Automatic tower crane inspection method and system based on unmanned aerial vehicle

Abstract

The invention provides a tower crane automatic inspection method and system based on an unmanned aerial vehicle, and the method comprises the steps: planning an inspection route based on the structure parameters of a tower crane and a three-dimensional model of the environment where the tower crane is located, and generating a preset route planning scheme based on a starting point; initializing the state of the tower crane; the unmanned aerial vehicle flies to a preset position and collects starting point data; generating an inspection path according to a preset route planning scheme and the collected starting point data; the unmanned aerial vehicle inspects key points on the tower crane according to the inspection path to obtain a key point video; performing defect identification according to the key point video; according to a defect identification result, performing safety assessment on the tower crane; the system is used for implementing the method. The safety and convenience of tower crane inspection are improved, and the tower crane can be inspected more comprehensively.

Description

Technical Field

[0001] The present invention relates to an automatic inspection method and system for tower cranes, and particularly to an automatic inspection method and system for tower cranes based on unmanned aerial vehicles (UAVs). Background Art

[0002] The information provided in this section is only background information related to the present disclosure and does not necessarily represent prior art.

[0003] With the development of the construction industry, tower cranes are widely used in various construction sites. At present, the detection methods for tower cranes are relatively traditional. Mainly, workers carry detection instruments and climb to the corresponding structural positions, and then check for obvious cracks, corrosion, missing parts, damage and other defects through the instruments or visually.

[0004] The above prior art has the following defects: manual operation is time-consuming and laborious, the safety risk of working at height is high, it is difficult for personnel to observe special parts, and it is difficult to avoid obstacles manually in a complex construction site environment.

[0005] It should be noted that the information disclosed in the above background art section is only used to enhance the understanding of the background of the present disclosure, and thus may include information that does not constitute prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] Object of the Invention: The technical problem to be solved by the present invention is to provide an automatic inspection method and system for tower cranes based on UAVs in view of the deficiencies of the prior art.

[0007] To solve the above technical problem, the present invention discloses an automatic inspection method and system for tower cranes based on UAVs, wherein the system includes the following steps:

[0008] Step 1: Based on the structural parameters of the tower crane and the three-dimensional model of its environment, plan the inspection route and generate a preset trajectory planning scheme based on the starting point;

[0009] Step 2: Initialize the state of the tower crane;

[0010] Step 3: The UAV flies to a preset position to collect starting point data;

[0011] Step 4: According to the preset trajectory planning scheme generated in Step 1 and the starting point data collected in Step 3, generate the inspection path;

[0012] Step 5: The UAV inspects key points on the tower crane according to the inspection path and obtains key point videos;

[0013] Step 6: Based on the key point videos, perform defect identification;

[0014] Step 7: Conduct a safety assessment of the tower crane based on the defect identification results.

[0015] Furthermore, the planned inspection route described in Step 1, that is, using the improved A* optimization method, generates a preset flight path based on the starting point according to the structural parameters of the tower crane and the three-dimensional model of its environment, specifically including:

[0016] Step 1-1: Construct a three-dimensional configuration space encoded by an octree, specifically as follows:

[0017] Based on the height of the tower crane, the length of the boom, the length of the counterweight arm, and the three-dimensional model of environmental obstacles, the space is defined as multiple cubic regions, and the side length of the cube is set to the neighborhood step size of the UAV;

[0018] Simulate the hierarchical space division of the octree, and use the 26-neighborhood connectivity model to represent the feasible motion primitives of the UAV, that is, the 26 motion directions of the UAV include:

[0019] 6 orthogonal motion directions, 12 diagonal edge motion directions, and 8 body diagonal motion directions;

[0020] The neighborhood step size of the UAV represents the minimum unit of the UAV's single-step motion;

[0021] Step 1-2: Set the starting point and key point data in the three-dimensional configuration space, specifically as follows:

[0022] The starting point data includes: the longitude and latitude of the starting point, the height relative to the ground, and the yaw angle of the UAV; set the coordinate system as follows: let the direction of the boom be the positive direction of the y-axis, the origin be the ground point at the center of the tower crane body, the three-dimensional coordinates of the starting point be (a, b, c), the yaw angle of the UAV be d°, and the UAV camera faces the side of the tower crane body;

[0023] The key point data includes:

[0024] The key point data of the jacking mechanism, in addition to the starting point data, also includes the height h1 of the jacking mechanism from the ground; the key point data of the jacking mechanism is (a, b, c + h1);

[0025] The key point data of the counterweight arm, in addition to the starting point data, also includes the height h2 of the counterweight arm from the ground, and the length e1 of the counterweight arm is increased in the direction of the yaw angle +90° relative to the starting point; the key point data of the counterweight arm is (a, b - e1, c + h2);

[0026] The key point data of the top of the boom, relative to the key point data of the counterweight arm, the total length e2 of the boom and the counterweight arm is increased in the direction of the yaw angle -90° relative to the starting point; the key point data of the top of the boom is (a, b - e1 + e2, c + h2);

[0027] The data of the key points at the top of the tower, in addition to the starting point data, also includes the height h3 of the top of the tower from the ground; the data of the key points at the top of the tower is (a, b, c + h3);

[0028] The above key point data is also key point data after being mirrored with respect to the z-axis height reference plane in the three-dimensional coordinate system;

[0029] Step 1-3, starting from the starting point, taking any key point as the target, and using a bidirectional search strategy to search until all key points are traversed to obtain a preset trajectory.

[0030] Furthermore, the search using the bidirectional search strategy in Step 1-3 includes:

[0031] Step 1-3-1, taking the current trajectory point and the next key point as the initial trajectory points respectively;

[0032] Step 1-3-2, starting the search based on the initial trajectory points, generating candidate trajectory point sets respectively, and selecting the one with the minimum evaluation function f(n) from the candidate trajectory points as the next trajectory point of the initial trajectory point;

[0033] Step 1-3-3, judging the distance between the next trajectory points of the two initial trajectory points. If it is less than 2 times the neighborhood step size, then execute Step 1-3-4 for path connectivity judgment, otherwise execute Step 1-3-5;

[0034] Step 1-3-4, if the path between the initial trajectory point and the next trajectory point is connected, then execute Step 1-3-5, otherwise discard the next trajectory point;

[0035] Step 1-3-5, putting the next trajectory point into the preset trajectory, and returning to Step 1-3-1, taking this trajectory point as the new current trajectory point.

[0036] Furthermore, the evaluation function f(n) in Step 1-3-2 is specifically as follows:

[0037] f(n) = wh(n) + g(n)

[0038] Among them, g(n) is the cost function, representing the actual cost from the current trajectory point to the next trajectory point n, h(n) is the heuristic function h(n), providing directional guidance from the current trajectory point to the next trajectory point, and w is the weight factor.

[0039] Furthermore, the heuristic function h(n) in Step 1-3-2 is specifically as follows:

[0040] h(n) = max(dx, dy, dz) + 0.414 * min(dx, dy, dz)

[0041] Among them, dx, dy, and dz are the absolute values of the coordinate differences between the current flight path point and the next flight path point where the drone is located in three-dimensional space, max() is the maximum value, and min() is the minimum value.

[0042] Furthermore, the cost function g(n) described in step 1-3-2 is specifically as follows:

[0043]

[0044] Among them, d k is the basic movement distance of the drone at the k-th step, α(θ k ) is the wind direction compensation coefficient, which is determined by the angle θ k between the moving direction of the drone and the wind speed direction, and β(v ⊥ , θ k ) is the crosswind stability compensation term, which is related to the crosswind component v ⊥ and the angle θ k .

[0045] Furthermore, the wind direction compensation coefficient α(θ k ) described in step 1-3-2 is specifically as follows:

[0046]

[0047] The crosswind stability compensation term β(v ⊥ , θ k ) is specifically as follows:

[0048]

[0049] Among them, λ is the wind resistance coefficient of the drone.

[0050] Furthermore, the initialization of the state of the tower crane described in step 2 includes:

[0051] Perform calibration of the tower crane attitude space coordinate system, drive the tower crane slewing mechanism, and make the axis of the boom form an orthogonal geometric relationship with the reference plane of the tower body introduction section.

[0052] Furthermore, the defect identification described in step 6 includes:

[0053] Adopt a pre-trained YOLOV8 recognition model to identify rust on the tower crane body, rust on the wire rope, looseness of the wire rope, and broken wires of the wire rope.

[0054] The present invention also proposes an automatic inspection system for tower cranes based on drones to implement the aforementioned automatic inspection method, including:

[0055] More than 1 drone conducts inspections on key points on the tower crane according to the flight route planned in the automatic inspection method;

[0056] The drone is connected to the cloud platform for data interaction;

[0057] The cloud platform conducts data analysis and processing, is connected to the mobile monitoring terminal, and sends the processed data to the mobile monitoring terminal.

[0058] Beneficial effects:

[0059] The present invention generates an improved automatic inspection route for drones through an improved A* optimization algorithm to achieve automatic inspection of tower cranes by drones. Through human-machine collaborative defect recognition and a fuzzy multi-level safety status evaluation method, the overall safety condition of the tower crane is quantitatively rated, specifically as follows:

[0060] 1. The present invention combines the principles of atmospheric fluid mechanics with a path search algorithm, making the path planning result conform to the laws of aerodynamics, and can reduce energy consumption by 15%-30% compared with traditional static cost models; extends the octile distance to three-dimensional space, and while maintaining the completeness of the algorithm, improves the node expansion efficiency by about 40%; combines a double-threshold termination condition (step_size×2 neighborhood determination) to effectively balance the contradiction between search accuracy and calculation efficiency;

[0061] 2. In the present invention, after the 4K images collected by the drone are pre-screened by the YOLOv8 model, the suspected defect areas are automatically marked, and the inspection personnel can review the defects in real time through the cloud API video transmission. The human-machine collaborative defect recognition makes the inspection more detailed, and the false alarm rate of defects is significantly reduced.

[0062] 3. In the present invention, the drone conducts non-contact automatic detection on tower cranes, achieving full coverage detection of key parts such as the connection points of tower crane standard sections, boom structural members, and jacking sleeve frames, significantly reducing the risk of high-altitude manual operations and improving the safety detection efficiency of tower cranes. Description of the drawings

[0063] The following further specifically describes the present invention in conjunction with the drawings and specific embodiments, and the above and / or other advantages of the present invention will become clearer.

[0064] Figure 1 It is a flowchart of the drone automatic inspection system.

[0065] Figure 2 It is the initial route and the optimized route in the three-dimensional environment of the tower crane.

[0066] Figure 3 It is a pose diagram of the tower crane inspection.

[0067] Figure 4 It is a live picture of the cloud API.

[0068] Figure 5a It is a schematic diagram of the YOLOV8 identifying the defect of the tower crane as rust.

[0069] Figure 5b It is a schematic diagram of the YOLOV8 identifying the defects of the tower crane as rusty, loose and broken wire ropes.

[0070] Figure 6 It is a flowchart for generating a prefabricated flight path framework.

[0071] Figure 7 It is a block diagram of the automatic inspection system for tower cranes. Specific implementation manners

[0072] The present invention provides an automatic inspection method and system for tower cranes based on unmanned aerial vehicles, aiming to generate an automatic inspection route for the unmanned aerial vehicle by combining an improved A* algorithm and centimeter-level spatial registration technology of an RTK module, detecting all defects of the tower crane by combining human-machine collaborative diagnosis technology, and quantitatively rating the overall safety condition of the tower crane by using a fuzzy multi-level safety state evaluation method, so as to realize the automatic inspection of the safety assessment of the whole tower crane.

[0073] Among them, the method is as Figure 1 shown, and mainly includes the following steps:

[0074] Step 1: Use an improved A* optimization algorithm to plan an improved inspection route based on the structural parameters of the tower crane and the three-dimensional model of the construction site, and generate a pre-set flight path framework based on the starting point of the flight route;

[0075] Among them, the improvement of the improved A* optimization algorithm is as follows:

[0076] Step 1-1: The improved A* optimization algorithm constructs a three-dimensional configuration space encoded by an octree, simulates the hierarchical space division of the octree by defining multiple cube regions (components), and uses a 26-neighborhood connectivity model to represent the feasible motion primitives of the unmanned aerial vehicle. The 26 motion directions include:

[0077] 6-direction orthogonal motion (±X, ±Y, ±Z axis directions)

[0078] 12-direction diagonal edge motion (such as ±X±Y, ±Y±Z, etc.)

[0079] 8-direction body diagonal motion (such as ±X±Y±Z)

[0080] The step size of each neighborhood is controlled by step_size = 5, which represents the minimum unit of the single-step motion of the unmanned aerial vehicle;

[0081] Step 1-2: The improved A* optimization algorithm uses a three-dimensional Octile distance heuristic function h(n). While maintaining the completeness of the algorithm, it improves the node expansion efficiency by approximately 40%. The function expression is as follows:

[0082] h(n) = max(dx, dy, dz) + 0.414 * min(dx, dy, dz)

[0083] where dx, dy, and dz are the absolute values of the coordinate differences between the current node and the next key point in three-dimensional space, max(dx, dy, dz) is the maximum difference in the three-axis directions, and min(dx, dy, dz) is the minimum difference in the three-axis directions;

[0084] A weight factor w is introduced in the heuristic function, and heuristic_weight = 1.2 is set to accelerate convergence;

[0085] Step 1-3: A bidirectional search strategy is introduced. One group is for the search starting from the current node, and the other group is for the search starting from the next key point. In each iteration, the node with the minimum evaluation value is selected for expansion from both directions. This evaluation value is the actual cost of the node plus the heuristic estimated cost to the target point. Combining the double-threshold termination condition (step_size × 2 neighborhood determination) effectively balances the contradiction between search accuracy and computational efficiency. The specific formula is as follows:

[0086] f(n) = wh(n) + g(n)

[0087] where f(n) is a very important evaluation function that determines the decision-making in the algorithm search process; w is the weight factor; g(n) is the actual cost from the current node to the next key point n, representing the path length or cost from the starting point to the current node; h(n) is the distance heuristic function; and the formula for g(n) is:

[0088]

[0089] where d k is the basic movement distance at the k-th step (default step size is 5 meters), α(θ k ) is the wind direction compensation coefficient, which is determined by the angle θ k between the movement direction and the wind speed direction, and β(v ⊥ , θ k ) is the crosswind stability compensation term, which is related to the crosswind component v ⊥ and the angle θ k ; the formula for α(θ k ) is:

[0090]

[0091] The cost increases against the wind, decreases with the wind, and is corrected according to the vertical component in the case of crosswind;

[0092] Among them, the formula for β(v ⊥ , θ k ) is:

[0093]

[0094] Among them, λ is the wind resistance coefficient of the UAV (determined by the model type. For example, λ = 0.1 s 2 / m 2 ), v ⊥ is the crosswind component, that is, the component of the wind speed perpendicular to the flight direction;

[0095] Steps 1-4: Establish a hydrodynamic model by introducing the three-dimensional wind speed vector influence factor (wind_effect), calculate the cosine value of the angle between the moving direction and the wind speed using the vector projection method, and implement a dynamic compensation mechanism for the cost of the flight segment, so that the path planning result conforms to the laws of aerodynamics. Compared with the traditional static cost model, the energy consumption can be reduced by 15%-30%. The mathematical model of the vector projection method:

[0096]

[0097] Among them, v wind is the wind speed vector, v move is the moving direction vector, and the result value is a scalar, representing the effective component of the wind speed in the moving direction (positive for downwind and negative for upwind);

[0098] Step 2: Perform calibration of the tower crane attitude space coordinate system, drive the tower crane slewing mechanism, and make the axis of the boom form an orthogonal geometric relationship with the reference plane of the tower body introduction section;

[0099] Among them, the calibration of the tower crane attitude space coordinate system means making the direction of the boom the same as the orientation of the tower body introduction section;

[0100] Step 3: The UAV flies to a specific position of the tower crane to collect the starting point of the flight path, and maps the starting point of the flight path to the flight path cloud control platform through centimeter-level spatial registration technology to obtain the starting point data;

[0101] Among them, the starting point of the flight path is that the UAV flies below the cab within the safe distance between the tower body and the ground, and the camera is directly facing one side of the tower body. The centimeter-level spatial registration technology RTK module accurately collects the longitude and latitude of the starting point and the yaw angle of the UAV;

[0102] Among them, the centimeter-level spatial registration technology RTK module is installed on the UAV, which can support network RTK and custom network RTK services to achieve centimeter-level positioning of the UAV;

[0103] Step 4: Generate an inspection path within the preset flight path framework based on the starting point coordinates. For key areas such as the pin connection surface of the tower cap and the hinge point of the balance arm, deploy zoom shooting points, and import the flight path to the flight path cloud control platform;

[0104] Among them, the key parts are the jacking mechanism, climbing claws, slewing bearing, cab, tower cap, wire rope winch, luffing trolley, pin connection surface of the tower cap, and hinge point of the balance arm. The optimal optical focal length is determined to be 2.0 times through human-machine flight video analysis;

[0105] The flight path cloud control platform is software developed for formulating flight paths, which can be connected to the drone to achieve the upload and download of flight paths;

[0106] Step 5: The drone performs fully automatic inspection of the tower crane;

[0107] Step 6: Adopt a human-machine collaborative defect diagnosis method. The inspection personnel analyze and locate defects in real time through the cloud API video backhaul technology, and use the yolov8 algorithm to analyze and identify defects in the backhaul video;

[0108] The defect recognition by the YOLOV8 algorithm in Step 6 is the optimal YOLOV8 model after training, which can recognize rust, wire rope rust, wire rope looseness, and wire rope broken wires;

[0109] Step 7: Adopt a fuzzy multi-level safety status evaluation method to conduct a safety assessment of the entire tower crane.

[0110] Among them, the fuzzy multi-level safety status evaluation method proposes a fuzzy multi-level analysis and calculation model. For example, by classifying equipment hidden dangers into three levels of risk categories (lightweight defects, structural abnormalities, core function failures) according to severity, and respectively configuring gradient-increasing deduction weightings (4 / 10 / 20 points). By accumulating the scores of each defect, the overall safety status of the tower crane can be quantitatively rated.

[0111] Example:

[0112] Step 1: Use the improved A* optimization algorithm to plan an improved inspection route based on the tower crane structure parameters and the 3D model of the construction site, and generate a preset flight path framework based on the starting point of the flight path;

[0113] As Figure 2 shown, the gray 3D model is a model established based on the tower crane structure parameters and building obstacles. The red dashed line is the original planned route, and the blue solid line is the route optimized by the improved A* optimization algorithm. Here, a preset flight path framework based on the starting point of the flight path is generated;

[0114] In this embodiment, it is assumed that the current starting waypoint of the UAV is n-1(1, 1, 1), and the next key point is n(9, 9, 9). At this time, some of the nodes are a(4, 5, 1) and b(5, 4, 1), and the step size d k = 5m. If the angle between the moving direction of node a and the wind direction is 60°, and the angle of node b is 30°, at this time the wind speed v = 4m / s, the crosswind component v ⊥ of side a = v·sin60°≈3.46m / s, and the crosswind component v ⊥ of side b = v·sin30° = 2m / s, and the wind resistance coefficient λ = 0.1;

[0115] Step 1-1: The improved A* optimization algorithm constructs a three-dimensional configuration space encoded by an octree. By defining multiple cube regions (components) to simulate the hierarchical space division of the octree, a 26-neighborhood connectivity model is used to represent the feasible motion primitives of the UAV. The 26 motion directions include:

[0116] 6-direction orthogonal motion (±X, ±Y, ±Z axis directions)

[0117] 12-direction diagonal edge motion (such as ±X±Y, ±Y±Z, etc.)

[0118] 8-direction body diagonal motion (such as ±X±Y±Z)

[0119] Each neighborhood step size is controlled by step_size = 5, representing the minimum unit of the UAV's single-step motion;

[0120] In step 1-2, the improved A* optimization algorithm uses a three-dimensional Octile distance heuristic function. While maintaining the completeness of the algorithm, it improves the node expansion efficiency by about 40%. The function expression is:

[0121] h(n) = max(dx, dy, dz)+0.414*min(dx, dy, dz)

[0122] where dx, dy, dz are the absolute values of the coordinate differences between the current node and the next key point in the three-dimensional space, max(dx, dy, dz) is the maximum difference in the three-axis directions, and min(dx, dy, dz) is the minimum difference in the three-axis directions;

[0123] Calculate:

[0124] h(a) = h(b) = max(dx, dy, dz)+0.414*min(dx, dy, dz) = 8 + 3.312 = 11.312

[0125] In the heuristic function, a weight factor w is introduced and heuristic_weight = 1.2 is set to accelerate convergence;

[0126] Steps 1 - 3: Introduce a bidirectional search strategy. One set is for the search starting from the current node, and the other set is for the search starting from the next key point. In each iteration, the algorithm selects the node with the minimum evaluation value for expansion from both directions. This evaluation value is the sum of the actual cost of the node and the heuristic estimated cost to the target point. Combining with the double - threshold termination condition (step_size×2 neighborhood determination), it effectively balances the contradiction between search accuracy and computational efficiency. The specific formula is as follows:

[0127] f(n) = wh(n)+g(n)

[0128] Where f(n) is a very important evaluation function that determines the decision - making in the algorithm's search process; w is the weight factor; g(n) is the actual movement cost from the current node to the next key point n, representing the path length or cost from the starting point to the current node; h(n) is the heuristic function;

[0129] Calculate:

[0130] f(a) = 1.2h(a)+g(a)=22.1044

[0131] f(b) = 1.2h(b)+g(b)=23.1044

[0132] f(b)>f(a)

[0133] Then the waypoint cost of point a is smaller than that of point b, and node a is preferentially selected;

[0134] Where the formula for g(n) is:

[0135]

[0136] Where d k The basic movement distance at the k - th step (default step size is 5 meters), α(θ k ) is the wind - direction compensation coefficient, which is determined by the angle θ k between the movement direction and the wind - speed direction, γ(v ⊥ ,θ k ) is the cross - wind stability compensation term, which is related to the cross - wind component v ⊥ and the angle θ k ;

[0137] Calculate:

[0138] g(a)=5*1.5 + 1.03=7.5 + 1.03=8.53

[0139] g(b)=5*1.866+0.2 = 9.53

[0140] Where the formula for α(θ k ) is:

[0141]

[0142] The cost increases against the wind, decreases with the wind, and is corrected according to the vertical component in the case of crosswind;

[0143] Calculation:

[0144] α a (60°) = 1 + cos 60° = 1 + 0.5 = 1.5

[0145] α b (30°) = 1 + cos 30° = 1 + 0.866 = 1.866

[0146] where β(v ⊥ , θ k ) The formula is:

[0147]

[0148] where λ is the wind resistance coefficient of the UAV (determined by the model type, for example, λ = 0.1 s 2 / m 2 ), v ⊥ is the crosswind component, that is, the component of the wind speed perpendicular to the flight direction;

[0149] Calculation:

[0150]

[0151] As Figure 6 shown, the preset flight path framework is a route framework constructed based on the three-dimensional structure model of the tower crane and the distribution of nearby building obstacles, and it generates a complete inspection flight route by fusing the optimal flight path point sequences under multiple constraints. The specific implementation process is as follows:

[0152] 1. Based on the tower crane parameters (crane height, boom length, counterweight arm length, etc.) and the distance between the building and the tower crane, establish a three-dimensional model. Adopt the hierarchical cube division strategy to decompose the space into discrete voxel units with the side length controlled by step_size = 5, and mark the voxel of the obstacle as an impassable area;

[0153] 2. Enter the starting point coordinates (longitude and latitude of the starting point, height relative to the ground, yaw angle of the UAV). Based on the starting point coordinates, determine the coordinates of key points using the tower crane parameters and functional relationships. The coordinates of the key inspection points of the tower crane are as follows, in order: Key point of the jacking mechanism: On the basis of the starting point coordinates, increase the height of the jacking mechanism from the ground, with the yaw angle of the UAV remaining unchanged; Key point of the counter jib: On the basis of the starting point coordinates, increase the height of the counter jib from the ground, and increase the length of the counter jib in the direction of the yaw angle +90° relative to the starting point; Key point at the top of the boom: Relative to the coordinates of the key point of the counter jib, increase the total length of the boom and the counter jib in the direction of the yaw angle -90° relative to the starting point; Key point of the tower top: On the basis of the starting point coordinates, increase the height of the tower top from the ground; At the same time, taking the vertical reference plane relative to the connection line of the boom and the counter jib as the standard, there are mirror-image key points on the other side of the tower crane;

[0154] 3. Start the search simultaneously from the current point of the flight path and the next key point, generate candidate waypoint sets respectively. Each time, select the node with the smallest f(n) from the candidate waypoints for expansion. By accumulating the movement cost (g(n)) from the current point to the next key point, including the basic step cost and wind speed compensation (increasing cost against the wind and decreasing cost with the wind), as well as the heuristic estimate (h(n)) of the three-dimensional Octile distance to estimate the remaining distance from the current waypoint to the target, and combining with the weight factor (1.2 times) to accelerate convergence. Each waypoint can move in the up-down, front-back, left-right, and diagonal directions (a total of 26 directions), with a fixed step length of 5 meters. When the distance between the waypoints of the two-way search ≤ 10 meters (i.e., 2 times the step length), it is determined that the path is connected, thus generating the flight path. In this way, a preset flight path framework can be generated through the starting point coordinates;

[0155] Step 2. Perform calibration of the tower crane attitude space coordinate system, drive the slewing mechanism of the tower crane to make the axis of the boom form an orthogonal geometric relationship with the reference plane of the tower body's insertion section;

[0156] As Figure 3 shown, for the calibration of the tower crane attitude space coordinate system described in Step 2, make the direction of the boom the same as the orientation of the tower body's insertion section;

[0157] Step 3. The UAV flies to a specific position of the tower crane to collect the starting point of the flight path, and maps the starting point of the flight path to the flight path cloud control platform through centimeter-level spatial registration technology to obtain the starting point data;

[0158] For the starting point of the flight path described in Step 3, the UAV flies 5 meters below the cab, 5 meters from the ground, and the camera is directly facing one side of the tower body. The longitude and latitude of the starting point and the yaw angle of the UAV are accurately collected by the centimeter-level spatial registration technology RTK module (an existing commercial RTK module can be used);

[0159] The centimeter-level spatial registration technology RTK module described in step 3-1 is installed on the DJI Mavic 3e and supports network RTK and custom network RTK services to achieve centimeter-level positioning of the drone.

[0160] In step 4, an automatic inspection route is generated within the preset flight path framework based on the starting point coordinates. Zoom shooting waypoints are deployed for key points, and the flight path is imported into the flight path cloud control platform.

[0161] The key parts mentioned above are the connection surfaces of the jacking mechanism, climbing claws, slewing bearing, cab, tower cap, wire rope hoist, luffing trolley, tower cap pin shaft, and balance arm hinge point. Through the analysis of the drone flight video, the optimal optical focal length is determined to be 2.0 times.

[0162] The flight path cloud control platform is software developed for formulating flight paths and can be connected to the drone to achieve the upload and download of flight paths.

[0163] In step 5, the drone conducts a full-automatic inspection of the tower crane according to the improved flight path.

[0164] In step 6, a human-machine collaborative defect diagnosis method is adopted. The inspection personnel analyze and locate defects in real time through the cloud API video backhaul technology, and use the YOLOV8 algorithm to analyze and identify defects in the backhaul video (reference: https: / / blog.csdn.net / virobotics / article / details / 130156212).

[0165] The cloud API mentioned above is a secondary development program for the drone in the cloud. Its interface is as Figure 4 shown. After connecting and debugging the cloud API, the video backhaul function is realized.

[0166] The YOLOV8 algorithm for defect recognition mentioned above is the optimal YOLOV8 model after training, which can recognize rust. The recognition effect is as Figure 5b shown. For wire rope rust, wire rope looseness, and wire rope broken wires, the recognition effect is as Figure 5a shown.

[0167] Regarding the fuzzy multi-level safety status evaluation method described in step 7, a fuzzy multi-level analysis and calculation model is proposed, including: Class A hidden dangers may involve minor surface damage or wear of the tower crane, with 4 points deducted for each item; Class B hidden dangers may include relatively serious structural damage or functional abnormalities of the tower crane, with 10 points deducted for each item; and Class C hidden dangers may be directly related to the safe operation of the tower crane, such as serious damage or failure of key components, with 20 points deducted for each item. By accumulating the scores of various defects, the overall safety status of the tower crane can be quantitatively rated; as shown in Table 1:

[0168] Table 1 Quantitative rating table

[0169]

[0170] As Figure 7 shown, it is a block diagram of an automatic inspection system for tower cranes. The system at least includes:

[0171] One or more drones are used to inspect key points on the tower crane according to the planned flight path in the foregoing method. The drones are data-connected to the cloud platform, which is used to control the drones, perform data analysis and processing, and finally send the processed data to the mobile monitoring terminal.

[0172] In specific implementation, the present application provides a computer storage medium and a corresponding data processing unit. Among them, the computer storage medium can store a computer program, and when the computer program is executed by the data processing unit, it can run the inventive content of a method and system for automatic inspection of tower cranes based on drones and some or all of the steps in each embodiment. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0173] Those skilled in the art can clearly understand that the technical solutions in the embodiments of the present invention can be implemented by means of a computer program and its corresponding general hardware platform. Based on such an understanding, the essence of the technical solutions in the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a computer program, that is, a software product. The computer program software product can be stored in a storage medium, including several instructions for causing a device (which can be a personal computer, a server, a single-chip microcomputer, an MCU, or a network device, etc.) including a data processing unit to execute the methods described in each embodiment or some parts of the embodiments of the present invention.

[0174] The present invention provides an idea and method for a method and system for automatic inspection of tower cranes based on drones. There are many methods and ways to specifically implement this technical solution. The foregoing is only the preferred implementation manner of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Each component not clearly defined in this embodiment can be implemented by the prior art.

Claims

1. An automatic inspection method for tower cranes based on drones, characterized in that, It includes the following steps: Step 1: Based on the structural parameters of the tower crane and the three-dimensional model of its environment, plan the inspection route and generate a preset trajectory planning scheme based on the starting point; Step 2: Initialize the state of the tower crane; Step 3: The UAV flies to the preset position to collect starting point data; Step 4: According to the preset trajectory planning scheme generated in Step 1 and the starting point data collected in Step 3, generate the inspection path; Step 5: The UAV inspects the key points on the tower crane according to the inspection path and obtains the key point videos; Step 6: Perform defect identification based on the key point videos; Step 7: Conduct a safety assessment of the tower crane according to the defect identification results.

2. The automatic inspection method of a tower crane based on a drone according to claim 1, characterized in that, The inspection route planning in Step 1, that is, using the improved A* optimization method, generates a preset trajectory based on the starting point according to the structural parameters of the tower crane and the three-dimensional model of its environment, specifically including: Step 1-1: Construct a three-dimensional configuration space encoded by an octree, specifically as follows: According to the height of the tower crane, the length of the boom, the length of the counterweight arm, and the three-dimensional model of environmental obstacles, the space is defined as multiple cubic regions, and the side length of the cube is set as the neighborhood step size of the UAV; Simulate the hierarchical space division of the octree, and use the 26-neighborhood connectivity model to express the feasible motion primitives of the UAV, that is, the 26 motion directions of the UAV include: 6 orthogonal motion directions, 12 diagonal edge motion directions, and 8 body diagonal motion directions; The neighborhood step size of the UAV represents the minimum unit of the UAV's single-step motion; Step 1-2: Set the starting point and key point data in the three-dimensional configuration space, specifically as follows: The starting point data includes: the longitude and latitude of the starting point, the height relative to the ground, and the yaw angle of the UAV; set the coordinate system as follows: let the direction of the boom be the positive direction of the y-axis, the origin be the ground point at the center of the tower crane body, the three-dimensional coordinates of the starting point be (a, b, c), the yaw angle be d°, and the UAV camera faces one side of the tower body; The key point data includes: The key point data of the jacking mechanism, in addition to the starting point data, also includes the height h1 of the jacking mechanism from the ground; the key point data of the jacking mechanism is (a, b, c + h1); The key point data of the counterweight arm, in addition to the starting point data, also includes the height h2 of the counterweight arm from the ground, and the length e1 of the counterweight arm is increased in the direction of the yaw angle +90° relative to the starting point; the key point data of the counterweight arm is (a, b - e1, c + h2); The key point data of the top of the boom, relative to the key point data of the counterweight arm, the total length e2 of the boom and the counterweight arm is increased in the direction of the yaw angle -90° relative to the starting point; the key point data of the top of the boom is (a, b - e1 + e2, c + h2); The key point data of the tower top, in addition to the starting point data, also includes the height h3 of the tower top from the ground; the key point data of the tower top is (a, b, c + h3); The above key point data is also key point data after being mirrored with respect to the z-axis height reference plane in the three-dimensional coordinate system; Steps 1-3: Starting from the starting point, taking any key point as the target, search using a bidirectional search strategy until all key points are traversed to obtain a preset flight path.

3. The automatic inspection method for tower cranes based on unmanned aerial vehicles according to claim 2, wherein, The search using the bidirectional search strategy described in Step 1-3 includes: Step 1-3-1: Take the current flight path point and the next key point as the initial flight path points respectively. Step 1-3-2: Start the search based on the initial flight path points, generate candidate flight point sets respectively, and select the one with the smallest evaluation function f(n) from the candidate flight points as the next flight path point of the initial flight path point. Step 1-3-3: Judge the distance between the next flight path points of the two initial flight path points. If it is less than twice the neighborhood step size, execute Step 1-3-4 for path connectivity judgment, otherwise execute Step 1-3-5. Step 1-3-4: If the path between the initial flight path point and the next flight path point is connected, execute Step 1-3-5, otherwise discard the next flight path point. Step 1-3-5: Put the next flight path point into the preset flight path, and return to Step 1-3-1, taking this flight path point as the new current flight path point.

4. The automatic inspection method for tower cranes based on unmanned aerial vehicles according to claim 3, characterized in that, The evaluation function f(n) described in Step 1-3-2 is specifically as follows: f(n) = wh(n) + g(n) where g(n) is the cost function, representing the actual cost from the current flight path point to the next flight path point n, h(n) is the heuristic function h(n), providing directional guidance from the current flight path point to the next flight path point, and w is the weight factor.

5. The automatic inspection method for tower cranes based on drones according to claim 4, characterized in that, The heuristic function h(n) described in Step 1-3-2 is specifically as follows: h(n) = max(dx, dy, dz) + 0.414 * min(dx, dy, dz) where dx, dy, dz are the absolute values of the coordinate differences in three-dimensional space between the current flight path point where the UAV is located and the next flight path point, max() is the maximum value, and min() is the minimum value.

6. The automatic inspection method for tower cranes based on unmanned aerial vehicles according to claim 5, characterized in that, The cost function g(n) described in Step 1-3-2 is specifically as follows: Among them, d k is the basic moving distance of the drone at the k-th step, and α(θ k ) is the wind direction compensation coefficient, which is determined by the angle θ k between the moving direction of the drone and the wind speed direction. β(v ⊥ , θ k ) is the crosswind stability compensation term, which is related to the crosswind component v ⊥ and the angle θ k .

7. The automatic inspection method of a tower crane based on a drone according to claim 6, characterized in that, The wind direction compensation coefficient α(θ k ) described in Step 1-3-2 is as follows: The crosswind stability compensation term β(v ⊥ , θ k ) is as follows: where λ is the UAV wind resistance coefficient.

8. The automatic inspection method of a tower crane based on an unmanned aerial vehicle according to claim 1, characterized in that, The initialization of the state of the tower crane described in Step 2 includes: Perform calibration of the tower crane attitude space coordinate system, drive the tower crane slewing mechanism, and make the axis of the boom form an orthogonal geometric relationship with the reference plane of the tower body introduction section.

9. The automatic inspection method for tower cranes based on unmanned aerial vehicles according to claim 1, characterized in that The defect identification described in Step 6 includes: Adopt a pre-trained YOLOV8 recognition model to identify the corrosion of the tower crane body, the corrosion of the wire rope, the looseness of the wire rope, and the broken wires of the wire rope.

10. An automatic inspection system for tower cranes based on drones, characterized in that, Using the automatic inspection method described in any one of Claims 1-9 includes: More than 1 UAV, according to the flight path planned in the automatic inspection method, inspect the key points on the tower crane; The UAV is connected to the cloud platform for data interaction; The cloud platform performs data analysis and processing, is connected to the mobile monitoring terminal, and sends the processed data to the mobile monitoring terminal.

Images