Intelligent Identification Method and Electronic Device for High-altitude Fall Risk Points on the Construction Floor Plan

Abstract

The present method discloses an intelligent identification method for high-altitude fall risk points on the plane of a construction floor, an electronic device, and a readable storage medium. Path simulation is carried out based on BIM images and depth data, and the edge of the opening is used as an obstacle to simulate the walking path of construction workers. The boundary features of the opening obtained through the path are used to train an unsupervised machine learning model to identify high-risk points, and this process is iterated to obtain all the risk points that need to add safety measures and the optimal walking path of construction workers in the work area. To achieve accurate identification and prediction of risk points before construction, and based on different site layout situations, combined with different construction stages, effectively predict and deploy the positions of safety protection measures.

Description

Technical Field

[0001] The present method relates to the technical field of construction safety information technology and computer machine learning composite technology, and particularly relates to an intelligent identification method, an electronic device and a readable storage medium for high-altitude fall risk points on the construction floor plane based on unsupervised machine learning and path simulation. Background Art

[0002] At present, the informatization level of construction enterprises in the field of safety is relatively low. The supervision means cannot track risk factors in a timely manner, so potential safety hazards cannot be effectively controlled. To reduce the probability of safety accidents, it is necessary to find solutions from the source and give early warnings of upcoming problems in a timely manner, so as to achieve the transformation from "handling after an accident occurs" to "safety prevention before an accident occurs". For this reason, there is an urgent need for an intelligent method for pre-identifying risks at the construction site.

[0003] At present, at home and abroad, it is mainly proposed to perform rule checking on BIM integrated with a large amount of design data. Since the development workload of rule checking is huge and the requirement for the detailed degree of rule description is high, it is commonly used for some risk types with strong regularity. For the characteristics of randomness and suddenness of high-altitude falls and other problems at the construction site, this method has poor applicability.

[0004] The risk visualization method based on data analysis can improve this problem. This method is mainly divided into two categories. One is the risk assessment system based on an expert system, and the other is the anomaly detection system based on machine learning and data mining. Compared with the risk identification method based on rule checking, whether it is an expert system or a machine learning method, it can flexibly discover risk problems, especially complex risks that are difficult to regularize. However, the disadvantage of this method is that it lacks the support of physical principles, so it is difficult to provide the mechanism of risk generation to help users deeply understand the risks.

[0005] Method content

[0006] To solve the above technical problems, in view of the deficiencies of the existing risk identification technologies at the construction site, the present invention provides an intelligent identification method, an electronic device and a readable storage medium for high-altitude fall risk points on the construction floor plane of a building project based on unsupervised machine learning and path simulation. To achieve accurate identification and prediction of risk points before construction, and to effectively predict and arrange the positions of safety protection measures according to different site layout situations and in combination with different construction stages.

[0007] To solve the above technical problems, the technical solutions provided by the present invention are as follows:

[0008] An intelligent identification method for high-altitude fall risk points on the construction floor plan, characterized in that: based on BIM images and depth data, path simulation is carried out, the edge of the opening is used as an obstacle to simulate the movement path of construction workers, and the unsupervised machine learning model is trained through the opening boundary features obtained from the path to identify high-risk points, and this unsupervised machine learning model is iteratively used to obtain all the risk points that need to add safety measures and the best movement path of construction workers in the work area.

[0009] In the above technical solution, the following steps are included:

[0010] S1: Obtain the BIM data of the completed building structure design, calculate the depth information of different floor plans in the structure by screening and processing the triangular faces in the data, and obtain the floor plan images containing depth information;

[0011] S2: Segment the floor plans of the building structure at different heights to obtain image data containing height information;

[0012] S3: Based on the image data obtained in step S2, use the method of obtaining the maximum connected subgraph in graph theory to separate the floor slab and the opening, and use the minimum circumscribed rectangle algorithm to obtain the size of the opening, and obtain the size and position information of the opening in the floor plan;

[0013] S4: Take the opening obtained in step S3 as a spatial obstacle, and calculate the discretized walking distance from the current position to the given end point or the designated work area position;

[0014] S5: Set that the discretized walking distance calculated in step S4 meets the shortest path requirement, and at the same time sample the possible movement paths in a random manner;

[0015] S6: Calculate the corresponding features of the opening boundary on the sampling path in step S5, and the opening boundary features include at least one of the average distance, the number of passes, and the depth of the opening boundary;

[0016] S7: Construct and train an unsupervised machine learning model based on the opening boundary features described in step S6, and construct a loss function oriented to safety cost to obtain the high-risk fall risk point information of the opening boundary;

[0017] S8: Add protective measures to the risk points identified in step S7 and update the obstacle distribution of the current plane, and repeat steps S4 to S7 until the path sampling result converges and no new risk points appear, then end the iterative work.

[0018] In the above technical solution, the image data in step S2 includes the position information, size information and depth information of the floor slab.

[0019] In the above technical solution, the discretized walking distance in step S4 refers to the geodesic distance in the manifold space composed of the floor slab and the obstacles.

[0020] In the above technical solution, in step S7, an unsupervised machine learning model is constructed based on the LEC evaluation method. The product of the average distance L of the opening boundary based on the sampling path, the number of passes E, and the depth C of the opening boundary is used to evaluate the possibility of a hazard source occurring, and a model is constructed using the Sigmoid function type in mathematics, that is:

[0021] ;

[0022] An evaluation curve of high-risk falling risk points of the opening boundary is obtained through this unsupervised machine learning model.

[0023] In the above technical solution, in step S8, the opening boundary is expanded to the risk point addition position, and steps S4 to S7 are repeated until the path sampling result converges. At this time, the path is the best path to ensure the safety of workers' progress.

[0024] In addition, the present invention proposes an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the intelligent identification method for high-altitude falling risk points on the construction floor plane as described above for the electronic device.

[0025] At the same time, the present invention proposes a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the intelligent identification method for high-altitude falling risk points on the construction floor plane as described above.

[0026] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0027] The present invention uses an image recognition method to post-process BIM model data and extract floor slabs at different heights and their depth information. Based on the information in the image data, a heuristic random walk algorithm that meets the shortest path requirement is developed according to the main body modeling idea. Path simulation is carried out based on the given site layout conditions and responsible work areas, and a machine learning model is trained using relevant features to identify high-risk points and add safety protection measures. Through iterative path simulation, risk point identification, and safety measure addition until the optimal path and all high-risk points that need to add protection measures are found. The present invention is based on building structure BIM data, comprehensively considers site layout and personnel factors, automatically outputs risk points and gives the safe progress route of personnel, and realizes the pre-event identification and analysis of risks.

[0028] The construction floor plane high - altitude fall risk point identification algorithm based on unsupervised machine learning and path simulation of the present invention obtains key data in the BIM model, extracts image information, depth information of the floor plane, and position dimensions of openings as high - altitude fall risk elements, etc. Information, and conducts path simulation of construction workers based on design data, ensuring high reliability of data sources and strong adaptability of simulation results.

[0029] While satisfying the principle of the shortest path, the idea of entity modeling is adopted to simulate the selection of travel routes by different construction workers, which conforms to the actual situation of the construction site. On this basis, the unsupervised machine learning method can flexibly identify complex risk points such as high - altitude falls that are difficult to regularize.

[0030] The present invention proposes a brand - new high - altitude fall pre - event intelligent identification algorithm based on unsupervised learning and construction worker path simulation, which has the functions of suggesting the best travel path for construction workers and flexibly giving the positions of effective safety protection measures at different construction stages according to different site layouts, realizing the accurate identification of risk points before construction, while the existing technologies generally cannot automatically identify in advance and cannot give effective warning information.

[0031] This algorithm is based on existing building model data and an automatic recognition mechanism for fall risk points according to the worker's responsible work area and site layout. Through this algorithm, high - risk fall risk points on the construction site can be identified in advance and the best path simulation can be carried out. It can not only be combined with intelligent construction progress management, but also provide different priority levels for the layout of construction prevention measures. This method provides technical support for the intelligent solution of construction safety management and can also provide reference for the pre - event identification and control of other types of risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The following will further illustrate the method in conjunction with the drawings. In the drawings:

[0033] Figure 1 is the flow chart of the construction floor plane high - altitude fall risk point identification method based on unsupervised machine learning and path simulation of the present invention.

[0034] Figure 2 is the floor depth information image obtained based on BIM data of the present invention.

[0035] Figure 3 is the image processing segmentation result diagram of floor slabs at different heights in the embodiment of the present invention.

[0036] Figure 4 is the opening image in the selected floor plane of the embodiment.

[0037] Figure 5 is the calculation result diagram of the discretized distance on the floor slab plane of the present invention.

[0038] Figure 6 It is a graph of the random path sampling results for the shortest path requirement satisfied by the present invention.

[0039] Figure 7 It is the path simulation convergence result after the present invention adds safety protection measures at all risk points.

[0040] Figure 8 It is a graph of the iterative process result of adding safety measures at risk points by the present invention. Detailed implementation manners

[0041] In order to make the purpose, technical solutions and advantages of the present method clearer, the present method will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present method and are not used to limit the present method.

[0042] The embodiments of the present application are described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application and should not be construed as a limitation of the present application.

[0043] The construction floor plane high-altitude fall risk point identification algorithm based on unsupervised machine learning and path simulation proposed by the present invention obtains key data in the BIM model, extracts image information and depth information of the floor plane, and information such as the position and size of the openings as high-altitude fall risk factors, and performs path simulation of construction workers based on the design data. While satisfying the shortest path principle, the idea of entity modeling is adopted to simulate the selection of different construction workers for the travel route. On this basis, an unsupervised machine learning method is used to identify high-altitude fall risk points.

[0044] Embodiment 1:

[0045] The construction floor plane high-altitude fall risk point identification method based on unsupervised machine learning and path simulation includes functions of suggesting the best travel route for construction workers and flexibly giving effective safety protection measure points in different construction stages according to different site layout situations, such as Figure 1 as shown, including:

[0046] S1: Obtain the BIM data of the completed building structure design, screen and process the triangular faces in the data to calculate the depth information of different floor planes in the structure, and obtain the floor plane images containing depth information.

[0047] S2: Segment the floor planes of the building structure at different heights to obtain the image data of different floor planes. Specifically, the image data information includes the plane coordinate position information and depth information of the floor slab.

[0048] S3: Based on the image data obtained in step S2, use the connected component algorithm in graph theory to separate the floor slab and the opening, and use the minimum bounding rectangle algorithm to obtain the size of the opening, so as to obtain the size and planar coordinate position information of the opening in the floor plan.

[0049] S4: Take the opening obtained in step S3 as a spatial obstacle, and calculate the discretized walking distance from the current position to the given end point (the designated work area position).

[0050] Specifically, the discretized walking distance refers to the discretization processing of the geodesic distance calculation method in the manifold space composed of the floor slab and the obstacle.

[0051] S5: When the distance calculated in step S4 meets the shortest path requirement, the algorithm samples the possible travel paths while ensuring randomness.

[0052] S6: Calculate the corresponding features of the opening boundary on the sampling path in step S5, including the average distance, the number of passes, the depth of the opening boundary, etc.

[0053] S7: Construct and train an unsupervised machine learning model based on the opening boundary features described in step S6, and construct a loss function oriented to safety cost. Evaluate the performance of the machine learning model, and if the evaluation is qualified, obtain the information of the high-risk fall risk points on the opening boundary.

[0054] S8: Add protective measures to the risk points identified in step S7 and update the obstacle distribution of the current plane. Specifically, expand the boundary of the opening (risk factor) to the position where the risk point is added. Repeat steps S4 to S7 until the path sampling result converges. At this time, the path is the best path to ensure the safety of workers' travel and all risk points that should take safety measures are identified.

[0055] Embodiment 2:

[0056] The intelligent method proposed by the present invention includes the following steps as Figure 1 shown:

[0057] S1: Obtain the BIM data of the designed building structure, and calculate the depth information of different floor planes in the structure by screening and processing the triangular faces in the data.

[0058] S2: Segment the floor planes of the building structure at different heights, including the image data with height information.

[0059] S3: Based on the image data obtained in step S2, use the method of obtaining the maximum connected subgraph in graph theory to separate the floor slab and the opening, and use the minimum bounding rectangle algorithm to obtain the pixel size of the opening. Calculate the actual size of the opening through the conversion of the pixel size and the actual size.

[0060] S4: Take the openings obtained in step S3 as spatial obstacles, and calculate the discretized walking distance from the current position to the given end point (designated work area location).

[0061] S5: While ensuring randomness to sample possible travel paths, the algorithm samples based on the distance calculated in step S4 that meets the shortest path requirement.

[0062] S6: Calculate the corresponding features of the opening boundary on the sampled path in step S5, including average distance, number of passes, depth of the opening boundary, etc.

[0063] S7: Construct and train an unsupervised machine learning model based on the opening boundary features described in step S6, and construct a loss function oriented to safety cost. Evaluate the performance of the machine learning model, and if the evaluation is qualified, obtain the information of high-risk falling points on the opening boundary.

[0064] S8: Add protective measures to the risk points identified in step S7 and update the obstacle distribution of the current plane. Repeat steps S4 to S7 until the path sampling results converge.

[0065] Example 3:

[0066] The method for identifying high-altitude falling risk points on the construction floor plane based on unsupervised machine learning and path simulation proposed by the present invention includes functions such as intelligent identification of high-altitude falling risk points in the floor plane, suggestion of the best travel path for construction personnel, and effective safety protection measure plans according to different site layout situations.

[0067] The following is a detailed introduction to the embodiments of the present invention with reference to the accompanying drawings:

[0068] In order to realize the pre-identification of high-altitude falling risk points in the construction site floor plane and automatically identify risk points based on existing design data, the present invention first proposes a new algorithm for identifying high-altitude falling risk points on the construction floor plane based on unsupervised machine learning algorithm and path simulation. The flow chart of the algorithm of the present invention is as Figure 1 shown, and the specific steps are as follows:

[0069] S1: Obtain the BIM data of the building structure of the Hubei Provincial Center for Disease Control and Prevention after design is completed, screen and process the triangular faces in the data to calculate the depth information of different floor planes in the structure, and obtain a floor plane image containing depth information, as Figure 2 shown.

[0070] S2: Segment the building structure floor planes at different heights to obtain image data of floor planes at different floor heights. Specifically, the image data contains the position information, size information, and depth information of the floor slabs, as Figure 3 shown.

[0071] S3: In this embodiment, only the floor slab at a height of 28.5 meters in the project is selected as the floor slab for the subsequent embodiments. Based on the image data obtained in step S2, the connected component algorithm in graph theory is used to separate the floor slab and the opening, and the minimum bounding rectangle algorithm is used to obtain the size of the opening, and the size and position information of the opening in the floor plan as shown in Figure 4 is obtained.

[0072] S4: Take the opening obtained in step S3 as a spatial obstacle, and calculate the discretized walking distance from the current position to the given end point (the specified work area position).

[0073] Specifically, the discretized walking distance refers to the geodesic distance in the manifold space composed of the floor slab and the obstacle. As shown in Figure 5 , the discretized walking distance is represented by color.

[0074] S5: When the distance calculated in step S4 meets the shortest path requirement, sample the possible travel paths while ensuring randomness. The sampling result is as shown in Figure 6 .

[0075] S6: Calculate the corresponding features of the opening boundary on the sampling path in step S5, including the average distance, the number of passes, the depth of the opening boundary, etc.

[0076] S7: Construct and train an unsupervised machine learning model based on the opening boundary features described in S6, and construct a loss function oriented to safety cost.

[0077] The model draws on the LEC evaluation method of Benjamin Graham (1894 - 1976) to construct an unsupervised machine learning model. The LEC evaluation method divides the risk level through the product of three hazard-related factors: the possibility of an accident occurring, the exposure frequency, and the severity of the consequences. The machine learning model in the present invention is inspired by the LEC method:

[0078] ;

[0079] Construct the product of features such as the average distance (L) of the opening boundary based on the sampling path, the number of passes (E), and the depth of the opening boundary (C) to evaluate the possibility of a hazard occurring, and use the Sigmoid function type in mathematics to construct the model, that is:

[0080] ;

[0081] Obtain the evaluation curve of high-risk fall risk points of the opening boundary through this unsupervised machine learning model.

[0082] S8: Add protective measures to the risk points identified in S7 and update the obstacle distribution of the current plane. Specifically, expand the boundary of the opening (risk factor) to the position where the risk points are added. Repeat steps S4 to S7 until the path sampling results converge. At this time, the path is the optimal path for the construction personnel to travel safely, as Figure 7 shown by the path line in Figure 8 . The results of the risk points that need to add safety measures are finally as shown by the orange points in

[0083] In summary, the present invention discloses a new method for identifying high-altitude fall risk points on the construction floor plane based on unsupervised machine learning algorithms and path simulation. The present invention uses BIM design data, combines image processing methods, unsupervised machine learning algorithms and the idea of main body modeling and simulation to build an algorithm for automatically identifying risk points on the floor plane, providing an intelligent solution for high-altitude fall risk control in construction safety management, and improving the risk control level and management efficiency.

[0084] So far, the ex-ante identification of high-altitude fall risk points on the floor plane and the optimal construction personnel travel path corresponding to the identified risk points have been completed.

[0085] To implement the above embodiments, an electronic device is proposed in an embodiment of the present application, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the algorithm for identifying high-altitude fall risk points on the construction floor plane based on unsupervised machine learning and path simulation as described in the method embodiment of the foregoing terminal device.

[0086] To implement the above embodiments, a computer-readable storage medium is proposed in an embodiment of the present application, on which a computer program is stored. When the program is executed by a processor, it implements the algorithm for identifying high-altitude fall risk points on the construction floor plane based on unsupervised machine learning and path simulation as described in the foregoing method embodiment.

[0087] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of this method.

Claims

1. An intelligent identification method for high-altitude fall risk points on the plane of a construction floor, characterized in that Path simulation is carried out based on BIM images and depth data. Taking the edge of the opening as an obstacle, the walking path of construction workers is simulated. The boundary features of the opening obtained from the path are used to train an unsupervised machine learning model to identify high-risk points, and this unsupervised machine learning model is iteratively used to obtain all the risk points that need to add safety measures and the optimal walking path of construction workers in the work area. The steps are as follows: S1: Obtain the BIM data of the designed building structure. By screening and processing the triangular faces in the data, calculate the depth information of different floor planes in the structure to obtain a floor plane image containing depth information. S2: Segment the floor planes of the building structure at different heights to obtain image data containing height information. S3: Based on the image data obtained in step S2, use the method of obtaining the maximum connected subgraph in graph theory to separate the floor slab and the opening, and use the minimum circumscribed rectangle algorithm to obtain the size of the opening, and obtain the size and position information of the opening in the floor plane. S4: Take the opening obtained in step S3 as a spatial obstacle and calculate the discretized walking distance from the current position to the given end point or the specified work area position. S5: Set that the discretized walking distance calculated in step S4 meets the shortest path requirement, and at the same time sample the possible walking paths in a random manner. S6: Calculate the corresponding features of the opening boundary on the sampled path in step S5. The opening boundary features include at least one of the average distance, the number of passes, and the depth of the opening boundary. S7: Construct and train an unsupervised machine learning model based on the opening boundary features described in step S6, and construct a loss function oriented to safety cost to obtain the information of high-risk falling points on the opening boundary. S8: Add protective measures to the risk points identified in step S7 and update the obstacle distribution of the current plane. Repeat steps S4 to S7 until the path sampling result converges and no new risk points appear, then end the iterative work.

2. The intelligent identification method for high-altitude fall risk points on the construction floor plan according to claim 1, characterized in that: The image data in step S2 contains the position information, size information, and depth information of the floor slab.

3. The intelligent identification method for high-altitude fall risk points on the construction floor plan according to claim 1, wherein: The discretized walking distance in step S4 refers to the geodesic distance in the manifold space composed of the floor slab and the obstacles.

4. The intelligent identification method for high-altitude fall risk points on the construction floor plan according to claim 1, characterized in that: In step S8, expand the opening boundary to the risk point addition position, and repeat steps S4 to S7 until the path sampling result converges. At this time, the path is the best path to ensure the safety of workers' walking.

5. An electronic device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the program, it implements the intelligent identification method for high-altitude falling risk points on the construction floor plane described in any one of claims 1-4 above.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that When the program is executed by the processor, it implements the intelligent identification method for high-altitude falling risk points on the construction floor plane described in any one of claims 1-4 above.

Images