Low-altitude economy-oriented vertical take-off and landing facility unmanned aerial vehicle image site selection method and system

Abstract

The invention provides a low-altitude economy-oriented vertical take-off and landing facility unmanned aerial vehicle image site selection method and system, and belongs to the technical field of image data processing and modeling. The method comprises the steps of performing oblique photogrammetry and laser radar scanning on a building roof through an unmanned aerial vehicle to obtain a high-overlapping-rate image and laser point cloud data, constructing a live-action three-dimensional grid model and an obstruction surface three-dimensional model after preprocessing, and screening a preliminary layout scheme according to the clearance height of a roof available area based on the models; and airspace safety simulation and noise propagation simulation are carried out on each scheme, a minimum spacing distance and noise intensity distribution diagram is calculated, and a final site selection scheme is screened according to airspace safety and noise constraints. According to the invention, automatic and precise site selection of the vertical take-off and landing facility in a complex roof environment is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of image data processing and modeling, and in particular relates to a method and system for image site selection of vertical take-off and landing (VTOL) facilities using drones (UAVs) for low-altitude economy. Background Art

[0002] With the rapid rise of the low-altitude economy, vertical take-off and landing (VTOL) facilities, critical infrastructure for scenarios like drone logistics and urban air traffic, face the technical challenges of complex spatial constraints and multi-dimensional safety regulations. Existing low-altitude facility site selection technologies generally suffer from a single-dimensional data collection model. Traditional measurement methods, such as total stations or satellite remote sensing, struggle to obtain high-precision three-dimensional spatial information in complex scenarios like building rooftops and urban canyons. This information is particularly missing from the representation of subtle structures like rooftop equipment and surrounding building outlines, leading to significant discrepancies between the basic site selection data and the actual scenario.

[0003] At the modeling and analysis level, traditional methods often rely on 2D CAD drawings or simplified 3D models, failing to effectively construct accurate digital twins that incorporate elements like obstruction surface morphology and airspace boundaries. This results in calculations of key parameters like clearance height and flight path safety distance deviating from actual requirements. Furthermore, existing technologies lack multi-dimensional collaborative assessment capabilities, often focusing solely on structural load-bearing capacity or airspace safety, failing to couple environmental constraints like noise propagation and the electromagnetic environment with aviation safety regulations. This often results in siting plans failing due to conflicting constraints.

[0004] Especially in urban built environments, the site selection of vertical take-off and landing facilities must simultaneously meet multiple constraints such as building structure feasibility, civil aviation safety standards, and urban planning requirements. However, in traditional technical processes, data collection, model construction, and safety assessment are separated from each other, and the degree of automation is low, making it difficult to achieve an efficient closed loop for the entire process from spatial feature extraction to site selection decision-making. Summary of the Invention

[0005] The present invention overcomes the deficiencies of the prior art and provides a method and system for vertical take-off and landing facility drone image site selection for low-altitude economy.

[0006] To achieve the above-mentioned purpose, the present invention adopts the following technical solution: a method for selecting a site for a vertical take-off and landing facility using drone images for low-altitude economy, comprising the following steps: Step S1: Obtain remote sensing data by performing oblique photogrammetry and laser radar scanning on the building roof using a drone; the remote sensing data includes high-overlapping imagery and laser point cloud data; Step S2: Preprocessing the remote sensing data, and establishing a real-scene 3D mesh model and a 3D obstacle surface model based on the preprocessed remote sensing data; the real-scene 3D mesh model is used to represent the surface morphology of the building roof; and the 3D obstacle surface model is used to define the airspace boundary for safe flight. Step S3: Based on the real-scene 3D mesh model and the 3D surface model of the obstruction, a preliminary layout plan is selected according to the clearance height of the available roof area; Step S4: For each preliminary layout plan, airspace safety simulation and noise propagation simulation are performed to obtain the minimum separation distance and noise intensity distribution map. The airspace safety simulation is based on distance detection between the flight trajectory and the real-world 3D mesh model and the 3D surface model of the obstruction. The noise propagation simulation is based on the calculation of the noise intensity distribution on the building roof. Step S5: Based on the minimum separation distance and the noise intensity distribution map, the site selection plan is screened according to airspace safety constraints and noise constraints.

[0007] Furthermore, the pre-processing in step S2 includes: Perform image distortion correction and aerial triangulation on images with high overlap ratio; Perform noise removal, data downsampling and coordinate registration on laser point cloud data.

[0008] Furthermore, the method for establishing the real-scene three-dimensional grid model includes: Dense matching is performed on the pre-processed high-overlap images to generate dense point clouds; Perform surface reconstruction on dense point clouds to generate triangular facet meshes; The texture information of high-overlapping images is projected onto a triangular mesh to generate a real-scene 3D mesh model.

[0009] Furthermore, based on the pre-processed high-overlapping image and laser point cloud data, three-dimensional features including plane features, contour features and elevation features are extracted; Based on three-dimensional features, the laser point cloud data is classified into roof plane laser point cloud, roof equipment laser point cloud and surrounding building laser point cloud through clustering and threshold judgment.

[0010] Furthermore, the method for establishing the three-dimensional model of the obstacle surface includes: Based on the classified laser point cloud of the rooftop equipment and the laser point cloud of the surrounding buildings, the system divides the points into several independent areas according to their spatial distribution. Each independent region was fitted into a plane using the least squares method; All fitting planes are spliced ​​according to their spatial positions to form a continuous three-dimensional model of the obstacle surface.

[0011] Furthermore, preliminary layout options were screened based on the headroom of the available roof area, including: Generate roof usable area vector surface based on real-life 3D mesh model; Based on the rooftop available area vector surface, a digital elevation model is constructed using the inverse distance weighted interpolation method, and the clearance height of each grid cell is calculated to generate a clearance height raster map; According to the landing pad size parameters, the roof available area vector surface is divided into regular grids; Grids with clearance height greater than the preset threshold are selected as preliminary layout plans.

[0012] Furthermore, airspace safety simulation includes: Based on the approach phase and climb phase of the route trajectory, an approach route trajectory model and a climb route trajectory model are established respectively; Calculate the minimum distance between the flight path and the vertex of the real-scene 3D mesh model surface; Calculate the minimum distance between the flight path and the three-dimensional model of the obstacle surface; The smaller of the two minimum distances is taken as the minimum separation distance.

[0013] Furthermore, the noise propagation simulation includes: Define the sound power level, spectral characteristics and spatial location of the noise source; Establish a noise propagation attenuation model including geometric divergence attenuation, air absorption attenuation and ground reflection attenuation; The noise intensity distribution map is generated based on the attenuation model calculation.

[0014] Furthermore, the method of screening site selection options based on airspace safety constraints and noise constraints includes: Set the minimum separation distance threshold and the peripheral area noise peak threshold; Screen out preliminary layout solutions that meet both the minimum separation distance threshold and the peripheral area noise peak threshold.

[0015] Another technical solution provided by the present invention is a low-altitude economic vertical take-off and landing facility drone image site selection system, based on the above method, comprising a data acquisition and processing module, a three-dimensional modeling module, a preliminary screening module, a simulation module, and a decision module connected in sequence; The data acquisition and processing module is used to obtain and pre-process remote sensing data; the 3D modeling module is used to build a real-scene 3D grid model and a 3D model of the obstacle surface; the preliminary screening module is used to screen the preliminary layout plan; the simulation module is used to perform airspace safety simulation and noise propagation simulation; the decision module is used to screen and output the site selection plan. The present invention solves the defects existing in the background technology and has the following beneficial effects: The present invention uses drones to perform oblique photogrammetry and lidar scanning on building roofs to obtain remote sensing data covering high-overlap images and laser point cloud data. After preprocessing, a real-scene 3D grid model and a 3D surface model of obstacles are constructed. This method can comprehensively capture the surface morphology of the building roof, the distribution of equipment, and the 3D spatial characteristics of the surrounding environment, providing a digital spatial benchmark based on real scenes for site selection analysis. This changes the traditional situation of relying on two-dimensional drawings or manual measurements that lead to information loss, realizes the transition from fuzzy qualitative analysis to precise quantitative modeling, effectively solves the problem of incomplete spatial information acquisition in complex roof environments, and frees site selection evaluation from subjective judgment bias.

[0016] Based on the three-dimensional model, the present invention conducts airspace safety simulation and noise propagation simulation on the preliminary layout plan. By calculating the minimum separation distance between the flight trajectory and the three-dimensional model and constructing a noise propagation model that includes geometric divergence attenuation, air absorption attenuation and ground reflection attenuation, the siting plan is quantitatively evaluated from the dual dimensions of aviation safety regulations and environmental noise impact, converting abstract safety requirements into calculable specific parameters. This breaks through the limitations of traditional methods that rely solely on experience or single-dimensional evaluation, realizes the scientific and standardized siting process, ensures that the siting plan has a verifiable technical basis in terms of structural safety and environmental adaptability, and solves the problem that traditional means cannot accurately quantify safety regulations.

[0017] The present invention combines three-dimensional modeling technology based on drone remote sensing with quantitative simulation methods for airspace safety and noise propagation. The three-dimensional model provides real spatial constraints for the simulation, and the simulation analysis infers the key safety parameters of the model, forming a complete technical chain from data collection, model construction to safety assessment. In the complex and confined space on the roof of an urban building, the entire process from spatial feature extraction to safety feasibility judgment can be completed quickly and automatically. Compared with the defects of disconnection between modeling and assessment or single dimension in the existing technology, the site selection efficiency and reliability are both improved, and the technical difficulty of traditional means in converting abstract regulations into specific engineering judgments in non-standard sites is effectively solved, so that aviation safety requirements are accurately matched with complex roof environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive efforts. Figure 1 It is a flow chart of the method for selecting the site of vertical take-off and landing facilities using drone images for low-altitude economy; Figure 2It is an architecture diagram of the drone image siting system for vertical take-off and landing facilities for low-altitude economy. DETAILED DESCRIPTION

[0019] 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.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0021] Exemplary methods: like Figure 1 As shown, a method for selecting a site for a vertical take-off and landing facility using drone images for low-altitude economy includes the following steps: Step S1: Obtain remote sensing data by performing oblique photogrammetry and laser radar scanning on the building roof using a drone; the remote sensing data includes high-overlapping imagery and laser point cloud data; Step S2: Preprocessing the remote sensing data, and establishing a real-scene 3D mesh model and a 3D obstacle surface model based on the preprocessed remote sensing data; the real-scene 3D mesh model is used to represent the surface morphology of the building roof; and the 3D obstacle surface model is used to define the airspace boundary for safe flight. Step S3: Based on the real-scene 3D mesh model and the 3D surface model of the obstruction, a preliminary layout plan is selected according to the clearance height of the available roof area; Step S4: For each preliminary layout plan, airspace safety simulation and noise propagation simulation are performed to obtain the minimum separation distance and noise intensity distribution map. The airspace safety simulation is based on distance detection between the flight trajectory and the real-world 3D mesh model and the 3D surface model of the obstruction. The noise propagation simulation is based on the calculation of the noise intensity distribution on the building roof. Step S5: Based on the minimum separation distance and the noise intensity distribution map, the site selection plan is screened according to airspace safety constraints and noise constraints.

[0022] As the core infrastructure for scenarios such as drone logistics and urban air traffic in the low-altitude economy, vertical take-off and landing facilities must be sited in complex spaces such as urban building rooftops, taking into account both structural bearing capacity and aviation safety regulations. The site selection of such facilities faces the technical requirements of densely distributed equipment on building rooftops, irregular spatial forms, and the need to meet multi-dimensional constraints such as clearance height, flight trajectory safety distance, and noise control. Efficient and accurate site selection solutions rely on digital mapping of the rooftop's three-dimensional environment and safety regulations. This embodiment proposes a site selection solution based on drone images. It obtains rooftop remote sensing data through oblique photography and lidar scanning, constructs a real-life three-dimensional model and an obstacle surface model, and combines airspace safety simulation with noise propagation analysis to achieve automated site selection decisions for vertical take-off and landing facilities in complex scenarios.

[0023] Below, each step will be described in detail.

[0024] For example, the drone is configured as a multi-rotor structure and equipped with an oblique camera and a lidar sensor. The oblique camera includes one vertical lens and four oblique lenses, which are used to simultaneously capture images of objects from different angles. The lidar sensor has a scanning frequency range of 100-500kHz and a ranging range of 0.5-200m, which is used to obtain three-dimensional position information of the surface of the object. The drone also has a global navigation satellite system receiver module that supports dual-mode positioning of the Beidou satellite navigation system and the Global Positioning System, which is used to simultaneously obtain the drone's position and time information during flight.

[0025] For example, high overlap images are earth observation images in which adjacent images have regional overlap, and are divided into orthophotos and oblique images; orthophotos are shot at a vertical angle, with the shooting angle controlled at 0±5° (based on the horizontal plane downward) to eliminate the influence of terrain undulations and projection differences; oblique images are shot at an oblique angle, with the shooting angle controlled at 30-60° (based on the horizontal plane downward) to supplement the side contour information of the ground objects.

[0026] Laser point cloud data is a data set composed of a large number of discrete three-dimensional points. Each discrete point contains three-dimensional coordinate information (X, Y, Z). The X and Y coordinates represent the position of the object on the horizontal plane, and the Z coordinate represents the elevation of the object. It fully represents the surface geometry of the object, including the undulations of building roofs, the spatial location of existing equipment, and the height contours of surrounding buildings.

[0027] For example, oblique photogrammetry plans and executes automated grid routes, controlling drones to fly along preset paths to ensure image overlap.

[0028] The method of planning and executing automated grid routes includes determining the route coverage area by combining the building roof and the surrounding area, including the entire building roof and the surrounding area of ​​50-200m. The route spacing is calculated by ,in is the route spacing (m); is the ground cover width of a single image (m); is the lateral overlap ratio (unitless, 0.6-0.8); it ensures that images collected from adjacent routes can form effective overlap to avoid data omission areas.

[0029] The planned flight path includes a flight altitude of 20-50 meters above building rooftops and a speed of 3-8 meters per second. The vertical and oblique lenses of the oblique camera collect images at intervals of 0.5-2 seconds, with a heading overlap ratio of 70-90% and a lateral overlap ratio of 60-80%. The heading overlap ratio refers to the overlap ratio between two adjacent images within the same route, while the lateral overlap ratio refers to the overlap ratio between corresponding images within adjacent routes.

[0030] After screening and eliminating invalid images that are motion blurred, overexposed or blocked, we obtain images with a high overlap rate including orthophotos and oblique images.

[0031] For example, the lidar scanning is performed by flying a drone at low altitude, controlling the lidar sensor to emit and receive laser beams, and calculating the three-dimensional coordinates of the building roof and the surrounding surface.

[0032] The flight parameters of the lidar scan include a flight altitude of 5-30m from the building roof, a flight speed of 1-5m / s, and a scanning angle controlled between -30 and 30° (based on the drone's flight direction).

[0033] The drone flies along a pre-set path, aligned with the oblique photogrammetry route, ensuring that the laser point cloud data matches the spatial extent of the high-overlap imagery. During flight, the lidar sensor continuously emits a laser beam, receives the beam reflected from the surface, and records the time difference between emission and reception.

[0034] Methods for calculating three-dimensional coordinates include using the formula Calculate the distance from the lidar sensor to the surface of the object, where Indicates the distance from the lidar sensor to the surface of the object, in meters (m); Indicates the propagation speed of laser in air, which is 3×10 8 m / s; It represents the time difference between the emission and reception of the laser beam, in seconds (s); combined with the real-time position and attitude information of the UAV (obtained by the global navigation satellite system receiving module), the distance is converted into three-dimensional coordinates (X, Y, Z) of the surface of the object. A large number of three-dimensional coordinate points are collected to form laser point cloud data.

[0035] High-overlap imagery and laser point cloud data utilize a unified time and spatial reference. The time reference is synchronized and timestamped by the Global Navigation Satellite System (GNSS) receiver module, with an accuracy of less than 1ms. This ensures that the time difference between the capture of the high-overlap imagery and the acquisition of the laser point cloud data is less than 1s. The National Geodetic Coordinate System 2000 (both horizontal and vertical coordinates are in meters) is used. The GNSS receiver module acquires the drone's 3D coordinates in real time and associates them with the high-overlap imagery and laser point cloud data. This ensures that the spatial location information of both remote sensing data types is based on the same coordinate system, preventing errors in subsequent data processing caused by coordinate system differences.

[0036] Exemplarily, the preprocessing method of high overlap images includes image distortion correction and aerial triangulation to eliminate errors in the original images.

[0037] Image distortion correction is based on the internal parameters of the oblique photography camera in step S1 (focal length, principal point coordinates, radial distortion coefficient, tangential distortion coefficient). The camera calibration model is used to perform geometric correction on high-overlap images to eliminate image deformation caused by lens optical distortion. After correction, the geometric deviation of a single image is controlled within 0.01m.

[0038] Aerial triangulation uses distortion-corrected, high-overlap images as input. Using the SIFT algorithm, the system identifies common feature points across images (such as stable features like roof edges and wall corners). Bundle adjustment is then used to calculate the extrinsic parameters (3D position and attitude) for each image, constructing a 3D control network covering the building roof and surrounding area. The resulting triangulation has a planar accuracy of at least 0.05m and an elevation accuracy of at least 0.1m, ensuring that the spatial positions of the images are consistent with the actual features.

[0039] For example, a real-scene 3D mesh model is constructed based on pre-processed high-overlapping images to intuitively represent the surface morphology of the building roof and surrounding objects. The steps are as follows: Using the image extrinsics obtained from aerial triangulation, dense matching is performed on the high-overlap images to generate a dense point cloud covering the building roof and surrounding areas (point density is 50-200 points / m 2 ), each point contains three-dimensional coordinates (X, Y, Z) and color information.

[0040] The dense point cloud is reconstructed and a continuous triangular mesh is generated using the Poisson surface reconstruction algorithm. The mesh resolution matches the density of the dense point cloud (0.05-0.2m), ensuring that details such as roof texture and equipment corners can be clearly expressed.

[0041] The texture information of high-overlap images is projected onto the triangular mesh, and the image exposure differences are eliminated through color balancing processing, so that the texture of the mesh model is visually consistent with the actual ground objects.

[0042] The final generated real-life 3D mesh model covers the entire roof of the building and the surrounding area of ​​50-100m.

[0043] Exemplarily, the preprocessing method of laser point cloud data includes noise point removal, data downsampling and coordinate registration performed in sequence.

[0044] Noise point removal uses a statistical filtering algorithm to calculate the average distance between each discrete point in the laser point cloud data and the surrounding 10-50 neighboring laser points. and standard deviation , set the filter threshold to to , remove outlier laser points whose distance exceeds the threshold (mainly measurement errors or bird interference points), and the removal ratio is controlled at 0.5%-5%.

[0045] Data downsampling uses a voxel grid method to divide the laser point cloud data into cubic voxels with a resolution of 0.05-0.2m. One laser point is retained in each voxel (the nearest point to the voxel center is taken), while reducing the data volume (retaining 30-50% of the original points) while maintaining the integrity of the geometric features.

[0046] For coordinate registration, select 3-10 feature points with the same name (which must be clearly identified in both the real-life 3D mesh model and the laser point cloud data, such as rooftop equipment vertices and corner points). Calculate the transformation matrix using the iterative closest point algorithm to align the laser point cloud data with the real-life 3D mesh model. After registration, the spatial position deviation is ≤0.3m, ensuring that the two types of data are based on the same coordinate system.

[0047] Exemplarily, the method for extracting three-dimensional features includes extracting plane features, contour features, and elevation features based on pre-processed high-overlapping images and laser point cloud data.

[0048] The plane features are fitted from the laser point cloud data using the random sampling consensus algorithm (RANSAC) to obtain parameters such as the slope and elevation of the roof plane. The plane fitting residual is ≤0.05m.

[0049] The contour feature uses edge detection (Canny operator) of high-overlap images to extract the contour lines of rooftop equipment and surrounding buildings (such as the rectangular boundaries of equipment and the eaves lines of buildings), with a contour accuracy of ≤0.1m.

[0050] The elevation feature is based on the Z coordinate of the laser point cloud data, and the maximum and minimum values ​​and distribution range of the elevation are counted to distinguish the elevation differences between the roof plane, protruding equipment and surrounding buildings.

[0051] Exemplarily, the classification method of laser point cloud data includes dividing the laser point cloud data into roof plane laser point cloud, roof equipment laser point cloud and surrounding building laser point cloud based on the extracted three-dimensional features through clustering and threshold judgment.

[0052] Roof plane laser point cloud refers to laser point cloud data with elevation within the main range of the building roof (within ±0.5m), uniform reflection intensity (coefficient of variation ≤ 0.1) and conforming to plane characteristics, corresponding to the flat area of ​​the roof.

[0053] Rooftop equipment laser point cloud refers to laser point cloud data with an elevation of more than 0.5m higher than the roof plane laser point cloud, high reflection intensity (10%-30% higher than the roof plane) and discrete clustered distribution, corresponding to rooftop air-conditioning units, communication antennas and other equipment.

[0054] The laser point cloud of surrounding buildings refers to the laser point cloud data whose elevation difference with the laser point cloud of the building roof plane is ≥3m and is located within 50m-100m around the building roof, corresponding to the roof or wall of the adjacent building.

[0055] An exemplary method for constructing a three-dimensional model of an obstruction surface includes constructing a three-dimensional model of the obstruction surface based on the classified laser point cloud of the rooftop equipment and the laser point cloud of the surrounding buildings to define the airspace boundary for safe flight, and the steps are as follows: The laser point cloud of the rooftop equipment and the laser point cloud of the surrounding buildings are divided into several independent areas (such as a single device, a section of wall) according to their spatial distribution. Each area is fitted into a plane using the least squares method. The fitting formula (slope equation) is: ,in , is the elevation of the obstacle surface at the plane coordinate (x, y) (m); , are the x- and y-direction slope coefficients of the k-th slice slope (unitless), satisfying ; is the constant term (m) of the kth slice slope; is the plane coordinate range (m×m) corresponding to the k-th slice slope.

[0056] All fitting planes must satisfy ( is the maximum allowable slope, ranging from 0.2 to 0.5), to ensure that the model can reflect the tilt of the obstacle.

[0057] All the sliced ​​planes are spliced ​​according to their spatial positions to form a continuous three-dimensional model of the obstacle surface. The deviation between the model boundary and the actual obstacle edge is ≤0.2m.

[0058] Step S3 generates a vector surface of the rooftop available area based on step S2, constructs a headroom height raster map, and selects a preliminary layout plan.

[0059] For example, the rooftop usable area vector surface is the usable area after excluding the space occupied by existing equipment on the roof. Specifically, a boundary point recognition algorithm based on K nearest neighbor is used to select 10-30 neighboring points (K value range) for each laser point in the roof plane laser point cloud, and calculate the average value of the normal vector angle between the laser point and its neighboring points. , and set the boundary point judgment threshold is 75°; when a laser point Greater than When the point is detected, it is determined to be a roof plane boundary point, and all boundary points are aggregated to form a roof plane boundary point set (point density 2-8 points / m), ensuring that the irregular edges of the roof plane are covered.

[0060] Based on the roof plane boundary point set, the Alpha Shapes algorithm is used to perform polygon fitting. (Value range 0.5-2m) Adjust the fitting effect: The smaller the value, the closer the fitted polygon is to the boundary point set, retaining more edge details; The larger the value, the smoother the polygon and the better at eliminating local noise. During fitting, ensure that a single-connected closed area is generated that completely contains all the roof plane laser point clouds from step S2, ultimately obtaining the initial roof plane area vector surface.

[0061] Extract the rooftop device laser point cloud classified in step S2. For each rooftop device laser point cloud, use the minimum bounding rectangle algorithm to calculate its projection range on the horizontal plane to obtain the device projection rectangle. The side length of the device projection rectangle must cover all horizontal coordinates (X, Y) of the corresponding rooftop device laser point cloud. Perform a Boolean difference operation on the initial roof plane area vector surface and all device projection rectangles: ,in is the roof available area vector surface (m 2 ); is the initial roof plane area vector surface (m 2 ); n is the number of laser point cloud clusters of rooftop equipment (unitless); is the horizontal projection rectangle of the i-th device point cloud group (m 2 ); Represents the union operation of multiple device projection rectangles (m 2 ); confirmation is required after operation Not empty, otherwise resize Or check the rooftop equipment laser point cloud classification results.

[0062] Illustratively, the clearance height grid map is used to represent the vertical airspace margin relative to the obstruction surface at each location within the rooftop usable area.

[0063] Specifically, after the rooftop usable area vector surface is generated, it is converted into a digital elevation model (DEM) in the form of a regular grid. The DEM grid resolution (res) is set to 1-2m. Square grid cells are divided according to this resolution, and the positions are marked by the plane coordinates (X, Y) of the center of the grid cells. The inverse distance weighted interpolation method (IDW) is used to calculate the elevation value of each grid cell based on the roof plane laser point cloud in step S2. The interpolation formula is: ,in is the elevation value of the grid cell at the coordinate (x, y) (m); m is the number of roof plane point clouds involved in interpolation (unitless), which is 10-20; is the elevation value of the j-th roof plane point cloud (m); is the horizontal distance (m) from the jth roof plane point cloud to the grid cell center (x, y), calculated as , 、 is the plane coordinate of the j-th roof plane point cloud); is the distance weight index (unitless, ranging from 1.5 to 3.0, the larger the value, the greater the influence of the neighboring points on the interpolation result).

[0064] After the DEM is constructed, a three-dimensional spatial difference operation is performed to calculate the clearance height of each grid cell. The clearance height is defined as the difference between the elevation (DEM elevation) of a certain location in the roof usable area and the elevation of the obstruction surface corresponding to that location. Based on the three-dimensional obstruction surface model generated in step S2, the obstruction surface segment area to which the center of each grid cell belongs is determined. , substitute the slope equation of the corresponding slice ( ), calculate the surface elevation of the obstacle at that location , and then calculate the clearance height by the formula ,in is the clearance height of the grid cell at coordinate (x, y) (m); is the DEM elevation of the location (m); is the surface elevation of the obstruction at that location (m).

[0065] Arrange the clearance heights of all grid cells according to plane coordinates to generate a clearance height grid map. Use gradient color scale to mark areas with clearance heights greater than 5m in green, areas with clearance heights between 0m and 5m in yellow, and areas with clearance heights less than 0m in red. Also, mark the coordinates, grid resolution, and clearance height units in the grid map.

[0066] Exemplarily, the preliminary layout options are screened based on the rooftop available area vector surface and vertical take-off and landing facility requirements.

[0067] Specifically, determine the landing pad size parameters of the vertical take-off and landing facility, including the landing pad length and width The value range is 10m-20m, and a square landing pad ( = ) to adapt to the take-off and landing postures of the aircraft in different directions.

[0068] Based on the set landing pad size parameters, the rooftop available area vector surface is divided into regular layout grids. The side length of each grid is equal to the side length of the landing pad, ensuring that the grid area can accommodate a landing pad; the center coordinate of the grid is The boundary of the layout grid, which serves as the center coordinate of the landing pad corresponding to the grid, must be completely inside the rooftop usable area vector surface.

[0069] Filter grids that meet the clearance height requirements and set the clearance height threshold 2-4m (the specific value can be adjusted according to the aircraft model, such as 2-3m for small aircraft and 3-4m for medium aircraft). If the clearance height of all grid cells in a grid is Both greater than , then the layout grid meets the clearance height requirement and is determined to be a valid layout position.

[0070] Convert the grids corresponding to all valid layout positions into a preliminary layout plan, and mark the coordinates of the center of the landing pad, the side length of the landing pad, and the clearance height range of the corresponding area (that is, the minimum and maximum clearance height range within the grid).

[0071] For example, the standard flight path of an aircraft includes an approach phase and a climb phase: the approach phase is the path of the aircraft descending from a high altitude to the take-off and landing pad, with the approach starting altitude range set to 50-100m, the approach speed range to 8-15m / s, the approach glide angle range to 3-5° (the glide angle is the angle between the route trajectory and the horizontal plane), and the end point of the approach route is the center coordinate of the take-off and landing pad in the preliminary layout plan. ,in is the landing pad elevation (obtained from the DEM data in step S3); the climbing phase is the path of the aircraft from the landing pad to high altitude takeoff, and the climbing starting height is set to The climbing end altitude range is 100-200m, the climbing speed range is 10-20m / s, the climbing elevation angle range is 8-12° (the elevation angle is the angle between the route trajectory and the horizontal plane), and the climbing route starting point is the center coordinate of the take-off and landing pad. At the same time, the horizontal deviation allowable range of the route trajectory is set to ±2m to ensure that small attitude deviations during aircraft operation are covered, thereby improving the comprehensiveness of simulation verification.

[0072] Centered around the landing pad A local coordinate system is established for the origin, with the X-axis oriented along the approach or climb direction, the Y-axis perpendicular to the X-axis (horizontally), and the Z-axis aligned with the Z-axis of the 2000 National Geodetic Coordinate System (vertically). Approach and climb trajectory models are established based on the approach and climb phases of the trajectory, respectively.

[0073] The approach trajectory model is , where t is the approach time (s), and its value range is 0- ; is the total approach time (s), Calculated from the approach start altitude and descent speed; is the approach speed (m / s); is the approach glide path angle (rad); is the approach starting altitude (m); is the maximum horizontal deviation during the approach phase (m), with a value range of ±2m; 、 、 are the three-dimensional coordinates (m) of the aircraft at any time t during the approach phase.

[0074] The climb trajectory model is , where t is the climbing time (s), and the value range is 0- ; is the total climbing time (s), Calculated from the climb termination altitude and the climb speed; is the climbing speed (m / s); is the climb pitch angle (rad); is the climb termination altitude (m); is the maximum horizontal deviation during the climbing phase (m), with a value range of ±2m; 、 、 are the three-dimensional coordinates (m) of the aircraft at any time t during the climb phase.

[0075] Extract the surface vertex set of the real-scene 3D mesh model generated in step S2 (n is the number of vertices), and the piecewise slope equation of the three-dimensional model of the obstacle surface ; At time intervals on approach and climb trajectories Take discrete points for 0.1-0.5s to form a trajectory point set (m is the number of trajectory points).

[0076] Calculate the minimum distance between the trajectory point and the real 3D mesh model, including for each trajectory point , calculate its relationship with all mesh vertices Euclidean distance ,in is the distance (m) from the i-th trajectory point to the j-th grid vertex. Take all The minimum value in is taken as the distance between the trajectory point and the real 3D mesh model. , then take all trajectory points The minimum value in (m).

[0077] Calculate the minimum distance between the trajectory point and the obstacle surface model, including for each trajectory point , determine the obstacle surface segment k to which it belongs, and substitute the corresponding slope equation to calculate the segment k Elevation , the vertical distance formula from the trajectory point to the slice is: ,in is the distance (m) from the i-th trajectory point to the k-th obstacle surface fragment. The minimum value among them is taken as the distance between the trajectory point and the obstacle surface model , then take all trajectory points The minimum value in (m).

[0078] Finally, and The smaller value is defined as the minimum separation distance of the preliminary layout plan, and a minimum separation distance report is generated, which includes the respective separation distances of the approach phase and the climb phase. 、 and minimum separation distance, and also mark the route parameters (speed, angle, altitude).

[0079] An exemplary noise source type is defined as a composite source of aerodynamic noise and mechanical noise. The sound power level range of the noise source is set to 105-120dB, the noise spectrum characteristic is a 1 / 3 octave spectrum, and the center frequency covers 250-8000Hz, of which the sound power level in the 250-2000Hz band accounts for 60-70% (dominated by aerodynamic noise), and the sound power level in the 2000-8000Hz band accounts for 30-40% (dominated by mechanical noise); the spatial position of the noise source is set to 1m-3m directly above the center of the take-off and landing pad in the preliminary layout plan (corresponding to the noise radiation center when the aircraft is hovering, taking off and landing), which is recorded as the noise source coordinates ,in is the noise source elevation (m), is the landing pad elevation (m), The radiation characteristics of the noise source are set to omnidirectional radiation, that is, the noise propagates evenly in the horizontal direction, and the vertical sound pressure level attenuation coefficient ranges from 0.5 to 1 dB / m (as the propagation height increases, the noise energy spreads horizontally, resulting in vertical attenuation).

[0080] A mathematical model of noise propagation attenuation is constructed. The total attenuation includes geometric divergence attenuation, air absorption attenuation and ground reflection attenuation. The noise intensity (sound pressure level) at each point is obtained by calculating the total attenuation.

[0081] Geometric divergence attenuation ( ) is the noise source, which is spherical wave radiation. During the propagation process, the energy diverges with the distance, which is expressed as ,in, is the geometric divergence attenuation (dB); is the straight-line distance from the point to the noise source (m), , is the three-dimensional coordinate of the point (m); is the reference distance (m), which is set to 1m (industry standard reference distance).

[0082] Air absorption attenuation ( ) is the energy loss caused by the vibration and friction of air molecules when noise propagates in the air, which is expressed as ,in is the air absorption attenuation (dB); is the air absorption coefficient (dB / m), ranging from 0.01 to 0.05 dB / m (typical value in the 250-8000 Hz frequency band at 25°C and standard atmospheric pressure); is the straight-line distance from the point to the noise source (m).

[0083] Ground reflection attenuation ( ) means the roof of the building is a hard surface (concrete or metal). The noise is reflected by the ground and superimposed on the direct sound, resulting in attenuation. The value range is 3-6dB (the reflection coefficient of hard ground is higher and the attenuation is smaller; the attenuation of soft ground is greater, so it is suitable for rooftop scenarios here).

[0084] Total attenuation ( ) is the sum of the three: .

[0085] Noise intensity (sound pressure level ) is calculated as: ,in is the noise sound pressure level at the point (dB), is the sound power level of the noise source (dB); 11 (dB) is the conversion constant between sound power level and sound pressure level under spherical wave radiation (corrected value at a reference distance of 1m).

[0086] The method for calculating the noise intensity distribution includes determining the entire roof of the building (the coverage of the roof available area vector surface in step S3) and the surrounding horizontal area of ​​50m-100m, and the vertical height range is -2 (ground, is the landing pad elevation, m) to +20 (above the roof); divided into regular three-dimensional grids with a grid resolution of 2-5m, and the noise propagation attenuation model is substituted into the center of each grid unit to calculate the noise sound pressure level When the grid cell is located on the roof plane ( ), needs to be combined with the corrected scope Calculate the distance from a point to the noise source: ,in Pick 、 、 Three typical values ​​are calculated separately, and the average of the three calculation results is taken as the final noise sound pressure level of the grid unit to ensure that all possible positions of the noise source in the vertical range are covered, thereby improving the comprehensiveness of the simulation results.

[0087] The noise sound pressure levels of all grid cells in the horizontal direction (i.e., the roof plane) are statistically analyzed and classified into levels according to the sound pressure level range: ≤60dB (low noise area), 60dB-70dB (medium noise area), 70dB-80dB (higher noise area), and >80dB (high noise area). Different colors (such as blue, green, yellow, and red) are used to mark the areas of each level to generate a horizontal noise intensity distribution map. At the same time, the noise sound pressure level variation curve in the vertical direction (i.e., the vertical line at the center of the take-off and landing pad) is extracted, and the noise attenuation patterns at different heights are marked as supplementary data for the distribution map. The output noise intensity distribution map must be marked with coordinates, grid resolution (m), noise sound pressure level unit (dB) and noise source parameters (sound power level, location), and accompanied by statistics on the area percentage of each noise level area.

[0088] Exemplarily, step S5 extracts parameter indicators for each preliminary layout solution from the simulation results of step S4, including: The minimum separation distance, directly derived from the output of the airspace safety simulation verification in step S4, is the closest distance (m) between the aircraft's standard flight path (approach and climb phases) and the real-world 3D mesh model and obstacle surface model. This metric, obtained directly from the distance calculation between the trajectory points and the 3D model in step S4, reflects airspace safety.

[0089] The peak noise level in the peripheral area is derived from the noise intensity distribution map extracted from the noise propagation simulation verification in step S4, encompassing the entire rooftop area and the surrounding 50-100m horizontal area. Step S4 already divides this area into grid cells with a resolution of 2-5m. In this step, the peripheral area is further defined as grid cells extending beyond 15m from the center of the landing pad (avoiding the strong noise core area directly below the landing pad to focus on the actual impact on the surrounding environment). The peak noise level in the peripheral area is determined by taking the maximum value of the noise pressure level of all grid cells within this range.

[0090] Exemplarily, the site selection scheme of the preliminary layout scheme is screened based on the above-mentioned minimum separation distance and the noise peak value of the peripheral area, including airspace safety constraints and noise constraints.

[0091] Among them, the airspace safety constraint sets a minimum separation distance of ≥7m, which is directly compared with the minimum separation distance. If it is not met, it will be excluded due to the risk of collision. The noise constraint sets a peak noise level in the peripheral area of ​​≤60dB, which is directly compared with the extracted peak noise level in the peripheral area. If it is not met, it will be excluded due to excessive interference with the surrounding environment. Only preliminary layout plans that meet both airspace safety constraints and noise constraints are included in the set of site selection plans.

[0092] After the screening is completed, the site selection plan set is sorted according to fixed rules, with priority given to sorting from large to small according to airspace safety redundancy (i.e. the difference between the minimum separation distance and 7m). The larger the difference, the more sufficient the airspace safety reserve. If the differences are the same, they are sorted from small to large according to the noise peak value of the peripheral area. The smaller the noise value, the less impact on the surrounding environment.

[0093] Finally, a site selection result report is generated, including a list of site selection options (labeled numbers, take-off and landing pad center coordinates, minimum separation distance, and noise peak values ​​in the peripheral area). In addition to the site selection list, you can also attach the following according to the labeling number: The real-life 3D mesh model (with the landing pad layout superimposed) is taken from the real-life 3D mesh model generated in step S2. The landing pad boundaries of each site selection plan are superimposed on the model (with the center coordinates and dimensions marked) to intuitively display the spatial location of the landing pad on the building roof and the surrounding environment.

[0094] The roof clearance height grid map (marking the siting plan area) is taken from the clearance height grid map generated in step S3. Special color blocks are used to mark the take-off and landing area ranges of each siting plan in the map. Combined with the color gradient of the grid map, the airspace safety redundancy corresponding to the siting plan is intuitively presented.

[0095] The noise intensity distribution map (associated with the noise peaks in the peripheral area) is taken from the noise intensity distribution map generated in step S4. The peripheral area 15 meters away from the center of the landing pad is marked with a dotted box in the map, and the noise peak positions corresponding to each site selection plan are marked. Combined with the color markings in the map, the noise impact range of different plans can be visualized.

[0096] Example systems: like Figure 2 As shown, a UAV image site selection system for vertical take-off and landing facilities for low-altitude economy includes a sequentially connected data acquisition and processing module, a three-dimensional modeling module, a preliminary screening module, a simulation module and a decision module.

[0097] The data acquisition and processing module is used to obtain remote sensing data of the building roof and surrounding areas and pre-process the remote sensing data, including: The data acquisition unit is configured to control the drone equipped with an oblique photography camera and a lidar sensor to perform oblique photogrammetry and lidar scanning on the building roof according to a preset route, obtain high-overlapping images and laser point cloud data, and synchronously record the time reference and space reference.

[0098] The data preprocessing unit is configured to perform distortion correction and aerial triangulation on high-overlap images, and to perform noise removal, downsampling, and coordinate registration on laser point cloud data to eliminate errors and unify the data benchmark.

[0099] The 3D modeling module is used to construct a real-world 3D model and obstruction surface model based on pre-processed remote sensing data, intuitively representing rooftop morphology and airspace boundaries, including: The real-scene 3D modeling unit is configured to perform dense matching on the pre-processed high-overlapping images to generate a dense point cloud, and to generate a real-scene 3D mesh model covering the roof and surrounding areas through surface reconstruction and texture projection.

[0100] The three-dimensional feature extraction unit is configured to extract plane features, contour features and elevation features from the pre-processed high-overlapping image and laser point cloud data.

[0101] The obstacle modeling unit is configured to classify the laser point cloud data based on three-dimensional features and construct a three-dimensional model of the obstacle surface to define the boundary of the safe flight airspace.

[0102] The preliminary screening module is used to calculate the headroom height of the rooftop usable area based on the 3D model and select preliminary layout plans that meet basic airspace requirements, including: The available area analysis unit is configured to generate a roof available area vector surface through boundary recognition and Boolean operation based on the roof plane point cloud and the equipment point cloud, excluding the space occupied by the equipment.

[0103] The clearance height calculation unit is configured to construct a digital elevation model based on the available area vector surface, and generate a clearance height raster map through inverse distance weighted interpolation and obstacle elevation calculation.

[0104] The layout screening unit is configured to divide the take-off and landing pad into regular grids according to the size parameters of the take-off and landing pad, and screen the grids with a clearance height greater than a preset threshold as the preliminary layout plan.

[0105] The simulation module is used to simulate airspace safety and noise propagation for preliminary layout plans, including: The airspace safety simulation unit is configured to establish approach and climb route trajectory models, calculate the minimum separation distance between the route and the real-life three-dimensional model and obstacle model, and assess the collision risk.

[0106] The noise propagation simulation unit is configured to define the characteristics of the noise source, construct a propagation model including geometric divergence, air absorption and ground reflection attenuation, calculate the noise intensity and generate a distribution map.

[0107] The decision module is used to select the final site selection plan based on the simulation results, including: The constraint screening unit is configured to set a minimum separation distance and a peripheral noise peak threshold, and screen solutions that meet both airspace safety and noise constraints.

[0108] The scheme sorting unit is configured to sort the screened schemes according to airspace safety redundancy and noise peak value to determine the priority.

[0109] The above description is based on the ideal embodiment of the present invention. Based on the above description, relevant personnel can make various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the content of the specification and must be determined according to the scope of the claims.

Claims

1. A method for selecting a site for vertical take-off and landing facilities using drone images for low-altitude economy, characterized by: The following steps are involved: Step S1: Obtain remote sensing data by performing oblique photogrammetry and laser radar scanning on the roof of the building using a drone; the remote sensing data includes high-overlapping images and laser point cloud data; Step S2: Preprocessing the remote sensing data, and establishing a real-scene 3D mesh model and an obstacle surface 3D model based on the preprocessed remote sensing data; the real-scene 3D mesh model is used to represent the surface morphology of the building roof; and the obstacle surface 3D model is used to define the airspace boundary for safe flight. Step S3: Based on the real-scene 3D grid model and the 3D surface model of the obstacle, a preliminary layout plan is selected according to the clearance height of the available roof area; Step S4: For each preliminary layout plan, airspace safety simulation and noise propagation simulation are performed to obtain the minimum separation distance and noise intensity distribution map; The airspace safety simulation is based on distance detection between the flight path and the real-scene three-dimensional grid model and the three-dimensional model of the obstacle surface; The noise propagation simulation is based on the calculation of the intensity distribution of noise on the roof of the building; Step S5: Based on the minimum separation distance and the noise intensity distribution map, the site selection plan is screened according to airspace safety constraints and noise constraints.

2. The method according to claim 1, characterized in that The pre-processing in step S2 includes: performing image distortion correction and aerial triangulation on the high overlap image; Noise point removal, data downsampling and coordinate registration are performed on the laser point cloud data.

3. The method according to claim 1, characterized in that The method for establishing the real-scene three-dimensional grid model includes: Dense matching is performed on the pre-processed high-overlap images to generate dense point clouds; Performing surface reconstruction on the dense point cloud to generate a triangular facet mesh; The texture information of the high-overlapping image is projected onto the triangular face mesh to generate a real-scene three-dimensional mesh model.

4. The method according to claim 1, wherein Extract 3D features including plane features, contour features, and elevation features based on pre-processed high-overlapping imagery and laser point cloud data; Based on the three-dimensional features, the laser point cloud data is classified into roof plane laser point cloud, roof equipment laser point cloud and surrounding building laser point cloud through clustering and threshold judgment.

5. The method according to claim 4, characterized in that The method for establishing the three-dimensional model of the obstacle surface includes: Based on the classified laser point cloud of the rooftop equipment and the laser point cloud of the surrounding buildings, the system divides the points into several independent areas according to their spatial distribution. Each independent region was fitted into a plane using the least squares method; All fitting planes are spliced ​​according to their spatial positions to form a continuous three-dimensional model of the obstacle surface.

6. The method according to claim 1, characterized in that The preliminary layout schemes selected based on the headroom height of the roof available area include: Generate roof usable area vector surface based on real-life 3D mesh model; Based on the roof available area vector surface, a digital elevation model is constructed by using an inverse distance weighted interpolation method, and the clearance height of each grid cell is calculated to generate a clearance height grid map; According to the landing pad size parameters, the roof available area vector surface is divided into regular grids; Grids with clearance height greater than the preset threshold are selected as preliminary layout plans.

7. The method according to claim 1, characterized in that The airspace safety simulation includes: Based on the approach phase and climb phase of the route trajectory, an approach route trajectory model and a climb route trajectory model are established respectively; Calculate the minimum distance between the flight path and the vertex of the real-scene 3D mesh model surface; Calculate the minimum distance between the flight path and the three-dimensional model of the obstacle surface; The smaller of the two minimum distances is taken as the minimum separation distance.

8. The method according to claim 1, characterized in that The noise propagation simulation includes: Define the sound power level, spectral characteristics and spatial location of the noise source; Establish a noise propagation attenuation model including geometric divergence attenuation, air absorption attenuation and ground reflection attenuation; A noise intensity distribution map is generated based on the attenuation model.

9. The method according to claim 1, characterized in that The method for screening site selection options based on airspace safety constraints and noise constraints includes: Set the minimum separation distance threshold and the peripheral area noise peak threshold; Screen out preliminary layout solutions that meet both the minimum separation distance threshold and the peripheral area noise peak threshold.

10. A vertical take-off and landing facility drone image site selection system for low-altitude economy, based on the method according to any one of claims 1 to 9, characterized in that: It includes a data acquisition and processing module, a three-dimensional modeling module, a preliminary screening module, a simulation module and a decision-making module connected in sequence; Among them, the data acquisition and processing module is used to obtain and pre-process remote sensing data; the three-dimensional modeling module is used to construct a real-scene three-dimensional grid model and a three-dimensional model of the obstacle surface; the preliminary screening module is used to screen preliminary layout plans; the simulation module is used to perform airspace safety simulation and noise propagation simulation; and the decision-making module is used to screen and output site selection plans.

Images