Power transmission tower anti-loose bolt loosening diagnosis method based on three-dimensional point cloud data and edge detection

Abstract

The application discloses a power transmission tower anti-loose bolt loosening diagnosis method based on three-dimensional point cloud data and edge detection, and belongs to the technical field of power transmission tower bolt connection node detection. Three-dimensional point cloud models of bolt connection nodes in initial and to-be-detected states are acquired, and bolt axes are adjusted to be parallel to preset coordinate axes; node plate images are obtained by projection in the axial direction, a target detection model is used to identify bolt positions, pixel coordinates are mapped to three-dimensional space to realize single bolt point cloud segmentation; single bolt point cloud is twice segmented, nut and screw head point clouds are extracted respectively and projected in the axial direction, and hexagonal projection images are obtained; edge detection and straight line detection are used to identify nut and screw head edge straight lines and included angles thereof; position angles of the nut and the screw head in the initial and to-be-detected states are calculated respectively, and a bolt loosening angle is determined through a difference value of rotation angles of the two. The application can overcome the influence of light and shielding, and effectively solve the problem of large loosening angle identification failure.

Description

Technical Field

[0001] This invention belongs to the field of detection technology for bolted connection nodes of transmission towers, specifically relating to a method for diagnosing the loosening of anti-loosening bolts in transmission towers based on three-dimensional point cloud data and edge detection. It is particularly suitable for quantitatively identifying the loosening angle of anti-loosening bolts in bolted connection nodes of transmission towers using three-dimensional point cloud models, point cloud processing, and edge detection. Background Technology

[0002] Electricity, as an indispensable energy source in modern society, relies on the safe operation of transmission lines for its stable supply. Transmission towers, as key infrastructure supporting these lines, directly impact the stable operation of the power grid. Bolted connections, one of the most common connection methods in transmission towers, are crucial for ensuring the smooth operation of transmission lines. During service, the local bolted connections of transmission towers are exposed to complex natural environments, subjected to the combined effects of wind loads, temperature differences, material aging, and dynamic loads such as pulsating winds. This causes the bolt preload to gradually decrease and even loosen, reducing the overall load-bearing capacity of the transmission tower structure. Therefore, timely and accurate diagnosis of bolt loosening in transmission tower bolted connections is of great significance for ensuring the safe operation of transmission lines.

[0003] Currently, the methods for detecting loose bolts on transmission towers can be mainly divided into the following categories.

[0004] The first category is manual inspection. Traditional bolt loosening detection mainly relies on regular manual inspections, with workers checking by visual observation or using simple tools. However, as transmission towers are tall structures, manual inspection suffers from problems such as low efficiency, high risk of falls from heights, and strong subjectivity. In addition, the number of node bolts on transmission towers is extremely large, and the number of node bolts on the entire transmission line is countless, making manual inspection time-consuming and labor-intensive, and difficult to achieve comprehensive and timely coverage.

[0005] The second category is detection methods based on smart sensors. This involves monitoring by installing hardware devices such as piezoelectric sensors, strain gauges, and vibration sensors on bolts or structures. For example, piezoelectric sensing elements can be installed on bolts to monitor changes in bolt tension, or fiber optic sensors can be deployed to detect bolt loosening. While this type of method can achieve a certain degree of automated monitoring, it has the following drawbacks: it requires installing a large number of sensors on the component being monitored, resulting in high hardware costs; sensor deployment and maintenance are complex; and it is easily affected by factors such as ambient temperature and electromagnetic interference. Furthermore, for existing transmission towers already in operation, the construction difficulty and cost of large-scale sensor installation are both high.

[0006] The third category comprises computer vision-based detection methods. With the rapid development of image processing technology and deep learning algorithms, computer vision-based bolt loosening detection methods have received widespread attention. These methods utilize drones or inspection robots to acquire bolt images and then employ image processing techniques and deep learning algorithms to detect bolt loosening. Specifically, existing vision-based bolt loosening detection methods mainly include the following technical approaches: (1) Bolt recognition and angle detection based on two-dimensional images: First, bolts are identified and bolt image regions are cropped based on deep learning; then, the angle between the nut edge line and the horizontal line is identified using Hough transform and Canny edge detection operator; finally, the bolt loosening angle is obtained by comparing and analyzing with the initial bolt state. This type of method has low cost and can quantitatively identify the bolt loosening angle, and has good development prospects. However, bolt recognition and loosening detection methods based on two-dimensional image processing also have obvious limitations: the acquired image is greatly affected by the lighting conditions, and the detection accuracy drops significantly under conditions of insufficient light, reflection or shadow; under large tilt angle shooting conditions, there is mutual occlusion between the screw and nut, which affects the complete extraction of the bolt edge; some methods require pre-pasting calibration icons on the bolt nodes to eliminate the influence of the shooting angle, which increases the complexity of on-site operation; for scenes where the bolt image texture is not obvious, the effect of bolt recognition based solely on two-dimensional RGB images is not ideal.

[0007] (2) Bolt loosening detection method based on 3D point cloud: With the development of 3D laser scanning technology, some studies have begun to use 3D point cloud data for bolt loosening detection. 3D point cloud data contains the spatial coordinate information of the object, can provide depth information that 2D images do not have, and is less affected by changes in illumination. Existing detection methods based on 3D point clouds mainly determine whether the bolt is loose by extracting the exposed length of the bolt or calculating the distance between the top surface of the bolt and the reference surface. However, when the existing 3D point cloud detection methods are applied to the identification of the loosening angle of anti-loosening bolts, there is a technical problem that has not been effectively solved: the existing formula for calculating the loosening angle of anti-loosening bolts is not applicable in the case of large loosening angles (such as exceeding 60°). Specifically, when the rotation angle of the nut of the anti-loosening bolt is large, due to the symmetry of the hexagonal structure of the nut (the same geometric shape is presented every 60°), the traditional method based on the identification of a single edge angle will produce angle ambiguity, resulting in a serious discrepancy between the calculated loosening angle and the actual loosening angle, causing misjudgment. This problem is particularly prominent in the actual operation and maintenance inspection of anti-loosening bolts, because bolt loosening is often a gradual process, and the appearance of a large loosening angle is an important early warning signal before bolt failure.

[0008] In summary, existing technologies for detecting loosening of anti-loosening bolts on transmission towers still have the following shortcomings: manual inspection is inefficient and unsafe; sensor methods are costly and difficult to implement; two-dimensional vision methods are greatly affected by lighting and occlusion, and are prone to failure at large angles; although existing three-dimensional point cloud methods can partially overcome the influence of lighting, they lack effective means to identify the relative rotation angle between the anti-loosening bolt nut and the screw, especially unable to accurately identify the true loosening state under large loosening angle conditions. Summary of the Invention

[0009] In view of this, the purpose of this invention is to provide a method for diagnosing the loosening of anti-loosening bolts on transmission towers based on three-dimensional point cloud data and edge detection. This method aims to overcome the shortcomings of existing bolt loosening detection technologies, which are greatly affected by light and shading, and are prone to failure in identifying large loosening angles of anti-loosening bolts. It aims to achieve accurate quantitative identification of the relative rotation angle between the nut and the bolt.

[0010] To achieve the above objectives, the present invention provides the following technical solution: A method for diagnosing loosening of anti-loosening bolts on transmission towers based on 3D point cloud data and edge detection includes the following steps: Step 1: Establish a 3D point cloud model of the bolt connection node of the transmission tower in the initial state and the state to be tested; adjust the attitude of the 3D point cloud model obtained in the initial state and the state to be tested respectively, so that the bolt axis direction is parallel to the preset coordinate axis; Step 2: Based on the 3D point cloud models obtained in the initial state and the state to be detected, perform point cloud segmentation for each individual bolt, including: 21) Project the 3D point cloud model along the bolt axis to obtain the node plate projection image; 22) The planar pixel coordinates of each bolt in the node plate projection image are identified using the target detection model, and the planar pixel coordinates of each bolt are converted into actual relative planar positions by combining the ratio coefficient between the actual length of the hexagonal side of the bolt and the pixel length. 23) Extract the three-dimensional coordinates of the center point of the hexagonal surface of any bolt in the three-dimensional point cloud model. Based on the difference between the planar coordinates in the three-dimensional coordinates and the actual relative planar position of any bolt, determine the planar coordinates of each bolt in the three-dimensional point cloud model. 24) Based on the planar coordinates of each bolt and the axial coordinates of any bolt, combined with the bolt dimensions, determine the three-dimensional spatial range of each bolt in the three-dimensional point cloud model, so as to segment out the point cloud of a single bolt; Step 3: For the segmented point cloud of a single bolt, cut out the point cloud of the bolt head region and the point cloud of the nut region respectively, and project them along the bolt axis to obtain the hexagonal projection image of the bolt head and the hexagonal projection image of the nut. Step 4: Perform edge detection and line detection on the hexagonal projection images of the screw head and the nut respectively, identify the key edge lines of the hexagonal structure of the screw head and the nut, and calculate the angle between each key edge line and the horizontal line. Step 5: Based on the angle between the key edge line of the screw head and nut and the horizontal line in the initial state and the state to be tested, calculate the rotation angle of the screw head and nut respectively, and determine the loosening angle of the bolt based on the difference in rotation angle between the screw head and nut.

[0011] Furthermore, in step one, a three-dimensional point cloud model of the transmission tower bolt connection node in the initial state and the state to be detected is established by using UAV oblique photogrammetry 3D modeling technology or LiDAR technology; the preset coordinate axis is any axis in the XYZ coordinate axis.

[0012] Furthermore, in step 21), the node plate projection image is a two-dimensional image obtained by projecting along the screw axis.

[0013] Furthermore, in step 22), the target detection model is a bolt target detection model based on the YOLO algorithm.

[0014] Furthermore, in step 23), the method for determining the planar coordinates of each bolt in the 3D point cloud model is as follows: Extract the coordinates of the center point of the hexagonal surface of any bolt in the 3D point cloud model. The bolt is any bolt selected from the projection image of the node plate; Calculate the actual relative planar position coordinates of the bolt. With coordinate values The difference is obtained as follows:

[0015] The actual relative planar position coordinates of each bolt Add each The planar coordinates of each bolt in the three-dimensional point cloud model are obtained. ; , This represents the total number of bolts in the projection image of the gusset plate.

[0016] Furthermore, in step three, the point cloud of the screw head region and the point cloud of the nut region are obtained by cutting along the screw axis direction from the point cloud of the single bolt according to the bolt size within a preset length range; the projection direction is from the exposed length side of the screw towards the screw head direction.

[0017] Furthermore, in step four, the Canny edge detection algorithm is used for edge detection, and the Hough line detection algorithm is used for line detection.

[0018] Furthermore, in step five, the calculation method for the rotation angle of the screw head and the nut is as follows:

[0019]

[0020] in: The initial position angle of the nut or screw head is determined by the angle between the straight lines of each key edge and the horizontal line in the initial state. The calculation yielded:

[0021] in: The rotation angle of the nut or screw head under the condition to be tested; The corrected version of the first [item] under the state to be detected The angle between the key edge line and the horizontal line; For the state to be detected, the first The angle between the key edge line and the horizontal line; This represents the number of key edge lines that have been identified. This is for the remainder operation.

[0022] Furthermore, in step five, the loosening angle of the bolt is:

[0023] in: This represents the loosening angle of the bolt; This represents the rotation angle of the nut. This represents the rotation angle of the screw head; To perform absolute value operations.

[0024] The beneficial effects of this invention are as follows: This invention provides a diagnostic method for loosening anti-loosening bolts on transmission towers based on 3D point cloud data and edge detection, achieving the following technical effects: (1) Strong environmental adaptability and perspective distortion elimination: Three-dimensional point cloud data is used to replace traditional two-dimensional images, which fundamentally avoids the impact of factors such as lighting changes, shadows, and reflections on detection accuracy; the perspective distortion problem caused by the shooting angle can be eliminated without pre-pasting calibration icons on the bolt nodes, which significantly reduces the complexity of on-site operation; (2) Precise automatic segmentation of single bolt point cloud: The node plate image is obtained by projecting along the bolt axis, the position of each bolt is located by using the target detection model, and the two-dimensional coordinates are mapped to three-dimensional space by combining the ratio coefficient of the actual length of the hexagonal side length and the pixel length. This realizes the automatic and accurate positioning and segmentation of single bolt point cloud in the three-dimensional point cloud model without manual intervention. (3) Independent angle recognition of nut and screw: The point cloud of a single bolt is divided into two parts, and the point clouds of the nut area and the screw head area are extracted separately. The independent hexagonal projection images are obtained by projecting them along the axial direction. The hexagonal structure edge line angles of the two are identified by edge detection and line detection respectively. The strategy of separate detection makes the rotation state of the nut and screw independently quantified, overcoming the limitation of existing methods that detect the bolt as a single whole. (4) Accurate identification of large loosening angle: The loosening angle of the anti-loosening bolt is calculated by the difference between the rotation angle of the nut and the rotation angle of the screw, which effectively avoids the problem of misjudgment of large loosening angle caused by the 60° periodic symmetry of the hexagonal nut, and realizes accurate quantitative diagnosis of the full range of 0°~360°.

[0025] (5) Non-contact automated inspection: The entire inspection process does not require contact with the bolts. Data collection can be completed by drones or inspection robots equipped with 3D scanning equipment. It is suitable for high-altitude and high-risk transmission tower operation and maintenance scenarios, improving inspection efficiency and safety. Attached Figure Description

[0026] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart of the method for diagnosing loose bolts on transmission towers based on three-dimensional point cloud data and edge detection, according to the present invention. Figure 2 This is a planar projection example of the point cloud model of a bolt connection node; Figure 3 Example images showing the point clouds of the bolt head region and nut region cut out from the bolt point cloud model; (a) the cut-out bolt head region and nut region; (b) the hexagonal projection image of the nut; (c) the hexagonal projection image of the bolt head. Figure 4 Example diagram for correct identification of the rotation angle of a single component; Figure 5 Example diagram for correctly identifying the relative rotation angle of bolts to prevent loosening. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0028] I. Diagnostic methods for loose bolts on transmission towers.

[0029] This invention aims to propose a diagnostic method for loosening anti-loosening bolts on transmission towers based on 3D point cloud data and edge detection. The method utilizes 3D scanning technology to construct 3D point cloud data of the bolt connection nodes. Then, a target detection algorithm is used to locate and cut the bolts in the projected image of the nodes. Finally, the point cloud data of the nut and bolt head are combined with the projection along the bolt shank axis. This method avoids the problem of existing image methods being greatly affected by lighting conditions and overcomes the deficiency of existing anti-loosening bolt loosening angle identification methods in identifying larger loosening angles, achieving accurate quantitative diagnosis of the loosening angle of anti-loosening bolts.

[0030] Specifically, such as Figure 1 As shown, the present invention provides a method for diagnosing loose bolts on transmission towers based on three-dimensional point cloud data and edge detection, comprising the following steps.

[0031] Step 1: Establish 3D point cloud models of the transmission tower bolt connection nodes in both the initial and test states. Adjust the orientation of the 3D point cloud models acquired in both states to ensure the bolt axis is parallel to the preset coordinate axes.

[0032] In this embodiment, UAV oblique photogrammetry 3D modeling technology or LiDAR technology are used to establish 3D point cloud models of the transmission tower bolt connection nodes in the initial state and the state to be detected, respectively; the preset coordinate axis is any axis in the XYZ coordinate axis.

[0033] (1) Establish a three-dimensional point cloud model in the initial state.

[0034] Obtain a 3D point cloud model of the bolted connection nodes of the transmission tower in their initial state (i.e., without loosening). This can be achieved using UAV oblique photogrammetry 3D modeling technology or LiDAR technology. 3D point cloud modeling technology is a mature technology in this field and will not be elaborated upon further here.

[0035] The initial 3D point cloud model of the bolt connection node is flipped and adjusted so that the bolt axis is parallel to one of the preset coordinate axes (e.g., the Z-axis) of the XYZ coordinate system. This coordinate system alignment operation provides a unified spatial reference for subsequent projection, segmentation, and measurement steps.

[0036] (2) Establish a three-dimensional point cloud model under the state to be detected.

[0037] A 3D point cloud model of the bolt connection node of the same transmission tower under the condition to be inspected is obtained, and the modeling method used is completely consistent with the 3D point cloud model under the initial condition. Similarly, the attitude of the 3D point cloud model under the condition to be inspected is adjusted so that the bolt axis direction is parallel to the same preset coordinate axis (e.g., Z-axis) as in the 3D point cloud model under the initial condition.

[0038] Step 2: Single Bolt Point Cloud Segmentation. Based on the 3D point cloud models acquired in the initial state and the state to be detected, single bolt point cloud segmentation is performed. In this embodiment, the 3D point cloud model is projected along the bolt axis (i.e., the preset coordinate axis direction, taking the Z-axis as an example; the same applies to other parallel axes) to obtain the node plate projection image, such as... Figure 2 As shown.

[0039] Specifically, the segmentation of the single bolt point cloud includes the following sub-steps.

[0040] 21) Project the three-dimensional point cloud model along the bolt axis to obtain the node plate projection image. In this embodiment, the node plate projection image is a two-dimensional image obtained by projecting along the bolt axis.

[0041] 22) The planar pixel coordinates of each bolt in the node plate projection image are identified using the target detection model, and the planar pixel coordinates of each bolt are converted into their actual relative planar positions by combining the ratio coefficient between the actual length of the bolt's hexagonal side and the pixel length. In this embodiment, the target detection model is a bolt target detection model based on the YOLO algorithm.

[0042] Specifically, based on the YOLO object detection model trained on the bolt images, the projection image of the node plate is processed to identify the position of each bolt in the projection image and record the planar pixel coordinates of each bolt. ,in , This represents the total number of bolts in the image.

[0043] Based on the ratio coefficient between the actual hexagonal side length and the pixel length of the bolt, the pixel coordinate value of each bolt is multiplied by this ratio coefficient to obtain the actual relative planar position of each bolt. ( , (This represents the total number of bolts in the image).

[0044] 23) Extract the three-dimensional coordinates of the center point of the hexagonal surface of any bolt in the three-dimensional point cloud model. Based on the difference between the planar coordinates in the three-dimensional coordinates and the actual relative planar position of any bolt, determine the planar coordinates of each bolt in the three-dimensional point cloud model.

[0045] In this embodiment, the method for determining the planar coordinate values ​​of each bolt in the three-dimensional point cloud model is as follows.

[0046] Extract the coordinates of the center point of the hexagonal surface of any bolt in the 3D point cloud model. Specifically, the bolt is any bolt selected from the projection image of the node plate.

[0047] Calculate the actual relative planar position coordinates of the bolt. With coordinate values The difference is obtained as follows:

[0048] The actual relative planar position coordinates of each bolt Add each The planar coordinates of each bolt in the three-dimensional point cloud model are obtained. ; , This represents the total number of bolts in the projection image of the gusset plate.

[0049] 24) Based on the planar coordinates of each bolt and the axial direction coordinates of any bolt, combined with the bolt dimensions, determine the three-dimensional spatial range of each bolt in the three-dimensional point cloud model, so as to segment out the point cloud of a single bolt.

[0050] Since the bolt axis of the transmission tower bolt connection node is parallel to the Z-axis, the hexagonal surfaces of each bolt are basically at the same Z-coordinate value. Therefore, the three-dimensional coordinate values ​​of the hexagonal surfaces of each bolt in the three-dimensional point cloud model can be expressed as follows: Based on the identified three-dimensional coordinates of the hexagonal surfaces of each bolt, combined with the bolt's dimensional parameters, the bolt point cloud range is selected, and each bolt is automatically segmented.

[0051] Step 3: Point Cloud Segmentation and Projection of Nut and Bolt Heads. For the segmented point cloud of a single bolt, the point clouds of the bolt head region and the nut region are separately cut out, and then projected along the bolt axis to obtain hexagonal projection images of the bolt head and nut. For example... Figure 3 As shown, in this embodiment, the point cloud of the screw head region and the point cloud of the nut region are obtained by cutting along the screw axis direction and within a single bolt point cloud according to the bolt size within a preset length range; the projection direction is from the exposed length side of the screw towards the screw head direction.

[0052] For the single bolt point cloud model segmented in step two, a second segmentation is performed to cut out the point cloud of the bolt head region (i.e., the nut region) and the point cloud of the nut region.

[0053] After cutting, the cut screw head area and nut area are projected along the bolt axis, with the projection direction from the exposed length of the screw towards the screw head (i.e., along the negative or positive Z-axis, depending on the coordinate system definition). The projection yields hexagonal projection images of the screw head (nut) and the nut, as shown below. Figure 3 As shown.

[0054] It should be noted that in this embodiment, "screw head" and "nut" are different names for the same component, both referring to the hexagonal head structure at the end of the bolt.

[0055] Step 4: Edge and line detection. Edge detection and line detection are performed on the hexagonal projection images of the screw head and the nut respectively to identify the key edge lines of the hexagonal structure of the screw head and the nut, and the angle between each key edge line and the horizontal line is calculated.

[0056] Edge detection and line detection are performed on the hexagonal projection images of the screw head (nut) and the nut, respectively. Specifically: The Canny edge detection algorithm is used to identify the hexagonal edges of the nut and screw head.

[0057] The Hough line detection algorithm is used to identify the key edge lines of the hexagonal structure of the nut and screw head, and the angle between each key edge line and the horizontal line is calculated.

[0058] Both the Canny edge detection algorithm and the Hough line detection algorithm are mature technologies in this field, and their specific implementation methods will not be elaborated here.

[0059] Step 5: Identify the loosening angle of the anti-loosening bolt. Based on the angles between the key edge lines of the screw head and nut and the horizontal line in the initial state and the state to be tested, calculate the rotation angles of the screw head and nut respectively. Determine the loosening angle of the bolt based on the difference between the rotation angles of the screw head and nut, and output the loosening angle of the bolt. Complete the loosening diagnosis.

[0060] In this embodiment, the calculation method for the rotation angle of the screw head and the nut is as follows.

[0061] In the initial state (no loosening), the position angle of the hexagonal head of the nut or screw can be expressed as:

[0062] Among them: Among them: This represents the initial position angle of the nut or screw head. For the initial state of the nut or screw hexagonal head The angle between the straight edge line and the horizontal line; The number of straight lines on the identified edge of the hexagonal head of the nut or screw; This is the remainder operator (i.e., calculating the remainder when the dividend is divided by the divisor).

[0063] Calculate the rotation angle of the nut or screw under the condition to be tested:

[0064] in: The rotation angle of the nut or screw head under the condition to be tested; The corrected version of the first [item] under the state to be detected The angle between the edge line and the horizontal line is determined as follows:

[0065] in: For the state to be detected, the first The angle between the key edge line and the horizontal line; This represents the number of key edge lines that have been identified. This is for the remainder operation.

[0066] Calculate the rotation angle of the nut under the test state respectively. Rotation angle of the screw head (nut) The loosening angle of the bolt is obtained as follows:

[0067] in: This represents the loosening angle of the bolt; This represents the rotation angle of the nut. This represents the rotation angle of the screw head; To perform absolute value operations.

[0068] II. Technical Principles and Effect Verification of the Method of the Invention

[0069] To more clearly illustrate the technical effectiveness of the method for diagnosing loose bolts on transmission towers proposed in this invention, the following specific examples will be used for verification.

[0070] Example 1: Correct identification of the rotation angle of a single component.

[0071] like Figure 4 As shown, assuming that in the initial state, the angles between the three edge lines of the nut and the horizontal line are respectively... , and The nut position angle is calculated using formula (1). for .

[0072] When the nut turns Afterwards, the angles between the three edge lines of the nut and the horizontal line become respectively... , and Calculate according to formulas (2-1) and (2-2): First, subtract the nut position angle under the baseline condition. The three corrected angles are as follows: , and Then respectively Taking the remainder and then averaging, we obtain the rotation angle of the nut as follows: This matches the actual situation (the nut did indeed turn). ).

[0073] Example 2: Correct identification of the relative rotation angle of the anti-loosening bolt.

[0074] like Figure 5 As shown, assuming the initial state, the angles between the three edge lines of the screw (nut) and the horizontal line are respectively... , and The screw position angle is calculated using formula (1). for The angles between the three edge lines of the nut and the horizontal line are respectively , and The nut position angle was calculated. for .

[0075] When the nut turns counterclockwise The screw rotates counterclockwise. When (i.e., both rotate in the same direction but with different amounts of rotation), the actual relative rotation angle of the anti-loosening bolt should be: .

[0076] If the existing method is used, that is, to calculate the position angles of the screw and nut in the initial state and the state to be tested separately, and then calculate the relative angle difference between the two, then the relative angle between the screw and nut in the initial state is: ( The relative angle between the screw and nut in the state to be tested is also... (Calculation process omitted) The conclusion is that the anti-loosening bolt did not loosen, which is obviously a misjudgment.

[0077] The method of this invention is as follows: For the nut: the initial angle is Rotate The angles of the last three edge lines are , , The rotation angle of the nut can be calculated using formulas (2-1) and (2-2). .

[0078] For the screw: the initial angle is Rotate The angles of the last three edge lines are , , The screw rotation angle can be calculated using formulas (2-1) and (2-2). .

[0079] The difference between the two is The relative rotation angle of the anti-loosening bolt is completely consistent with the actual rotation angle, which correctly identifies the loose state of the bolt.

[0080] The above verification results show that the anti-loosening bolt loosening angle identification formula proposed in this invention can effectively overcome the technical defects of traditional methods in identifying failures when the loosening angle is large or when the nut and screw rotate in the same direction by different amounts.

[0081] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for diagnosing loosening of anti-loosening bolts on transmission towers based on 3D point cloud data and edge detection, characterized in that: Includes the following steps: Step 1: Establish a 3D point cloud model of the bolt connection node of the transmission tower in the initial state and the state to be tested; adjust the attitude of the 3D point cloud model obtained in the initial state and the state to be tested respectively, so that the bolt axis direction is parallel to the preset coordinate axis; Step 2: Based on the 3D point cloud models obtained in the initial state and the state to be detected, perform point cloud segmentation for each individual bolt, including: 21) Project the 3D point cloud model along the bolt axis to obtain the node plate projection image; 22) The planar pixel coordinates of each bolt in the node plate projection image are identified using the target detection model, and the planar pixel coordinates of each bolt are converted into actual relative planar positions by combining the ratio coefficient between the actual length of the hexagonal side of the bolt and the pixel length. 23) Extract the three-dimensional coordinates of the center point of the hexagonal surface of any bolt in the three-dimensional point cloud model. Based on the difference between the planar coordinates in the three-dimensional coordinates and the actual relative planar position of any bolt, determine the planar coordinates of each bolt in the three-dimensional point cloud model. 24) Based on the planar coordinates of each bolt and the axial coordinates of any bolt, combined with the bolt dimensions, determine the three-dimensional spatial range of each bolt in the three-dimensional point cloud model, so as to segment out the point cloud of a single bolt; Step 3: For the segmented point cloud of a single bolt, cut out the point cloud of the bolt head region and the point cloud of the nut region respectively, and project them along the bolt axis to obtain the hexagonal projection image of the bolt head and the hexagonal projection image of the nut. Step 4: Perform edge detection and line detection on the hexagonal projection images of the screw head and the nut respectively, identify the key edge lines of the hexagonal structure of the screw head and the nut, and calculate the angle between each key edge line and the horizontal line. Step 5: Based on the angle between the key edge line of the screw head and nut and the horizontal line in the initial state and the state to be tested, calculate the rotation angle of the screw head and nut respectively, and determine the loosening angle of the bolt based on the difference in rotation angle between the screw head and nut.

2. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step one, UAV oblique photogrammetry 3D modeling technology or LiDAR technology are used to establish 3D point cloud models of the transmission tower bolt connection nodes in the initial state and the state to be tested, respectively; the preset coordinate axis is any axis in the XYZ coordinate axis.

3. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step 21), the node plate projection image is a two-dimensional image obtained by projecting along the screw axis.

4. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step 22), the target detection model is a bolt target detection model based on the YOLO algorithm.

5. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step 23), the method for determining the planar coordinates of each bolt in the 3D point cloud model is as follows: Extract the coordinates of the center point of the hexagonal surface of any bolt in the 3D point cloud model. The bolt is any bolt selected from the projection image of the node plate; Calculate the actual relative planar position coordinates of the bolt. With coordinate values The difference is obtained as follows: The actual relative planar position coordinates of each bolt Add each The planar coordinates of each bolt in the three-dimensional point cloud model are obtained. ; , This represents the total number of bolts in the projection image of the gusset plate.

6. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step three, the point cloud of the screw head region and the point cloud of the nut region are obtained by cutting along the screw axis direction and within the single bolt point cloud according to the bolt size within a preset length range; the projection direction is from the exposed length side of the screw towards the screw head direction.

7. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step four, the Canny edge detection algorithm is used for edge detection, and the Hough line detection algorithm is used for line detection.

8. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step five, the calculation method for the rotation angle of the screw head and the nut is as follows: in: The initial position angle of the nut or screw head is determined by the angle between the straight lines of each key edge and the horizontal line in the initial state. The calculation yielded: in: The rotation angle of the nut or screw head under the condition to be tested; The corrected version of the first [item] under the state to be detected The angle between the key edge line and the horizontal line; The first one under the state to be detected The angle between the key edge line and the horizontal line; This represents the number of key edge lines that have been identified. This is for the remainder operation.

9. The method for diagnosing loosening of transmission tower anti-loosening bolts based on three-dimensional point cloud data and edge detection according to claim 1, characterized in that: In step five, the bolt loosening angle is: in: This represents the loosening angle of the bolt. This represents the rotation angle of the nut. This represents the rotation angle of the screw head; To perform absolute value operations.

Images