Unmanned aerial vehicle path planning method for safety inspection of hoisting machinery

Abstract

The invention discloses an unmanned aerial vehicle path planning method for safety inspection of hoisting machinery, and relates to the technical field of safety inspection, and the method comprises the steps: constructing a construction site three-dimensional point cloud model based on construction site data; based on the three-dimensional point cloud model of the construction site, analyzing the digital surface geometric characteristics of each hoisting machine, and generating a full-coverage scanning path of an inspection target point for each hoisting machine; establishing a target function by taking minimization of the total transfer distance between the hoisting machinery as an optimization target, solving and generating a transfer path between the hoisting machinery by adopting an improved dung beetle optimization algorithm, and introducing good point set initialization, a dynamic disturbance factor and Cauchy Gaussian mixture variation on the dung beetle optimization algorithm; and a global fitness function is constructed, and an improved dung beetle optimization algorithm is adopted to solve and generate a global path. According to the method, the global optimality, convergence stability and inspection efficiency of path planning in a complex construction site environment can be greatly improved, the safety and efficiency of automatic inspection of the hoisting machinery are effectively guaranteed, and the method has important engineering application value.

Description

Technical Field

[0001] This invention relates to the field of safety inspection technology, and in particular to a method for unmanned aerial vehicle (UAV) path planning for safety inspection of lifting machinery. Background Technology

[0002] With the acceleration of urbanization, super high-rise buildings and large-scale infrastructure projects are emerging, and the scale and structural complexity of modern construction sites are growing exponentially. The safe operation of construction lifting machinery is directly related to the lives of construction workers and the progress of the project. Manual inspection is limited by the height and shape of construction lifting machinery, making it easy to miss blind spots and dead ends, and high-altitude operations have high safety risks and low work efficiency.

[0003] Drone technology offers a revolutionary solution for the safety inspection of construction lifting machinery, but its efficiency is highly dependent on the performance of the underlying path planning method. In complex, unstructured construction site environments, planning a safe and efficient flight path for drones that can cover critical parts of the lifting machinery, such as the tower, boom, and tie rods, without blind spots and while avoiding obstacles, is the core challenge and bottleneck for achieving automated inspection. Existing path planning methods face significant limitations in such scenarios: graph search-based algorithms have a large number of computational nodes in three-dimensional sparse space, resulting in low planning efficiency; paths generated by random sampling algorithms are often highly random and lack smoothness, failing to meet the requirements of drone flight control; and traditional swarm intelligence algorithms are prone to getting trapped in local optima due to loss of population diversity or inappropriate search strategies, leading to lengthy planned paths, collision risks, or incomplete coverage of all inspection points.

[0004] The Dung Beetle Optimizer (DBO), as an emerging metaheuristic algorithm, simulates multiple natural behaviors such as rolling, reproduction, foraging, and stealing, demonstrating good global exploration capabilities and convergence speed when solving complex optimization problems. This provides a potential solution to the aforementioned path planning problem. However, directly applying the original DBO algorithm to the refined inspection path planning of construction crane machinery still has significant limitations: First, the completely random nature of its initial population generation makes it difficult to quickly form a high-quality solution set in a complex three-dimensional constrained space, resulting in poor convergence starting points and long warm-up times; second, the algorithm lacks sufficient development capabilities in the later stages, with weak local fine-grained search capabilities for current elite solutions, easily leading to a final path that is not optimal in detail, containing unnecessary turns or undulations; finally, the algorithm model does not fully consider practical requirements such as path smoothness and UAV dynamic constraints, and the planned path may be "theoretically shortest" but "practically difficult to fly." These inherent defects severely restrict the planning effect and application potential of DBO in the high-precision scenario of UAV inspection.

[0005] Therefore, in response to the specific needs of safety inspection of construction lifting machinery and the technical difficulties of three-dimensional path planning, it is urgent to make targeted improvements to the original DBO algorithm to solve key problems such as long planning time, poor path quality, and premature convergence. This will improve the intelligence level and operational reliability of UAV autonomous inspection, and ultimately meet the normalized and refined requirements of safety monitoring in modern large-scale projects. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a UAV path planning method for safety inspection of lifting machinery, which improves the intelligence level and operational reliability of UAV autonomous inspection based on an improved dung beetle optimization algorithm.

[0007] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0008] This invention provides a UAV path planning method for safety inspection of lifting machinery, comprising:

[0009] Acquire construction site data, including lifting machinery, and construct a 3D point cloud model of the construction site based on the construction site data;

[0010] Based on the inspection task, the inspection target points of each crane are marked on the three-dimensional point cloud model of the construction site;

[0011] Based on the construction site 3D point cloud model, analyze the digital surface geometric characteristics of each lifting machine, and generate a full-coverage scanning path for each lifting machine's inspection target points.

[0012] An objective function is constructed with minimizing the total transfer distance between lifting machines as the optimization objective. This objective function is then used as the fitness function, and an improved dung beetle optimization algorithm is employed to generate the transfer paths between the lifting machines. The improved dung beetle optimization algorithm introduces optimal point set initialization, dynamic perturbation factor, and Cauchy-Gaussian hybrid mutation into the dung beetle optimization algorithm.

[0013] A global fitness function is constructed based on the full-coverage scanning path of the lifting machinery and the transfer path between the lifting machinery. An improved dung beetle optimization algorithm is used to solve and generate the global path.

[0014] Optionally, acquiring construction site data, including lifting machinery, and constructing a 3D point cloud model of the construction site based on the construction site data includes:

[0015] The drone is controlled to fly around the construction site, and the 3D coordinates of feature points on the lifting machinery are obtained through the RTK positioning module on the drone; the image data of the construction site is obtained through the camera on the drone.

[0016] A preliminary map of the construction site is constructed based on the image data using a visual SLAM algorithm, and feature points of the lifting machinery are associated on the preliminary map.

[0017] By combining map point reprojection error with feature point prior error, a joint cost function is constructed. :

[0018]

[0019] In the formula, For the first in the image data Camera pose of keyframes For the first in the preliminary map The three-dimensional coordinates of a map point For camera projection model functions, for exist Image observations, for The covariance matrix, For robust kernel functions; The set of feature points that were successfully associated. For the first The three-dimensional coordinates of each feature point For the first The three-dimensional coordinates of the map points corresponding to each feature point for The covariance matrix;

[0020] The minimum value of the joint cost function is found using the gradient descent method. The poses of all cameras and the coordinates of map points are then adjusted to obtain the final 3D point cloud model of the construction site.

[0021] Optionally, the step of marking the inspection target points for each crane on the 3D point cloud model of the construction site according to the inspection task includes:

[0022] Based on the inspection task, determine all the lifting machinery that needs to be inspected, and mark a limited number of inspection target points on the three-dimensional point cloud model of the construction site according to the model and safety specifications of each lifting machinery.

[0023] Optionally, the step of analyzing the digital surface geometric characteristics of each crane based on the 3D point cloud model of the construction site and generating a full-coverage scanning path for the inspection target points of each crane includes:

[0024] The skeleton of the lifting machinery in the three-dimensional point cloud model of the construction site is extracted, and the spatial main axis of the lifting machinery is marked based on the skeleton extraction result;

[0025] Based on the main spatial axis, the lifting machinery is decomposed into multiple structural components, and the type of each structural component is identified through geometric calculations.

[0026] A scanning strategy is developed for each structural component based on its type, and an initial scanning path for the lifting machinery is generated based on the scanning strategy.

[0027] On the initial scan path, a camera viewpoint is simulated, and path points are manually added for structural components that are prone to blind spots to ensure full scan coverage.

[0028] Optionally, the step of developing a scanning strategy based on the type of each structural component includes:

[0029] For structural components of columnar type, a scanning strategy of spiral ascent with equal radius is adopted;

[0030] For structural components of the beam type, a surface scanning strategy along its main axis is adopted;

[0031] For structural components that are complex aggregates, a scanning strategy is adopted to generate multi-view observation points along multiple surface normal directions of their outer envelope.

[0032] Optionally, the objective function constructed with minimizing the total transfer distance between lifting machines as the optimization objective includes:

[0033] Each drone takeoff point corresponding to a lifting machine is considered a task point, and the set of task points is denoted as . , For the number of lifting machinery;

[0034] Set the inspection sequence of lifting machinery based on the task point set. , To minimize the total transfer distance between lifting machines, an objective function is constructed. :

[0035]

[0036] In the formula, For inspection sequence The Middle There are 10 task points to be inspected, and d is the distance function.

[0037] Optionally, the process of solving and generating transfer paths between lifting machinery includes:

[0038] Set the maximum number of iterations for the improved dung beetle optimization algorithm. and population size ;

[0039] Based on population size Initialize the dung beetle population using a set of optimal points;

[0040] After each iteration, the optimal fitness value of an individual dung beetle is calculated, and different dung beetles update their positions according to their respective strategies.

[0041] If the optimal fitness value remains unchanged for multiple consecutive times, the improved dung beetle optimization algorithm enters a stagnation state, and the position of the dung beetle population is updated using Cauchy-Gaussian mixture mutation.

[0042] When the maximum number of iterations is reached Then, the dung beetle individual corresponding to the optimal fitness value is obtained, and the transfer path between the lifting machinery is generated.

[0043] Optionally, the different dung beetles updating their positions according to their respective strategies includes:

[0044] The rolling dung beetle updates its position when there are no obstacles:

[0045]

[0046] In the formula, This represents the current iteration number. For the first Only one dung beetle in the first Position information at the next iteration The constant representing the deflection coefficient, , This is the natural coefficient, assigned a value of 1 or -1; It is a constant. , This is the worst position globally at present;

[0047] The rolling dung beetle updates its position when there are obstacles:

[0048]

[0049] In the formula, For deflection angle, , Take 0, or At that time, the dung beetle does not update its position;

[0050] Breeding dung beetles update their locations when there are obstacles, including:

[0051] Reproductive dung beetles search for suitable oviposition areas around dung balls to lay their eggs. The boundaries of the oviposition area are defined as follows:

[0052]

[0053] In the formula, These are the lower and upper boundaries of the spawning area, respectively. This is the current local optimum position. As an intermediate variable, , These are the lower and upper bounds of the search space, respectively;

[0054] After determining the boundaries of the oviposition area, the breeding dung beetle lays its eggs, and the location of the egg balls is updated as follows:

[0055]

[0056] In the formula, For the first The egg sphere in the first Position information at the next iteration Let each be an independent random vector of size 1×D, where D is the dimension of the optimization problem;

[0057] Dung beetles update their location when there are obstacles, including:

[0058] Small dung beetles are used to mimic foraging behavior and mark out foraging areas:

[0059]

[0060] In the formula, These are the lower and upper boundaries of the foraging area, respectively. This is the current globally optimal position;

[0061] The location of the dung beetle has been updated to:

[0062]

[0063] In the formula, For random numbers that follow a normal distribution, for A random vector within a certain range;

[0064] The thieving dung beetle updates its location when there are obstacles:

[0065]

[0066] In the formula, It is a random vector of size 1×D that follows a normal distribution; This is a dynamic perturbation factor that reflects the intensity of random perturbations by the dung beetle.

[0067] Optionally, the dynamic disturbance factor for:

[0068]

[0069] In the formula, To improve the maximum number of iterations in the dung beetle optimization algorithm.

[0070] Optionally, the construction of the global fitness function based on the full-coverage scanning path of the lifting machinery and the transfer path between the lifting machinery includes:

[0071] Cost function for constructing path length :

[0072]

[0073] In the formula, Number of waypoints For the first The three-dimensional coordinates of each waypoint;

[0074] Constructing the cost function of energy loss :

[0075]

[0076] In the formula, For the first From the first waypoint to the second Vectors of waypoints;

[0077] Constructing the cost function of collision risk :

[0078]

[0079] In the formula, For drones in the The minimum distance from each waypoint to the surface of the obstacle. For drones in the The cost of threats at each waypoint For the drone's wingspan;

[0080] According to the cost function Constructing the global fitness function :

[0081]

[0082] In the formula, These are the weighting coefficients, , .

[0083] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0084] This invention provides a UAV path planning method for safety inspection of lifting machinery. Compared with the original dung beetle optimization algorithm, this method significantly enhances the algorithm's global exploration capability and local development accuracy by introducing multiple strategies such as optimal point set initialization, dynamic perturbation factor, and Cauchy-Gaussian hybrid mutation. It effectively overcomes the shortcomings of the original algorithm, such as premature convergence, poor initial solution quality, and weak later optimization. This allows the improved algorithm to quickly plan shorter, smoother, and safer flight paths in complex 3D construction site environments. It not only greatly improves the planning efficiency and path quality of UAV inspection but also ensures the safety and economy of the flight process, providing an effective solution to the UAV path planning problem in complex construction site environments. It effectively solves the inherent drawbacks of traditional manual inspection, such as high safety risks, low work efficiency, and the presence of blind spots, achieving automated, full-coverage, and high-precision unmanned inspection. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating the UAV path planning method for safety inspection of lifting machinery provided in an embodiment of the present invention.

[0086] Figure 2 This is a schematic diagram illustrating the implementation of the UAV path planning method for safety inspection of lifting machinery provided in this embodiment of the invention.

[0087] Figure 3 This is a diagram illustrating the effect of the improved population initialization strategy for optimal point sets provided in this embodiment of the invention.

[0088] Figure 4 This is a diagram illustrating the effect of the disturbance factor improvement strategy provided in this embodiment of the invention.

[0089] Figure 5 This is a flowchart illustrating the improved dung beetle optimization algorithm provided in an embodiment of the present invention;

[0090] Figure 6 This is a three-dimensional view of the comparative simulation results of the UAV under working condition A provided in the embodiment of the present invention;

[0091] Figure 7 This is a top view of the comparative simulation results of the UAV under working condition A provided in the embodiment of the present invention;

[0092] Figure 8 This is a three-dimensional view of the comparative simulation results of the UAV under working condition B provided in the embodiment of the present invention;

[0093] Figure 9 This is a top view of the comparative simulation results of the UAV under working condition B provided in the embodiment of the present invention;

[0094] Figure 10 This is a three-dimensional view of the comparative simulation results of the UAV under working condition C provided in the embodiment of the present invention;

[0095] Figure 11 This is a top view of the comparative simulation results of the UAV under working condition C provided in the embodiment of the present invention. Detailed Implementation

[0096] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0097] Example 1

[0098] like Figure 1 and Figure 2 As shown, this embodiment of the invention provides a UAV path planning method for safety inspection of lifting machinery, including the following steps:

[0099] Step S1: Obtain construction site data, including cranes, and construct a 3D point cloud model of the construction site based on the construction site data.

[0100] Before the inspection task begins, a 3D reconstruction of the construction site environment and obstacle mapping are first carried out, specifically including:

[0101] 1.1 Control the drone to fly around the construction site, obtain the three-dimensional coordinates of feature points on the lifting machinery through the RTK positioning module on the drone; and obtain image data of the construction site through the camera on the drone.

[0102] Feature points, such as the apex of the crane boom, the apex of the counterweight boom, and the apex of the lifting machinery, can be set according to actual needs.

[0103] 1.2. A preliminary map of the construction site is constructed based on the image data using a visual SLAM algorithm, and feature points of the lifting machinery are associated on the preliminary map;

[0104] 1.3. Combine map point reprojection error with feature point prior error to construct a joint cost function. :

[0105]

[0106] In the formula, For the first in the image data Camera pose of keyframes For the first in the preliminary map The three-dimensional coordinates of a map point For camera projection model functions, for exist Image observations, for The covariance matrix, For robust kernel functions; The set of feature points that were successfully associated. For the first The three-dimensional coordinates of each feature point For the first The three-dimensional coordinates of the map points corresponding to each feature point for The covariance matrix; the joint cost function The two terms are summed. The first term is the sum of the reprojection errors of all map points, and the second term is the core constraint introduced in this invention. The feature point constraint term acts as a strong constraint, which can reduce the error between the map generated by visual SLAM and the real construction site. After optimization and convergence, a high-precision 3D environment map with global correction is directly output.

[0107] 1.4. Use the gradient descent method to find the minimum value of the joint cost function, adjust the poses of all cameras and the coordinates of map points to obtain the final 3D point cloud model of the construction site.

[0108] An accurate 3D point cloud model of the construction site includes the static environment, including the structure of the crane itself, surrounding buildings, and overhead cables. This serves as the spatial basis for subsequent path planning, and is used to identify obstacles and delineate safe flight airspace.

[0109] Step S2: Based on the inspection task, mark the inspection target points for each crane on the 3D point cloud model of the construction site; specifically including:

[0110] Based on the inspection task, determine all the lifting machinery that needs to be inspected, and mark a limited number of inspection target points on the three-dimensional point cloud model of the construction site according to the model and safety specifications of each lifting machinery.

[0111] Step S3: Analyze the digital surface geometry of each crane based on the 3D point cloud model of the construction site, and generate a full-coverage scan path for the inspection target points of each crane; specifically including:

[0112] 3.1 Extract the skeleton of the 3D point cloud model of the crane in the 3D point cloud model of the construction site, and mark the spatial main axis of the crane based on the skeleton extraction results, such as the central axis of the tower body and the extension direction of the boom.

[0113] 3.2. Based on the spatial main axis, the lifting machinery is decomposed into multiple structural components, and the type of each structural component is identified through geometric calculations; structural components include tower body, boom, counterweight boom, etc.

[0114] 3.3. Develop a scanning strategy for each structural component based on its type, and generate a preliminary scanning path for the lifting machinery based on the scanning strategy;

[0115] For structural components of columnar type (such as tower body), a scanning strategy of spiral ascent with equal radius is adopted;

[0116] For structural components of the beam type (such as booms and counterweights), a surface scanning strategy along their main axis is adopted;

[0117] For structural components that are complex assemblies (such as cabs, slewing bearings, and lifting mechanisms), a scanning strategy is adopted to generate multi-view observation points along the normal directions of multiple surfaces of their outer envelope.

[0118] 3.4. On the initial scanning path, simulate the camera's perspective and manually add path points for structural components that are prone to blind spots to ensure full scan coverage;

[0119] For structures such as lifting mechanisms and driver's cabs that are prone to blind spots, manually add sub-path points for "detour" or "lift" to ensure full coverage of the inspection.

[0120] After generating a full-coverage scan path for the lifting machinery, camera actions can be added to each sub-path based on the characteristics of different structures on the construction lifting machinery. These actions include gimbal yaw angle, gimbal pitch angle, camera zoom, taking photos, and recording videos. The complete action sequence for controlling the drone to complete the inspection of the entire equipment will then be written into the drone's executable file.

[0121] Step S4: Construct an objective function with the goal of minimizing the total transfer distance between lifting machines. Use the objective function as the fitness function and solve the transfer path between lifting machines using the improved dung beetle optimization algorithm. The improved dung beetle optimization algorithm introduces optimal point set initialization, dynamic perturbation factor, and Cauchy-Gaussian hybrid mutation into the dung beetle optimization algorithm.

[0122] (1) Initialization of the optimal point set. To avoid uneven initial population distribution due to random generation, the dung beetle population is initialized based on the optimal point set. The specific formula for the optimal point set is as follows:

[0123]

[0124] In the formula, For best results, The fractional portion is the portion that is extracted. For the optimal number of points, , for The unit cube in Euclidean space, the set of optimal points The following constraint must be satisfied:

[0125]

[0126] In the formula, For only with and ( (where is any positive number) and related constants. The specific construction rules are shown in the following formula:

[0127]

[0128] In the formula, To meet The smallest prime number. Mapping the optimal point set to the actual search space initializes the population. Specifically, the location of the dung beetle population is shown in the following equation:

[0129]

[0130] In the formula, For the first In the population under the dimension, the _th The position of each individual and These represent the lower and upper bounds of the search boundary in the corresponding dimension, respectively.

[0131] Based on population size Dimension , the smallest prime number For example, the population distribution is as follows: Figure 3 As shown.

[0132] (2) Dynamic perturbation factor. The perturbation factor, as a parameter reflecting the intensity of random perturbation in the dung beetle optimization algorithm, is a constant with a value of 0.5 in the original algorithm. Static parameter design cannot adapt to the needs of different stages in the optimization process. Therefore, it is proposed to dynamically adjust the perturbation factor using the exponential decay formula. Specifically, The calculation formula is as follows:

[0133]

[0134] In the formula, To improve the maximum number of iterations in the dung beetle optimization algorithm.

[0135] Compared to static parameters, the perturbation factor under a dynamic strategy can enhance search capability with high perturbation in the early stages and improve convergence accuracy with low perturbation in the later stages. For example... Figure 4 As shown, the improvement effect of the perturbation factor is demonstrated.

[0136] (3) Cauchy-Gaussian Mutation. To address the issue that traditional optimization algorithms are prone to getting trapped in local optima when encountering multimodal problems, a dynamic mutation mechanism based on a Cauchy-Gaussian mixture distribution is introduced. Specifically, the new solution after mutation is shown in the following equation:

[0137]

[0138] In the formula, For the first The first generation of the population The current solution for each individual. This is the new solution after mutation. Weighting coefficients. and This constitutes a regulatory mechanism that evolves over time. This represents the current iteration number. This represents the maximum number of iterations. It enhances global exploration capabilities through the long-tail characteristic of Cauchy mutation while retaining the local development advantages of Gaussian mutation.

[0139] Specifically, the objective function constructed with minimizing the total transfer distance between lifting machines as the optimization objective includes:

[0140] Each drone takeoff point corresponding to a lifting machine is considered a task point, and the set of task points is denoted as . , For the number of lifting machinery;

[0141] Set the inspection sequence of lifting machinery based on the task point set. , To minimize the total transfer distance between lifting machines, an objective function is constructed. :

[0142]

[0143] In the formula, For inspection sequence The Middle There are 10 task points to be inspected, where d is a distance function, such as the Euclidean distance function.

[0144] like Figure 5 As shown, solving for the transfer path between lifting machinery includes:

[0145] 4.1 Set the maximum number of iterations for the improved dung beetle optimization algorithm. and population size ;

[0146] 4.2. Based on population size Initialize the dung beetle population using a set of optimal points;

[0147] 4.3. After each iteration, calculate the optimal fitness value of the dung beetle individual, and update the position of different dung beetles according to their respective strategies;

[0148] 4.4 If the optimal fitness value remains unchanged for multiple consecutive times, the improved dung beetle optimization algorithm enters a stagnation state, and the position of the dung beetle population is updated using Cauchy-Gaussian mixture mutation.

[0149] 4.5. Upon reaching the maximum number of iterations Then, the dung beetle individual corresponding to the optimal fitness value is obtained, and the transfer path between the lifting machinery is generated.

[0150] Among them, different dung beetles update their positions according to their own strategies, including:

[0151] The rolling dung beetle updates its position when there are no obstacles:

[0152]

[0153] In the formula, This represents the current iteration number. For the first Only one dung beetle in the first Position information at the next iteration The constant representing the deflection coefficient, , This is the natural coefficient, assigned a value of 1 or -1; It is a constant. , This is the worst position globally at present;

[0154] The rolling dung beetle updates its position when there are obstacles:

[0155]

[0156] In the formula, For deflection angle, , Take 0, or At that time, the dung beetle does not update its position;

[0157] Breeding dung beetles update their locations when there are obstacles, including:

[0158] Reproductive dung beetles search for suitable oviposition areas around dung balls to lay their eggs. The boundaries of the oviposition area are defined as follows:

[0159]

[0160] In the formula, These are the lower and upper boundaries of the spawning area, respectively. This is the current local optimum position. As an intermediate variable, , These are the lower and upper bounds of the search space, respectively;

[0161] After determining the boundaries of the oviposition area, the breeding dung beetle lays its eggs, and the location of the egg balls is updated as follows:

[0162]

[0163] In the formula, For the first The egg sphere in the first Position information at the next iteration Let each be an independent random vector of size 1×D, where D is the dimension of the optimization problem;

[0164] Dung beetles update their location when there are obstacles, including:

[0165] Small dung beetles are used to mimic foraging behavior and mark out foraging areas:

[0166]

[0167] In the formula, These are the lower and upper boundaries of the foraging area, respectively. This is the current globally optimal position;

[0168] The location of the dung beetle has been updated to:

[0169]

[0170] In the formula, For random numbers that follow a normal distribution, for A random vector within a certain range;

[0171] The thieving dung beetle updates its location when there are obstacles:

[0172]

[0173] In the formula, It is a random vector of size 1×D that follows a normal distribution; This is a dynamic perturbation factor that reflects the intensity of random perturbations by the dung beetle.

[0174] Step S5: Construct a global fitness function based on the full-coverage scanning path of the lifting machinery and the transfer path between the lifting machinery, and use the improved dung beetle optimization algorithm to solve and generate the global path.

[0175] The full-coverage scan path can be considered as an intra-equipment path, and the transfer path between lifting machinery can be considered as an extra-equipment path. The global fitness function is constructed based on the full-coverage scan path and the transfer path between lifting machinery, including:

[0176] In a 3D environment, the cost function of path length is constructed by calculating the Euclidean distance between adjacent waypoints as the cost of the path length between devices. :

[0177]

[0178] In the formula, Number of waypoints For the first The three-dimensional coordinates of each waypoint;

[0179] When drones turn, larger yaw angles often result in greater energy consumption. Therefore, smaller yaw angles along the inspection path can effectively improve the drone's endurance. This allows for the construction of a cost function for energy consumption. :

[0180]

[0181] In the formula, For the first From the first waypoint to the second Vectors of waypoints;

[0182] To avoid safety hazards caused by drones getting too close to obstacles or equipment to be inspected, a distance-based collision risk assessment system is proposed, and a cost function for collision risk is constructed. :

[0183]

[0184] In the formula, For drones in the The minimum distance from each waypoint to the surface of the obstacle. For drones in the The cost of threats at each waypoint For the drone's wingspan;

[0185] According to the cost function Constructing the global fitness function :

[0186]

[0187] In the formula, These are the weighting coefficients, , .

[0188] The improved dung beetle optimization algorithm is used again to iteratively optimize the encoded solution. By simulating behaviors such as rolling a ball, dancing, and foraging, the algorithm fine-tunes the waypoint parameters of the paths within and between devices, and finally outputs a complete inspection plan that is optimal on the comprehensive fitness function and seamlessly connects the paths inside and outside the devices.

[0189] After determining the optimal global inspection plan, the plan is converted into a standard KML / KMZ mission file format that the DJI drone platform can recognize and execute, suitable for the target drone's flight control system. This mission file contains complete flight instructions: an ordered sequence of waypoints (including all in-equipment waypoints and inter-equipment transfer waypoints), action instructions for each waypoint (drone actions, gimbal actions), and flight parameters (flight speed, turning mode). The generated standard mission file is uploaded to the drone via the ground station control software, DJI Pilot 2. After takeoff, the drone automatically loads and executes the pre-planned path, completing the automated inspection of the entire construction site without manual remote control intervention.

[0190] In this embodiment, a simulated construction site environment is built to test the performance of the improved algorithm. The construction site size is set to 500×500×200, and columnar structures are placed on the map to simulate obstacles and no-fly zones encountered on the construction site. Due to the complexity and diversity of the construction site environment, three different environments are constructed as follows:

[0191] (1) Scenario A: This scenario contains 500 obstacles and 2 no-fly zones with a size of 100×100×150. The starting coordinates are (80,250,100) and the ending coordinates are (80,210,100). The UAV needs to start from the starting point, traverse 5 preset waypoints in sequence, and reach the ending point.

[0192] (2) Scenario B: This scenario contains 30 obstacles, one no-fly zone of size 100×100×100 and one no-fly zone of size 150×150×150. The starting coordinates are (10,200,1) and the ending coordinates are (10,100,1). The drone needs to start from the starting point, traverse the three construction cranes and reach the ending point.

[0193] (3) Condition C: This scenario contains 200 obstacles and a no-fly zone of size 100×100×100. The starting coordinates are (10,210,1) and the ending coordinates are (10,190,1). The drone needs to start from the starting point, traverse 7 construction cranes and reach the ending point.

[0194] In this embodiment, the improved IDBO and the original DBO were run 10 times each under three simulation conditions, and the optimal and average values ​​of the fitness function were calculated as shown in Table 1.

[0195] Table 1. Results of the simulation under three different environments

[0196]

[0197] The 3D view of the UAV path planning under condition A is as follows: Figure 6 As shown, the top view is as follows Figure 7As shown. Obstacles are represented by gray semi-transparent lines, solid orange lines represent paths planned by IDBO, and dashed red lines represent paths planned by DBO. (Refer to Table 1 and...) Figure 6 , Figure 7 It can be seen that IDBO paths exhibit smoother spatial adaptability when avoiding obstacles, while DBO paths have more turns. Compared to DBO, IDBO path planning reduces the average fitness function by 11.47%, verifying IDBO's optimization capability in complex 3D environments.

[0198] The 3D view of the UAV path planning under condition B is as follows: Figure 8 As shown, the top view is as follows Figure 9 As shown in the figure, the gray solid line represents the preset inspection path. The drone will automatically enter the inspection path after arriving at the tower crane. The results in Table 1 also show that, under operating condition B, IDBO is 7.23% lower than DBO in terms of the average fitness function.

[0199] The 3D view of the UAV path planning under operating condition C is as follows: Figure 10 As shown, the top view is as follows Figure 11 As shown in Table 1, in environments with more obstacles, DBO exhibits greater path redundancy and significantly greater vertical fluctuations than IDBO. Table 1 also shows that under condition C, IDBO reduces the average fitness function by 9.17% compared to DBO.

[0200] Combining the path planning performance across three different scenarios, as the number of obstacles increases and the environment becomes more complex, DBO exhibits significantly increased redundancy in the vertical direction, leading to additional energy consumption and flight instability. In contrast, IDBO shows less trajectory fluctuation in the vertical direction and demonstrates greater optimization in scenarios with a higher proportion of obstacles.

[0201] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0202] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0203] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0204] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0205] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for UAV path planning for safety inspection of lifting machinery, characterized in that, include: Acquire construction site data, including lifting machinery, and construct a 3D point cloud model of the construction site based on the construction site data; Based on the inspection task, the inspection target points of each crane are marked on the three-dimensional point cloud model of the construction site; Based on the construction site 3D point cloud model, analyze the digital surface geometric characteristics of each lifting machine, and generate a full-coverage scanning path for each lifting machine's inspection target points. An objective function is constructed with minimizing the total transfer distance between lifting machines as the optimization objective. This objective function is then used as the fitness function, and an improved dung beetle optimization algorithm is employed to generate the transfer paths between the lifting machines. The improved dung beetle optimization algorithm introduces optimal point set initialization, dynamic perturbation factor, and Cauchy-Gaussian hybrid mutation into the dung beetle optimization algorithm. A global fitness function is constructed based on the full-coverage scanning path of the lifting machinery and the transfer path between the lifting machinery. An improved dung beetle optimization algorithm is used to solve and generate the global path.

2. The UAV path planning method for safety inspection of lifting machinery according to claim 1, characterized in that, The step of acquiring construction site data, including lifting machinery, and constructing a 3D point cloud model of the construction site based on the construction site data includes: The drone is controlled to fly around the construction site, and the 3D coordinates of feature points on the lifting machinery are obtained through the RTK positioning module on the drone; the image data of the construction site is obtained through the camera on the drone. A preliminary map of the construction site is constructed using the visual SLAM algorithm based on the image data, and feature points of the lifting machinery are associated on the preliminary map. By combining map point reprojection error with feature point prior error, a joint cost function is constructed. : ； In the formula, For the first in the image data Camera pose of keyframes For the first in the preliminary map The three-dimensional coordinates of a map point For camera projection model functions, for exist Image observations, for The covariance matrix, For robust kernel functions; The set of feature points that were successfully associated. For the first The three-dimensional coordinates of each feature point For the first The three-dimensional coordinates of the map points corresponding to each feature point for The covariance matrix; The minimum value of the joint cost function is found using the gradient descent method. The poses of all cameras and the coordinates of map points are then adjusted to obtain the final 3D point cloud model of the construction site.

3. The UAV path planning method for safety inspection of lifting machinery according to claim 1, characterized in that, The step of marking the inspection target points for each crane on the 3D point cloud model of the construction site according to the inspection task includes: Based on the inspection task, determine all the lifting machinery that needs to be inspected, and mark a limited number of inspection target points on the three-dimensional point cloud model of the construction site according to the model and safety specifications of each lifting machinery.

4. The UAV path planning method for safety inspection of lifting machinery according to claim 1, characterized in that, The step of analyzing the digital surface geometric characteristics of each crane based on the 3D point cloud model of the construction site, and generating a full-coverage scanning path for the inspection target points of each crane, includes: The skeleton of the lifting machinery in the three-dimensional point cloud model of the construction site is extracted, and the spatial main axis of the lifting machinery is marked based on the skeleton extraction result; Based on the main spatial axis, the lifting machinery is decomposed into multiple structural components, and the type of each structural component is identified through geometric calculations. A scanning strategy is developed for each structural component based on its type, and an initial scanning path for the lifting machinery is generated based on the scanning strategy. On the initial scan path, a camera viewpoint is simulated, and path points are manually added for structural components that are prone to blind spots to ensure full scan coverage.

5. The UAV path planning method for safety inspection of lifting machinery according to claim 4, characterized in that, The process of developing a scanning strategy based on the type of each structural component includes: For structural components of columnar type, a scanning strategy of spiral ascent with equal radius is adopted; For structural components of the beam type, a surface scanning strategy along its main axis is adopted; For structural components that are complex aggregates, a scanning strategy is adopted to generate multi-view observation points along multiple surface normal directions of their outer envelope.

6. The UAV path planning method for safety inspection of lifting machinery according to claim 1, characterized in that, The objective function constructed with minimizing the total transfer distance between lifting machines as the optimization objective includes: Each drone takeoff point corresponding to a lifting machine is considered a task point, and the set of task points is denoted as . , For the number of lifting machinery; Set the inspection sequence of lifting machinery based on the task point set. , To minimize the total transfer distance between lifting machines, an objective function is constructed. : ； In the formula, For inspection sequence The Middle There are 10 task points to be inspected, and d is the distance function.

7. The UAV path planning method for safety inspection of lifting machinery according to claim 1 or 6, characterized in that, The solution generates the transfer paths between lifting machines, including: Set the maximum number of iterations for the improved dung beetle optimization algorithm. and population size ; Based on population size Initialize the dung beetle population using a set of optimal points; After each iteration, the optimal fitness value of an individual dung beetle is calculated, and different dung beetles update their positions according to their respective strategies. If the optimal fitness value remains unchanged for multiple consecutive times, the improved dung beetle optimization algorithm enters a stagnation state, and the position of the dung beetle population is updated using Cauchy-Gaussian mixture mutation. When the maximum number of iterations is reached Then, the dung beetle individual corresponding to the optimal fitness value is obtained, and the transfer path between the lifting machinery is generated.

8. The UAV path planning method for safety inspection of lifting machinery according to claim 7, characterized in that, The different dung beetles update their positions according to their respective strategies, including: The rolling dung beetle updates its position when there are no obstacles: ； In the formula, This represents the current iteration number. For the first Only one dung beetle in the first Position information at the next iteration The constant representing the deflection coefficient, , The natural coefficient is assigned a value of 1 or -1; It is a constant. , This is the worst position globally at present; The rolling dung beetle updates its position when there are obstacles: ； In the formula, For deflection angle, , Take 0, or At that time, the dung beetle does not update its position; Breeding dung beetles update their locations when there are obstacles, including: Reproductive dung beetles search for suitable oviposition areas around dung balls to lay their eggs. The boundaries of the oviposition area are defined as follows: ； In the formula, These are the lower and upper boundaries of the spawning area, respectively. This is the current local optimal position. As an intermediate variable, , These are the lower and upper bounds of the search space, respectively; After determining the boundaries of the oviposition area, the breeding dung beetle lays its eggs, and the location of the egg balls is updated as follows: ； In the formula, For the first The egg sphere in the first Position information at the next iteration Let each be an independent random vector of size 1×D, where D is the dimension of the optimization problem; Dung beetles update their location when there are obstacles, including: Small dung beetles are used to mimic foraging behavior and mark out foraging areas: ； In the formula, These are the lower and upper boundaries of the foraging area, respectively. This is the current globally optimal position; The location of the dung beetle has been updated to: ； In the formula, For random numbers that follow a normal distribution, for A random vector within a certain range; The thieving dung beetle updates its location when there are obstacles: ； In the formula, It is a random vector of size 1×D that follows a normal distribution; This is a dynamic perturbation factor that reflects the intensity of random perturbations by the dung beetle.

9. The UAV path planning method for safety inspection of lifting machinery according to claim 1, characterized in that, The dynamic disturbance factor is: ； In the formula, Let be the dynamic perturbation factor at the t-th iteration. To improve the maximum number of iterations in the dung beetle optimization algorithm, exp is an exponential function.

10. The UAV path planning method for safety inspection of lifting machinery according to claim 1, characterized in that, The global fitness function is constructed based on the full-coverage scanning path of the lifting machinery and the transfer path between the lifting machinery, including: Cost function for constructing path length : ； In the formula, Number of waypoints For the first The three-dimensional coordinates of each waypoint; Constructing the cost function of energy loss : ； In the formula, For the first From the first waypoint to the second Vectors of waypoints; Constructing the cost function of collision risk : ； In the formula, For drones in the The minimum distance from each waypoint to the surface of the obstacle. For drones in the The cost of threats at each waypoint For the drone's wingspan; According to the cost function Constructing the global fitness function : ； In the formula, These are the weighting coefficients, , .

Images