A tower crane cable-stayed inclined hoisting intelligent detection method, terminal, device and medium

Abstract

The application belongs to the technical field of image processing, and discloses a tower crane inclined pulling and inclined hoisting intelligent detection method, a terminal, equipment and a medium. The tower crane inclined pulling and inclined hoisting intelligent detection method comprises the following steps: acquiring a real-time image of a hook, and using a pre-constructed and trained yolov5 hook detection model to identify the hook based on the real-time image of the hook, and calculating the offset distance and the inclined pulling angle of the hook based on the pre-calibrated initial position of the hook to determine the inclined pulling of the hook. The application takes the tower crane hook as a marker to solve the problem of the influence of the installation position of the camera and the change of the magnification on the detection accuracy. The tower crane inclined pulling and inclined hoisting intelligent detection method of the application does not need to additionally increase markers and is more suitable for the working conditions of tower crane construction operations. The tower crane inclined pulling and inclined hoisting intelligent detection method of the application can be suitable for the detection of the tower crane inclined pulling and inclined hoisting under different working conditions, and has high detection accuracy.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and particularly relates to an intelligent detection, terminal, device and medium for tower crane cable-stayed cranes. Background Technology

[0002] Currently, tower cranes (hereinafter referred to as tower cranes) are mainly used for the vertical lifting and horizontal transportation of materials and the installation of building components in construction. To ensure safety, tower crane operating procedures explicitly prohibit "oblique pulling and lifting." However, in practice, for convenience, tower crane operators often engage in oblique pulling and lifting operations, leading to serious safety accidents such as tower crane overturning. Therefore, real-time monitoring of the tower crane's lifting status to prevent such violations is crucial. Since the steel ropes are in a swinging state during tower crane lifting, traditional methods for detecting oblique pulling and lifting of the tower crane hook rely on mechanical detection methods. For example, patent CN215558465U describes a tower crane anti-oblique pulling detection device, which designs a mechanical structure to detect oblique pulling and lifting of the hook. However, traditional methods are difficult to implement. With the development of artificial intelligence and machine vision, visual image recognition of whether the hook is being pulled at an angle shows great promise.

[0003] The prior art 1 (CN101428740A A method for preventing eccentric pull in trolley lifting operations) involves placing a video camera above the hook on the luffing trolley of the tower crane, adding a fixed marker pole below the camera, and then using image recognition to determine the angle of the hook pull.

[0004] However, the above method has two problems: (1) It is necessary to install a marker pole on the trolley. According to the working principle of the tower crane, the hook is always swinging during the hoisting operation. This method may cause the hook steel rope to wear and collide with the marker, resulting in safety accidents such as falling off; (2) The original calculation method requires that the center position of the marker pole, the center of the camera, and the initial position of the hook (i.e., the steel rope is vertically downward and the hook is not inclined) be consistent. However, in actual engineering, due to the limited installation position, it is difficult for the camera installation position to be consistent with the initial position of the hook. At the same time, if the center position of the marker pole is consistent with the initial position of the hook, it is easy to cause the steel rope to collide with the marker pole. Therefore, the existing technology cannot solve the engineering problem in actual work.

[0005] Based on the above analysis, the problems and defects of the existing technology are as follows: the existing detection of inclined cable and inclined suspension is relatively difficult, the detection accuracy is not high, additional marker poles need to be set up, and it cannot be applied to actual scenarios. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides an intelligent detection method, terminal, equipment, and medium for tower crane inclined cable pulling and lifting.

[0007] This invention is implemented as follows: an intelligent detection method for inclined tower cranes with cable-stayed and inclined lifting mechanisms, the method comprising:

[0008] Real-time images of the hook are acquired, and the hook is identified based on the pre-built and trained YOLOv5 hook detection model. At the same time, the hook offset distance and angle are calculated based on the pre-calibrated initial position of the hook to determine the hook's angled pull.

[0009] Furthermore, the intelligent detection method for tower crane cable-stayed cranes includes the following steps:

[0010] Step 1: Prepare a hook sample set by acquiring images of the hook, construct a YOLOv5 hook detection model, and train the YOLOv5 hook detection model based on the hook sample set; this model is used to identify tower crane hooks in video images.

[0011] Step two: By setting the actual hook size, an image of the hook's initial position is obtained and the hook is identified. Based on this, the installation offset distance of the hook's initial position is calculated, and the initial position of the hook is calibrated. The setting of markers in the image recognition is completed, which solves the problem of identifying the hook being pulled or suspended at an angle when the hook installation position is uncertain, and improves the applicability of the method of the present invention.

[0012] Step 3 involves acquiring a real-time image of the hook, identifying the hook, and calculating the ratio between the actual hook size and the pixel size at the current video magnification. This process establishes the offset distance between the current and initial positions of the hook in the real-time image, enabling the determination of hook oblique pull angle. This solves the problem of calculating the hook offset distance under varying tower crane magnification conditions and improves the accuracy of hook oblique pull detection.

[0013] Furthermore, step one, which involves preparing a hook sample set by acquiring images of the hook, constructing a YOLOv5 hook detection model, and training the YOLOv5 hook detection model based on the hook sample set, includes:

[0014] (1) Use a camera device installed above the hook at the bottom of the trolley of the tower crane to acquire images of the hook in real time;

[0015] (2) LabelImg is used to annotate the collected hook images to construct a hook sample set containing hook images in different states;

[0016] (3) Construct a YOLOv5 hook detection model; train the YOLOv5 hook detection model until the model converges, and set the trained network structure and parameters into the YOLOv5 target detection network to obtain the trained YOLOv5 hook detection model.

[0017] Furthermore, the YOLOv5 hook detection model includes:

[0018] The backbone feature extraction network is used to generate five feature layers for the hook using a deep residual network: 320*320*64, 160*160*128, 80*80*256, 40*40*512, and 20*20*1024.

[0019] A multi-scale feature fusion structure is used to fuse extracted network features using FPN+PAN to obtain an effective feature layer that fuses feature information at different scales.

[0020] The classifier and regressor are used to perform a convolution process on the obtained effective feature layers to obtain the YoloHead corresponding to the three effective feature layers respectively. CIOU_Loss is used as the loss function for the bounding box to perform target recognition on the image and regress the original grid coordinates of the features.

[0021] Furthermore, in step two, by setting the long side dimension of the actual hook, an image of the hook's initial position is obtained and the hook is identified. Based on this, the installation offset distance of the hook's initial position is calculated to complete the calibration of the hook's initial position, including:

[0022] First, the actual long side dimension of the hook is determined using the hook as a marker; when the hook is in its initial position, an image of the hook's initial position is obtained.

[0023] Secondly, the trained YOLOv5 hook detection model is used to identify the hook in the initial position image and regress the position of the hook in the image to obtain the vertices of the rectangular box for hook target recognition.

[0024] Then, based on the identified vertices of the hook rectangle, the pixel coordinates of the hook center in the image when the hook is in its initial position are obtained;

[0025] Finally, based on the vertices of the bounding box identified by the hook target and the actual long side size of the hook, the ratio between the actual size and the pixel size, as well as the actual offset between the center of the hook's initial position and the center of the camera device, are calculated:

[0026] ;

[0027] ;

[0028] ;

[0029] in, k 0 This indicates the ratio between the actual size and the pixel size; a This indicates the actual length of the hook; （x10 ， y 10 ） , （x 20 ，y 20 ） The vertices of the rectangle representing the target identification of the hook; w、h This indicates the initial position of the hook image (Im0 pixels in size). （O x0 ，O y0 ） This indicates the center of the hook when it is in its initial position. O Pixel coordinates in the image; x 0 and y 0 The center of the hook's initial position O Center of the camera A The actual offset.

[0030] Furthermore, the initial position of the hook is when the steel rope hangs naturally, and the hook is not in a position where it is pulled or lifted at an angle.

[0031] Furthermore, step three, which involves acquiring a real-time image of the hook, identifying the hook, and calculating the ratio between the actual hook size and pixel size at the current video magnification, the positional offset distance in the real-time image of the hook, the distance between the current position and the initial position of the hook, and the hook's angled pull angle to determine the hook's oblique pull, includes:

[0032] 1) Use video equipment to acquire real-time video of the hook during the tower crane's operation, and extract frames every 5 frames of the video image stream to obtain the current image of the hook;

[0033] 2) Use the trained hook detection model to identify the hook in the current image, regress the position of the hook in the current image, and obtain the vertices of the hook rectangle. （x 1 ，y 1 ） as well as （x 2 ，y 2 ） And calculate the center of the hook when the hook is in its current position. O pixel coordinates （O x ，O y ） :

[0034] ;

[0035] 3) Based on the vertices of the rectangle in the current image of the hook. （x 1 ，y 1 ） , （x 2 ，y 2 ） and the actual length of the hook a Calculate the ratio between the actual size and the pixel size at the current video magnification. k :

[0036] ;

[0037] 4) Calculate the position of the hook in the current image relative to the center of the camera. A actual offset distance x and y :

[0038] ;

[0039] 5) Based on the initial position of the hook and the center of the camera A offset distance x 0 、y 0 And the current position of the hook relative to the center of the camera. A offset distance x, y Calculate the distance between the current hook position and the initial position. s :

[0040] ;

[0041] 6) Calculate the angle of attack between the initial position and the current position of the hook. θ: θ = arcsin(s / L) ;in, L This indicates the current length of the hook wire rope;

[0042] 7) When the angle of the hook is diagonal θ≤θ 0 If the hook is not in a diagonal pulling state, it is determined that the hook is in a diagonal pulling state; otherwise, it is determined that the hook is in a diagonal pulling state and an alarm is triggered. θ 0 This indicates the threshold for determining the angled pull of the hook;

[0043] 8) Repeat steps 1) to 7) to monitor the status of the tower crane hook's oblique pulling and lifting in real time.

[0044] Another object of the present invention is to provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, the computer program being executed by the processor causing the processor to perform the steps of the intelligent detection method for tower crane inclined cable pulling and lifting.

[0045] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the intelligent detection method for tower crane inclined cable pulling and lifting.

[0046] Another objective of this invention is to provide an information data processing terminal for executing the intelligent detection method for tower crane inclined cable pulling and lifting.

[0047] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0048] First, addressing the technical problems existing in the prior art and the difficulty in solving them, this paper closely analyzes, in conjunction with the technical solution to be protected by this invention and the results and data obtained during the research and development process, how the technical solution of this invention solves the technical problems, and the inventive technical effects brought about by solving these problems. The specific description is as follows:

[0049] This invention utilizes the YOLOv5 target detection algorithm to achieve hook image detection, and solves the problem of the influence of camera installation position on the detection accuracy of tower crane hooks at angled pulls and lifts by calibrating the initial position of the hook.

[0050] Second, considering the technical solution as a whole or from a product perspective, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows:

[0051] This invention uses image recognition technology to detect the oblique pulling and lifting of tower crane hooks. The tower crane hook is used as a marker, eliminating the need for additional markers. Furthermore, the detection algorithm addresses the impact of camera zoom on the image. This invention is more suitable for tower crane construction operations.

[0052] Third, as supporting evidence of the inventiveness of this invention, it is also reflected in the following important aspects:

[0053] (1) The expected benefits and commercial value of the technical solution of this invention after transformation are as follows:

[0054] With social development and a people-centered approach, safe production has become a key focus for the entire society. Inadequate monitoring of tower crane inclined lifting operations and operator violations can lead to serious accidents such as tower crane overturning. This invention enables monitoring of tower crane lifting operations, effectively preventing violations and improving the safety of tower crane operations, demonstrating significant commercial value. The construction industry is a vital part of my country's national economy, with 500,000 tower cranes in operation nationwide. The technical solution of this invention, once transformed into a product, will generate substantial economic benefits.

[0055] (2) The technical solution of the present invention solves a technical problem that people have long wanted to solve but have never been able to solve successfully:

[0056] To improve the safety of tower crane operations, the detection of the angled pulling and lifting of tower crane hooks is a key concern. However, existing detection methods are difficult to implement, unsuitable for tower crane operating conditions, and have low detection accuracy. The method of this invention is easy to install and improves detection accuracy. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the intelligent detection method for tower crane inclined cable pulling and lifting provided in an embodiment of the present invention;

[0058] Figure 2 This is a flowchart of the intelligent detection method for tower crane inclined cable pulling and lifting provided in an embodiment of the present invention;

[0059] Figure 3 This is a schematic diagram of the YOLOv5 hook detection model provided in an embodiment of the present invention;

[0060] Figure 4 This is a flowchart of the initial position calibration of the hook provided in an embodiment of the present invention;

[0061] Figure 5 This is a pixel diagram of the initial position image of the hook provided in an embodiment of the present invention;

[0062] Figure 6 This is a flowchart of the YOLOv5 hook detection model construction and training process provided in this embodiment of the invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0064] I. Explanation and Description of Embodiments. To enable those skilled in the art to fully understand how the present invention is specifically implemented, this section provides an explanation and description of the embodiments that elaborate on the technical solutions.

[0065] like Figure 1 , Figure 2 As shown, the intelligent detection method for tower crane inclined cable-stayed cranes provided in this embodiment of the invention includes the following steps:

[0066] S101, prepare a hook sample set by acquiring images of the hook, construct a YOLOv5 hook detection model, and train the YOLOv5 hook detection model based on the hook sample set;

[0067] S102, the initial position of the hook is calibrated by setting the long side dimension of the actual hook, obtaining an image of the initial position of the hook, identifying the hook, and calculating the offset distance of the initial position of the hook;

[0068] S103, by acquiring a real-time image of the hook, performing hook recognition and calculating the ratio between the actual size of the hook and the pixel size at the current video magnification, the position offset distance in the real-time image of the hook, the distance between the current position and the initial position of the hook, and the hook angle, the hook is judged to be inclined.

[0069] The intelligent detection method for inclined tower crane pulling and lifting provided in this embodiment of the invention includes: hook detection model training, hook initial position calibration, and hook inclined pulling and lifting detection. The specific steps are as follows:

[0070] Step 1: Train the YOLOv5 hook detection model

[0071] Reference Figure 1 This is a flowchart of the training process for the hook detection model.

[0072] 1-1: Acquiring images of the hook

[0073] A video camera is installed above the hook at the bottom of the tower crane's trolley. The camera captures images of the hook in real time, and one hook image is extracted every 5 frames from the image stream.

[0074] 1-2: Creation of a Sample Set of Lifting Hooks

[0075] The labelImg software was used to annotate the collected hook images, and a sample set of 1606 tower crane hooks under different conditions was established.

[0076] 1-3: Building a YOLOv5 object detection model

[0077] Reference Figure 3 The YOLOv5 hook detection model is constructed by the YOLOv5 network, which consists of three parts: a backbone feature extraction network, a multi-scale feature fusion structure, and a classifier and regressor.

[0078] The backbone feature extraction network uses a deep residual network to generate five feature layers: 320*320*64, 160*160*128, 80*80*256, 40*40*512, and 20*20*1024. This network has high accuracy and extracts a lot of image data feature information, which is beneficial for target detection in complex scenes.

[0079] The multi-scale feature fusion structure (Neck) uses FPN+PAN to fuse network features. For example... Figure 3 Neck extracted three feature layers from the backbone feature extraction network: 80*80*256, 40*40*512, and 20*20*1024. The three feature layers were upsampled to achieve feature fusion, and then the feature layers were downsampled to achieve feature fusion. Finally, an effective feature layer that integrates feature information of different scales was obtained.

[0080] The classifier and regressor (Head model) perform a convolution operation on the three effective feature layers 20*20*1024, 40*40*512 and 80*80*256 respectively to obtain the corresponding YoloHead for the three feature layers. CIOU_Loss is used as the loss function for the bounding box to perform target recognition on the image and regress the original grid coordinates of the features.

[0081] 1-4: Model Training

[0082] The labeled hook image sample set is imported into the YOLOv5 object detection model for training until the model converges. The trained network structure and parameters are then set into the YOLOv5 object detection network to obtain the YOLOv5 hook detection model.

[0083] Step 2: Initial position calibration of the hook

[0084] The initial position of the hook is when the steel rope hangs naturally and there is no oblique pulling or lifting. Considering that the installation position of the hook camera and the center position of the hook may be offset to varying degrees under different working conditions, the initial position calibration of the hook determines the offset distance between the initial position of the hook and the center of the camera, which facilitates subsequent accurate calculation of the oblique pulling or lifting of the hook. (Refer to...) Figure 4 This is the procedure for calibrating the initial position of the hook in the method of the present invention.

[0085] 2-1: Set the actual length of the hook a

[0086] The method of this invention uses the hook as a marker. The image of the hook in the tower crane trolley video is set to the actual long side dimension of the hook. a .

[0087] 2-2: Obtaining the image of the initial position of the hook

[0088] When the hook is in its initial position, acquire the initial position image Im0 of the hook.

[0089] 2-3: Hook Identification

[0090] Using the hook detection model trained in step 1, the hook in the initial position image Im0 is detected, and the position of the hook in the image is regressed to obtain the vertices of the rectangular box for hook target recognition. （x 10 ，y 10 ） , （x 20 ，y 20 ） .

[0091] Figure 5 This is a pixel diagram of the initial position image Im0 of the hook. The pixel size of the image is [missing information]. w*h The top left corner is the pixel origin. （0,0） The rectangle in the image represents the location of the detected hook target. The location is determined based on the vertices of the identified hook rectangle. （x 10 ， y 10 ） , （x 20 ，y 20 ） This allows us to obtain the center of the hook when it is in its initial position. O Pixel coordinates in the image （O x0 ，O y0 ） :

[0092] .

[0093] 2-4: Calculation of the offset of the initial position of the hook

[0094] Based on the top of the hook in the image （x 10 ，y 10 ） , （x 20 ，y 20 ） and the actual length of the hook a Calculate the ratio between the actual size and the pixel size. k 0 :

[0095]

[0096] To handle magnification changes in the tower crane hook video camera, it is necessary to calculate the center of the hook's initial position. O Center of the camera A actual offset x 0 and y 0 (like Figure 5 ):

[0097] .

[0098] Step 3: Hook oblique pull and oblique lifting test

[0099] refer to Figure 6 The flowchart for testing the angled pull and lifting of a crane hook is as follows:

[0100] 3-1: Acquire real-time images of the hook

[0101] The crane uses a trolley camera to capture real-time video of the hook during tower crane operation. Frames are extracted every 5 frames in the image stream to obtain the current hook image (Im). 。

[0102] 3-2: Hook Identification

[0103] Using the hook detection model trained in step 1, identify the hook in the current image (Im), regress the hook's position in the current image, and obtain the vertices of the hook's bounding box. （x 1 ，y 1 ） as well as （x 2 ，y 2 ） Calculate the center of the hook when the hook is in its current position. O pixel coordinates （O x ，O y ） :

[0104] .

[0105] 3-3: Calculate the ratio between the actual size and the pixel size at the current video magnification. k

[0106] Based on the vertices of the rectangle in the current image of the hook （x 1 ，y1 ） , （x 2 ，y 2 ） and the actual length of the hook a Calculate the ratio between the actual size and the pixel size at the current video magnification. k：

[0107] .

[0108] 3-4: Calculate the position offset distance of the hook in the current image. x and y

[0109] Calculate the position of the hook in the current image relative to the center of the camera. A actual offset distance x and y :

[0110] .

[0111] 3-5: Calculate the distance between the current position and the initial position of the hook.

[0112] Based on the initial position of the hook and the center of the camera A offset distance x 0 , y 0 And the current position of the hook relative to the center of the camera. A offset distance x , y Calculate the distance between the current hook position and the initial position. s :

[0113] .

[0114] 3-6: Calculate the hook angle

[0115] Calculate the angle of attack between the initial position and the current position of the hook. θ: θ = arcsin(s / L)

[0116] in, L This is the current length of the hook wire rope;

[0117] 3-7: Hook oblique pull judgment alarm

[0118] If the hook's angle of attack θ≤θ 0 The system determines that the hook is not in a diagonal pulling state; otherwise, it determines that the hook is in a diagonal pulling state and issues an alarm. θ 0The threshold for judging the angled pull of the hook can be set to 5 degrees.

[0119] The cycle steps 3-1 to 3-7 can detect the state of the tower crane hook being pulled or lifted at an angle in real time.

[0120] II. Application Examples. To demonstrate the inventiveness and technical value of the present invention, this section provides application examples of the technical solution on specific products or related technologies.

[0121] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0122] III. Evidence of the Relevant Effects of the Embodiments. The embodiments of the present invention have achieved some positive effects during research and development or use, and indeed possess significant advantages compared to existing technologies. The following description, in conjunction with data, charts, and other materials from the experimental process, illustrates these advantages.

[0123] The intelligent detection method for tower crane cable-stayed and inclined lifting provided in this embodiment of the invention was applied to a construction site, and the specific data are as follows:

[0124] Long side of tower crane hook a =1.15 meters, current length of hook wire rope L=10 Meters; the pixel count of the hook camera is w*h= 1280*720 Using the trained hook detection model, the initial position of the tower crane hook is calibrated and calculated to obtain the initial center position of the hook. o Offset from the center of the camera x 0 =-0.35360962 , y 0 =-1.14775401 ;

[0125] Real-time acquisition of images of the tower crane hook during operation; using a hook recognition model to regress the vertices of the rectangle representing the hook's position. （598,529） as well as （815,553） ; Calculate the current proportional system as k=0.005299539 Get the current hook center o Offset from the center of the camera x=-0.35241935 , y=-0.94596774 ;

[0126] Calculate the distance between the current position and the initial position of the hook. s=0.19481129 Meters; the length of the hook steel rope is: L=10 Meters; the calculated angle of the hook pull is as follows: θ=1.12° Due to the calculated hook angle θ=1.12° Less than the threshold for judging the angled pull of the hook θ 0 =5° Therefore, it can be determined that the hook is currently in a non-oblique pulling state.

[0127] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A smart detection method for inclined tower cranes, characterized in that, The intelligent detection method for tower crane inclined pull and inclined suspension includes: Real-time images of the hook are acquired, and the hook is identified based on the pre-built and trained YOLOv5 hook detection model. At the same time, the hook offset distance and angle are calculated based on the pre-calibrated initial position of the hook to determine the hook angle. The intelligent detection method for tower crane inclined cable-stayed cranes includes the following steps: Step 1: Prepare a hook sample set by acquiring images of the hook, construct a YOLOv5 hook detection model, and train the YOLOv5 hook detection model based on the hook sample set; Step two: The initial position of the hook is calibrated by setting the actual long side dimension of the hook, acquiring an image of the hook's initial position, identifying the hook, and calculating the offset distance of the hook's initial position. Step 3: Obtain a real-time image of the hook, identify the hook, and calculate the ratio between the actual size of the hook and the pixel size at the current video magnification, the position offset distance in the real-time image of the hook, the distance between the current position and the initial position of the hook, and the hook angle to determine the hook's oblique pull. The YOLOv5 hook detection model includes: The backbone feature extraction network is used to generate five feature layers for the hook using a deep residual network: 320*320*64, 160*160*128, 80*80*256, 40*40*512, and 20*20*1024. A multi-scale feature fusion structure is used to fuse extracted network features using FPN+PAN to obtain an effective feature layer that fuses feature information at different scales. The classifier and regressor are used to perform a convolution process on the obtained effective feature layers to obtain the YoloHead corresponding to the three effective feature layers respectively. CIOU_Loss is used as the loss function for the bounding box to perform target recognition on the image and regress the original grid coordinates of the features. Step two, which involves calibrating the initial position of the hook by setting the long side dimension of the actual hook, acquiring an image of the hook's initial position, identifying the hook, and calculating the offset distance of the hook's initial position, includes: First, the actual long side dimension of the hook is determined using the hook as a marker; when the hook is in its initial position, an image of the hook's initial position is obtained. Secondly, the trained YOLOv5 hook detection model is used to identify the hook in the initial position image and regress the position of the hook in the image to obtain the vertices of the rectangular box for hook target recognition. Then, based on the identified vertices of the hook rectangle, the pixel coordinates of the hook center in the image when the hook is in its initial position are obtained; Finally, based on the vertices of the bounding box identified by the hook target and the actual long side size of the hook, the ratio between the actual size and the pixel size, as well as the actual offset between the center of the hook's initial position and the center of the camera device, are calculated: ； ； ； in, k 0 represents the ratio between the actual size and the pixel size; a represents the actual length of the hook; x 10 , y 10 ), ( x 20 , y 20 () represents the vertex of the rectangular bounding box for hook target identification; w , h This indicates the initial position image size Im0 pixels; O x0 , O y0 This indicates the pixel coordinates of the hook center O in the image when the hook is in its initial position; x 0 and y 0 indicates the center of the initial position of the hook. O Center of the camera A The actual offset.

2. The intelligent detection method for tower crane inclined cable pulling and lifting as described in claim 1, characterized in that, Step one, which involves acquiring images of the hooks to prepare a hook sample set, constructing a YOLOv5 hook detection model, and training the YOLOv5 hook detection model based on the hook sample set, includes: (1) Use a camera device installed above the hook at the bottom of the trolley of the tower crane to acquire images of the hook in real time; (2) LabelImg is used to annotate the collected hook images to construct a hook sample set containing hook images in different states; (3) Construct a YOLOv5 hook detection model; import the hook sample set into the YOLOv5 hook detection model for training until the model converges, and set the trained network structure and parameters into the YOLOv5 target detection network to obtain the trained YOLOv5 hook detection model.

3. The intelligent detection method for tower crane inclined-stayed and inclined-lift cranes as described in claim 1, characterized in that, The initial position of the hook is when the steel rope hangs naturally, and the hook is not in a position where it is pulled or lifted at an angle.

4. The intelligent detection method for inclined tower cranes as described in claim 1, characterized in that, Step three involves determining the hook's angled pull by acquiring a real-time image of the hook, identifying the hook, and calculating the ratio between the actual hook size and pixel size at the current video magnification, the positional offset distance in the real-time image of the hook, the distance between the current and initial positions of the hook, and the hook's angled pull angle. 1) Use video equipment to acquire real-time video of the hook during the tower crane's operation, and extract frames every 5 frames of the video image stream to obtain the current image of the hook; 2) Use the trained hook detection model to identify the hook in the current image, regress the position of the hook in the current image, and obtain the vertex of the hook rectangle ( x 1, y 1) and ( x 2, y 2), and calculate the center of the hook when the hook is in its current position. O pixel coordinates ( O x , O y ): ； 3) Based on the vertices of the rectangle in the current image of the hook ( x 1, y 1), ( x 2, y 2), and the actual length of the hook. a Calculate the ratio between the actual size and the pixel size at the current video magnification. k : ； 4) Calculate the position of the hook in the current image relative to the center of the camera. A actual offset distance x and y : ； 5) Based on the initial position of the hook and the center of the camera A offset distance x 0、 y 0, and the current position of the hook relative to the center of the camera. A offset distance x , y Calculate the distance between the current hook position and the initial position. s : ； 6) Calculate the angle of attack between the initial position and the current position of the hook. θ : θ = arcsin ( s / L );in, L This indicates the current length of the hook wire rope; 7) When the angle of the hook is diagonal θ ≤ θ 0 indicates the hook is not in a diagonal pulling state; otherwise, the hook is in a diagonal pulling state and an alarm is triggered. θ 0 indicates the threshold for judging the angled pull of the hook; 8) Repeat steps 1) to 7 to monitor the state of the tower crane hook's oblique pulling and lifting in real time.

5. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the steps of the intelligent detection method for tower crane inclined cable pulling and lifting as described in any one of claims 1-4.

6. A computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the intelligent detection method for tower crane inclined cable-stayed cranes as described in any one of claims 1-4.

7. An information data processing terminal, characterized in that, The information data processing terminal is used to execute the intelligent detection method for tower crane cable-stayed cranes as described in any one of claims 1-4.

Images