A machine vision-based method for determining the rebar cage lowering process

Abstract

This invention discloses a machine vision-based method for determining the rebar cage lowering process, comprising the following steps: Step S1: Installing a camera at the construction site to collect data on the rebar cage construction process, creating a dataset, and training a model; Step S2: Receiving video data from the camera, using the model to detect the rebar cage target in the video, obtaining the coordinate information of the rebar cage detection frame, and transmitting the coordinate information in real time to the rebar cage process determination algorithm for process determination, identifying four processes: rebar cage hoisting start, rebar cage sleeve connection start, rebar cage sleeve connection end, and rebar cage hoisting end, and recording the process time; Step S3: While processing the coordinate data in real time, performing noise reduction processing on the data and counting the number of rebar cages lowered. This invention can effectively control the placement time of the rebar cage, reduce the probability of borehole collapse accidents, and improve the digitalization and intelligence level of bored pile construction.

Description

Technical Field

[0001] This invention relates to the field of bored pile construction. More specifically, this invention relates to a machine vision-based method for determining the rebar cage lowering process. Background Technology

[0002] Currently, the pile formation process for bored cast-in-place piles involves numerous steps and lacks effective on-site management and process control methods. For example, failure to promptly place the reinforcing cage after cleaning the borehole or leaving it in place for too long occasionally leads to borehole collapse accidents. To solve this problem, it is necessary to achieve automatic recording and control of the reinforcing cage placement time. Summary of the Invention

[0003] The purpose of this invention is to propose a machine vision-based method for identifying the rebar cage lowering process. To achieve automatic identification of the rebar cage construction process, this invention uses cameras installed at the construction site to film the placement of the rebar cage. The YOLOv7 algorithm is used to track and identify the rebar cage target in the filmed footage, obtaining the movement trajectory and coordinates of the hoisted rebar cage. A process identification algorithm is then used to analyze the changing trends of the rebar cage's trajectory coordinates in real time, identifying four process nodes during the lowering process: the start of rebar cage hoisting, the start of rebar cage sleeve connection, the end of rebar cage sleeve connection, and the end of rebar cage hoisting. The time of each node is automatically recorded. Through this method, automatic identification, recording, and timeout warnings for the rebar cage lowering process are achieved, thereby improving the level of on-site construction progress control.

[0004] The purpose of this invention is to provide a machine vision-based method for judging the rebar cage lowering process, which can effectively control the placement time of the rebar cage, reduce the probability of hole collapse accidents, and improve the digitalization and intelligence level of bored pile construction.

[0005] The technical solution adopted by this invention to solve this technical problem is: a machine vision-based method for judging the process of lowering a steel cage, comprising the following steps:

[0006] Step S1: Install cameras at the construction site to collect data on the construction process of the rebar cage, and create a dataset and train the model.

[0007] Step S2: Receive camera video data and use the model to detect the rebar cage target in the video, obtain the coordinate information of the rebar cage detection box, and transmit the coordinate information to the rebar cage process discrimination algorithm in real time for process discrimination. The algorithm will identify four processes: rebar cage hoisting start, rebar cage sleeve connection start, rebar cage sleeve connection end, and rebar cage hoisting end, and record the process time.

[0008] Step S3: While processing coordinate data in real time, perform noise reduction on the data and count the number of steel cages lowered.

[0009] As a further aspect of the present invention, step S1 specifically includes:

[0010] S11. Install high-definition cameras on site to capture images of the steel cage construction site, extract images of the steel cage hoisting from the captured images, mark the steel cages in the images to obtain tag data, save the tag data to a file to form a dataset;

[0011] S12. Following the method in S11, create n datasets, divide the datasets into training and validation sets in a ratio of a:1, and store the images and labels of the training and validation sets according to the directory structure.

[0012] S13. Train the YOLOv7 model.

[0013] As a further aspect of the present invention, step S2 specifically includes:

[0014] S21. After the model training is completed, the best.pt file containing the model's weight parameters is obtained. The trained weights best.pt are written into the weights section of the test.py program and used as the detection model parameters of the test.py program. Then, g images are randomly selected from the captured image for testing to test the detection effect.

[0015] S22 and best.pt performed well in terms of weights, so best.pt was continued to be used as the weight for detecting the rebar cage target in test.py. The system was called to connect to the on-site camera and acquire the image, the upper left coordinate (x1, y1) and lower right coordinate (x2, y2) of the detection box. x1, y1, x2, y2 were input into the process discrimination algorithm and saved to a file.

[0016] As a further aspect of the present invention, step S2, which involves transmitting coordinate information in real time to the rebar cage process discrimination algorithm for process discrimination, specifically includes:

[0017] S2a1. Determine the start of hoisting, including: when the crane lifts the steel cage and the angle between the steel cage and the ground is greater than 35°, the YOLOv7 model can detect the steel cage target. When the coordinates of the detection box change from no coordinate value (0, 0, 0, 0) to having coordinate value, and the detection box exists for more than 10 frames, it is determined that the hoisting of the steel cage has started, and the hoisting start time of the steel cage is recorded.

[0018] S2a2, Determine the start of sleeve connection, including: the crane hoists the steel cage to the top of the pile hole and suspends it, and makes sleeve connection with the steel cage in the pile hole. At this time, the coordinates x1,x2 and y1,y2 of the steel cage change from the changing state when moving to the relatively stationary state. When the coordinates remain stationary or change within the threshold 10 for more than 60 frames, determine the start of the steel cage sleeve connection and record the start time of the steel cage sleeve connection.

[0019] S2a3, Determine the end of the sleeve connection, including: After the sleeve connection of the rebar cage is completed, the crane lowers the rebar cage. At this time, the x1, x2, y2 coordinates of the rebar cage detection frame are still relatively stationary, but the upper left vertical coordinate y1 of the coordinate frame will continue to drop. If the y1 coordinate continues to drop for more than 20 frames, it is determined that the rebar cage sleeve connection is over, the rebar cage enters the lowering stage, and the end time of the rebar cage sleeve connection is recorded.

[0020] S2a4. Determine the end of hoisting, including: after the steel cage is lowered into the pile hole, the detection frame disappears. If the disappearance state lasts for more than 30 frames, the hoisting of the steel cage is determined to be over, and the hoisting end time of the steel cage is recorded.

[0021] As a further aspect of the present invention, the method for determining that the upper left vertical coordinate y1 is continuously decreasing is as follows:

[0022] After determining that the rebar cage sleeve connection has started, an empty list y1_value=[] is created to store the acquired y1 values. The length of the y1_value list is fixed at 10, and y1_value=y1_value[-10:] is calculated. The average value of y1_value is then obtained. The formula is as follows:

[0023]

[0024] Among them, when At that time, the average value is calculated starting from the 10th coordinate data point. If the state persists for more than 20 frames, it is judged as a drop.

[0025] As a further aspect of the present invention, step S3, which involves noise reduction processing of the data, specifically includes:

[0026] S3a1. Create two empty lists X_values ​​= [] and Y_values ​​= [] to store the obtained x1 and y1 coordinates respectively. Use the x1 coordinate to determine the left and right movement of the steel cage, and the y1 coordinate to determine the up and down movement of the steel cage.

[0027] S3a2. Control the number of list data to g. When the lengths of X_values ​​and Y_values ​​are greater than g, use slicing operations: X_values ​​= X_values[-100:], Y_values ​​= Y_values[-100:], and the algorithm calculates the average of X_values ​​and Y_values ​​once every p iterations. and ;

[0028] S3a3, Each time new coordinates x1, y1 are obtained and and Compare, if the current value x1 is... The difference between the current value y1 and the image width is greater than 1%, and is determined to be interference coordinates; If the difference is greater than 2% of the image height, it is determined to be an interference coordinate.

[0029] As a further aspect of the present invention, step S3, which involves counting the number of steel cages lowered, specifically includes:

[0030] S3b1. Set the total number of steel cages to be lowered into the pile hole as x and the initial value i=0, and create an empty dictionary dict_cage = {} to record the process of each steel cage;

[0031] S3b2. When the start time of rebar cage hoisting is recorded, the dictionary is searched for "end time of hoisting the {i}th rebar cage". If it exists, i = i + 1 is executed, and "start time of hoisting the {i}th rebar cage: {cage_start_datetime}" is saved to the dictionary dict_cage = {}, and the process proceeds to the next step of the process judgment. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0032] S3b3. When determining the start of the rebar cage sleeve connection and recording the start time of the rebar cage sleeve connection, dict_cage.get searches the dictionary for "start time of hoisting the {i}th rebar cage". If it exists, "start time of connection of the {i}th rebar cage sleeve: {cage_start_datetime}" is saved to the dictionary dict_cage = {}, and the process proceeds to the next process judgment. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0033] S3b4. When it is determined that the rebar cage sleeve connection is completed and the start time of the rebar cage sleeve connection is recorded, dict_cage.get searches the dictionary for "start time of the {i}th rebar cage sleeve connection"; if it exists, "end time of the {i}th rebar cage sleeve connection: {cage_start_datetime}" is saved to the dictionary dict_cage = {}, and the process proceeds to the next process judgment; if it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0034] S3b5. When the hoisting of the reinforcing cage is determined to be completed and the hoisting end time is recorded, dict_cage.get searches the dictionary for "end time of connection of the {i}th reinforcing cage sleeve". If it exists, save "end time of hoisting of the {i}th reinforcing cage: {cage_start_datetime}" to the dictionary dict_cage = {}. Check if i is equal to x. If they are not equal, proceed to the process judgment process of the next reinforcing cage. If they are equal, the process judgment process of the reinforcing cage of the pile hole ends, and the entire contents of dict_cage are saved as an Excel file using pandas.to_excel. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0035] As a further aspect of the present invention, step S3 further includes a timeout warning, specifically including: in step S3b5, when the program enters the process of determining the next steel cage, the timer is started in another new thread with cage_start_datetime as the starting time. When the time exceeds 30 minutes and the hoisting of the next steel cage has not been determined, a warning is issued, and the warning information is uploaded to the visualization cloud platform via HTTP.

[0036] The present invention has at least the following beneficial effects: The present invention improves the intelligence level of the rebar cage placement process, realizes contactless collection of rebar cage movement data, automatically identifies, records and provides timeout warnings for the rebar cage placement process without affecting construction, and reminds technicians to control behaviors such as rebar cage placement stagnation and timeouts while monitoring the entire rebar cage construction process, thereby improving the overall management level of bored pile construction and reducing the risk of hole collapse.

[0037] This invention relates to a machine vision-based method for identifying the rebar cage placement process. By placing a monitoring camera next to the pile hole and using machine vision and data processing methods, the method identifies the rebar cage placement process and automatically records the placement time of each rebar cage section. When the placement interval of a certain rebar cage section is too long, an early warning message is transmitted to a visualization cloud platform via HTTP protocol, notifying technicians to go to the construction site to understand the situation.

[0038] This method requires no modification to the engineering machinery. It uses a camera and target detection technology to acquire dynamic data of the rebar cage in a non-contact manner, and achieves the purpose of process identification through data analysis. During the detection of the rebar cage and the identification of the process, some data may be inaccurate due to external interference. This can be resolved by using an interference data processing mechanism.

[0039] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the dataset directory structure of this invention;

[0041] Figure 2 This is a schematic diagram of the detection results of the model of the present invention;

[0042] Figure 3 This is a schematic diagram of the steel cage inspection of the present invention;

[0043] Figure 4 This is a schematic diagram of the steel cage process discrimination result of the present invention;

[0044] Figure 5 This is a flowchart of the steel cage counting and discrimination process of the present invention;

[0045] Figure 6 This is a flowchart of the rebar cage construction process discrimination method based on machine vision according to the present invention. Detailed Implementation

[0046] The present invention will now be described in detail and completely with reference to the accompanying drawings. Those skilled in the art will be able to implement the present invention based on these descriptions. Before describing the present invention with reference to the accompanying drawings, it should be particularly noted that the technical solutions and features provided in various parts of the present invention, including the following description, can be combined with each other without conflict.

[0047] Furthermore, the embodiments of the present invention described below are generally only some, not all, of the embodiments of the present invention. Therefore, all other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

[0048] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific implementation process is as follows:

[0049] A machine vision-based method for determining the lowering process of a rebar cage includes the following steps:

[0050] Step S1: Install cameras at the construction site to collect data on the construction process of the rebar cage, and create a dataset and train the model.

[0051] Step S2: Receive camera video data and use the model to detect the rebar cage target in the video, obtain the coordinate information of the rebar cage detection box, and transmit the coordinate information to the rebar cage process discrimination algorithm in real time for process discrimination. The algorithm will identify four processes: rebar cage hoisting start, rebar cage sleeve connection start, rebar cage sleeve connection end, and rebar cage hoisting end, and record the process time.

[0052] Step S3: While processing coordinate data in real time, perform noise reduction on the data and count the number of steel cages lowered.

[0053] This technical solution may also include the following technical details to better achieve the technical effect: Step S1 specifically includes:

[0054] S11. Install high-definition cameras on site to capture images of the rebar cage construction site, extract images of the rebar cage hoisting from the captured images, and use the software Labelimg to mark the rebar cages in the images to obtain tag data. Save the tag data to a txt file to form a dataset.

[0055] S12. Following the method in S11, create n datasets. Divide the datasets into training and validation sets in a : 1 ratio, and then arrange the images and labels in the training and validation sets as follows: Figure 1 The directory structure shown is used for storage; in this embodiment, n is 300 and a is 4.

[0056] S13. Train the YOLOv7 model. Specifically, create a RebarCage.yaml file, write the paths to the training and validation sets, set the label class parameter nc to 1, replace the default yaml file in train.py with RebarCage.yaml, use yolov7_training.pt for weights, set the batch size to 12, the number of epochs to 200, and execute the training program file python train.py to train the model.

[0057] This technical solution may also include the following technical details to better achieve the technical effect: Step S2 specifically includes:

[0058] S21. After the model training is complete, the best.pt file containing the model's weight parameters is obtained. The trained weights best.pt are written into the weights section of the test.py program and used as the detection model parameters for the test.py program. Then, g images are randomly selected from the captured image for testing. In this embodiment, g is 100 images. The test results are good, and some detection results are as follows: Figure 2 As shown;

[0059] S22 and best.pt performed well in terms of weights, so best.pt was continued to be used as the weight for detecting the rebar cage target in test.py. The system called the connection to the on-site camera and captured the image at a frame rate of 3FPS. The upper left coordinate (x1, y1) and lower right coordinate (x2, y2) of the detection box were obtained. x1, y1, x2, y2 were input into the process discrimination algorithm and saved to a txt file.

[0060] This technical solution may also include the following technical details to better achieve the technical effect: The real-time transmission of coordinate information to the rebar cage process discrimination algorithm in step S2 specifically includes:

[0061] S2a1, Determine the start of hoisting, including: such as Figure 3 As shown, when the crane lifts the steel cage and the angle between the steel cage and the ground is greater than 35°, the YOLOv7 model can detect the steel cage target. When the coordinates of the detection box change from no coordinate value (0, 0, 0, 0) to having coordinate values, and the detection box exists for more than 10 frames, it is determined that the steel cage hoisting has started, and the start time of the steel cage hoisting is recorded.

[0062] S2a2, Determine the start of sleeve connection, including: the crane hoists the steel cage to the top of the pile hole and suspends it, and makes sleeve connection with the steel cage in the pile hole (except for the first steel cage). At this time, the coordinates x1,x2 and y1,y2 of the steel cage change from the changing state when moving to the relatively stationary state. When the coordinates remain stationary or change within the threshold 10 for more than 60 frames, determine the start of the steel cage sleeve connection and record the start time of the steel cage sleeve connection.

[0063] S2a3, Determine the end of the sleeve connection, including: After the sleeve connection of the rebar cage is completed, the crane lowers the rebar cage. At this time, the x1, x2, y2 coordinates of the rebar cage detection frame are still relatively stationary, but the upper left vertical coordinate y1 of the coordinate frame will continue to drop. If the y1 coordinate continues to drop for more than 20 frames, it is determined that the rebar cage sleeve connection is over, the rebar cage enters the lowering stage, and the end time of the rebar cage sleeve connection is recorded.

[0064] S2a4. Determine the end of hoisting, including: after the steel cage is lowered into the pile hole, the detection frame disappears. If the disappearance state lasts for more than 30 frames, the hoisting of the steel cage is determined to be over, and the hoisting end time of the steel cage is recorded.

[0065] S2a5. Visualize the coordinate data saved to the txt file using the plt plotting tool, such as... Figure 4 As shown, based on the recorded time and coordinate data, the four process steps were identified at the marked locations in the figure, and the results indicate that the identification positions are relatively accurate.

[0066] This technical solution may also include the following technical details to better achieve the technical effect: The method for determining that the upper left vertical coordinate y1 is continuously decreasing is as follows:

[0067] After determining that the rebar cage sleeve connection has started, an empty list y1_value=[] is created to store the acquired y1 values. The length of the y1_value list is fixed at 10, and y1_value=y1_value[-10:] is calculated. The average value of y1_value is then obtained. The formula is as follows:

[0068]

[0069] Among them, when At that time, the average value is calculated starting from the 10th coordinate data point. If the state persists for more than 20 frames, it is judged as a drop.

[0070] This technical solution may also include the following technical details to better achieve the technical effect: In step S2, the rebar cage construction process discrimination algorithm is a real-time processing algorithm. During the process of receiving the detection frame coordinate data, the detection frame may become unstable due to factors such as light and obstructions. At this time, the coordinate points will change drastically. Such coordinate data may affect the accuracy of the process discrimination algorithm. The above problem can be solved by the following noise reduction method:

[0071] Step S3, which involves noise reduction of the data, specifically includes:

[0072] S3a1. Create two empty lists X_values ​​= [] and Y_values ​​= [] to store the obtained x1 and y1 coordinates respectively. Because during the discrimination process, it was found that the x1 and y1 coordinates are more stable and have obvious characteristics. The x1 coordinate is mainly used to determine the left and right movement of the steel cage, and the y1 coordinate is used to determine the up and down movement of the steel cage.

[0073] S3a2. To reduce computational burden, we limit the number of list data items to g=100. When the lengths of X_values ​​and Y_values ​​exceed g=100, we use slicing operations: X_values ​​= X_values[-100:], Y_values ​​= Y_values[-100:], and the algorithm calculates the average of X_values ​​and Y_values ​​every p=3 iterations. and The formula for calculating the average value is as follows:

[0074] ;

[0075] S3a3, Each time new coordinates x1, y1 are obtained and and Compare, if the current value x1 is... If the difference between the current value y1 and the image width is greater than 1%, it is determined to be an interference coordinate, and the current discrimination operation is terminated using `break`; the current value y1 and... If the difference between x1 and y1 is greater than 2% of the image height, it is determined to be an interference coordinate, and the current discrimination operation is terminated using the `break` function. This method effectively solves the problem of interference coordinate data affecting the process discrimination algorithm. Any difference between x1 and y1 greater than a specified value is considered an interference coordinate.

[0076] This technical solution may also include the following technical details to better achieve the technical effect: based on the process-oriented rebar cage construction procedure discrimination algorithm, it simultaneously realizes the counting of rebar cages being lowered, and designs a relatively rigorous discrimination process to effectively avoid misjudging a certain procedure. The process is as follows: Figure 5 As shown, step S3, which involves counting the number of steel cages lowered, specifically includes:

[0077] S3b1. Set the total number of steel cages to be lowered into the pile hole as x and the initial value i=0, and create an empty dictionary dict_cage = {} to record the process of each steel cage;

[0078] S3b2. When the start of rebar cage hoisting is determined, and the start time of rebar cage hoisting is recorded, search the dictionary for "end time of hoisting the {i}th rebar cage" (excluding the first rebar cage). If it exists, execute i=i+1, and save "start time of hoisting the {i}th rebar cage: {cage_start_datetime}" to the dictionary dict_cage = {}, and proceed to the next process judgment process. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and continue to execute the current process judgment algorithm.

[0079] S3b3. When determining the start of the rebar cage sleeve connection and recording the start time of the rebar cage sleeve connection, dict_cage.get searches the dictionary for "start time of hoisting the {i}th rebar cage". If it exists, "start time of connection of the {i}th rebar cage sleeve: {cage_start_datetime}" is saved to the dictionary dict_cage = {}, and the process proceeds to the next process judgment. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0080] S3b4. When it is determined that the rebar cage sleeve connection is completed and the start time of the rebar cage sleeve connection is recorded, dict_cage.get searches the dictionary for "start time of the {i}th rebar cage sleeve connection"; if it exists, "end time of the {i}th rebar cage sleeve connection: {cage_start_datetime}" is saved to the dictionary dict_cage = {}, and the process proceeds to the next process judgment; if it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0081] S3b5. When the hoisting of the reinforcing cage is determined to be completed and the hoisting end time is recorded, dict_cage.get searches the dictionary for "end time of connection of the {i}th reinforcing cage sleeve". If it exists, save "end time of hoisting of the {i}th reinforcing cage: {cage_start_datetime}" to the dictionary dict_cage = {}. Check if i is equal to x. If they are not equal, proceed to the process judgment process of the next reinforcing cage. If they are equal, the process judgment process of the reinforcing cage of the pile hole ends, and the entire contents of dict_cage are saved as an Excel file using pandas.to_excel. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed.

[0082] This technical solution may also include the following technical details to better achieve the technical effect: Step S3 also includes a timeout warning, specifically including: In step S3b5, when the program enters the process judgment flow of the next steel cage, the timer is started in another new thread with cage_start_datetime as the starting time. When the time exceeds 30 minutes and the hoisting of the next steel cage has not been determined, a warning is issued. The warning information will be uploaded to the visualization cloud platform via HTTP to remind technicians to check the situation in the corresponding pile hole area.

[0083] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. A machine vision-based method for determining the rebar cage lowering process, characterized in that, Includes the following steps: Step S1: Install cameras at the construction site to collect data on the construction process of the rebar cage, and create a dataset and train the model. Step S2: Receive camera video data and use the model to detect the rebar cage target in the video, obtain the coordinate information of the rebar cage detection box, and transmit the coordinate information to the rebar cage process discrimination algorithm in real time for process discrimination. The algorithm will identify four processes: rebar cage hoisting start, rebar cage sleeve connection start, rebar cage sleeve connection end, and rebar cage hoisting end, and record the process time. Step S3: While processing the coordinate data in real time, the algorithm will also perform noise reduction processing on the data and count the number of rebar cages lowered. Step S2 specifically includes: S21. After the model training is completed, the best.pt file containing the model's weight parameters is obtained. The trained weights best.pt are written into the weights section of the test.py program and used as the detection model parameters of the test.py program. Then, g images are randomly selected from the captured image for testing to test the detection effect. S22 and best.pt performed well in terms of weights, so best.pt was continued to be used as the weight for detecting the rebar cage target in test.py. The connection to the on-site camera was called to acquire the image and obtain the upper left coordinate (x1, y1) and lower right coordinate (x2, y2) of the detection box. x1, y1, x2, y2 were input into the process discrimination algorithm and saved to a file. The step S2, which involves transmitting coordinate information in real time to the rebar cage process discrimination algorithm for process discrimination, specifically includes: S2a1. Determine the start of hoisting, including: when the crane lifts the steel cage and the angle between the steel cage and the ground is greater than 35°, the YOLOv7 model can detect the steel cage target. When the coordinates of the detection box change from no coordinate value (0, 0, 0, 0) to having coordinate value, and the detection box exists for more than 10 frames, it is determined that the hoisting of the steel cage has started, and the hoisting start time of the steel cage is recorded. S2a2, Determine the start of sleeve connection, including: the crane hoists the steel cage to the top of the pile hole and suspends it, and makes sleeve connection with the steel cage in the pile hole. At this time, the coordinates x1,x2 and y1,y2 of the steel cage change from the changing state when moving to the relatively stationary state. When the coordinates remain stationary or change within the threshold 10 for more than 60 frames, determine the start of the steel cage sleeve connection and record the start time of the steel cage sleeve connection. S2a3, Determine the end of the sleeve connection, including: After the sleeve connection of the rebar cage is completed, the crane lowers the rebar cage. At this time, the x1, x2, y2 coordinates of the rebar cage detection frame are still relatively stationary, but the upper left vertical coordinate y1 of the coordinate frame will continue to drop. If the y1 coordinate continues to drop for more than 20 frames, it is determined that the rebar cage sleeve connection is over, the rebar cage enters the lowering stage, and the end time of the rebar cage sleeve connection is recorded. S2a4. Determine the end of hoisting, including: after the steel cage is lowered into the pile hole, the detection frame disappears. If the disappearance state lasts for more than 30 frames, the hoisting of the steel cage is determined to be over, and the hoisting end time of the steel cage is recorded.

2. The machine vision-based method for determining the rebar cage lowering process as described in claim 1, characterized in that, Step S1 specifically includes: S11. Install high-definition cameras on site to capture images of the steel cage construction site, extract images of the steel cage hoisting from the captured images, mark the steel cages in the images to obtain tag data, save the tag data to a file to form a dataset; S12. Following the method in S11, create n datasets, divide the datasets into training and validation sets in a ratio of a:1, and store the images and labels of the training and validation sets according to the directory structure. S13. Train the YOLOv7 model.

3. The method for determining the rebar cage lowering process based on machine vision as described in claim 1, characterized in that, The method for determining that the top-left ordinate y1 is continuously decreasing is as follows: After determining that the rebar cage sleeve connection has started, an empty list y1_value=[] is created to store the acquired y1 values. The length of the y1_value list is fixed at 10, and y1_value=y1_value[-10:] is calculated. The average value of y1_value is then obtained. The formula is as follows: Among them, when At that time, the average value is calculated starting from the 10th coordinate data point. If the state persists for more than 20 frames, it is judged as a drop.

4. The machine vision-based method for judging the rebar cage lowering process as described in claim 1, characterized in that, Step S3, which involves noise reduction of the data, specifically includes: S3a1. Create two empty lists X_values ​​= [] and Y_values ​​= [] to store the obtained x1 and y1 coordinates respectively. Use the x1 coordinate to determine the left and right movement of the steel cage, and the y1 coordinate to determine the up and down movement of the steel cage. S3a2. Control the number of list data to g. When the lengths of X_values ​​and Y_values ​​are greater than g, use slicing operations: X_values ​​= X_values[-100:], Y_values ​​= Y_values[-100:], and the algorithm calculates the average of X_values ​​and Y_values ​​once every p iterations. and ; S3a3, Each time new coordinates x1, y1 are obtained and and Compare, if the current value x1 is... The difference between the current value y1 and the image width is greater than 1%, and is determined to be interference coordinates; If the difference is greater than 2% of the image height, it is determined to be an interference coordinate.

5. The machine vision-based method for determining the rebar cage lowering process as described in claim 1, characterized in that, The specific steps of step S3, which involves counting the number of steel cages lowered, include: S3b1. Set the total number of steel cages to be lowered into the pile hole as x and the initial value i=0, and create an empty dictionary dict_cage= {} to record the process of each steel cage; S3b2. When the start time of rebar cage hoisting is recorded, the dictionary is searched for "end time of hoisting the {i}th rebar cage". If it exists, i = i + 1 is executed, and "start time of hoisting the {i}th rebar cage: {cage_start_datetime}" is saved to the dictionary dict_cage = {}, and the process proceeds to the next step of the process judgment. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm continues to be executed. S3b3. When determining the start of the rebar cage sleeve connection and recording the start time of the rebar cage sleeve connection, dict_cage.get searches the dictionary for "start time of hoisting the {i}th rebar cage" (excluding the first rebar cage). If it exists, save "start time of connection of the {i}th rebar cage sleeve: {cage_start_datetime}" to the dictionary dict_cage = {} and proceed to the next process judgment flow. If it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and continue to execute the current process judgment algorithm. S3b4. When it is determined that the rebar cage sleeve connection is completed and the start time of the rebar cage sleeve connection is recorded, dict_cage.get searches the dictionary for "start time of the {i}th rebar cage sleeve connection"; if it exists, save "end time of the {i}th rebar cage sleeve connection: {cage_start_datetime}" to the dictionary dict_cage = {} and proceed to the next process judgment flow; if it does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and continue to execute the current process judgment algorithm; S3b5. When the hoisting of the reinforcing cage is determined to be completed and the hoisting end time is recorded, dict_cage.get searches the dictionary for "end time of connection of the {i}th reinforcing cage sleeve". If it exists, "end time of hoisting of the {i}th reinforcing cage: {cage_start_datetime}" is saved to the dictionary dict_cage = {}. It is then checked whether i is equal to x. If they are not equal, the process judgment process for the next reinforcing cage is entered. If they are equal, the process judgment process for the reinforcing cage of this pile hole ends, and all contents of dict_cage are saved to an Excel file using pandas.to_excel. If the process judgment does not exist in the dictionary, it means that there is a misjudgment in the process judgment, and the current process judgment algorithm is continued.

6. The machine vision-based method for determining the rebar cage lowering process as described in claim 5, characterized in that, Step S3 also includes a timeout warning, specifically: In step S3b5, when the program enters the process of determining the next steel cage, a timer is started in another new thread with cage_start_datetime as the starting time. If the time exceeds 30 minutes and the hoisting of the next steel cage has not been determined, a warning is issued, and the warning information is uploaded to the visualization cloud platform via HTTP.

Images