incomplete information in single-frame images under complex lighting, occlusion, or distortion conditions. This system enhances the integrity and robustness of the visual representation of the landing area, laying the foundation for subsequent high-precision identification; ensures the system possesses an executable baseline control strategy, avoiding energy and time waste caused by blind searches; identifies landing areas that meet engineering safety specifications, solving the misjudgment problem caused by the lack of smoothness semantic understanding in traditional methods; overcomes the limitation of existing solutions ignoring wind field-structure coupling effects; achieves refined coordinated control of multi-rotor thrust, enabling flight response to actively compensate for attitude disturbances caused by turbulent airflow, improving flight stability under heavy load conditions; and ensures that the UAV can still complete landing maneuvers accurately and smoothly under complex weather conditions, meeting the requirements of industrial-grade operations. The system meets stringent requirements for safety and reliability. Furthermore, it compares the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial internet field with the enhanced panoramic image to obtain multi-dimensional comparison results, including shape, size, and distribution continuity. Based on this, target sub-regions are selected, and multiple candidate landing points are extracted based on the boundary range and center coordinates of each target sub-region. The terrain flatness and coverage of each candidate landing point are further calculated, and finally, the candidate landing point locations that simultaneously meet the preset terrain flatness and coverage conditions are determined as qualified landing point locations. This solves the technical bottleneck of existing solutions that struggle to simultaneously consider both flatness semantic understanding and spatial continuity judgment. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating the visual recognition-based landing point identification and processing method for heavy-load UAVs provided in an embodiment of the present invention.

[0043] Figure 2 This is a schematic diagram illustrating a specific implementation of the visual recognition-based landing point identification and processing method for heavy-duty unmanned aerial vehicles provided in an embodiment of the present invention.

[0044] Figure 3 This is a schematic diagram of the structure of a visual recognition-based heavy-duty UAV landing point identification and processing system provided in an embodiment of the present invention. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely 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.

[0046] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0047] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0048] To address the technical challenge of achieving high-precision and high-stability autonomous landing of heavy-duty drones in scenarios such as power line inspection, emergency material delivery, and industrial logistics, where there is no fixed take-off and landing platform and there is turbulent airflow, existing solutions based on the fusion of single-frame vision and inertial navigation are limited by incomplete image information and static control logic, which cannot accurately restore the microscopic flatness of the landing surface, and are also difficult to cope with the dynamic coupling effect of aerodynamic disturbance and airframe response.

[0049] To address this, this invention constructs a robust panoramic visual representation by stitching and enhancing multiple frames of images, overcoming the blind spots in recognition of single-frame images in complex environments. It then introduces a standardized landing qualification feature dataset from the industrial internet field for structured comparison, accurately selecting landing points that meet engineering safety requirements. Simultaneously, by combining real-time airflow disturbance data with digital twin technology, it dynamically simulates the fuselage stress state during landing, thereby driving precise adjustments to the multi-rotor lift distribution. This achieves proactive adaptation to airflow disturbances and high-precision closed-loop control of the heavy-load landing process, solving the problems of inaccurate recognition, delayed response, and coarse control in existing solutions under complex weather and unstructured site conditions.

[0050] Example 1

[0051] Embodiment 1 of the present invention provides a visual recognition-based landing point identification and processing method for heavy-load unmanned aerial vehicles (UAVs), and a flowchart of a specific implementation is shown below. Figure 1 As shown, the method includes:

[0052] Step 101: Obtain airflow speed, direction, and airflow change data above the landing area of ​​the heavy-load UAV, as well as real-time flight data and landing area images. Then, stitch and enhance the landing area images to obtain an enhanced panoramic image.

[0053] In this step, "heavy-load UAV" refers to an unmanned aerial vehicle with a large payload capacity and capable of performing landing missions; "above the landing area" refers to the space area directly above and around the pre-set landing area; "airflow speed and direction" refers to the specific direction of airflow above the landing area; "airflow change data" refers to relevant information on the changes in airflow speed and direction over time above the landing area; "real-time flight data" refers to the operational status data such as the aircraft attitude and position collected in real time during the flight of the heavy-load UAV; "landing area image" refers to an image containing the pre-set landing area captured by the visual acquisition equipment carried by the heavy-load UAV; and "enhanced panoramic image" refers to an image that completely covers the pre-set landing area after stitching and integrating multiple landing area images and improving detail recognition through image enhancement processing.

[0054] In this embodiment of the invention, airflow speed, direction and airflow change data above the landing area are collected by an airflow sensor carried by a heavy-duty UAV, real-time flight data is collected by a flight status sensor, and images of the landing area are captured by a visual acquisition device. Multiple images of the landing area are stitched together according to the pixel overlap area to fill the gaps in the images to form a preliminary panoramic image. Then, the image is enhanced by adjusting the image contrast and sharpening details to finally obtain an enhanced panoramic image.

[0055] Step 102: Generate the initial landing trajectory based on the initial positioning of the heavy-duty UAV, the initial coordinates of the preset landing area, and the flight mission parameters.

[0056] In this step, the initial positioning position refers to the spatial coordinate position of the heavy-load UAV when it begins to perform the landing mission; the initial coordinates of the preset landing area refer to the coordinate information of the boundary and center of the planned landing area of ​​the heavy-load UAV; the flight mission parameters refer to the various performance limitations and requirements parameters of the heavy-load UAV when performing the landing mission; and the initial landing trajectory refers to the preliminary flight path of the heavy-load UAV from the starting position to the preset landing area, planned based on the initial positioning position and the initial coordinates of the preset landing area.

[0057] Step 103: Compare the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial internet field with the enhanced panoramic image to determine the location of the qualified landing point.

[0058] In this step, the visual inspection dataset for landing qualification features refers to a pre-compiled dataset in the industrial internet field that contains surface feature standards for the safe landing of heavy-duty drones; the qualified landing point location refers to the specific landing coordinates selected from the enhanced panoramic image that meet the requirements of landing qualification features.

[0059] Step 104: Based on the qualified landing point location, the enhanced panoramic image, the airflow speed and direction, and the airflow change data, simulate the stress on the fuselage of the heavy-load UAV during the landing process using digital twin technology.

[0060] In this step, the fuselage stress simulation results refer to the relevant information on the various forces acting on the fuselage of a heavy-load UAV during landing, simulated using digital twin technology.

[0061] Step 105: Based on the qualified landing point location and the fuselage force simulation results, calculate the lift distribution ratio of each rotor of the heavy-load UAV to correct the real-time flight data and obtain the corrected flight data.

[0062] In this step, the lift distribution ratio of each rotor refers to the proportion of lift output by each of the multiple rotors of the heavy-load UAV to the total lift; the corrected flight data refers to the real-time attitude, position and other operational status data of the fuselage after lift adjustment.

[0063] Step 106: Based on the qualified landing point location and the corrected flight data, optimize the initial landing trajectory to obtain the optimized landing trajectory, so as to realize the landing control of the heavy-load UAV.

[0064] In this step, the optimized landing trajectory refers to the final flight path adjusted by combining the corrected flight data and the qualified landing point location.

[0065] The embodiments of the present invention enable precise landing of heavy-load UAVs in complex airflow environments, solve the landing deviation problem caused by airflow disturbance, and improve the stability and accuracy of the landing process.

[0066] This invention provides a specific embodiment. Step 102 involves generating an initial landing trajectory based on the initial positioning of the heavy-load UAV, the initial coordinates of the preset landing area, and flight mission parameters. This specifically includes the following steps:

[0067] Step 201: Calculate the straight-line distance and relative orientation between the initial positioning position of the heavy-load UAV and the initial coordinates of the preset landing area.

[0068] In this step, the straight-line distance refers to the spatial straight-line length between the initial positioning position and the initial coordinates of the preset landing area; the relative orientation refers to the horizontal angle between the initial positioning position and the initial coordinates of the preset landing area.

[0069] In this embodiment of the invention, the three-dimensional coordinates of the initial positioning position are obtained through the positioning module. The three-dimensional coordinates of the initial coordinates of the preset landing area The straight-line distance and relative orientation between the two are calculated based on three-dimensional coordinates. , , where arctan is the arctangent function, used to calculate the direction angle on the horizontal plane.

[0070] Step 202: Based on the straight-line distance and the relative orientation, and combined with the maximum flight speed, maximum descent speed and allowable flight altitude range in the flight mission parameters of the heavy-load UAV, set multiple trajectory nodes for the heavy-load UAV.

[0071] In this step, maximum flight speed refers to the highest flight rate allowed when the heavy-load UAV is performing a mission; maximum descent speed refers to the highest vertical descent rate allowed during the landing of the heavy-load UAV; allowed flight altitude range refers to the altitude range allowed when the heavy-load UAV is flying; and trajectory nodes refer to key points with clear coordinates and altitudes marked on the initial landing trajectory.

[0072] In this embodiment of the invention, the approximate flight time of the heavy-load UAV from the initial positioning position to the initial coordinates of the preset landing area is determined based on the calculated straight-line distance and relative orientation, combined with the maximum flight speed in the flight mission parameters. The vertical altitude change rhythm is determined by combining the maximum descent speed and the allowable flight altitude range. Multiple trajectory nodes are set near the line connecting the initial positioning position and the initial coordinates of the preset landing area at uniform time intervals or distance intervals. Each trajectory node contains corresponding three-dimensional coordinates and altitude information.

[0073] Step 203: Based on the turning limit angle in the flight mission parameters, calculate the flight direction and speed of the heavy-load UAV from the previous trajectory node to the next trajectory node, and generate the initial flight path by combining the arrangement order of the trajectory nodes.

[0074] In this step, the turning limit angle refers to the maximum turning angle allowed when the heavy-load UAV is flying; the previous trajectory node refers to a certain trajectory node in the trajectory node sequence; the next trajectory node refers to the adjacent trajectory node in the trajectory node sequence that follows the previous trajectory node; the flight direction refers to the horizontal direction in which the heavy-load UAV flies from the current trajectory node to the next trajectory node; the flight speed refers to the speed at which the heavy-load UAV flies between two adjacent trajectory nodes; and the initial flight path refers to the continuous flight route formed by connecting all trajectory nodes in sequence and combining the flight direction and flight speed between each node.

[0075] In this embodiment of the invention, the turning limit angle in the flight mission parameters is extracted. For each pair of adjacent preceding and following trajectory nodes, the horizontal angle difference between them is calculated to ensure that the angle difference does not exceed the turning limit angle. The flight direction is determined based on the angle difference. The flight speed is calculated by combining the horizontal distance between the preceding and following trajectory nodes and the preset flight time. According to the arrangement order of the trajectory nodes, all trajectory nodes are sequentially connected in series with the corresponding flight direction and flight speed to form a continuous initial flight path without turning limit exceedance.

[0076] Step 204: Based on the maximum flight speed, the node distance and node height difference between adjacent trajectory nodes in the initial flight path, calculate the descent rate parameters of each trajectory node in the initial flight path, and generate the initial landing trajectory by combining the initial flight path.

[0077] In this step, node distance refers to the straight-line distance between two adjacent trajectory nodes; node height difference refers to the altitude difference between two adjacent trajectory nodes; and descent rate parameter refers to the vertical descent rate of the heavy-load UAV when flying between two adjacent trajectory nodes.

[0078] In this embodiment of the invention, the node distance and node height difference of each group of adjacent trajectory nodes in the initial flight path are calculated. ,in, The coordinates of the previous trajectory node. The coordinates of the next trajectory node. The flight time between adjacent trajectory nodes is calculated by combining the maximum flight speed and node distance, and the descent rate parameter is calculated. The descent rate parameter = node altitude difference ÷ flight time. The descent rate parameters of all trajectory nodes are integrated with the initial flight path to form an initial landing trajectory that includes coordinates, altitude, flight direction, flight speed and descent rate parameters.

[0079] The embodiments of the present invention provide a stable and reliable basic path for the subsequent optimization and adjustment of the landing process, thereby improving the rationality and feasibility of the landing trajectory.

[0080] For example, the initial positioning position of heavy-duty drone A is The initial coordinates of the preset landing area are: The flight mission parameters include a maximum flight speed of 15 m / s, a maximum descent speed of 4 m / s, a permissible flight altitude range of 0 to 100 m, and a limited turning angle. First calculate , Based on the above parameters, set 5 trajectory nodes every 50m: Node 1 (850m, 550m, 80m), Node 2 (900m, 600m, 60m), Node 3 (940m, 640m, 40m), Node 4 (970m, 670m, 20m), and Node 5 (990m, 690m, 10m). Calculate the horizontal angle difference between adjacent nodes to ensure it does not exceed [the specified value]. The flight direction between each node is determined to be around Fine-tuning was performed, and the flight speed was controlled between 12 and 15 m / s based on the node distance calculation. The initial flight path was formed by connecting the nodes. The node distance and node height difference between adjacent nodes were calculated. For example, the node distance between node 1 and node 2 is ≈70.7m, the node height difference is 20m, the flight time is ≈5.9s, and the descent rate parameter is ≈3.4m / s. All parameters were integrated to generate the initial landing trajectory.

[0081] This invention provides a specific embodiment, such as... Figure 2 As shown, step 103 involves comparing the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial internet field with the enhanced panoramic image to determine the location of the qualified landing point. This specifically includes the following steps:

[0082] Step 301: Compare the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial Internet field with the enhanced panoramic image to obtain the comparison results. The comparison results include shape comparison results, size comparison results, and distribution continuity comparison results.

[0083] In this step, the comparison result refers to the comprehensive information generated after comparing the landing qualification features of heavy-duty UAVs in the industrial internet field with the enhanced panoramic image; the shape comparison result refers to the comparison information between the landing qualification shape features in the visual inspection dataset and the shape features of each region in the enhanced panoramic image; the size comparison result refers to the comparison information between the landing qualification size features in the visual inspection dataset and the size features of each region in the enhanced panoramic image; and the distribution continuity comparison result refers to the comparison information between the landing qualification distribution continuity features in the visual inspection dataset and the distribution continuity features of each region in the enhanced panoramic image.

[0084] In this embodiment of the invention, the preset landing qualification shape standard, size standard and distribution continuity standard are extracted from the visual detection dataset. These standards are compared one by one with the shape, size and distribution continuity features of all identifiable regions in the enhanced panoramic image. The relevant information of each region in terms of shape matching degree, size fit degree and distribution continuity is recorded. This information is integrated to form a comparison result that includes shape comparison result, size comparison result and distribution continuity comparison result.

[0085] Step 302: Based on the comparison results, select the target sub-region from the enhanced panoramic image.

[0086] In this step, the target sub-region refers to the local area selected from the enhanced panoramic image that meets the landing qualification requirements; the preset conditions refer to the pre-set indicators for judging whether the sub-region meets the landing requirements.

[0087] In this embodiment of the invention, the corresponding shape qualification threshold, size qualification threshold, and distribution continuity qualification threshold are retrieved from the visual detection dataset. The shape comparison result is compared with the shape qualification threshold, the size comparison result is compared with the size qualification threshold, and the distribution continuity comparison result is compared with the distribution continuity qualification threshold. Regions in which the comparison results of the three dimensions all reach or exceed the corresponding qualification thresholds are selected, and these regions are defined as target sub-regions.

[0088] Step 303: Determine multiple candidate landing points based on the boundary range and center coordinates of each target sub-region.

[0089] In this step, the boundary range refers to the spatial range covered by the outline boundary of the target sub-region; the center coordinates refer to the three-dimensional coordinates of the geometric center of the target sub-region; and the candidate landing point refers to the specific location with landing potential determined based on the target sub-region.

[0090] In this embodiment of the invention, the boundary range of the target sub-region is determined by identifying the pixel coordinates of the contour edge of the target sub-region, and then the geometric center coordinates of the boundary range are calculated. The point corresponding to the center coordinates of each target sub-region is determined as a candidate landing point.

[0091] Step 304: Calculate the terrain flatness and coverage of each candidate landing point, and select the candidate landing point whose terrain flatness and coverage both meet the preset conditions as the qualified landing point location.

[0092] In this step, terrain flatness refers to the degree of surface smoothness of the target sub-region where the candidate landing point is located; coverage area refers to the horizontal projected area of ​​the target sub-region where the candidate landing point is located.

[0093] In this embodiment of the invention, the elevation data of multiple sampling points within the target sub-region where the candidate landing point is located are obtained, and the terrain flatness is calculated. The coverage area is calculated using the horizontal projection coordinates of the boundary range. The terrain flatness and coverage area are compared with the flatness threshold and coverage area threshold in the preset conditions, respectively. The center coordinates of the candidate landing points that meet the corresponding threshold requirements for both parameters are determined as qualified landing point locations.

[0094] This invention eliminates areas that do not meet landing conditions, providing safe and reliable landing targets for heavy-duty UAVs and improving the accuracy and safety of landing point identification.

[0095] This invention provides a specific embodiment, step 302, which involves selecting a target sub-region from the enhanced panoramic image based on the comparison result, specifically including the following steps:

[0096] Step 311: Based on the comparison results, and combined with the shape qualification threshold, size qualification threshold, and distribution continuity qualification threshold in the visual detection dataset, calculate the qualification dimension ratio of each sub-region to be detected in the enhanced panoramic image.

[0097] In this step, the shape qualification threshold refers to the preset critical value in the visual detection dataset for judging whether the shape of the sub-region to be detected meets the landing requirements; the size qualification threshold refers to the preset critical value in the visual detection dataset for judging whether the size of the sub-region to be detected meets the landing requirements; the distribution continuity qualification threshold refers to the preset critical value in the visual detection dataset for judging whether the distribution continuity of the sub-region to be detected meets the landing requirements; the sub-region to be detected refers to the local region in the enhanced panoramic image that is divided out for feature comparison; the qualified dimension ratio refers to the proportion of the number of dimensions in the sub-region to be detected that meet the qualification threshold requirements to the total number of dimensions compared.

[0098] In this embodiment of the invention, the enhanced panoramic image is divided into multiple sub-regions to be detected according to a grid of fixed size. The comparison results corresponding to each sub-region to be detected are retrieved. The shape comparison results are compared with the shape qualification threshold, the size comparison results are compared with the size qualification threshold, and the distribution continuity comparison results are compared with the distribution continuity qualification threshold. The number of dimensions in each sub-region to be detected that meet the corresponding qualification threshold requirements is counted, and the qualification dimension ratio is calculated. Qualified dimension ratio = number of qualified dimensions ÷ total number of compared dimensions, where the total number of compared dimensions is the total number of dimensions, thereby obtaining the qualified dimension ratio of each sub-region to be detected.

[0099] Step 312: Select the sub-regions to be detected with a qualified dimension ratio greater than or equal to the preset qualified dimension ratio threshold as the target sub-regions. If there are sub-regions to be detected with overlapping regions, select the sub-region with the larger region as the target sub-region.

[0100] In this step, the preset qualified dimension ratio threshold refers to the pre-set critical value for determining whether the sub-region to be detected can be used as the target sub-region in terms of qualified dimension ratio; the overlapping of regions refers to the situation where the boundary ranges of two or more sub-regions to be detected partially or completely overlap.

[0101] In this embodiment of the invention, a preset qualified dimension ratio threshold is retrieved, and the qualified dimension ratio of each sub-region to be detected is compared with the threshold. Sub-regions to be detected with a qualified dimension ratio greater than or equal to the threshold are selected as preliminary candidate target sub-regions. By comparing the boundary range coordinates of each preliminary candidate target sub-region, it is determined whether there is an overlap of the region range. If there is, the sub-region to be detected with the larger horizontal projection area corresponding to the boundary range of the overlapping region is extracted. If there is no overlap, it is directly retained. Finally, all sub-regions to be detected that meet the requirements are determined as target sub-regions.

[0102] The embodiments of the present invention achieve accurate screening of target sub-regions, avoiding misscreening or omissions caused by single-dimensional judgment, improving the reliability and rationality of target sub-region screening, and providing a high-quality candidate region basis for subsequent determination of qualified landing point locations.

[0103] This invention provides a specific embodiment. Step 104 involves using digital twin technology to simulate the stress on the fuselage of the heavy-load UAV during landing, based on the qualified landing point location, the enhanced panoramic image, the airflow velocity direction, and the airflow change data. This simulation includes the following steps:

[0104] Step 401: Based on the qualified landing point location and the terrain features of the enhanced panoramic image, construct a digital twin simulation scene.

[0105] In this step, terrain features refer to the terrain-related attributes such as the surface undulations and obstacle distribution presented in the enhanced panoramic image; the digital twin simulation scene refers to a virtual scene that uses the actual landing environment as a prototype and recreates key elements such as terrain and qualified landing point locations through virtual simulation technology.

[0106] In this embodiment of the invention, terrain feature-related information such as elevation data and obstacle outline data of the ground surface are extracted from the enhanced panoramic image, and the terrain undulation shape and obstacle position are restored in proportion. The coordinates of the qualified landing point position are mapped to the corresponding spatial position of the virtual scene to construct a digital twin simulation scene that is highly consistent with the actual landing environment.

[0107] Step 402: Convert the airflow velocity direction and the airflow change data into airflow parameters that can be recognized by the digital twin simulation scene, and set the airflow action parameters of the heavy-load UAV in combination with the airflow obstruction and disturbance law and the terrain features.

[0108] In this step, airflow parameters refer to airflow-related data adapted to the data format of the digital twin simulation scenario; airflow obstruction and disturbance laws refer to the inherent laws of obstruction, bypassing, or disturbance generated when airflow encounters different terrains; airflow action parameters refer to parameters used to quantify the magnitude and direction of the force exerted by airflow on the heavy-load UAV; and the qualified landing point location in the digital twin simulation scenario refers to the virtual coordinates corresponding to the qualified landing point location in the digital twin simulation scenario.

[0109] In this embodiment of the invention, the airflow velocity direction is converted into vector data according to the coordinate system of the virtual scene, and the airflow change data is converted into scene-identifiable dynamic data according to the time series, together forming airflow parameters. Combined with information such as the slope of the terrain undulation and the height of obstacles in the terrain features, the influence coefficient of different terrain areas on the airflow is determined according to the airflow obstruction and disturbance law, and the airflow action parameters are calculated. The airflow action parameters = velocity value in the airflow parameters × terrain influence coefficient × time decay coefficient, thereby setting the airflow action parameters when the heavy-load UAV flies in different terrain areas.

[0110] Step 403: Based on the initial landing trajectory, the qualified landing point location in the digital twin simulation scenario, and the airflow parameters, combined with the fuselage structure parameters and flight performance parameters of the heavy-load UAV, calculate the airflow force, gravity, and air resistance experienced by the heavy-load UAV in each landing phase to form a fuselage force simulation result.

[0111] In this step, the airframe structural parameters refer to structural data such as the external dimensions, airframe weight, number of rotors, and layout of the heavy-load UAV; the flight performance parameters refer to flight performance data such as the maximum lift and maximum wind resistance of the heavy-load UAV; the landing phase refers to the flight segments of the heavy-load UAV from its initial positioning position to the qualified landing point position in the digital twin simulation scenario, divided by altitude range; the airflow force refers to the mechanical force generated by the airflow on the heavy-load UAV corresponding to the airflow parameters; gravity refers to the vertical downward force generated by the Earth's gravity on the heavy-load UAV; and air resistance refers to the force exerted by the air on the heavy-load UAV in flight that hinders its movement.

[0112] In this embodiment of the invention, the initial landing trajectory is divided into multiple landing stages based on the altitude difference. The fuselage structure parameters and flight performance parameters of the heavy-load UAV are retrieved. Combined with the flight speed of the initial landing trajectory, the coordinates of the qualified landing point in the digital twin simulation scenario, and the airflow parameters corresponding to each landing stage, the airflow force on the heavy-load UAV, the Earth's gravity on the heavy-load UAV, and the air resistance on the heavy-load UAV in each landing stage are calculated. Gravity = fuselage mass × gravitational acceleration, where gravitational acceleration is a constant g, with a value... The airflow force = airflow parameters × rotor frontal area, where the rotor frontal area refers to the total surface area of ​​the fuselage structure that can withstand the airflow force after all rotors are deployed. Wherein, air density is an inherent environmental constant, and its value is taken under standard conditions. Flight speed refers to the flight rate during the corresponding landing phase in the initial landing trajectory, and drag coefficient refers to the preset air resistance correlation coefficient in the fuselage structural parameters. The calculation results of each landing phase are integrated to form the fuselage stress simulation results.

[0113] The embodiments of the present invention accurately simulate the stress on the fuselage during the landing process of a heavy-load UAV, providing a reliable basis for subsequent adjustment of the rotor lift distribution ratio and correction of flight data, and addressing the stress interference problem caused by both terrain and airflow.

[0114] For example, a heavy-duty UAV A has determined a suitable landing point location (X=970m, Y=670m, Z=0m). The enhanced panoramic image shows that the landing area's terrain features a gentle slope on the west side and flat, unobstructed ground on the east side. Based on this terrain feature and the suitable landing point location, a scaled-down digital twin simulation scene is constructed, where the coordinates of the suitable landing point match the actual location. The airflow velocity data collected by the heavy-duty UAV A is horizontal westward, with airflow variation data showing a velocity fluctuation of ±0.5m / s every 3 seconds. This data is converted into vector airflow parameters recognizable by the scene. Based on the dynamic fluctuation parameters (3s, ±0.5m / s) and the airflow obstruction and disturbance law in the gentle slope area, the terrain influence coefficient is determined to be 1.1 for the gentle slope area and 1.0 for the flat area. The calculated airflow effect parameter for the gentle slope area is 3m / s × 1.1 × 1.0 = 3.3 N / m, and for the flat area it is 3.0 N / m. The initial landing trajectory is divided into two landing stages according to altitude: Z = 80m - Z = 40m and Z = 40m - Z = 0m. Using the fuselage structural parameters (future mass 500kg, rotor frontal area 2m²) and flight performance parameters (maximum wind resistance 5m / s), the first stage, the gentle slope area... , gravity=500kg×g=4900N, Phase Two Gravity and air resistance are calculated based on the corresponding flight speed at each stage, ultimately forming a fuselage stress simulation result containing force data for both stages.

[0115] This invention provides a specific embodiment. Step 105 involves calculating the lift distribution ratio of each rotor of the heavy-load UAV based on the qualified landing point location and the fuselage force simulation results, in order to correct the real-time flight data and obtain corrected flight data. The specific steps include:

[0116] Step 501: Analyze the simulation results of the airframe forces to determine the trend of the combined forces acting on the heavy-load UAV during each landing phase and their impact on the deviation of the real-time flight data.

[0117] In this step, the combined force refers to the resultant force of airflow, gravity, and air resistance in each landing phase; the deviation influence trend refers to the direction and change pattern of real-time flight data deviating from the ideal state due to the combined force.

[0118] In this embodiment of the invention, the airflow force, gravity and air resistance of each landing stage are extracted from the fuselage force simulation results. The combined force of each landing stage is calculated according to the vector synthesis rule. The correlation between the magnitude and direction of the combined force and the changes in fuselage attitude and position in real-time flight data is analyzed. The deviation trend of the combined force causing the fuselage to tilt in which direction and the fuselage to shift in which direction is the real-time position affected.

[0119] Step 502: Calculate the positional deviation between the real-time fuselage position and the qualified landing point position in the real-time flight data, and calculate the attitude deviation between the real-time fuselage attitude and the preset standard landing attitude in the real-time flight data.

[0120] In this step, the real-time fuselage position refers to the current three-dimensional coordinates of the heavy-load UAV recorded in the real-time flight data; the position deviation refers to the spatial distance difference between the real-time fuselage position and the qualified landing point position; the real-time fuselage attitude refers to the current pitch angle, roll angle, and yaw angle of the heavy-load UAV recorded in the real-time flight data; the preset standard landing attitude refers to the reference values ​​of pitch angle, roll angle, and yaw angle that the heavy-load UAV should maintain when landing safely; and the attitude deviation refers to the angular difference between the real-time fuselage attitude and the preset standard landing attitude.

[0121] In this embodiment of the invention, the real-time position and attitude of the fuselage are extracted from real-time flight data, and the position deviation is calculated. The differences between the pitch angle, roll angle, and yaw angle in the real-time attitude of the computer body and the corresponding angles of the preset standard landing attitude are used to calculate the attitude deviation. The position deviation and attitude deviation are obtained.

[0122] Step 503: Based on the deviation influence trend, and in combination with the position deviation and the attitude deviation, determine the lift change direction and lift change amplitude that each rotor of the heavy-load UAV needs to be adjusted.

[0123] In this step, the direction of lift change refers to the adjustment direction in which the lift of each rotor of the heavy-load UAV needs to be increased or decreased; the magnitude of lift change refers to the specific proportion or value by which the lift of each rotor of the heavy-load UAV needs to be increased or decreased.

[0124] In this embodiment of the invention, the direction of interference of the combined force on the fuselage is determined by combining the trend of deviation influence. If the position deviation shows that the fuselage is offset in the positive X-axis direction and the attitude deviation shows that the roll angle of the fuselage is too large, the direction of lift change is determined based on the characteristics of the rotor layout. The corresponding rotor needs to increase lift and the other side needs to maintain or reduce lift. According to the magnitude of the position deviation, the angle of the attitude deviation and the intensity of the combined force, the lift change range that each rotor needs to be adjusted is calculated to ensure that the adjustment can offset the interference of the combined force and correct the deviation.

[0125] Step 504: Calculate the lift distribution ratio of each rotor of the heavy-load UAV based on the direction and magnitude of the lift change.

[0126] In this embodiment of the invention, the initial lift distribution ratio of each rotor of the heavy-load UAV is retrieved. Based on the direction and magnitude of the lift change, the initial lift distribution ratio of each rotor is adjusted, and the lift distribution ratio of each rotor is calculated. The lift distribution ratio = initial lift distribution ratio × (1 + lift change magnitude) to ensure that the sum of the lift distribution ratios of all rotors is 1, thus obtaining the final lift distribution ratio of each rotor.

[0127] The embodiments of the present invention achieve precise correction of deviations in real-time flight data, offset the interference caused by airflow and terrain, improve the stability of flight attitude and position of heavy-load UAVs, and lay the foundation for subsequent trajectory optimization.

[0128] This invention provides a specific embodiment. Step 106 involves optimizing the initial landing trajectory based on the qualified landing point location and the corrected flight data to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV. The specific steps include:

[0129] Step 601: Based on the corrected real-time position of the fuselage in the corrected flight data and the initial landing trajectory, determine the target trajectory node corresponding to the heavy-load UAV in the initial landing trajectory and the target preset standard landing attitude of the target trajectory node.

[0130] In this step, the corrected real-time fuselage position refers to the current three-dimensional coordinates of the heavy-load UAV after lift adjustment; the target trajectory node refers to the key point in the initial landing trajectory that is closest to the corrected real-time fuselage position; and the target preset standard landing attitude refers to the baseline landing attitude that the heavy-load UAV corresponding to the target trajectory node should maintain.

[0131] In this embodiment of the invention, the corrected real-time position of the fuselage is extracted from the corrected flight data, all trajectory nodes in the initial landing trajectory are traversed, the spatial distance between the corrected real-time position of the fuselage and each trajectory node is calculated, the trajectory node with the smallest distance is determined as the target trajectory node, and the preset reference landing attitude of the target trajectory node is retrieved as the target preset standard landing attitude.

[0132] Step 602: Calculate the position deviation between the corrected real-time fuselage position and the target trajectory node, and calculate the attitude deviation between the corrected real-time fuselage attitude and the target preset standard landing attitude.

[0133] In this step, the position deviation value refers to the spatial distance difference between the real-time position of the corrected fuselage and the target trajectory node; the real-time attitude of the corrected fuselage refers to the pitch angle, roll angle and yaw angle of the heavy-load UAV after lift adjustment correction; the attitude deviation value refers to the angle difference between the real-time attitude of the corrected fuselage and the target preset standard landing attitude.

[0134] In this embodiment of the invention, the position deviation value is calculated. The differences between the pitch angle, roll angle, and yaw angle in the real-time attitude of the corrected fuselage and the corresponding angles of the target's preset standard landing attitude are calculated to obtain the attitude deviation value. The position deviation value and attitude deviation value are obtained.

[0135] Step 603: Based on the qualified landing point location, adjust the position of subsequent trajectory nodes in the initial landing trajectory that are located after the corrected real-time position of the fuselage to obtain the adjusted trajectory nodes.

[0136] In this step, subsequent trajectory nodes refer to all trajectory nodes in the initial landing trajectory that are located after the corrected real-time position of the aircraft and after the target trajectory node; adjusted trajectory nodes refer to subsequent trajectory nodes after position adjustment.

[0137] In this embodiment of the invention, the coordinates of the qualified landing point are determined. Taking the real-time position of the corrected fuselage as the starting point and the qualified landing point as the ending point, the distribution pattern of subsequent trajectory nodes is replanned. The coordinates of subsequent trajectory nodes in the initial landing trajectory are translated or offset according to the newly planned path to ensure that the line connecting the adjusted trajectory nodes always points to the qualified landing point, thus obtaining the adjusted trajectory nodes.

[0138] Step 604: Based on the position deviation value and the attitude deviation value, correct the flight direction and flight speed corresponding to the subsequent trajectory nodes to obtain the corrected flight direction and corrected flight speed.

[0139] In this step, the corrected flight direction refers to the horizontal direction in which the heavy-load UAV flies between the adjusted trajectory nodes after deviation correction; the corrected flight speed refers to the speed at which the heavy-load UAV flies between the adjusted trajectory nodes after deviation correction.

[0140] In this embodiment of the invention, the direction in which the fuselage deviates from the target trajectory node is determined based on the position deviation value, and the fuselage attitude tilt is determined in combination with the attitude deviation value. The original flight direction of the subsequent trajectory node is then finely adjusted to ensure that the finely adjusted flight direction can offset the deviation and allow the fuselage to smoothly approach the adjusted trajectory node. The flight speed is then appropriately adjusted according to the magnitude of the position deviation value and the distance between the adjusted trajectory nodes so that the fuselage can reach the adjusted trajectory node within a preset time, thus obtaining the corrected flight direction and the corrected flight speed.

[0141] Step 605: Based on the surface features of the qualified landing point location, optimize the descent rate parameters of the last trajectory node in the subsequent trajectory nodes to obtain the optimized descent rate parameters.

[0142] In this step, surface features refer to the surface properties of the area where the qualified landing point is located; the last trajectory node refers to the trajectory node that is closest to the qualified landing point among the subsequent trajectory nodes; and the optimized descent rate parameter refers to the vertical descent rate from the last trajectory node to the qualified landing point after optimization.

[0143] In this embodiment of the invention, surface features of the qualified landing point location are extracted. If the surface features are soft terrain, the descent rate needs to be reduced to avoid sinking; if the surface features are hard and flat terrain, a reasonable descent rate can be maintained. Combining the height difference and horizontal distance between the end trajectory node and the qualified landing point location, the optimized descent rate parameters are calculated. The optimized descent rate parameters were obtained.

[0144] Step 606: Integrate the adjusted trajectory nodes, the corrected flight direction, the corrected flight speed, and the optimized descent rate parameters to form an optimized landing trajectory, and send the optimized landing trajectory to the flight control unit of the heavy-load UAV to realize the landing control of the heavy-load UAV.

[0145] In this step, the flight control unit refers to the core component of a heavy-duty UAV that receives flight trajectory commands and controls the flight of the aircraft.

[0146] In this embodiment of the invention, the adjusted trajectory nodes are arranged in sequence, and each adjusted trajectory node is matched with a corresponding corrected flight direction and corrected flight speed. The optimized descent rate parameter is associated with the last trajectory node, and these parameters are integrated to form a complete optimized landing trajectory. The optimized landing trajectory is sent to the flight control unit through the data transmission module. The flight control unit controls the rotor speed and flight attitude of the heavy-load UAV according to the trajectory command to achieve landing control.

[0147] The embodiments of the present invention achieve precise optimization of the initial landing trajectory, enabling the trajectory to adapt to the actual situation of the corrected flight data and the qualified landing point location, thereby improving the accuracy and stability of the landing of heavy-load UAVs.

[0148] Example 2

[0149] Figure 3 This is a schematic diagram of a specific implementation of the vision-based heavy-load UAV landing point identification and processing system provided in this invention. (Refer to...) Figure 3 The system may include:

[0150] The acquisition module 21 is used to acquire the airflow speed, direction and airflow change data above the landing area of ​​the heavy-load UAV, as well as real-time flight data and landing area images, and to stitch and enhance the landing area images to obtain an enhanced panoramic image.

[0151] The generation module 22 is used to generate an initial landing trajectory based on the initial positioning position of the heavy-load UAV, the initial coordinates of the preset landing area, and the flight mission parameters.

[0152] The comparison module 23 is used to compare the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial Internet field with the enhanced panoramic image to determine the location of the qualified landing point.

[0153] Simulation module 24 is used to simulate the stress on the fuselage of the heavy-load UAV during landing, based on the qualified landing point location, the enhanced panoramic image, the airflow speed and direction, and the airflow change data, using digital twin technology.

[0154] The correction module 25 is used to calculate the lift distribution ratio of each rotor of the heavy-load UAV based on the qualified landing point position and the fuselage force simulation results, so as to correct the real-time flight data and obtain the corrected flight data.

[0155] The optimization module 26 optimizes the initial landing trajectory based on the qualified landing point location and the corrected flight data to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV.

[0156] The visual recognition-based heavy-load UAV landing point identification and processing system of this invention is used to implement the aforementioned visual recognition-based heavy-load UAV landing point identification and processing method. Therefore, the specific implementation of the visual recognition-based heavy-load UAV landing point identification and processing system can be found in the embodiment section of the visual recognition-based heavy-load UAV landing point identification and processing method above. The specific implementation can be referred to the description of the corresponding embodiment, and will not be repeated here.

[0157] The present invention also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described vision-based heavy-load UAV landing point identification processing method.

[0158] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described vision-based heavy-load UAV landing point identification and processing methods.

[0159] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0160] The above provides a detailed description of the visual recognition-based landing point identification and processing method and system for heavy-duty UAVs provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and core ideas of this invention. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the protection scope of this invention.

Claims

1. A method for identifying and processing landing points of heavy-load unmanned aerial vehicles based on visual recognition, characterized in that, include: Data on airflow speed, direction, and changes in airflow over the landing area of ​​a heavy-duty UAV, as well as real-time flight data and landing area images, are acquired. The landing area images are then stitched together and enhanced to obtain an enhanced panoramic image. The initial landing trajectory is generated based on the initial positioning of the heavy-duty UAV, the initial coordinates of the preset landing area, and the flight mission parameters. The visual detection dataset of landing qualification features of heavy-duty drones in the field of industrial internet is compared with the enhanced panoramic image to determine the location of qualified landing points. Based on the qualified landing point location, the enhanced panoramic image, the airflow speed and direction, and the airflow change data, the simulation results of the fuselage force of the heavy-load UAV during the landing process are simulated using digital twin technology. Based on the qualified landing point location and the stress simulation results of the fuselage, the lift distribution ratio of each rotor of the heavy-load UAV is calculated to correct the real-time flight data and obtain the corrected flight data. Based on the qualified landing point location and the corrected flight data, the initial landing trajectory is optimized to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV.

2. The method according to claim 1, characterized in that, Based on the initial positioning of the heavy-load UAV, the initial coordinates of the preset landing area, and flight mission parameters, an initial landing trajectory is generated, including: Calculate the straight-line distance and relative orientation between the initial positioning position of the heavy-load UAV and the initial coordinates of the preset landing area; Based on the straight-line distance and the relative orientation, and combined with the maximum flight speed, maximum descent speed and allowable flight altitude range in the flight mission parameters of the heavy-load UAV, multiple trajectory nodes of the heavy-load UAV are set. Based on the turning limit angle in the flight mission parameters, the flight direction and speed of the heavy-load UAV from the previous trajectory node to the next trajectory node are calculated, and the initial flight path is generated by combining the arrangement order of the trajectory nodes. Based on the maximum flight speed, the distance between adjacent trajectory nodes in the initial flight path, and the height difference between nodes, the descent rate parameters of each trajectory node in the initial flight path are calculated, and the initial landing trajectory is generated by combining the initial flight path.

3. The method according to claim 1, characterized in that, The visual detection dataset of landing qualification features of heavy-duty drones in the industrial internet field is compared with the enhanced panoramic image to determine the location of qualified landing points, including: The visual detection dataset of landing qualification features of heavy-duty UAVs in the field of industrial Internet is compared with the enhanced panoramic image to obtain the comparison results, which include shape comparison results, size comparison results and distribution continuity comparison results. Based on the comparison results, target sub-regions are selected from the enhanced panoramic image; Based on the boundary range and center coordinates of each target sub-region, multiple candidate landing points are determined; The terrain flatness and coverage of each candidate landing point are calculated, and the location of the candidate landing point that meets the preset conditions in terms of both terrain flatness and coverage is taken as the qualified landing point location.

4. The method according to claim 3, characterized in that, Based on the comparison results, target sub-regions are selected from the enhanced panoramic image, including: Based on the comparison results, and combined with the shape qualification threshold, size qualification threshold, and distribution continuity qualification threshold in the visual detection dataset, the qualification dimension ratio of each sub-region to be detected in the enhanced panoramic image is calculated. The sub-regions to be detected with a qualified dimension ratio greater than or equal to the preset qualified dimension ratio threshold are designated as target sub-regions. If there are sub-regions to be detected with overlapping regions, the sub-region with the larger region is designated as the target sub-region.

5. The method according to claim 1, characterized in that, Based on the qualified landing point location, the enhanced panoramic image, the airflow velocity and direction, and the airflow change data, digital twin technology is used to simulate the fuselage forces of the heavy-load UAV during landing, including: Based on the qualified landing point location and the terrain features of the enhanced panoramic image, a digital twin simulation scene is constructed; The airflow velocity direction and the airflow change data are converted into airflow parameters that can be recognized by the digital twin simulation scene. Combined with the airflow obstruction and disturbance law and the terrain features, the airflow action parameters of the heavy-load UAV are set. Based on the initial landing trajectory, the qualified landing point location in the digital twin simulation scenario, and the airflow parameters, combined with the fuselage structure parameters and flight performance parameters of the heavy-load UAV, the airflow force, gravity, and air resistance experienced by the heavy-load UAV in each landing phase are calculated to form the fuselage force simulation results.

6. The method according to claim 1, characterized in that, Based on the qualified landing point location and the fuselage stress simulation results, the lift distribution ratio of each rotor of the heavy-load UAV is calculated to correct the real-time flight data, resulting in corrected flight data, including: The results of the fuselage stress simulation are analyzed to determine the trend of the combined forces acting on the heavy-load UAV during each landing phase and their impact on the deviation of the real-time flight data. Calculate the positional deviation between the real-time fuselage position and the qualified landing point position in the real-time flight data, and calculate the attitude deviation between the real-time fuselage attitude and the preset standard landing attitude in the real-time flight data; Based on the trend of the deviation, and in combination with the position deviation and the attitude deviation, the direction and magnitude of the lift change that each rotor of the heavy-load UAV needs to be adjusted are determined. Based on the direction and magnitude of the lift change, the lift distribution ratio of each rotor of the heavy-load UAV is calculated.

7. The method according to claim 1, characterized in that, Based on the qualified landing point location and the corrected flight data, the initial landing trajectory is optimized to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV, including: Based on the corrected real-time position of the fuselage in the corrected flight data and the initial landing trajectory, determine the target trajectory node corresponding to the heavy-load UAV in the initial landing trajectory and the target preset standard landing attitude of the target trajectory node; Calculate the position deviation between the corrected real-time fuselage position and the target trajectory node, and calculate the attitude deviation between the corrected real-time fuselage attitude and the target preset standard landing attitude; Based on the qualified landing point location, the positions of subsequent trajectory nodes in the initial landing trajectory that are located after the corrected real-time fuselage position are adjusted to obtain the adjusted trajectory nodes; Based on the position deviation value and the attitude deviation value, the flight direction and flight speed corresponding to the subsequent trajectory nodes are corrected to obtain the corrected flight direction and corrected flight speed. Based on the surface features of the qualified landing point location, the descent rate parameters of the last trajectory node in the subsequent trajectory nodes are optimized to obtain the optimized descent rate parameters; The adjusted trajectory nodes, the corrected flight direction, the corrected flight speed, and the optimized descent rate parameters are integrated to form an optimized landing trajectory. The optimized landing trajectory is then sent to the flight control unit of the heavy-load UAV to achieve landing control of the heavy-load UAV.

8. A visual recognition-based landing point identification and processing system for heavy-duty unmanned aerial vehicles (UAVs), characterized in that, include: The acquisition module is used to acquire airflow speed, direction and airflow change data above the landing area of ​​the heavy-load UAV, as well as real-time flight data and landing area images. The landing area images are then stitched and enhanced to obtain an enhanced panoramic image. The generation module is used to generate the initial landing trajectory based on the initial positioning of the heavy-duty UAV, the initial coordinates of the preset landing area, and the flight mission parameters. The comparison module is used to compare the visual detection dataset of landing qualification features of heavy-duty UAVs in the industrial Internet field with the enhanced panoramic image to determine the location of qualified landing points. The simulation module is used to simulate the stress on the fuselage of the heavy-load UAV during the landing process using digital twin technology, based on the qualified landing point location, the enhanced panoramic image, the airflow speed and direction, and the airflow change data. The correction module is used to calculate the lift distribution ratio of each rotor of the heavy-load UAV based on the qualified landing point location and the stress simulation results of the fuselage, so as to correct the real-time flight data and obtain the corrected flight data. The optimization module optimizes the initial landing trajectory based on the qualified landing point location and the corrected flight data to obtain an optimized landing trajectory, thereby achieving landing control of the heavy-load UAV.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the visual recognition-based heavy-load UAV landing point identification processing method as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the visual recognition-based landing point identification processing method for heavy-duty unmanned aerial vehicles as described in any one of claims 1 to 7.

Images