Building construction safety management method based on image recognition

Abstract

The invention relates to the technical field of construction management, and particularly discloses a building construction safety management method based on image recognition, and the method comprises the steps: deploying a plurality of image collection devices, collecting the image data of a construction site according to a preset time interval or an event triggering mechanism, and carrying out the preprocessing; constructing a classification model of the personnel and the equipment, and performing training; based on the identified personnel and equipment information, detecting whether the personnel have an unsafe behavior and whether the equipment has an abnormal state; evaluating the safety risk of the construction site according to the type, severity and influence range of the potential safety hazard, and establishing a risk evaluation model; the risk level is determined by comprehensively considering the probability of occurrence of hidden dangers and possible consequence factors; and when the potential safety hazard is detected and the risk level reaches an early warning threshold value, early warning information is sent out. According to the method, multiple factors are comprehensively considered, so that risk assessment is more scientific and accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of construction management, and particularly relates to a building construction safety management method based on image recognition. Background Art

[0002] The construction environment of buildings is extremely complex, with a large number of construction workers, various types of equipment, and many construction links converging therein. In such a complex environment, potential safety hazards are everywhere. The traditional building construction safety management mode mainly relies on manual inspections. However, manual inspections have significant drawbacks. Their efficiency is relatively low, and they are extremely vulnerable to human factors, such as the fatigue and negligence of staff. This results in difficulties in comprehensively and timely discovering and handling safety hazards, and many safety hazards may thus be overlooked.

[0003] With the booming development of computer technology and artificial intelligence technology, image recognition technology has been widely applied in many fields. In the aspect of building construction safety management, with the help of image recognition technology, real-time monitoring of the construction site can be carried out and intelligent analysis can be performed. In this way, safety hazards can be quickly and accurately discovered, greatly improving the efficiency and accuracy of safety management. However, the existing safety management methods based on image recognition currently expose many problems in practical applications. For example, the recognition accuracy cannot reach the ideal state, the real-time performance is poor, and the adaptability to complex construction scenarios is also insufficient. These defects make it impossible to effectively meet the actual needs of building construction safety management and urgently need further optimization and improvement. Summary of the Invention

[0004] The purpose of the present invention is to provide a building construction safety management method based on image recognition to solve the above problems existing in the prior art, improve the level of building construction safety management, and reduce the risk of safety accidents.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A building construction safety management method based on image recognition includes the following steps:

[0007] Deploy a plurality of image acquisition devices at the building construction site, collect image data of the construction site at preset time intervals or event trigger mechanisms, and preprocess the collected image data;

[0008] Use deep learning algorithms to identify personnel and construction equipment in the preprocessed images, construct classification models of personnel and equipment, and train them with the labeled image data;

[0009] Based on the identified personnel and equipment information, conduct safety hazard detection on the image, and according to the set corresponding thresholds and rules, detect whether there are unsafe behaviors of personnel and whether there are abnormal states of equipment;

[0010] Evaluate the safety risks at the construction site according to the type, severity, and impact scope of the safety hazards, and establish a risk assessment model; comprehensively consider the probability of the occurrence of the hazards and the factors of possible consequences, and determine the risk level; when a safety hazard is detected and the risk level reaches the warning threshold, send a warning message.

[0011] As a further solution of the present invention: the deep learning algorithm is a convolutional neural network CNN, the different personnel include workers and management personnel, and the various construction equipment includes cranes, tower cranes, and excavators.

[0012] As a further solution of the present invention: in the risk assessment and warning, the risk level R of the safety hazard is jointly determined by the probability P of the occurrence of the hazard, the severity C of the possible consequences, and the impact scope S, and the calculation formula is:

[0013] R = w1P + w2C + w3S;

[0014] Among them, w1, w2, and w3 are the weight coefficients of the probability of the occurrence of the hazard, the severity of the consequences, and the impact scope respectively, and w1 + w2 + w3 = 1.

[0015] As a further solution of the present invention: the probability P of the occurrence of the hazard is obtained through historical data statistics. In the past n times of the same or similar construction scenarios, the number of times this hazard occurred is m, then P = m / n; the severity C of the consequences is quantified according to the possible economic losses and casualties caused by the accident.

[0016] As a further solution of the present invention: the impact scope S is quantified according to the area A of the construction area that the hazard may affect and the number N of affected personnel, and the expression is S = αA + βN, where α and β are the corresponding weight coefficients, and α + β = 1.

[0017] As a further solution of the present invention: it also includes recording the identified safety hazards, risk assessment results, and warning messages, establishing a safety management database, and regularly analyzing the data in the database. The specific steps are as follows:

[0018] Let the statistical time period be T, and divide this time period evenly into k time intervals Δt i , i = 1, 2,..., k, and the number of safety hazards occurring in each time interval is l i , then the total number of occurrences of safety hazards within the time period T The average occurrence frequency f of safety hazards = Q / T.

[0019] As a further solution of the present invention: the method of sending the early warning information includes notifying relevant management personnel by text message, displaying the warning information on a display screen at the construction site, and sounding an alarm.

[0020] A construction safety management system based on image recognition, comprising:

[0021] Data acquisition module: used to deploy multiple image acquisition devices at the construction site to collect image data of the construction site according to preset time intervals or event trigger mechanisms;

[0022] Preprocessing module: used to preprocess the collected image data, including image denoising, contrast enhancement and brightness adjustment;

[0023] Recognition module: Use deep learning algorithms to identify personnel and construction equipment in pre-processed images, build classification models for personnel and equipment, and train them with a large amount of annotated image data so that the model can identify different personnel and various types of construction equipment at the construction site;

[0024] Safety hazard detection module: Based on the identified personnel and equipment information, the image is detected for safety hazards. The personnel are detected for unsafe behaviors such as not wearing safety helmets, not wearing safety belts, and illegally entering dangerous areas. The equipment is detected for abnormal conditions. The corresponding thresholds and rules are set to judge the status of personnel and equipment in the image and identify potential safety hazards.

[0025] Risk assessment and early warning module: Assess the safety risks of the construction site according to the type, severity and impact scope of the safety hazards, establish a risk assessment model, comprehensively consider the probability of hidden dangers and possible consequences, and determine the risk level; when a safety hazard is detected and the risk level reaches the early warning threshold, issue a warning message.

[0026] Beneficial effects of the present invention:

[0027] The present invention uses image recognition technology to achieve real-time, automated monitoring of construction sites. Compared with traditional manual inspections, it can discover potential safety hazards more quickly and comprehensively, greatly improving the efficiency of safety management. It uses deep learning algorithms to identify personnel and equipment and detect potential safety hazards with high recognition accuracy, which can effectively avoid missing potential safety hazards due to human negligence.

[0028] A comprehensive risk assessment and early warning mechanism has been established to issue early warnings in a timely manner according to the severity of potential safety hazards, so that managers can take prompt measures to reduce the risk of safety accidents. The risk assessment model uses quantitative mathematical formulas to comprehensively consider multiple factors, making risk assessment more scientific and accurate.

[0029] Recording and analyzing safety management data provides data support for formulating targeted safety management measures, helps continuously optimize safety management strategies, and improves the overall level of construction safety management. The statistical formula for the occurrence frequency of safety hazards can intuitively reflect the occurrence of safety hazards and provide a strong basis for management decisions. Brief Description of the Drawings

[0030] The present invention will be further described below with reference to the accompanying drawings.

[0031] Figure 1 It is a schematic flow chart of a construction safety management method based on image recognition according to the present invention. Detailed Embodiments

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0033] Construction sites have always been complex and challenging. A large number of construction workers shuttle through them, construction equipment is dazzling, and various construction links are intertwined, with one move affecting the whole situation. Such a complex situation makes safety accident hazards follow like a shadow, and serious consequences may occur with a slight carelessness. In the past, construction safety management mainly relied on manual inspections, but this method has obvious shortcomings. On the one hand, the efficiency of manual inspections is not high. Facing a large construction site, it is difficult for staff to complete a comprehensive inspection in a short time. On the other hand, the interference of human factors is extremely large. Once the staff is fatigued, their attention will decline, or due to carelessness, it is very easy to miss many safety hazards and it is difficult to detect and handle them comprehensively and in a timely manner.

[0034] Nowadays, computer technology and artificial intelligence technology have advanced by leaps and bounds. Image recognition technology has emerged in many fields, and this technology has also been introduced into the field of construction safety management. By using image recognition technology, it is possible to implement 24-hour real-time monitoring of the construction site, and it can quickly and accurately identify safety hazards through intelligent analysis, greatly improving the efficiency and accuracy of safety management. However, the ideal is very plump, but the reality is very skinny. In the process of practice, the current safety management method based on image recognition still exposes many problems. For example, in complex and changeable construction scenarios, the recognition accuracy is not satisfactory, and the real-time performance is also difficult to guarantee, and it is impossible to timely feedback hazard information, which restricts its actual application effect in construction safety management and urgently needs to be deeply optimized and improved to meet the real needs of construction safety management.

[0035] Please refer to Figure 1 As shown, the present invention is a construction safety management method based on image recognition, including the following steps:

[0036] Data collection and preprocessing: Deploy multiple image collection devices, such as high-definition cameras, at the construction site, and collect image data of the construction site according to a preset time interval or event trigger mechanism. Perform preprocessing on the collected image data, including operations such as image denoising, contrast enhancement, and brightness adjustment, to improve the quality of the images and facilitate subsequent feature extraction and recognition analysis.

[0037] In a specific implementation, high-definition cameras can be installed at each key position of the construction site, such as entrances and exits, construction areas, and high-altitude operation points, and their shooting angles and coverage ranges can be reasonably set to ensure that comprehensive image data reflecting the construction site situation can be collected. Set the image collection device to automatically collect images at regular intervals (such as every 5 minutes), or trigger collection when abnormal actions of personnel or equipment are detected.

[0038] Personnel and equipment recognition: Use deep learning algorithms, such as convolutional neural networks (CNNs), to identify personnel and construction equipment in the preprocessed images. Build classification models for personnel and equipment, and train them with a large amount of labeled image data so that the models can accurately identify different personnel (such as workers, managers) and various construction equipment (such as cranes, tower cranes, excavators, etc.) at the construction site.

[0039] In this embodiment, for the collected image data, use an image denoising algorithm to remove noise interference, adopt methods such as histogram equalization to enhance the image contrast, and make the details in the image clearer by adjusting the brightness. Input the preprocessed image into a pre-trained convolutional neural network model, which has been trained with a large amount of image data containing different personnel and equipment, and can accurately identify the personnel and equipment categories in the image.

[0040] Among them, the specific steps for building classification models for personnel and equipment include:

[0041] Collect a large number of construction site images containing personnel and various construction equipment. Images can be collected by installing multiple high-definition cameras at the construction site at regular intervals. At the same time, some data can also be obtained from public image datasets and image resource libraries in related industries to enrich the diversity of the dataset;

[0042] Annotate the collected images to clarify the categories of personnel and equipment in the images. Professional image annotation tools such as LabelImg and RectLabel can be used. The annotation content includes the identity of personnel (such as workers, managers), the type of equipment (such as cranes, tower cranes, excavators, etc.), and the annotation format usually adopts XML, JSON, etc.;

[0043] Divide the annotated dataset into a training set, a validation set, and a test set, generally divided according to the ratio of 70%-15%-15%. The training set is used for model training, the validation set is used to evaluate the performance of the model during training and adjust the hyperparameters of the model, and the test set is used to finally evaluate the generalization ability of the model;

[0044] To increase the diversity of data and the generalization ability of the model, data augmentation operations can be performed on the training set. Common data augmentation methods include image flipping (horizontal flipping, vertical flipping), rotation, scaling, brightness adjustment, contrast adjustment, etc. For example, these operations can be easily implemented using the OpenCV library in Python;

[0045] According to the selected model architecture, define the structure of the model using a deep learning framework; common CNN models include AlexNet, VGG, ResNet, Inception, etc. Among them, ResNet solves the problems of gradient disappearance and gradient explosion during the training of deep neural networks by introducing residual blocks, can train deeper networks, and thus extract more complex features, achieving good results in image classification tasks;

[0046] For classification tasks, the commonly used loss function is the cross-entropy loss function (CrossEntropyLoss). Optimizers can choose Stochastic Gradient Descent (SGD), Adam, etc.;

[0047] Input the data of the training set into the model and perform multiple iterative trainings. In each iteration, calculate the loss between the output of the model and the true label, and then use the optimizer to update the parameters of the model;

[0048] During the training process, regularly use the validation set to evaluate the model, calculate metrics such as the accuracy, recall rate, and F1 value of the model, and adjust the hyperparameters of the model (such as the learning rate, batch size, etc.) according to the evaluation results to prevent the model from overfitting or underfitting;

[0049] After the model training is completed, use the test set to perform a final evaluation of the model to obtain the performance metrics of the model on unseen data and evaluate the generalization ability of the model.

[0050] Safety hazard detection: Based on the identified personnel and equipment information, the image is subjected to safety hazard detection. For personnel, it is detected whether there are unsafe behaviors such as not wearing a safety helmet, not fastening the seat belt, and illegally entering the dangerous area. For equipment, it is detected whether the equipment is in an abnormal state, such as the abnormal angle of the boom of a crane, the wear of the ropes of a tower crane, and the loosening of equipment components. By setting corresponding thresholds and rules, the states of personnel and equipment in the image are judged to identify potential safety hazards. Taking the detection of the abnormal angle of the crane boom as an example, let the normal angle range of the crane boom be [θ min , θ max , and the actually detected boom angle is θ actual . When θ actual < θ min or θ actual > θ max , it is determined that the angle of the crane boom is abnormal. To more accurately judge the degree of abnormality, an abnormality index D θ is defined:

[0051]

[0052] It can be understood that in the detection of personnel not wearing a safety helmet, by analyzing the characteristics of the head area of the personnel in the image, it is judged whether there are the shape and color characteristics of the safety helmet. If the corresponding characteristics are not detected, it is determined that the person is not wearing a safety helmet. For the detection of not fastening the seat belt, after identifying the personnel working at height, it is checked whether there are connection marks of the seat belt around their bodies. When detecting the abnormal state of equipment, for example, for a crane, by identifying information such as the angle, length of the boom and the positions of key components on the equipment, and comparing with the parameters in the normal operating state, it is judged whether it is abnormal using a formula. When the calculated D θ is not 0, it is determined that the angle of the crane boom is abnormal, and the degree of abnormality is evaluated according to the value of D θ .

[0053] Risk assessment and early warning: According to the type, severity and influence scope of safety hazards, the safety risks at the construction site are evaluated. A risk assessment model is established, and various factors are comprehensively considered, such as the probability of the occurrence of hazards and the possible consequences, to determine the risk level. Let the risk level R of the safety hazard be jointly determined by the probability P of the occurrence of the hazard, the severity C of the possible consequences, and the influence scope S, and the risk level is calculated by the method of weighted summation:

[0054] R = w1P + w2C + w3S;

[0055] Among them, w1, w2, and w3 are the weight coefficients of the probability of potential hazard occurrence, the severity of consequences, and the scope of influence respectively, and satisfy w1 + w2 + w3 = 1. The probability P of potential hazard occurrence can be obtained through statistical analysis of historical data. For example, in the past n times of the same or similar construction scenarios, the number of occurrences of this potential hazard is m, then P = m / n. The severity of consequences C can be quantified according to the possible economic losses, casualties, etc. caused by the accident. The scope of influence S can be quantified according to factors such as the area of the construction area that may be affected by the potential hazard and the number of affected personnel. For example, if the area of the affected construction area is A and the number of affected personnel is N, a comprehensive index S = αA + βN is constructed, where α and β are the corresponding weight coefficients and α + β = 1.

[0056] Safety hazards are classified into different levels, such as minor hazards, general hazards, and major hazards. For minor hazards such as not wearing a safety helmet, a lower risk level is set; for hazards such as abnormal crane boom angles that may cause serious consequences, a higher risk level is set. When a safety hazard is detected and the risk level reaches the warning threshold, a warning message containing the location, type, and risk level of the hazard is sent to relevant personnel such as project managers and safety administrators through a short message platform; a hierarchical structure model is constructed to decompose the risk assessment problem into an objective layer (risk level), a criterion layer (probability of potential hazard occurrence, severity of consequences, scope of influence), and a scheme layer (different types of safety hazards). By comparing the relative importance between each criterion, a judgment matrix is constructed to calculate the weights of each criterion.

[0057] When a safety hazard is detected and the risk level reaches the warning threshold, a warning message is sent in a timely manner. The warning message can be sent in various ways, such as notifying relevant management personnel by text message, displaying warning messages on the display screens at the construction site, and sounding alarm sounds.

[0058] Data recording and analysis: Record the identified safety hazards, risk assessment results, and warning messages, and establish a safety management database. Regularly analyze the data in the database to summarize information such as the occurrence patterns, common types, and high-incidence areas and time periods of safety hazards. Let the statistical time period be T (for example, one month), and divide this time period into k small time intervals Δt i i = 1, 2,..., k, and the number of safety hazards occurring in each time interval is l i , then the total number of occurrences of safety hazards within the time period T The average occurrence frequency f of safety hazards = Q / T.

[0059] Store the safety hazard information identified each time, including the time, location, type, personnel and equipment involved in the hazard, as well as the risk assessment results and early warning records, into the safety management database. Calculate the total occurrence times and average occurrence frequencies according to the above formula, and regularly (such as weekly or monthly) conduct statistical analysis on the data in the database, draw charts showing the change of safety hazard occurrence frequencies over time and location, and identify the high-incidence areas and periods of safety hazards, providing a basis for formulating more effective safety management measures.

[0060] Based on the data analysis results, formulate targeted safety management measures, such as strengthening the monitoring of specific areas and periods, providing special safety training for workers, etc., to continuously optimize the safety management strategy.

[0061] Dynamic adjustment and optimization: As the construction progress advances and the construction site conditions change, adjust the position and parameters of the image acquisition equipment in real time to ensure that the image data of the key areas can be comprehensively and accurately collected. According to the types and characteristics of newly emerging safety hazards, update the recognition models of personnel and equipment and the safety hazard detection rules in a timely manner to improve the accuracy and adaptability of image recognition. At the same time, optimize the risk assessment model and early warning threshold according to the implementation effect of the safety management measures to make the safety management more scientific and effective.

[0062] According to the change of construction progress, such as the gradual rise of the building or the transfer of the construction area, adjust the position and shooting angle of the image acquisition equipment in a timely manner to ensure continuous monitoring of the key areas. When new safety hazards brought by new construction equipment or construction technologies appear, collect relevant image data, update and optimize the recognition model and detection rules to improve the adaptability and accuracy of the system. At the same time, adjust the parameters in the risk assessment model and the early warning threshold according to the feedback of the implementation effect of the safety management measures to make the safety management more scientific and reasonable.

[0063] The above has described an embodiment of the present invention in detail, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the patent coverage scope of the present invention.

Claims

1. A building construction safety management method based on image recognition, characterized in that, It includes the following steps: Deploy multiple image acquisition devices at the construction site, collect image data of the construction site at preset time intervals or according to an event trigger mechanism, and preprocess the collected image data; Use deep learning algorithms to identify personnel and construction equipment in the preprocessed images, construct classification models for personnel and equipment, and train them with the labeled image data; Based on the identified personnel and equipment information, detect potential safety hazards in the images, and according to the set corresponding thresholds and rules, detect whether there are unsafe behaviors of personnel and abnormal states of equipment; Evaluate the safety risks of the construction site according to the type, severity and scope of influence of potential safety hazards, establish a risk assessment model; comprehensively consider the probability of occurrence of hazards and the possible consequence factors to determine the risk level; When a potential safety hazard is detected and the risk level reaches the warning threshold, send a warning message.

2. The method for construction safety management based on image recognition according to claim 1, characterized in that, The deep learning algorithm is a convolutional neural network CNN. The different personnel include workers and management personnel, and the various construction equipment includes cranes, tower cranes and excavators.

3. The method for construction safety management based on image recognition according to claim 1, characterized in that, In risk assessment and early warning, the risk level R of potential safety hazards is jointly determined by the probability P of occurrence of hazards, the severity C of possible consequences, and the scope of influence S. The calculation formula is: R = w1P + w2C + w3S; Among them, w1, w2, and w3 are the weight coefficients of the probability of occurrence of hazards, the severity of consequences, and the scope of influence respectively, and w1 + w2 + w3 = 1.

4. The method for safety management of building construction based on image recognition according to claim 3, characterized in that, The probability P of occurrence of hazards is obtained through historical data statistics. In the past n times of the same or similar construction scenarios, the number of times this hazard occurred is m, then P = m / n; the severity C of consequences is quantified according to the possible economic losses and casualties caused by the accident.

5. The method for construction safety management based on image recognition according to claim 3, characterized in that, The scope of influence S is quantified according to the area A of the construction area that the hazard may affect and the number of people N affected. The expression is S = αA + βN, where α and β are the corresponding weight coefficients, and α + β = 1.

6. The method for construction safety management based on image recognition according to claim 1, characterized in that, It also includes recording the identified potential safety hazards, risk assessment results and warning messages, establishing a safety management database, and regularly analyzing the data in the database. The specific steps are as follows: Let the statistical time period be T, and divide this time period evenly into k time intervals Δt i , where i = 1, 2,..., k, and the number of safety hazards occurring in each time interval is l i , then the total number of occurrences of safety hazards within the time period T The average occurrence frequency f of safety hazards is f = Q / T.

7. A method for construction safety management based on image recognition according to claim 1, characterized in that, The sending methods of the warning messages include sending text messages to notify relevant management personnel, displaying warning messages on the display screen at the construction site, and emitting alarm sounds.

8. An image recognition-based construction safety management system, characterized in that, It includes: Data acquisition module: used to deploy multiple image acquisition devices at the construction site and collect image data of the construction site at preset time intervals or according to an event trigger mechanism; Preprocessing module: used to preprocess the collected image data. The preprocessing includes image denoising, contrast enhancement and brightness adjustment; Recognition module: use deep learning algorithms to identify personnel and construction equipment in the preprocessed images, construct classification models for personnel and equipment, and train them with a large amount of labeled image data so that the model can identify different personnel and various construction equipment at the construction site; Safety hazard detection module: Based on the identified personnel and equipment information, perform safety hazard detection on the image. For personnel, detect whether there are unsafe behaviors such as not wearing a safety helmet, not fastening the seat belt, and illegally entering the dangerous area. For equipment, detect whether there is an abnormal state; set corresponding thresholds and rules to judge the status of personnel and equipment in the image and identify potential safety hazards; Risk assessment and early warning module: According to the type, severity, and scope of influence of safety hazards, assess the safety risks at the construction site, establish a risk assessment model, comprehensively consider the probability of the occurrence of hazards and the factors that may cause consequences, and determine the risk level; When a safety hazard is detected and the risk level reaches the early warning threshold, send out an early warning message.

Images