A method for monitoring the rotation angle of a tower crane

Abstract

This invention provides a method for monitoring the rotation angle of a tower crane, comprising: using a drone equipped with a camera to collect image data of the tower crane's operating area; using a YOLOv8 model to identify the tower crane target in the collected raw images, filtering bounding boxes that meet the confidence requirements to obtain candidate tower crane region images; refining the tower crane region using an adaptive threshold segmentation algorithm to obtain a binary image of the tower crane; performing skeleton extraction and rotation segment detection on the binary image of the tower crane to obtain the skeleton of the boom and pivot, determining the rotation center and the centroid coordinates of the boom skeleton curve; and finally calculating the rotation angle. This invention uses drone vertical photography to achieve non-contact data acquisition, avoiding on-site installation interference. Through a dual refinement strategy of YOLO recognition and threshold segmentation, interference factors are effectively eliminated, ensuring accurate extraction of the tower crane region. Based on the core structural features of the tower crane, high-precision calculation of the rotation angle is achieved, adaptable to different construction scenarios and tower crane models.

Description

Technical Field

[0001] This invention belongs to the field of engineering monitoring and computer vision technology, and in particular relates to a method for monitoring the rotation angle of a tower crane. Background Technology

[0002] In construction, accurate monitoring of tower crane rotation angle is crucial for ensuring construction safety, avoiding collisions, and improving work efficiency. Traditional methods mainly rely on physical sensors (such as encoders and tilt sensors) installed at the slewing bearing. These methods suffer from problems such as complex installation, significant construction interference, high maintenance costs, susceptibility to environmental factors like vibration and dust leading to measurement drift, and poor compatibility with older or leased equipment.

[0003] Therefore, there is an urgent need for a method that can accurately segment the tower crane structure and calculate the rotation angle without the need for sensor installation, in order to overcome the problems of cumbersome installation, weak anti-interference and limited monitoring range of existing technologies, and achieve efficient and reliable tower crane rotation angle monitoring. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for monitoring the rotation angle of tower cranes. This method is an automated monitoring method based on the fusion of UAV vertical photography, YOLOv recognition, and threshold segmentation. It solves the technical pain points in existing tower crane rotation angle monitoring. Through modular design, it integrates key steps such as image acquisition, accurate recognition, segmentation and purification, and angle calculation. The synergistic effect of each step ensures monitoring accuracy and practicality.

[0005] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0006] A method for monitoring the rotation angle of a tower crane includes the following process:

[0007] Step 1: Use a drone equipped with a camera to collect images of the tower crane's operating area. At the same time, use the drone's built-in sensors to record equipment and environmental parameters, and use a calibration board to perform calibration.

[0008] Step 2: Use the YOLOv8 model to identify tower crane targets in the acquired raw images, and filter the bounding boxes that meet the confidence requirements to obtain the final tower crane candidate region image;

[0009] Step 3: Within the candidate tower crane region obtained in Step 2, a purified binary image of the tower crane is obtained through grayscale processing, adaptive threshold segmentation processing, and morphological processing.

[0010] Step 4: Perform skeleton recognition on the purified binary image of the tower crane to determine the rotation center, the rotation line segment representing the direction of the boom, and the centroid coordinates of the boom skeleton curve.

[0011] Step 5: Calculate the tower crane rotation angle based on the data obtained in Step 4.

[0012] Furthermore, in step 1, the UAV adopts a vertical photography mode to ensure that the shooting angle of the UAV is perpendicular to the tower crane arm; the equipment parameters include camera intrinsic parameters and UAV extrinsic parameters. The camera intrinsic parameters include focal length, pixel size, and principal point coordinates, while the UAV extrinsic parameters include flight altitude, latitude and longitude, and roll angle; the environmental parameters include light intensity and weather conditions; all parameters are stored in association with the image file.

[0013] Furthermore, in step 2, a tower crane vertical photography dataset containing different construction scenarios, lighting conditions, and tower crane models is first constructed. Then, the main area of ​​the tower crane is labeled, and the YOLOv8 model is used for training. Then, the original images collected by the UAV are input into the trained YOLOv8 model for recognition. The model outputs the coordinates of the tower crane bounding box, and then the bounding boxes with a confidence level of 0.85 are selected to obtain the final tower crane candidate area image.

[0014] Furthermore, the specific process of step 3 is as follows:

[0015] First, the candidate region image of the tower crane is converted into a grayscale image, and the optimal segmentation threshold T is calculated using the Otsu's method. Then, based on T, the grayscale image is converted into a binary image. Pixels with grayscale values ​​greater than T are identified as background, and pixels with grayscale values ​​less than or equal to T are identified as foreground. The foreground is marked as black with a pixel value of 0, and the background is marked as white with a pixel value of 1, achieving pixel-level separation between the tower crane and the background. Finally, the binary image is subjected to morphological processing of erosion followed by dilation to remove small noise points and holes, while filling in the broken parts within the tower crane region, resulting in a binary image of the tower crane. The erosion operation uses a 3×3 rectangular structuring element and iterates once, while the dilation operation uses a 5×5 rectangular structuring element and iterates once.

[0016] Furthermore, the specific process of step 4 is as follows:

[0017] First, the Zhang-Suen thinning algorithm is used to skeletonize the binary image of the tower crane, obtaining skeleton lines that reflect the structural form of the tower crane. Then, the pivot skeleton and the boom skeleton are identified from the skeleton lines, and their centroids are calculated. The pivot centroid is the rotation center. The rotation center is connected to the boom centroid, and the resulting line segment is used as the rotation line segment representing the direction of the boom. Finally, the coordinates of the rotation center (xo, yo) and the coordinates of the boom skeleton curve centroid (xa, ya) are recorded to provide key parameters for subsequent rotation angle calculation.

[0018] Furthermore, the specific process of step 5 is as follows:

[0019] In the image coordinate system established with the top left corner of the image as the origin, the positive x-axis pointing horizontally to the right, and the positive y-axis pointing vertically downwards, with the rotation center (xo, yo) as the vertex, calculate the angle between the boom direction line segment and the positive x-axis in the two different states of the tower crane. This angle is the rotation angle.

[0020] The included angle θ1 in state 1 is: θ1 = arctan2 ((ya-yo),(xa-xo));

[0021] The included angle θ2 in state 2 is: θ2 = arctan2 ((yb-yo),(xb-xo));

[0022] Where (xa, ya) and (xb, yb) correspond to the centroid coordinates of the crane boom frame curve in the two states, respectively. The rotation angle of the tower crane from state 1 to state 2 is the difference θ between the two included angles, θ = θ2 - θ1.

[0023] The present invention has the following beneficial effects:

[0024] Compared to traditional physical sensor monitoring methods, this invention optimizes the entire process from "vertical photography acquisition—YOLOv precise identification—threshold segmentation and purification—structural feature extraction—calibration and conversion," eliminating the need to disassemble tower crane components. It utilizes UAV vertical photography for non-contact data acquisition, avoiding on-site installation interference. Through a dual purification strategy of YOLO identification and threshold segmentation, it effectively eliminates interference factors such as background environment and lighting changes, maintaining stable identification even in complex construction scenarios. This ensures accurate extraction of the tower crane area and strong anti-interference capabilities. Based on the core structural features of the tower crane, it achieves high-precision calculation of rotation angles, adapting to different construction scenarios and tower crane models, providing technical support for construction safety monitoring. Furthermore, the UAV can flexibly adjust the monitoring position, covering the entire operating range of the tower crane and adapting to different tower crane models and complex construction sites. The automated monitoring process significantly reduces manual intervention and improves the efficiency of construction safety management. Attached Figure Description

[0025] Figure 1 This is a flowchart of the tower crane rotation angle monitoring method described in this invention;

[0026] Figure 2 A schematic diagram of the candidate region image finally identified by the YOLOv8 model;

[0027] Figure 3 This is a schematic diagram of the binary image of the tower crane after processing by the adaptive threshold segmentation algorithm.

[0028] Figure 4 A schematic diagram showing the results of tower crane frame extraction and rotating line segment detection;

[0029] Figure 5This is a schematic diagram for calculating the rotation angle.

[0030] In the diagram: 1-Spindle frame; 2-Lifting boom frame; 3-Slewing center; 4-Center of crane boom; 5-Rotation segment; Detailed Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0032] The method for monitoring the rotation angle of a tower crane as described in this invention is as follows: Figure 1 As shown, the process includes the following:

[0033] Step 1: Drone vertical photography and image acquisition;

[0034] Before the drone takes flight, OpenCV is used to calibrate the intrinsic parameters (focal length, pixel size, principal point coordinates) of the camera on board to ensure the accuracy of the imaging parameters. Then, the flight altitude is set according to the height of the tower crane (in this embodiment, the tower crane height is 80m and the flight altitude is set to 120m), and the flight path is planned to ensure that the drone can hover directly above the tower crane. The vertical photography mode is used to acquire images, that is, the camera optical axis is perpendicular to the ground (pitch angle 0°), and RGB color images are acquired to ensure that the tower crane occupies a reasonable proportion in the acquired images.

[0035] Once ready, the drone is launched to collect image data of the tower crane's operating area. The drone's built-in sensors simultaneously record equipment parameters such as flight altitude and latitude and longitude, as well as environmental parameters. All parameters are stored in association with the image files, and calibration is performed using a checkerboard-patterned square calibration plate to calculate the actual dimensions.

[0036] Step 2: Tower crane target identification based on YOLOv;

[0037] After the drone equipped with a camera collects images, they are uploaded to a computer for subsequent analysis and processing. First, a vertical photography dataset of tower cranes with several images (5000 in this example) in different scenes is constructed. The main area of ​​the tower crane is labeled using the LabelImg tool, and the YOLOv8 model is trained with 300 iterations.

[0038] Then, the original image acquired in step 1 is input into the trained YOLOv8 model for recognition. The model outputs the coordinates of the tower crane bounding box. Bounding boxes with a confidence level of 0.85 are then selected to obtain the final candidate image of the tower crane region (e.g., ...). Figure 2 (As shown), to ensure the accuracy of target identification.

[0039] Step 3: Candidate region threshold segmentation and tower crane purification;

[0040] Within the candidate regions identified by YOLOv8, the tower crane area is further refined using an adaptive threshold segmentation algorithm:

[0041] First, the candidate region image is converted to a grayscale image, and the optimal segmentation threshold T is calculated using the Otsu's algorithm (maximum inter-class variance method). Then, based on the threshold T, the grayscale image is converted to a binary image. Pixels with grayscale values ​​greater than T are identified as background, and pixels with grayscale values ​​less than or equal to T are identified as the foreground of the crane. The foreground of the crane is marked as black (pixel value 0), and the background is marked as white (pixel value 1), achieving pixel-level separation between the crane and the background. Finally, the binary image undergoes morphological processing of erosion followed by dilation to remove small noise points and holes, while filling in broken parts within the crane area, ultimately obtaining a complete and clean binary image of the crane (e.g., ...). Figure 3 As shown in the figure, the erosion operation uses a 3×3 rectangular structure element and iterates once, while the dilation operation uses a 5×5 rectangular structure element and iterates once.

[0042] Step 4: Extraction of key structural features of the tower crane;

[0043] Based on the purified binary image of the tower crane, we focused on extracting two key structural features: the center of rotation and the centroid of the boom.

[0044] First, the Zhang-Suen thinning algorithm is used to skeletonize the binary image of the tower crane, obtaining skeleton lines that reflect the structural form of the tower crane. Then, to determine the direction of the boom, the pivot and boom sections are identified from the skeleton (i.e., pivot skeleton 1 and boom skeleton 2 are identified separately), and their centroids are calculated. The pivot centroid is the rotation center 3. The rotation center 3 is connected to the boom centroid 4, and the resulting line segment is used as the rotation segment 5 representing the direction of the boom. The result is as follows: Figure 4 As shown; finally, record the coordinates of the slewing center (xo, yo) and the centroid coordinates of the crane boom frame curve (xa, ya) to provide key parameters for subsequent rotation angle calculations.

[0045] Step 5: Calculate the rotation angle;

[0046] In an image coordinate system established with the top left corner of the image as the origin, the positive x-axis pointing horizontally to the right, and the positive y-axis pointing vertically downwards, the angle between the boom direction line segment and the positive x-axis is calculated with the rotation center (xo, yo) as the vertex, under different states of the tower crane (i.e., the initial state and the state after rotation); as detailed below:

[0047] The formula for calculating the included angle θ1 in state 1 is: θ1 = arctan2((ya-yo),(xa-xo)), and the formula for calculating the included angle θ2 in state 2 is: θ2 = arctan2((yb-yo),(xb-xo)), where (xa, ya) and (xb, yb) correspond to the centroid coordinates of the crane boom frame curve in the two states, respectively. The rotation angle of the tower crane from state 1 to state 2 is the difference θ between the two included angles, θ = θ2 - θ1. The result is as follows: Figure 5 As shown.

[0048] The core of this invention lies in: using UAV vertical photography to achieve non-contact data acquisition, avoiding interference from on-site installation; ensuring accurate extraction of the tower crane area through dual purification of YOLO recognition and threshold segmentation, with strong anti-interference capabilities; and achieving high-precision calculation of rotation angle based on the core structural features of the tower crane, adapting to different construction scenarios and tower crane models, and providing technical support for building construction safety monitoring.

[0049] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for monitoring the rotation angle of a tower crane, characterized in that, The process includes the following: Step 1: Use a drone equipped with a camera to collect images of the tower crane's operating area. At the same time, use the drone's built-in sensors to record equipment and environmental parameters, and use a calibration board to perform calibration. Step 2: Use the YOLOv8 model to identify tower crane targets in the acquired raw images, and filter the bounding boxes that meet the confidence requirements to obtain the final tower crane candidate region image; Step 3: Within the candidate tower crane region obtained in Step 2, a purified binary image of the tower crane is obtained through grayscale processing, adaptive threshold segmentation processing, and morphological processing. Step 4: Perform skeleton recognition on the purified binary image of the tower crane to determine the rotation center (3), the rotation line segment (5) representing the direction of the boom, and the centroid coordinates of the boom skeleton curve; Step 5: Calculate the tower crane rotation angle based on the data obtained in Step 4.

2. The method for monitoring the rotation angle of a tower crane according to claim 1, characterized in that, In step 1, the drone adopts a vertical photography mode to ensure that the drone's shooting angle is perpendicular to the tower crane arm; the equipment parameters include camera intrinsic parameters and drone extrinsic parameters. The camera intrinsic parameters include focal length, pixel size, and principal point coordinates, while the drone extrinsic parameters include flight altitude, latitude and longitude, and roll angle; the environmental parameters include light intensity and weather conditions; all parameters are stored in association with the image file.

3. The method for monitoring the rotation angle of a tower crane according to claim 1, characterized in that, In step 2, a vertical photography dataset of tower cranes is first constructed, which includes different construction scenarios, lighting conditions, and tower crane models. Then, the main area of ​​the tower crane is labeled, and the YOLOv8 model is used for training. Then, the original images collected by the UAV are input into the trained YOLOv8 model for recognition. The model outputs the coordinates of the tower crane bounding box, and then the bounding boxes with a confidence level of 0.85 are selected to obtain the final candidate images of the tower crane area.

4. The method for monitoring the rotation angle of a tower crane according to claim 1, characterized in that, The specific process of step 3 is as follows: First, the candidate region image of the tower crane is converted into a grayscale image, and the optimal segmentation threshold T is calculated using the Otsu's method. Then, based on T, the grayscale image is converted into a binary image. Pixels with grayscale values ​​greater than T are identified as background, and pixels with grayscale values ​​less than or equal to T are identified as foreground. The foreground is marked as black with a pixel value of 0, and the background is marked as white with a pixel value of 1, achieving pixel-level separation between the tower crane and the background. Finally, the binary image is subjected to morphological processing of erosion followed by dilation to remove small noise points and holes, while filling in the broken parts within the tower crane region, resulting in a binary image of the tower crane. The erosion operation uses a 3×3 rectangular structuring element and iterates once, while the dilation operation uses a 5×5 rectangular structuring element and iterates once.

5. The method for monitoring the rotation angle of a tower crane according to claim 1, characterized in that, The specific process of step 4 is as follows: First, the Zhang-Suen thinning algorithm is used to skeletonize the binary image of the tower crane to obtain skeleton lines that reflect the structural form of the tower crane. Then, the pivot skeleton (1) and the boom skeleton (2) are identified from the skeleton lines and their centroids are calculated. The pivot centroid is the rotation center (3). The rotation center (3) is connected to the boom centroid (4), and the resulting line segment is used as the rotation line segment (5) representing the direction of the boom. Finally, the coordinates of the rotation center (xo, yo) and the coordinates of the boom skeleton curve centroid (xa, ya) are recorded to provide key parameters for subsequent rotation angle calculation.

6. The method for monitoring the rotation angle of a tower crane according to claim 5, characterized in that, The specific process of step 5 is as follows: In the image coordinate system established with the top left corner of the image as the origin, the positive x-axis pointing horizontally to the right, and the positive y-axis pointing vertically downwards, with the rotation center (xo, yo) as the vertex, calculate the angle between the boom direction line segment and the positive x-axis in the two different states of the tower crane. This angle is the rotation angle. The included angle θ1 in state 1 is: θ1 = arctan2 ((ya-yo),(xa-xo)); The included angle θ2 in state 2 is: θ2 = arctan2 ((yb-yo),(xb-xo)); Where (xa, ya) and (xb, yb) correspond to the centroid coordinates of the crane boom frame curve in the two states, respectively. The rotation angle of the tower crane from state 1 to state 2 is the difference θ between the two included angles, θ = θ2 - θ1.

Images