Method and device for monitoring the safety of a tower construction

Abstract

The application relates to the technical field of electric power and computer vision, in particular to a tower construction safety monitoring method and device, which acquires real-time tower construction site video streams of each position and synchronously collects current weather. The construction stage, construction atmosphere, dangerous construction behavior and personnel construction experience are identified based on the video streams. The initial dangerous coefficient of the dangerous construction behavior is corrected by a second correction coefficient of the construction stage, the construction atmosphere, personnel experience and weather to obtain a first dangerous coefficient, and the video streams are arranged in descending order according to the first dangerous coefficient. The second dangerous coefficient of the initial dangerous coefficient of the construction stage is corrected by a fourth correction coefficient of the construction atmosphere, personnel experience and weather, and the video streams are arranged in descending order according to the second dangerous coefficient, and the dangerous behavior video is in the front. The application can more accurately evaluate the actual dangerous degree of the current dangerous construction behavior, more accurately evaluate the potential dangerous degree of the current tower construction and greatly shorten the response link from risk occurrence to intervention.

Description

Technical Field

[0001] This application relates to the fields of power and computer vision technology, and in particular to a method and device for monitoring the safety of tower construction. Background Technology

[0002] To monitor tower construction safety, ground-mounted cameras or camera-equipped drones are currently used for safety monitoring, and the monitoring videos are simultaneously transmitted to a backend monitor for display. The monitoring videos from different camera positions are displayed on the backend monitor in a fixed arrangement. This fixed arrangement makes it difficult for monitoring personnel to quickly identify dangerous behaviors and potential construction risks during tower construction, thus failing to shorten the response chain from risk emergence to intervention. Furthermore, when different camera positions exhibit different dangerous behaviors, monitoring personnel struggle to quickly identify the actual degree of danger between different behaviors; conversely, when no dangerous behaviors occur at different camera positions and different construction stages are underway, monitoring personnel struggle to quickly identify the potential construction risks at different stages. Moreover, due to differing perceptions of the actual danger of different behaviors among monitoring personnel, the lack of objective and quantifiable risk assessment standards leads to resource misallocation—for example, excessive intervention resources are allocated to low-risk behaviors, while insufficient attention is paid to high-risk behaviors, resulting in low regulatory efficiency. Summary of the Invention

[0003] In view of the above-mentioned technical problems, the purpose of this application is to provide a method and device for monitoring the safety of tower construction, which aims to solve the aforementioned technical problems.

[0004] In a first aspect, embodiments of this application provide a method for monitoring the safety of tower construction, including:

[0005] Real-time acquisition of video streams from the construction sites of each tower;

[0006] Based on the video streams of each tower construction site, the current construction stage, current construction atmosphere, current dangerous construction behavior, and current tower construction experience of the construction personnel are identified.

[0007] Simultaneously acquire the current weather conditions at each tower construction site;

[0008] For video streams containing dangerous construction activities, a second correction coefficient for the current dangerous construction activity is obtained from a preset first association table to correct the initial risk coefficient of the current dangerous construction activity, resulting in a first risk coefficient; wherein the second correction coefficient includes: the first correction coefficients for the current construction stage, the current construction atmosphere, the current tower construction experience of the construction personnel, and the current weather conditions;

[0009] The monitoring videos of each machine location exhibiting dangerous construction activities are arranged in descending order of the highest hazard factor.

[0010] For video streams without dangerous construction activities, the fourth correction coefficient for the current construction stage is obtained from the preset second association table to correct the initial risk coefficient for the current construction stage, thus obtaining the second risk coefficient. The fourth correction coefficient includes the third correction coefficient based on the current construction atmosphere, the current construction personnel's experience in tower construction, and the current weather conditions.

[0011] The monitoring videos of each machine position that do not involve dangerous construction activities are arranged in descending order of the second hazard factor.

[0012] All sorted surveillance videos are output to the backend monitoring screen for display; among them, surveillance videos with dangerous construction activities are arranged before surveillance videos without dangerous construction activities.

[0013] Furthermore, the step of obtaining a second correction coefficient for the current hazardous construction behavior from a preset first association table to correct the initial hazard coefficient of the current hazardous construction behavior, and obtaining the first hazard coefficient, includes:

[0014] The initial risk factor of the current hazardous construction behavior is corrected according to the following model;

[0015] R1 = a1 × a2 × a3 × a4 × R0;

[0016] Wherein, R1 is the first risk factor, R0 is the initial risk factor of the current dangerous construction behavior, a1 is the first correction factor of the current construction stage, a2 is the first correction factor of the current construction atmosphere, a3 is the first correction factor of the current construction personnel's tower construction experience, and a4 is the first correction factor of the current weather.

[0017] Furthermore, the step of obtaining the fourth correction coefficient for the current construction stage from the preset second association table to correct the initial hazard coefficient for the current construction stage and obtain the second hazard coefficient includes:

[0018] The initial risk factor for the current construction stage is corrected according to the following model to obtain the second risk factor;

[0019] R4 = a5 × a6 × a7 × R3;

[0020] Among them, R4 is the second risk factor, R3 is the initial risk factor of the current construction stage, a5 is the third correction factor of the current construction atmosphere, a6 is the third correction factor of the current construction personnel's tower construction experience, and a7 is the third correction factor of the current weather.

[0021] Furthermore, the first correction coefficient is calculated using the following method:

[0022] Collect the first construction history data of the tower; wherein, the first construction history data of the tower includes the identification of dangerous construction behavior, the initial risk coefficient of dangerous construction behavior, the construction stage, the construction atmosphere, the tower construction experience of construction personnel, the weather, and the corresponding first actual risk results;

[0023] The first construction history data of the tower is preprocessed, including quantifying the first actual risk result and grouping it according to the combination of dangerous construction behavior type, construction stage, construction atmosphere, tower construction experience of construction personnel and weather conditions.

[0024] The first multiple regression model is constructed, which includes taking the quantified first actual risk result as the dependent variable and taking the first correction coefficient of the construction stage, the first correction coefficient of the construction atmosphere, the first correction coefficient of the construction personnel's tower construction experience, and the first correction coefficient of the weather as independent variables.

[0025] The preprocessed historical data of the first construction of the tower is divided into a training set and a test set. The first multivariate regression model is trained using the training set. The first correction coefficients that minimize the sum of squared residuals between the predicted value of the first multivariate regression model and the quantified first actual risk result are solved by the least squares method.

[0026] The first multivariate regression model trained is validated using a test set. If the prediction error is within a preset range, the first correction coefficients obtained are determined as the final first correction coefficients. If the prediction error is not within the preset range, the data samples are adjusted until the first correction coefficients that meet the error requirements are obtained.

[0027] Furthermore, the third correction coefficient is calculated using the following method:

[0028] Collect second construction history data for the towers; wherein, the second construction history data for the towers includes the construction stage, the initial risk factor of the construction stage, the construction atmosphere, the tower construction experience of the construction personnel, the weather, and the corresponding second actual risk results;

[0029] The second construction history data of the tower is preprocessed, including quantifying the second actual risk result and grouping it according to the combination of construction stage, construction atmosphere, tower construction experience of construction personnel and weather conditions.

[0030] A second multiple regression model was constructed, which included using the quantified second actual risk result as the dependent variable and the third correction coefficient of the construction atmosphere, the third correction coefficient of the construction personnel's tower construction experience, and the third correction coefficient of the weather as independent variables.

[0031] The preprocessed second construction history data of the tower is divided into a training set and a test set. The second multivariate regression model is trained using the training set. The third correction coefficients that minimize the sum of squared residuals between the predicted value of the second multivariate regression model and the quantified second actual risk result are solved by the least squares method.

[0032] The trained second multiple regression model is validated using a test set. If the prediction error is within a preset range, the obtained third correction coefficients are determined as the final third correction coefficients. If the prediction error is not within a preset range, the data samples are adjusted until the third correction coefficients that meet the error requirements are obtained.

[0033] Furthermore, the steps for identifying the current construction stage based on the video stream at the tower construction site include:

[0034] The video stream from the tower construction site is preprocessed to obtain a preprocessed tower video stream;

[0035] Extract keyframes from the preprocessed tower video stream; the keyframe extraction includes filtering out frames with significant motion changes as keyframes by detecting motion changes between video frames using optical flow;

[0036] The extracted keyframes are input into a pre-trained spatiotemporal feature extraction model to obtain spatiotemporal features; wherein, the spatiotemporal feature extraction model includes a cascaded 3D convolutional neural network and a bidirectional long short-term memory network;

[0037] By using the Transformer cross-attention mechanism, the spatiotemporal features are fused with the IoT data from the tower construction site in the channel dimension to obtain the fused features;

[0038] The fused feature vector is input into the Softmax classifier, which outputs the current tower construction stage category. The category includes at least: foundation excavation and grounding trench construction, tower material ground pre-assembly and transportation, tower leg / bottom section assembly, tower body segment / segment high-level hoisting, crossarm overall hoisting, permanent guy wire tensioning and pole removal, bolt final tightening and cleaning.

[0039] Furthermore, the steps for identifying the current construction atmosphere based on the video stream at the tower construction site include:

[0040] The current construction atmosphere score is calculated based on the following model;

[0041]

[0042] b jσ =a jσ ×β+δ;

[0043]

[0044] Where X is the current construction atmosphere score, and Q is... ij For the current construction status score of the i-th and j-th construction worker, m is the number of construction workers identified, n is the number of evaluations, and b is the number of evaluations. jσ Let a be the probability value of the σ-th emotion corresponding to the j-th construction worker. jσ Let w be the σ-th feature value extracted from the face of the j-th construction worker, where σ = 1, 2, ..., s, corresponding to the recognition dimensions of s kinds of emotions, β and δ are parameters that adjust the linear relationship between emotional features and construction worker emotions. jσ Let f(·) be the weight value of the σ-th facial feature of the j-th construction worker; f(·) represents the mapping function used to map the result to a preset construction status score.

[0045] The current construction atmosphere score is mapped to a construction atmosphere type.

[0046] Secondly, embodiments of this application provide a tower construction safety monitoring device, the device comprising:

[0047] The video stream acquisition module is used to acquire video streams from the construction site of each tower in real time.

[0048] The identification module is used to identify the current construction stage, current construction atmosphere, current dangerous construction behavior, and current tower construction experience of the construction personnel based on the video stream of each tower construction site.

[0049] The weather acquisition module is used to synchronously acquire the current weather conditions at the construction sites of each tower location;

[0050] The first correction module is used to obtain a second correction coefficient for the current dangerous construction behavior from a preset first association table for video streams containing dangerous construction behavior, and to correct the initial danger coefficient of the current dangerous construction behavior to obtain a first danger coefficient; wherein the second correction coefficient includes: the first correction coefficient of the current construction stage, the current construction atmosphere, the current tower construction experience of the construction personnel, and the current weather.

[0051] The first sorting module is used to sort the monitoring videos of each machine position that have dangerous construction behavior in descending order according to the first hazard factor;

[0052] The second correction module is used to obtain the fourth correction coefficient of the current construction stage from the preset second association table for video streams without dangerous construction behavior, and to correct the initial danger coefficient of the current construction stage to obtain the second danger coefficient; wherein, the fourth correction coefficient includes: the current construction atmosphere, the current tower construction experience of the construction personnel and the current weather conditions.

[0053] The second sorting module is used to sort the monitoring videos of each machine position that do not have dangerous construction behavior in descending order according to the second hazard factor;

[0054] The output module is used to output all sorted surveillance videos to the background monitoring screen for display; among them, surveillance videos with dangerous construction activities are arranged before surveillance videos without dangerous construction activities.

[0055] In this embodiment, the video streams from each camera position are analyzed to determine their current construction stage, current construction atmosphere, current hazardous construction behavior, the experience of the current construction personnel, and current weather conditions. For video streams exhibiting hazardous construction behavior, a first correction coefficient for the current construction stage, the first correction coefficient for the current construction atmosphere, the first correction coefficient for the current construction personnel's experience in pole construction, and the initial risk coefficient for the current hazardous construction behavior are obtained from a preset first association table. This first correction coefficient, along with the first correction coefficients for the current construction stage, the current construction atmosphere, the current construction personnel's experience in pole construction, and the current weather conditions, is then used to correct the initial risk coefficient of the current hazardous construction behavior, resulting in a first risk coefficient. This multi-dimensional correction of the initial risk coefficient for hazardous construction behavior allows for a more accurate assessment of the actual risk level of the current hazardous construction behavior. The monitoring videos from each camera position exhibiting hazardous construction behavior are arranged in descending order of the first risk coefficient; that is, for camera positions where violations have occurred, those with higher risk levels are ranked higher, enabling monitoring personnel to quickly identify the actual risk levels between different hazardous behaviors and to quickly identify high-risk operations.

[0056] For video streams without dangerous construction activities, a fourth correction coefficient for the current construction stage is obtained from a pre-set second association table to correct the initial hazard coefficient of the current construction stage, resulting in a second hazard coefficient. This fourth correction coefficient includes a third correction coefficient based on the current construction atmosphere, the current construction personnel's experience in tower construction, and the current weather conditions. This multi-dimensional correction of the initial hazard coefficient for potential construction stages allows for a more accurate assessment of the potential hazard level. Monitoring videos from each camera position without dangerous construction activities are arranged in descending order of the second hazard coefficient; that is, the higher the potential hazard level, the earlier it appears. This allows monitoring personnel to quickly identify the potential hazard levels at different construction stages. All sorted monitoring videos are then output to a backend monitoring screen, with videos exhibiting dangerous construction activities placed before those without. This transforms the entire monitoring screen into a real-time updated "risk heat map"—the top left corner shows the most urgent, already violated scenes, while the right side shows normal operations with the highest potential hazard. Monitoring personnel do not need to cycle through each screen; they can simply handle issues from top to bottom according to the order of arrangement, greatly shortening the response chain from risk emergence to intervention and improving regulatory efficiency. Attached Figure Description

[0057] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0058] Figure 1 This is a schematic flowchart of a method for monitoring the safety of tower construction provided in an embodiment of this application;

[0059] Figure 2 This is a schematic diagram of the structure of a tower construction safety monitoring device provided in one embodiment of this application. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0061] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, modules, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, modules, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any modules and all combinations of one or more associated listed items.

[0062] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0063] Please see Figure 1 This application provides a method for monitoring the safety of tower construction, including steps S1-S8:

[0064] S1. Real-time acquisition of video streams from the construction sites of each tower location;

[0065] S2. Based on the video streams of each tower construction site, identify the current construction stage, current construction atmosphere, current dangerous construction behavior, and current tower construction experience of the construction personnel.

[0066] S3. Simultaneously acquire the current weather conditions at each tower construction site;

[0067] S4. For video streams containing dangerous construction activities, a second correction coefficient for the current dangerous construction activity is obtained from a preset first association table to correct the initial risk coefficient of the current dangerous construction activity, resulting in a first risk coefficient. The second correction coefficient includes the first correction coefficients for the current construction stage, the current construction atmosphere, the current tower construction experience of the construction personnel, and the current weather conditions.

[0068] S5. The monitoring videos of each machine location exhibiting dangerous construction behavior are arranged in descending order of the highest hazard factor;

[0069] S6. For video streams without dangerous construction activities, the fourth correction coefficient of the current construction stage is obtained from the preset second association table to correct the initial danger coefficient of the current construction stage, thus obtaining the second danger coefficient. The fourth correction coefficient includes the third correction coefficient of the current construction atmosphere, the tower construction experience of the current construction personnel, and the current weather.

[0070] S7. Arrange the monitoring videos of each machine position that do not have dangerous construction behavior in descending order of the second hazard factor;

[0071] S8. Output all sorted surveillance videos to the backend monitoring screen for display; among them, surveillance videos with dangerous construction activities are arranged before surveillance videos without dangerous construction activities.

[0072] In step S1, multiple cameras are deployed at each tower construction site to obtain video streams from different tower construction sites and video streams from different camera positions at the same tower construction site, i.e., to obtain video streams from each tower construction site.

[0073] In steps S2 and S3, based on the video streams of each tower construction site, the current construction stage, current construction atmosphere, current hazardous construction behavior, and current tower construction experience of the construction personnel are identified. Specifically, the tower construction stage includes foundation excavation and grounding trench construction, ground pre-assembly and transportation of tower materials, tower leg / base section assembly, high-level hoisting of tower body sections / segments, overall hoisting of crossarms, permanent guy wire tensioning and pole removal, final bolt tightening and cleaning / defect elimination. The construction atmosphere includes excellent, good, average, and poor. Hazardous construction behaviors include not wearing safety belts and not wearing safety helmets. The tower construction experience of the construction personnel includes experts, skilled workers, and novices. The weather includes good, normal, and severe. The tower construction experience of the construction personnel can be determined by facial recognition to identify the personnel's identity information. The experience of the tower construction personnel is obtained from the identity information. When multiple construction personnel are identified, the experience of the construction personnel with the lowest experience is used as the tower construction experience of the current construction personnel in that video stream. Obtain the current weather conditions at each tower construction site. Specifically, this includes collecting current weather data, such as temperature, rainfall, humidity, and wind speed, through weather acquisition devices at the tower construction site. Based on the current weather data, assess whether the current weather is good, normal, or severe.

[0074] In step S4, the first association table includes at least: a hazardous construction behavior identifier field, used to store unique identifiers for various hazardous construction behaviors; an initial hazard coefficient field, used to store the initial hazard coefficient for each type of hazardous construction behavior; construction stage level field, construction atmosphere level field, tower construction experience level field for construction personnel, and weather level field, respectively used to store the optional level corresponding to the hazardous construction behavior; and a first correction coefficient field, used to store the first correction coefficient corresponding to each level. The first association table is illustrated in Table 1 below:

[0075] Table 1 First Association Table

[0076]

[0077] In step S6, the second association table includes at least: a construction stage identifier field, used to store the unique identifier of each construction stage; a construction stage initial hazard coefficient field, used to store the initial hazard coefficient for each construction stage; a construction atmosphere level field, a tower construction experience level field for construction personnel, and a weather level field, used to store the optional levels corresponding to the construction stage; and a third correction coefficient field, used to store the third correction coefficient corresponding to each level. The second association table is illustrated in Table 1 below:

[0078] The second association table is shown in Table 2 below as an illustration:

[0079] Table 2 Second Relationship Table

[0080]

[0081] In this embodiment, the video streams from each camera position are analyzed to determine their current construction stage, current construction atmosphere, current hazardous construction behavior, the experience of the current construction personnel, and current weather conditions. For video streams exhibiting hazardous construction behavior, a first correction coefficient for the current construction stage, the first correction coefficient for the current construction atmosphere, the first correction coefficient for the current construction personnel's experience in pole construction, and the initial risk coefficient for the current hazardous construction behavior are obtained from a preset first association table. This first correction coefficient, along with the first correction coefficients for the current construction stage, the current construction atmosphere, the current construction personnel's experience in pole construction, and the current weather conditions, is then used to correct the initial risk coefficient of the current hazardous construction behavior, resulting in a first risk coefficient. This multi-dimensional correction of the initial risk coefficient for hazardous construction behavior allows for a more accurate assessment of the actual risk level of the current hazardous construction behavior. The monitoring videos from each camera position exhibiting hazardous construction behavior are arranged in descending order of the first risk coefficient; that is, for camera positions where violations have occurred, those with higher risk levels are ranked higher, enabling monitoring personnel to quickly identify the actual risk levels between different hazardous behaviors and to quickly identify high-risk operations.

[0082] For video streams without dangerous construction activities, a fourth correction coefficient for the current construction stage is obtained from a pre-set second association table to correct the initial hazard coefficient of the current construction stage, resulting in a second hazard coefficient. This fourth correction coefficient includes a third correction coefficient based on the current construction atmosphere, the current construction personnel's experience in tower construction, and the current weather conditions. This multi-dimensional correction of the initial hazard coefficient for potential construction stages allows for a more accurate assessment of the potential hazard level. Monitoring videos from each camera position without dangerous construction activities are arranged in descending order of the second hazard coefficient; that is, the higher the potential hazard level, the earlier it appears. This allows monitoring personnel to quickly identify the potential hazard levels at different construction stages. All sorted monitoring videos are then output to a backend monitoring screen, with videos exhibiting dangerous construction activities placed before those without. This transforms the entire monitoring screen into a real-time updated "risk heat map"—the top left corner shows the most urgent, already violated scenes, while the right side shows normal operations with the highest potential hazard. Monitoring personnel do not need to cycle through each screen; they can simply handle issues from top to bottom according to the order of arrangement, greatly shortening the response chain from risk emergence to intervention and improving regulatory efficiency.

[0083] In one embodiment, when multiple dangerous construction behaviors are identified at the same camera position, the monitoring videos are sorted according to the highest first danger coefficient as the first danger coefficient of that camera position.

[0084] In one embodiment, the step of obtaining a second correction coefficient for the current hazardous construction behavior from a preset first association table to correct the initial hazard coefficient of the current hazardous construction behavior, and obtaining a first hazard coefficient, includes:

[0085] The initial risk factor of the current hazardous construction behavior is corrected according to the following model;

[0086] R1 = a1 × a2 × a3 × a4 × R0;

[0087] Wherein, R1 is the first risk factor, R0 is the initial risk factor of the current dangerous construction behavior, a1 is the first correction factor of the current construction stage, a2 is the first correction factor of the current construction atmosphere, a3 is the first correction factor of the current construction personnel's tower construction experience, and a4 is the first correction factor of the current weather.

[0088] In this embodiment of the application, the initial risk coefficient of the current dangerous construction behavior is corrected by the model, and the resulting first risk coefficient can more accurately reflect the degree of danger of the current dangerous construction behavior.

[0089] In one embodiment, the step of obtaining the fourth correction coefficient of the current construction stage from a preset second association table to correct the initial hazard coefficient of the current construction stage and obtain the second hazard coefficient includes:

[0090] The initial risk factor for the current construction stage is corrected according to the following model to obtain the second risk factor;

[0091] R4 = a5 × a6 × a7 × R3;

[0092] Among them, R4 is the second risk factor, R3 is the initial risk factor of the current construction stage, a5 is the third correction factor of the current construction atmosphere, a6 is the third correction factor of the current construction personnel's tower construction experience, and a7 is the third correction factor of the current weather.

[0093] In this embodiment of the application, the initial hazard factor of the current construction stage is corrected by the model, and the resulting second hazard factor can more accurately reflect the potential hazards of the current construction stage.

[0094] In one embodiment, the first correction coefficient is calculated by the following method:

[0095] Collect the first construction history data of the tower; wherein, the first construction history data of the tower includes the identification of dangerous construction behavior, the initial risk coefficient of dangerous construction behavior, the construction stage, the construction atmosphere, the tower construction experience of construction personnel, the weather, and the corresponding first actual risk results;

[0096] The first construction history data of the tower is preprocessed, including quantifying the first actual risk result and grouping it according to the combination of dangerous construction behavior type, construction stage, construction atmosphere, tower construction experience of construction personnel and weather conditions.

[0097] The first multiple regression model is constructed, which includes taking the quantified first actual risk result as the dependent variable and taking the first correction coefficient of the construction stage, the first correction coefficient of the construction atmosphere, the first correction coefficient of the construction personnel's tower construction experience, and the first correction coefficient of the weather as independent variables.

[0098] The preprocessed historical data of the first construction of the tower is divided into a training set and a test set. The first multivariate regression model is trained using the training set. The first correction coefficients that minimize the sum of squared residuals between the predicted value of the first multivariate regression model and the quantified first actual risk result are solved by the least squares method.

[0099] The first multivariate regression model trained is validated using a test set. If the prediction error is within a preset range, the first correction coefficients obtained are determined as the final first correction coefficients. If the prediction error is not within the preset range, the data samples are adjusted until the first correction coefficients that meet the error requirements are obtained.

[0100] In this embodiment, the first actual risk outcome is quantified. Specifically, based on the actual losses caused by dangerous behaviors during tower construction, the risk outcome is divided into five levels (minor consequences, general consequences, relatively serious consequences, serious consequences, and extremely serious consequences), each level corresponding to a fixed score. The higher the score, the more severe the outcome. The constructed multiple regression model is as follows:

[0101] Y1=k1×(a1×a2×a3×a4×R0);

[0102] Where Y1 is the first actual risk result after quantification, R0 is the initial risk coefficient of the current dangerous construction behavior, k1 is the first regression coefficient, which is a constant term, a1 is the first correction coefficient of the current construction stage, a2 is the first correction coefficient of the current construction atmosphere, a3 is the first correction coefficient of the current construction personnel's tower construction experience, and a4 is the first correction coefficient of the current weather.

[0103] In this embodiment, by collecting historical construction data of the towers and quantifying the actual risk results, the originally vague risk concept can be transformed into specific numerical values, laying the foundation for subsequently obtaining the first correction coefficients. By comprehensively considering the impact of four factors on hazardous construction behavior—construction stage, construction atmosphere, tower construction experience of construction personnel, and weather—the accuracy of risk assessment is improved. A first multiple regression model is constructed using historical data, and the optimal first correction coefficients are solved using the least squares method. This data-driven approach allows the model to be trained and optimized based on actual conditions, improving its accuracy and reliability. As the data sample is continuously increased and updated, the model can be continuously optimized and improved. By dividing the data into training and testing sets, the model is trained and validated. If the prediction error is not within the preset range, the data sample is adjusted until the first correction coefficients that meet the error requirements are obtained. This mechanism ensures the effectiveness and stability of the model, enabling timely identification and improvement of model problems, and ensuring the accuracy of risk assessment.

[0104] In one embodiment, the third correction coefficient is calculated by the following method:

[0105] Collect second construction history data for the towers; wherein, the second construction history data for the towers includes the construction stage, the initial risk factor of the construction stage, the construction atmosphere, the tower construction experience of the construction personnel, the weather, and the corresponding second actual risk results;

[0106] The second construction history data of the tower is preprocessed, including quantifying the second actual risk result and grouping it according to the combination of construction stage, construction atmosphere, tower construction experience of construction personnel and weather conditions.

[0107] A second multiple regression model was constructed, which included using the quantified second actual risk result as the dependent variable and the third correction coefficient of the construction atmosphere, the third correction coefficient of the construction personnel's tower construction experience, and the third correction coefficient of the weather as independent variables.

[0108] The preprocessed second construction history data of the tower is divided into a training set and a test set. The second multivariate regression model is trained using the training set. The third correction coefficients that minimize the sum of squared residuals between the predicted value of the second multivariate regression model and the quantified second actual risk result are solved by the least squares method.

[0109] The trained second multiple regression model is validated using a test set. If the prediction error is within a preset range, the obtained third correction coefficients are determined as the final third correction coefficients. If the prediction error is not within a preset range, the data samples are adjusted until the third correction coefficients that meet the error requirements are obtained.

[0110] In this embodiment, the second actual risk outcome is quantified. Specifically, based on the actual losses caused by dangerous behaviors during tower construction, the risk outcome is divided into five levels (minor consequences, general consequences, relatively serious consequences, serious consequences, and extremely serious consequences), each level corresponding to a fixed score. The higher the score, the more severe the outcome. The constructed multiple regression model is as follows:

[0111] Y2 = k2 × (a4 × a5 × a6 × R3);

[0112] Wherein, Y2 is the quantified second actual risk result, k2 is the second regression coefficient, which is a constant term, R3 is the initial risk coefficient of the current construction stage, a5 is the third correction coefficient of the current construction atmosphere, a6 is the third correction coefficient of the current construction personnel's tower construction experience, and a7 is the third correction coefficient of the current weather.

[0113] In this embodiment, by collecting historical data on tower construction and quantifying the second actual risk result, the originally vague risk concept can be transformed into concrete numerical values, laying the foundation for subsequently obtaining the third correction coefficients. By comprehensively considering the impact of four factors on the construction stage—construction atmosphere, tower construction experience of construction personnel, and weather—the accuracy of risk assessment is improved. A second multiple regression model is constructed using historical data, and the optimal third correction coefficient is solved using the least squares method. This data-driven approach allows the model to be trained and optimized based on actual conditions, improving its accuracy and reliability. As the data sample is continuously increased and updated, the model can be continuously optimized and improved. By dividing the data into training and testing sets, the model is trained and validated. If the prediction error is not within the preset range, the data sample is adjusted until the first correction coefficients that meet the error requirements are obtained. This mechanism ensures the effectiveness and stability of the model, enabling timely identification and improvement of model problems, and ensuring the accuracy of risk assessment.

[0114] In one embodiment, the step of identifying the current construction stage based on the video stream of the tower construction site includes:

[0115] S31. Preprocess the video stream of the tower construction site to obtain the preprocessed tower video stream;

[0116] S32. Extract key frames from the preprocessed tower video stream; wherein, extracting key frames includes filtering out frames with significant motion changes as key frames by detecting motion changes between video frames using optical flow method;

[0117] S33. Input the extracted keyframes into a pre-trained spatiotemporal feature extraction model to obtain spatiotemporal features; wherein, the spatiotemporal feature extraction model includes a cascaded 3D convolutional neural network and a bidirectional long short-term memory network;

[0118] S34. By using the Transformer cross-attention mechanism, the spatiotemporal features are fused with the IoT data from the tower construction site in the channel dimension to obtain the fused features;

[0119] S35. Input the fused feature vector into the Softmax classifier and output the current tower construction stage category. The category includes at least: foundation excavation and grounding trench construction, tower material ground pre-assembly and transportation, tower leg / bottom section assembly, tower body segment / segment high-level hoisting, crossarm overall hoisting, permanent guy wire tensioning and pole removal, bolt final tightening and cleaning.

[0120] In step S31, the video stream of the tower construction site is preprocessed, specifically including noise reduction, contrast enhancement, and lens distortion removal.

[0121] In step S32, motion changes between video frames are detected using optical flow, and frames with significant motion changes are selected as keyframes. Specifically, optical flow vectors between adjacent frames are calculated using optical flow. For the optical flow vector of each pixel, its magnitude is calculated as the motion intensity. A motion intensity threshold is set based on actual construction scenarios and experience. When the motion intensity of a frame exceeds this threshold, the frame is considered to contain significant motion changes and is marked as a keyframe. By extracting keyframes, the amount of data is reduced, processing efficiency is improved, and key information in the construction process is preserved.

[0122] In step S33, the spatiotemporal feature extraction model is trained as follows: A large amount of video data from multiple tower construction sites at different construction stages is collected. These stages include foundation excavation and grounding trench construction, tower material pre-assembly and transportation on the ground, tower leg / base section assembly, high-level hoisting of tower body segments / sections, overall hoisting of crossarms, permanent guy wire tensioning and pole dismantling, final bolt tightening, and cleaning and defect elimination. Then, the video data is labeled to indicate the construction stage corresponding to each video segment. Keyframes are extracted from each video segment and input into a cascaded spatiotemporal feature extraction model using a 3D convolutional neural network and a bidirectional long short-term memory network to extract the spatiotemporal features of the tower construction. By cascading the 3D convolutional neural network and the bidirectional long short-term memory network, the model can simultaneously capture local features and global dynamics, providing a more comprehensive spatiotemporal feature representation. There are numerous dynamic changes during tower construction, such as the hoisting of tower materials and the installation of crossarms. The combination of 3D-CNN and Bi-LSTM can effectively capture these dynamic changes and improve the model's recognition ability. Furthermore, the combination of 3D convolutional neural networks and bidirectional long short-term memory networks can improve the model's adaptability to complex construction environments, such as changes in lighting and occlusion. Then, through the Transformer cross-attention mechanism, the spatiotemporal features extracted during tower construction are fused with IoT data from the tower construction site in the channel dimension, enhancing the model's adaptability and robustness to complex construction environments, resulting in fused features. The fused feature vector is then input into a Softmax classifier, which outputs the predicted probability distribution corresponding to the current tower construction stage. Based on the one-hot encoding of the real labels and the predicted probability distribution output by the model, the cross-entropy loss is calculated. Based on the cross-entropy loss, the model parameters are updated through the backpropagation algorithm until the training stop condition is met, resulting in a trained spatiotemporal feature extraction model.

[0123] In step S34, IoT data refers to data collected from various sensors and devices at the tower construction site. This data provides real-time status information of the construction site, which, combined with video data, enhances the accuracy of classification of construction stages. For example, IoT data includes the slewing angle and lifting height of the tower crane. When observing the segmented hoisting of the tower body in the video, combining the tower crane's slewing angle and lifting height from the IoT data can more accurately determine the specific stage of tower body hoisting. For instance, when the video shows a tower body segment being lifted and moved to the installation position, the IoT data shows that the tower crane's slewing angle and lifting height match this action, confirming that the current stage is the high-level hoisting of the tower body segment, thus improving the accuracy of classification. As another example, IoT data includes excavator operation data. When observing foundation excavation in the video, combining the excavator's operation data from the IoT data can more accurately determine the specific stage of foundation excavation. For instance, when the video shows the excavator digging, the IoT data shows that the excavator's operating status is "working" and the operating time matches the excavation task, confirming that the current stage is the foundation excavation stage.

[0124] In step S34, the extracted spatiotemporal features during tower construction are fused with IoT data from the tower construction site at the channel dimension using the Transformer cross-attention mechanism. This demonstrates high robustness to environmental disturbances under three extreme scenarios: sudden changes in lighting, heavy fog, and partial occlusion. Furthermore, compared to single-video processing, the addition of IoT-cross-attention fusion improves the accuracy of construction phase classification.

[0125] In one embodiment, to accurately identify the construction atmosphere, this application embodiment includes the following steps for identifying the current construction atmosphere based on the video stream of the tower construction site:

[0126] The current construction atmosphere score is calculated based on the following model;

[0127]

[0128] b jσ =a jσ ×β+δ;

[0129]

[0130] Where X is the current construction atmosphere score, and Q is... ij For the current construction status score of the i-th and j-th construction worker, m is the number of construction workers identified, n is the number of evaluations, and b is the number of evaluations. jσ Let a be the probability value of the σ-th emotion corresponding to the j-th construction worker. jσLet w be the σ-th feature value extracted from the face of the j-th construction worker, where σ = 1, 2, ..., s, corresponding to the recognition dimensions of s kinds of emotions, β and δ are parameters that adjust the linear relationship between emotional features and construction worker emotions. jσ Let f(·) be the weight value of the σ-th facial feature of the j-th construction worker; f(·) represents the mapping function used to map the result to a preset construction status score.

[0131] The current construction atmosphere score is mapped to a construction atmosphere type.

[0132] In this embodiment of the application, the score of the current construction atmosphere is mapped to the construction atmosphere type according to a preset rule.

[0133] like Figure 2 As shown in the figure, this application embodiment also provides a tower construction safety monitoring device, the device comprising:

[0134] Video stream acquisition module 1 is used to acquire video streams from the construction site of each tower in real time;

[0135] The identification module 2 is used to identify the current construction stage, current construction atmosphere, current dangerous construction behavior, and current tower construction experience of the construction personnel based on the video stream of each tower construction site.

[0136] Meteorological acquisition module 3 is used to synchronously acquire the current weather conditions at the construction sites of each tower location;

[0137] The first correction module 4 is used to obtain a second correction coefficient for the current dangerous construction behavior from a preset first association table for video streams containing dangerous construction behavior, and to correct the initial danger coefficient of the current dangerous construction behavior to obtain a first danger coefficient; wherein the second correction coefficient includes: the first correction coefficient of the current construction stage, the current construction atmosphere, the current tower construction experience of the construction personnel, and the current weather.

[0138] The first arrangement module 5 is used to arrange the monitoring videos of each machine position with dangerous construction behavior in descending order according to the first hazard coefficient;

[0139] The second correction module 6 is used to obtain the fourth correction coefficient of the current construction stage from the preset second association table for video streams without dangerous construction behavior, and to correct the initial danger coefficient of the current construction stage to obtain the second danger coefficient; wherein, the fourth correction coefficient includes: the current construction atmosphere, the current construction personnel's tower construction experience and the current weather conditions.

[0140] The second arrangement module 7 is used to arrange the monitoring videos of each machine position that do not have dangerous construction behavior in descending order according to the second hazard coefficient;

[0141] Output module 8 is used to output all sorted monitoring videos to the background monitoring screen for display; among them, monitoring videos with dangerous construction behavior are arranged before monitoring videos without dangerous construction behavior.

[0142] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media provided in this application and in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0143] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0144] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for monitoring the safety of tower construction, characterized in that, The method includes: Real-time acquisition of video streams from the construction sites of each tower; Based on the video streams of each tower construction site, the current construction stage, current construction atmosphere, current dangerous construction behavior, and current tower construction experience of the construction personnel are identified. Simultaneously acquire the current weather conditions at each tower construction site; For video streams containing dangerous construction activities, a second correction coefficient for the current dangerous construction activity is obtained from a preset first association table to correct the initial risk coefficient of the current dangerous construction activity, resulting in a first risk coefficient; wherein the second correction coefficient includes: the first correction coefficients for the current construction stage, the current construction atmosphere, the current tower construction experience of the construction personnel, and the current weather conditions; The monitoring videos of each machine location exhibiting dangerous construction activities are arranged in descending order of the highest hazard factor. For video streams without dangerous construction activities, the fourth correction coefficient for the current construction stage is obtained from the preset second association table to correct the initial risk coefficient for the current construction stage, thus obtaining the second risk coefficient. The fourth correction coefficient includes the third correction coefficient based on the current construction atmosphere, the current construction personnel's experience in tower construction, and the current weather conditions. The monitoring videos of each machine position that do not involve dangerous construction activities are arranged in descending order of the second hazard factor. All sorted surveillance videos are output to the backend monitoring screen for display; among them, surveillance videos with dangerous construction activities are arranged before surveillance videos without dangerous construction activities. The steps for identifying the current construction stage based on the video stream at the tower construction site include: The video stream from the tower construction site is preprocessed to obtain a preprocessed tower video stream; Extract keyframes from the preprocessed tower video stream; the keyframe extraction includes filtering out frames with significant motion changes as keyframes by detecting motion changes between video frames using optical flow; The extracted keyframes are input into a pre-trained spatiotemporal feature extraction model to obtain spatiotemporal features; wherein, the spatiotemporal feature extraction model includes a cascaded 3D convolutional neural network and a bidirectional long short-term memory network; By using the Transformer cross-attention mechanism, the spatiotemporal features are fused with the IoT data from the tower construction site in the channel dimension to obtain the fused features; The fused feature vector is input into the Softmax classifier, which outputs the current tower construction stage category. The category includes at least: foundation excavation and grounding trench construction, tower material ground pre-assembly and transportation, tower leg / bottom section assembly, tower body segment / segment high-level hoisting, crossarm overall hoisting, permanent guy wire tensioning and pole removal, bolt final tightening and cleaning.

2. The method for monitoring the safety of tower construction according to claim 1, characterized in that, The steps of obtaining the first hazard coefficient by retrieving the second correction coefficient of the current hazardous construction behavior from the preset first association table to correct the initial hazard coefficient of the current hazardous construction behavior include: The initial risk factor of the current hazardous construction behavior is corrected according to the following model; ； in, The highest risk factor is... This represents the initial risk factor for the current hazardous construction activity. This is the first correction factor for the current construction phase. The first correction factor for the current construction atmosphere. The first correction factor for the current construction workers' experience in tower construction. This is the first correction factor for the current weather conditions.

3. The method for monitoring the safety of tower construction according to claim 1, characterized in that, The step of obtaining the fourth correction coefficient for the current construction stage from the preset second association table to correct the initial hazard coefficient for the current construction stage and obtain the second hazard coefficient includes: The initial risk factor for the current construction stage is corrected according to the following model to obtain the second risk factor; ； in, The second highest risk factor. This represents the initial risk factor for the current construction phase. The third correction factor for the current construction atmosphere. The third correction factor for the current construction workers' experience in tower construction. This is the third correction factor for the current weather conditions.

4. The method for monitoring the safety of tower construction according to claim 1, characterized in that, The first correction factor is calculated using the following method: Collect the first construction history data of the tower; wherein, the first construction history data of the tower includes the identification of dangerous construction behavior, the initial risk coefficient of dangerous construction behavior, the construction stage, the construction atmosphere, the tower construction experience of construction personnel, the weather, and the corresponding first actual risk results; The first construction history data of the tower is preprocessed, including quantifying the first actual risk result and grouping it according to the combination of dangerous construction behavior type, construction stage, construction atmosphere, tower construction experience of construction personnel and weather conditions. The first multiple regression model is constructed, which includes taking the quantified first actual risk result as the dependent variable and taking the first correction coefficient of the construction stage, the first correction coefficient of the construction atmosphere, the first correction coefficient of the construction personnel's tower construction experience, and the first correction coefficient of the weather as independent variables. The preprocessed historical data of the first construction of the tower is divided into a training set and a test set. The first multivariate regression model is trained using the training set. The first correction coefficients that minimize the sum of squared residuals between the predicted value of the first multivariate regression model and the quantified first actual risk result are solved by the least squares method. The first multivariate regression model trained is validated using a test set. If the prediction error is within a preset range, the first correction coefficients obtained are determined as the final first correction coefficients. If the prediction error is not within the preset range, the data samples are adjusted until the first correction coefficients that meet the error requirements are obtained.

5. The method for monitoring the safety of tower construction according to claim 1, characterized in that, The third correction factor is calculated using the following method: Collect second construction history data for the towers; wherein, the second construction history data for the towers includes the construction stage, the initial risk factor of the construction stage, the construction atmosphere, the tower construction experience of the construction personnel, the weather, and the corresponding second actual risk results; The second construction history data of the tower is preprocessed, including quantifying the second actual risk result and grouping it according to the combination of construction stage, construction atmosphere, tower construction experience of construction personnel and weather conditions. A second multiple regression model was constructed, which included using the quantified second actual risk result as the dependent variable and the third correction coefficient of the construction atmosphere, the third correction coefficient of the construction personnel's tower construction experience, and the third correction coefficient of the weather as independent variables. The preprocessed second construction history data of the tower is divided into a training set and a test set. The second multivariate regression model is trained using the training set. The third correction coefficients that minimize the sum of squared residuals between the predicted value of the second multivariate regression model and the quantified second actual risk result are solved by the least squares method. The trained second multiple regression model is validated using a test set. If the prediction error is within a preset range, the obtained third correction coefficients are determined as the final third correction coefficients. If the prediction error is not within a preset range, the data samples are adjusted until the third correction coefficients that meet the error requirements are obtained.

6. The method for monitoring the safety of tower construction according to claim 1, characterized in that, The steps for identifying the current construction atmosphere based on video stream at the tower construction site include: The current construction atmosphere score is calculated based on the following model; ； ； ； ； Where X is the current construction atmosphere score, The current construction status score is given to the i-th and j-th construction worker, where m is the number of identified construction workers and n is the number of evaluations. Let be the probability value of the σ-th emotion corresponding to the j-th construction worker. Let σ be the σ-th feature value extracted from the face of the j-th construction worker, where σ = 1, 2, ..., s, corresponding to the recognition dimensions of s kinds of emotions. and It is a parameter that adjusts the linear relationship between emotional characteristics and construction emotions. Let σ be the weight value of the σ-th facial feature of the j-th construction worker; This represents a mapping function used to map the results to a preset construction status score. The current construction atmosphere score is mapped to a construction atmosphere type.

7. A safety monitoring device for tower construction, characterized in that, The device includes: The video stream acquisition module is used to acquire video streams from the construction site of each tower in real time. The identification module is used to identify the current construction stage, current construction atmosphere, current dangerous construction behavior, and current tower construction experience of the construction personnel based on the video stream of each tower construction site. The weather acquisition module is used to synchronously acquire the current weather conditions at the construction sites of each tower location; The first correction module is used to obtain a second correction coefficient for the current dangerous construction behavior from a preset first association table for video streams containing dangerous construction behavior, and to correct the initial danger coefficient of the current dangerous construction behavior to obtain a first danger coefficient; wherein the second correction coefficient includes: the first correction coefficient of the current construction stage, the current construction atmosphere, the current tower construction experience of the construction personnel, and the current weather. The first sorting module is used to sort the monitoring videos of each machine position that have dangerous construction behavior in descending order according to the first hazard factor; The second correction module is used to obtain the fourth correction coefficient of the current construction stage from the preset second association table for video streams without dangerous construction behavior, and to correct the initial danger coefficient of the current construction stage to obtain the second danger coefficient; wherein, the fourth correction coefficient includes: the current construction atmosphere, the current tower construction experience of the construction personnel and the current weather conditions. The second sorting module is used to sort the monitoring videos of each machine position that do not have dangerous construction behavior in descending order according to the second hazard factor; The output module is used to output all sorted surveillance videos to the background monitoring screen for display; among them, surveillance videos with dangerous construction activities are arranged before surveillance videos without dangerous construction activities; The identification of the current construction stage based on the video stream at the tower construction site includes: The video stream from the tower construction site is preprocessed to obtain a preprocessed tower video stream; Extract keyframes from the preprocessed tower video stream; the keyframe extraction includes filtering out frames with significant motion changes as keyframes by detecting motion changes between video frames using optical flow; The extracted keyframes are input into a pre-trained spatiotemporal feature extraction model to obtain spatiotemporal features; wherein, the spatiotemporal feature extraction model includes a cascaded 3D convolutional neural network and a bidirectional long short-term memory network; By using the Transformer cross-attention mechanism, the spatiotemporal features are fused with the IoT data from the tower construction site in the channel dimension to obtain the fused features; The fused feature vector is input into the Softmax classifier, which outputs the current tower construction stage category. The category includes at least: foundation excavation and grounding trench construction, tower material ground pre-assembly and transportation, tower leg / bottom section assembly, tower body segment / segment high-level hoisting, crossarm overall hoisting, permanent guy wire tensioning and pole removal, bolt final tightening and cleaning.

Images