Method and system for analyzing and monitoring engineering safety project by using intelligent image

Abstract

The invention relates to a method and a system for analyzing and monitoring an engineering safety project by using an intelligent image, and relates to the technical field of image analysis. Determining a plurality of historical engineering classes, an engineering safety project corresponding to each historical engineering class and a dependency coefficient between any two engineering safety projects, and determining similar historical engineering classes of the current engineering; determining engineering safety projects of a plurality of construction nodes of the current engineering based on the engineering safety projects corresponding to the similar historical engineering classes and the dependency coefficient between any two engineering safety projects, and further determining image acquisition schemes of the plurality of construction nodes corresponding to the current engineering; the construction site images acquired by the plurality of image acquisition devices are analyzed based on the engineering safety projects of the plurality of construction nodes of the current engineering, the safety prompt information is generated, and the method has the advantages of improving the real-time performance, the accuracy and the comprehensiveness of the monitoring of the engineering safety projects.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a method and system for monitoring engineering safety projects using intelligent image analysis. Background Art

[0002] Project safety management is an important means to ensure the safety of the construction process and protect the lives of workers. Effective engineering safety projects include formulating strict safety specifications, conducting regular safety training, implementing on-site safety monitoring, conducting regular safety inspections, establishing emergency plans, and strengthening safety culture construction. Among them, formulating strict safety specifications is the foundation, which involves formulating reasonable work procedures and operating processes to ensure that each link can follow the safe operation guidelines. Through clear specifications, accidents can be effectively prevented and controlled, which is the key to ensuring the effective implementation of safety management. Among them, on-site safety monitoring is a key link in timely discovering and solving safety hazards. Safety management personnel need to conduct regular on-site inspections to check whether the actual situation of the operation meets the safety specifications, correct unsafe behaviors in a timely manner, register the safety hazards on the site, and implement hidden danger rectification measures. However, the monitoring of engineering safety projects on site manually has poor real-time performance and is prone to overlooking many safety hazards.

[0003] Therefore, there is a need to provide a method and system for monitoring engineering safety projects using intelligent image analysis, which can be used to improve the real-time, accuracy and comprehensiveness of monitoring engineering safety projects using intelligent image analysis. Summary of the invention

[0004] The present invention provides a method for monitoring engineering safety projects using intelligent image analysis, comprising: obtaining first relevant data of multiple historical projects, wherein the first relevant data of the historical projects at least include construction plan data, construction environment data, construction personnel data and engineering safety risk data of multiple construction nodes of the historical projects; classifying the multiple historical projects based on the first relevant data of the multiple historical projects to determine multiple historical project classes; for each of the historical project classes, determining the engineering safety projects corresponding to the historical project class and the dependency coefficient between any two engineering safety projects based on the first relevant data of the historical projects included in the historical project class; obtaining second relevant data of the current project, wherein the second relevant data of the current project at least include construction plan data, construction environment data and construction personnel data of the current project. personnel data; based on the second relevant data of the current project and the first relevant data of the multiple historical projects, determine a similar historical project class of the current project from the multiple historical project classes; based on the engineering safety projects corresponding to the similar historical project classes of the current project and the dependency coefficient between any two engineering safety projects, determine the engineering safety projects of multiple construction nodes of the current project; based on the engineering safety projects of the multiple construction nodes of the current project, determine the image acquisition scheme of the multiple construction nodes corresponding to the current project; based on the image acquisition scheme of the multiple construction nodes corresponding to the current project, set up multiple image acquisition devices at the construction site of the current project; based on the engineering safety projects of the multiple construction nodes of the current project, analyze the construction site images acquired by the multiple image acquisition devices to generate safety prompt information.

[0005] Furthermore, based on the first relevant data of multiple historical projects, the multiple historical projects are classified to determine multiple historical project categories, including: numbering each of the historical projects; for any two historical projects, based on the construction plan data of the two historical projects, calculating the construction plan similarity of the two historical projects; based on the construction environment data of the two historical projects, calculating the construction environment similarity of the two historical projects; based on the construction personnel data of the two historical projects, calculating the construction personnel similarity of the two historical projects; based on the construction plan similarity, construction environment similarity and construction personnel similarity of the two historical projects, calculating the comprehensive similarity of the two historical projects; for each historical project, based on the comprehensive similarity between the historical project and other historical projects and the number of each historical project, generating a similarity sequence corresponding to the historical project; for any two historical projects, based on the similarity sequence corresponding to each of the historical projects, calculating the sequence similarity of the two historical projects; and classifying the multiple historical projects according to the sequence similarity of any two historical projects by a hierarchical clustering algorithm to determine multiple historical project categories.

[0006] Furthermore, based on the first relevant data of the historical projects included in the historical engineering class, the engineering safety projects corresponding to the historical engineering class and the dependence coefficient between any two engineering safety projects are determined, including: for each historical project included in the historical engineering class, based on the engineering safety risk data of multiple construction nodes of the historical project, the engineering safety projects of multiple construction nodes of the historical project are determined; based on the engineering safety projects of multiple construction nodes of each historical project included in the historical engineering class, the engineering safety projects corresponding to the historical engineering class are determined; for each engineering safety project corresponding to the historical engineering class, based on the engineering safety risk data of each historical project included in the historical engineering class, the project risk sequence of the engineering safety project is determined; for any two engineering safety projects corresponding to the historical engineering class, based on the project risk sequences of the two engineering safety projects, the dependence coefficient between the two engineering safety projects is calculated.

[0007] Furthermore, based on the second relevant data of the current project and the first relevant data of the multiple historical projects, a similar historical project class of the current project is determined from the multiple historical project classes, including: S11, for each of the historical project classes, a plurality of historical projects are sampled from the historical project class as sampled historical projects, based on the construction plan data of the sampled historical projects and the construction plan data of the current project, the construction plan similarity between the sampled historical projects and the current project is calculated, based on the construction environment data of the sampled historical projects and the construction environment data of the current project, the construction environment similarity between the sampled historical projects and the current project is calculated, based on the construction personnel data of the sampled historical projects and the construction personnel data of the current project, the construction personnel similarity between the sampled historical projects and the current project is calculated, based on the construction plan similarity, construction environment similarity and construction personnel similarity between the sampled historical projects and the current project, the similarity is calculated. The comprehensive similarity between the sampled historical project and the current project, based on the comprehensive similarity between the current project and each of the sampled historical projects, calculate the engineering similarity between the current project and the historical engineering class; S12, determine whether the current sampling round is greater than the sampling round threshold, if not, execute S11, if so, execute S13; S13, based on the engineering similarity between the current project and each of the historical engineering classes in each round of sampling, calculate the global fluctuation parameter sequence, wherein the global fluctuation parameter sequence includes the global fluctuation parameters corresponding to any two adjacent sampling rounds; S14, based on the global fluctuation parameter sequence, determine whether to perform the next round of sampling, if so, execute S11, if not, execute S15; S15, based on the global fluctuation parameter sequence, determine the target sampling round; S16, based on the engineering similarity between the current project and each of the historical engineering classes in the target sampling round, determine the similar historical engineering class of the current project from the multiple historical engineering classes.

[0008] Furthermore, based on the engineering safety projects corresponding to similar historical engineering classes of the current project and the dependency coefficient between any two engineering safety projects, the engineering safety projects of multiple construction nodes of the current project are determined, including: establishing a safety project prediction model corresponding to each of the historical engineering classes; retrieving a target safety project prediction model based on similar historical engineering classes of the current project; extracting characteristic data of the current project from second related data of the current project; determining the first engineering safety project of multiple construction nodes of the current project based on the characteristic data of the current project through the target safety project prediction model; determining the second engineering safety project of multiple construction nodes of the current project according to the dependency coefficient between any two engineering safety projects and the first engineering safety project of multiple construction nodes of the current project.

[0009] Furthermore, a safety project prediction model corresponding to each of the historical engineering classes is established, including: calculating the clustering parameters of any two of the historical engineering classes according to the engineering safety projects corresponding to each of the historical engineering classes and the dependence coefficient between any two engineering safety projects; clustering the multiple historical engineering classes according to the clustering parameters of any two of the historical engineering classes through a hierarchical clustering algorithm to determine multiple historical engineering class clusters; for each of the historical engineering class clusters, an initial prediction model corresponding to the cluster center of the historical engineering class cluster is established, a training sample corresponding to the cluster center of the historical engineering class cluster is generated, and the initial prediction model is trained using the training sample corresponding to the cluster center of the historical engineering class cluster to generate a trained initial prediction model corresponding to the cluster center of the historical engineering class cluster; for each historical engineering class included in the historical engineering class cluster, the trained initial prediction model corresponding to the cluster center of the historical engineering class cluster is adjusted, a safety project prediction model corresponding to the historical engineering class is established, a training sample corresponding to the historical engineering class is generated, and the safety project prediction model corresponding to the historical engineering class is trained using the training sample corresponding to the historical engineering class to generate a safety project prediction model corresponding to the historical engineering class.

[0010] Furthermore, based on the engineering safety items of the multiple construction nodes of the current project, an image acquisition scheme for the multiple construction nodes corresponding to the current project is determined, including: for each construction node of the current project, based on the engineering safety items of the construction nodes, multiple image monitoring positions are determined; based on the multiple image monitoring positions of each construction node of the current project, multiple candidate image acquisition schemes for the multiple construction nodes corresponding to the current project are generated; a scheme evaluation system is established; and according to the scheme evaluation system, multiple candidate image acquisition schemes for the multiple construction nodes corresponding to the current project are screened to determine the image acquisition scheme for the multiple construction nodes corresponding to the current project.

[0011] Furthermore, the construction site images collected by the multiple image acquisition devices are analyzed based on the engineering safety items of the multiple construction nodes of the current project to generate safety warning information, including: for each of the engineering safety items, an image analysis model corresponding to the engineering safety item is established; based on the engineering safety items of the multiple construction nodes of the current project, multiple target image analysis models are retrieved; for each of the target image analysis models, the construction site image corresponding to the target image analysis model is extracted from the construction site images collected by the multiple image acquisition devices, and the construction site image corresponding to the target image analysis model is analyzed by the target image analysis model to generate safety warning information.

[0012] Furthermore, for each of the engineering safety projects, an image analysis model corresponding to the engineering safety project is established, including: for any two of the engineering safety projects, determining the image features corresponding to the two engineering safety projects, and determining the similarity of the image features of the two engineering safety projects; clustering the multiple engineering safety projects according to the similarity of the image features of any two engineering safety projects through a hierarchical clustering algorithm to determine multiple engineering safety project clusters; for each of the engineering safety project clusters, an initial image analysis model corresponding to the cluster center of the engineering safety project cluster is established, a training sample corresponding to the cluster center of the engineering safety project cluster is generated, and the initial image analysis model corresponding to the cluster center of the engineering safety project cluster is trained using the initial image analysis model corresponding to the cluster center of the engineering safety project cluster to generate the trained initial image analysis model corresponding to the cluster center of the engineering safety project cluster; for each engineering safety project included in the engineering safety project cluster, the trained initial image analysis model corresponding to the cluster center of the engineering safety project cluster is adjusted, an image analysis model corresponding to the engineering safety project is established, a training sample corresponding to the engineering safety project is generated, and the image analysis model corresponding to the engineering safety project is trained using the training sample corresponding to the engineering safety project to generate the image analysis model corresponding to the engineering safety project.

[0013] The present invention provides a system for monitoring engineering safety projects using intelligent image analysis, which is used to execute the above-mentioned method for monitoring engineering safety projects using intelligent image analysis, including: a data analysis module, which is used to obtain first relevant data of multiple historical projects, wherein the first relevant data of the historical projects at least include construction plan data, construction environment data, construction personnel data and engineering safety risk data of multiple construction nodes of the historical projects, and is also used to classify the multiple historical projects based on the first relevant data of the multiple historical projects, determine multiple historical project classes, and for each of the historical project classes, determine the engineering safety projects corresponding to the historical project class and the dependence coefficient between any two engineering safety projects based on the first relevant data of the historical projects included in the historical project class; a data acquisition module, which is used to obtain second relevant data of the current project, wherein , the second relevant data of the current project at least includes the construction plan data, construction environment data and construction personnel data of the current project; a project determination module, which is used to determine the similar historical engineering class of the current project from the multiple historical engineering classes based on the second relevant data of the current project and the first relevant data of the multiple historical projects, and is also used to determine the engineering safety projects of multiple construction nodes of the current project based on the engineering safety projects corresponding to the similar historical engineering classes of the current project and the dependency coefficient between any two engineering safety projects; a scheme determination module, which is used to determine the image acquisition scheme of multiple construction nodes corresponding to the current project based on the engineering safety projects of multiple construction nodes of the current project, and is also used to set up multiple image acquisition devices at the construction site of the current project based on the image acquisition scheme of multiple construction nodes corresponding to the current project;

[0014] The image analysis module is used to analyze the construction site images collected by the multiple image acquisition devices based on the engineering safety items of multiple construction nodes of the current project to generate safety prompt information.

[0015] Compared with the prior art, the method and system for monitoring engineering safety projects using intelligent image analysis provided in this specification have at least the following beneficial effects:

[0016] 1. Through detailed analysis of historical project data, especially the construction plan, environment, personnel, and safety risk data of each construction node, it is possible to more accurately identify the safety risks that the current project may face, so as to take targeted preventive measures and significantly reduce the incidence of safety accidents. Based on the classification of historical project data and the determination of dependency coefficients, it is possible to identify key engineering safety projects in different engineering categories and their correlations. This makes safety management more precise, that is, to formulate personalized safety monitoring and management strategies for different engineering types and characteristics. By comparing the similarities between the current project and the historical project, determining similar historical engineering categories, and setting the construction node safety projects and image acquisition schemes for the current project accordingly, it is possible to avoid over-allocation or under-allocation of resources and achieve optimal utilization of resources. Setting up an image acquisition device at the construction site can monitor the construction process in real time. Once a potential safety risk is discovered, safety prompt information is immediately generated through image analysis to achieve immediate response, effectively reducing the occurrence of safety accidents and the losses that may result. The entire process realizes automation and intelligence from data collection, analysis to safety monitoring, greatly reducing the time and cost of manual intervention and judgment, and improving engineering management efficiency;

[0017] 2. Through multiple rounds of sampling and calculation of similarities in multiple dimensions such as construction plans, construction environments, and construction personnel, and comprehensive consideration of these similarities to obtain comprehensive similarities, the similarity between the current project and the historical project can be evaluated more comprehensively and accurately. This helps to reduce the deviation that may be caused by the similarity evaluation of a single dimension. By calculating the global fluctuation parameter sequence and judging whether to continue sampling, a certain stability can be ensured in the similarity evaluation process. When the global fluctuation parameters show that the evaluation results have stabilized, further sampling can be stopped, thereby improving efficiency while ensuring the accuracy of the evaluation. Through multiple rounds of sampling and judgments based on global fluctuation parameters, the sampling rounds can be automatically adjusted to avoid unnecessary oversampling. This helps to optimize resource utilization and reduce time and cost in data collection and processing. Determining similar historical engineering categories based on the engineering similarities obtained in the target sampling rounds can ensure the reliability and accuracy of the decision-making basis. This helps to make more scientific and reasonable decisions in subsequent safety project determination, image acquisition plan formulation, and other links;

[0018] 3. Cluster multiple historical engineering classes through hierarchical clustering algorithms to form multiple historical engineering clusters. This cluster analysis helps to discover the commonalities and differences between different engineering classes, providing a basis for establishing more targeted prediction models. For each historical engineering cluster, an initial prediction model of the cluster center is first established and trained with training samples. Then, each historical engineering class is fine-tuned to meet the specific needs of different engineering classes. This hierarchical model building and adjustment process is more efficient than directly training each historical engineering class. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0020] Figure 1 is a flow chart of a method for monitoring engineering safety projects using intelligent image analysis shown in an embodiment of the present application;

[0021] Figure 2 is a flowchart of determining a similar historical engineering class of a current engineering from multiple historical engineering classes shown in an embodiment of the present application;

[0022] Figure 3 It is a module diagram of a system for monitoring engineering safety projects using intelligent image analysis shown in one embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly introduces the drawings required for describing the embodiments.

[0024] Figure 1 FIG. 1 is a flow chart of a method for monitoring engineering safety projects using intelligent image analysis in one embodiment of the present application. Figure 1 As shown, a method for monitoring engineering safety projects using intelligent image analysis may include the following process.

[0025] Step 110: Acquire first relevant data of a plurality of historical projects.

[0026] The first relevant data of the historical project at least includes the construction plan data, construction environment data, construction personnel data and engineering safety risk data of multiple construction nodes of the historical project.

[0027] Specifically, the construction plan data details the various phases, time nodes and key tasks of the project. For example:

[0028] Start date and completion date: clearly state when the project starts and ends, e.g., if the project starts on a certain date in a certain month in a certain year, it is planned to be completed on a certain date in a certain month in the following year;

[0029] Construction cycle and duration: the total construction cycle and the duration division of each construction node, such as foundation construction node, main structure construction node, decoration and renovation construction node, etc., the specific start and end time of each construction node and the estimated duration;

[0030] Construction sequence and process: List the sequence and logical relationship of multiple construction nodes in detail;

[0031] Resource allocation: including allocation plans for human, material and financial resources, such as the number of personnel required for each construction node, the types and quantities of materials, equipment configuration, etc.

[0032] Construction environment data records the natural conditions and social environment of the project site and various factors that may affect the construction. For example, construction environment data includes:

[0033] Geological conditions: geological conditions on both sides of the river and on the riverbed, such as soil type, rock distribution, groundwater level, etc.;

[0034] Climate conditions: meteorological data such as temperature, precipitation, wind speed, etc. at the project location;

[0035] Hydrological conditions: hydrological data such as river water level, flow velocity, and flow.

[0036] Social environment: population distribution, traffic conditions, etc. at the project site.

[0037] Construction personnel data records the basic information and division of responsibilities of various personnel involved in the project construction. For example, construction personnel data includes:

[0038] Personnel list: List all personnel involved in the construction, including project managers, technical leaders, construction team leaders, workers of various trades, etc.;

[0039] Qualification Certificate: record the professional qualification certificates of construction personnel, such as builder certificate, electrician certificate, welder certificate, etc.;

[0040] Attendance records: record the attendance of construction personnel, including working hours, leave, etc.;

[0041] Training records: record the content and time of safety education, skills training, etc. for construction workers.

[0042] Step 120: classify the plurality of historical projects based on the first relevant data of the plurality of historical projects to determine a plurality of historical project categories.

[0043] Specifically include:

[0044] Numbering each historical project, wherein the number of each historical project is digital and unique, for example, one historical project is numbered 1, another historical project is numbered 2, and so on;

[0045] For any two historical projects, based on the construction plan data of the two historical projects, the similarity of the construction plans of the two historical projects is calculated; based on the construction environment data of the two historical projects, the similarity of the construction environments of the two historical projects is calculated; based on the construction personnel data of the two historical projects, the similarity of the construction personnel of the two historical projects is calculated; based on the similarity of the construction plans of the two historical projects, the similarity of the construction environments and the similarity of the construction personnel, the comprehensive similarity of the two historical projects is calculated;

[0046] For each historical project, based on the comprehensive similarity between the historical project and other historical projects and the number of each historical project, a similarity sequence corresponding to the historical project is generated, wherein one element of the similarity sequence corresponds to one number, and the value of the element is the comprehensive similarity between the historical project and the historical project with the number, and the numbers corresponding to the multiple elements of the similarity sequence are sorted from small to large;

[0047] For any two historical projects, the sequence similarity of the two historical projects is calculated based on the similarity sequence corresponding to each historical project;

[0048] Through the hierarchical clustering algorithm, multiple historical projects are classified according to the sequence similarity of any two historical projects, and multiple historical project classes are determined.

[0049] Specifically, key features are extracted from the construction plan data of the two historical projects, such as the main construction phases, key tasks, construction schedule, resource allocation, etc. These features are converted into quantifiable forms, such as the construction schedule can be converted into the start and end time points of each phase, and the resource allocation can be converted into the usage of each type of work or equipment. The feature data of each construction plan is converted into a vector, where each dimension of the vector represents a feature and the value of the dimension is the quantized value of the feature. If the features of the two construction plans are not exactly the same, zero values ​​can be introduced into the vector to fill the missing features, or new and more representative features can be created through feature engineering. The cosine similarity between the vectors corresponding to the construction plan data of the two historical projects is calculated as the construction plan similarity of the two historical projects. The method of calculating the construction personnel similarity of the two historical projects and calculating the comprehensive similarity of the two historical projects is similar to the method of calculating the construction plan similarity of the two historical projects, which will not be repeated here.

[0050] Preferably, the sequence similarity of two historical projects can be calculated according to the following formula:

[0051]

[0052] Among them, S (i,j) is the sequence similarity between the i-th historical project and the j-th historical project, a 11 、a 12 and a13 is the preset weight, a 11 、a 12 and a 13 Greater than 0, a 11 +a 12 +a 13 =1, V (i,n) is the value of the nth element of the similarity sequence corresponding to the i-th historical project, V (j,n) is the value of the nth element of the similarity sequence corresponding to the jth historical project, N is the total number of elements in the similarity sequence, that is, the total number of historical projects, V (i,m) is the value of the mth element sampled from the similarity sequence corresponding to the i-th historical project, V (j,m) is the value of the mth element sampled from the similarity sequence corresponding to the jth historical project, and M is the total number of sampled elements.

[0053] Step 130 : for each historical engineering class, based on the first relevant data of the historical engineering included in the historical engineering class, determine the engineering safety items corresponding to the historical engineering class and the dependency coefficient between any two engineering safety items.

[0054] Specifically include:

[0055] For each historical project included in the historical project category, based on the engineering safety risk data of multiple construction nodes of the historical project, determine the engineering safety items of multiple construction nodes of the historical project;

[0056] Based on the engineering safety items of multiple construction nodes of each historical project included in the historical project category, determine the engineering safety items corresponding to the historical project category;

[0057] For each engineering safety project corresponding to the historical engineering class, a project risk sequence of the engineering safety project is determined based on the engineering safety risk data of each historical project included in the historical engineering class, wherein an element of the project risk sequence corresponds to a historical project included in the historical engineering class, and if a safety accident corresponding to the engineering safety project occurs in the corresponding historical project, the value of the element is 1, otherwise it is 0;

[0058] For any two engineering safety projects corresponding to the historical engineering category, the dependency coefficient between the two engineering safety projects is calculated based on the project risk sequences of the two engineering safety projects.

[0059] Specifically, the engineering safety projects of multiple construction nodes of the historical project can be determined based on the engineering safety risk data of multiple construction nodes of the historical project by manual (for example, experts, etc.) or machine learning models. For example, for high-altitude work nodes, the engineering safety projects can include safety belt detection projects, safety net detection projects, edge protection projects, and hole protection projects.

[0060] As a preferred method, the dependency coefficient between two engineering safety items can be calculated according to the following formula:

[0061]

[0062] Among them, r (e,f) is the dependency coefficient between the e-th engineering safety item and the f-th engineering safety item corresponding to the historical engineering category, is the nth item in the project risk sequence of the eth engineering safety item corresponding to the historical engineering category 2 The value of the element, is the nth item in the project risk sequence of the fth engineering safety item corresponding to the historical engineering category 2 The value of the element, N 2 The total number of historical projects included for the Historical Projects category.

[0063] Step 140, obtaining second related data of the current project.

[0064] The second relevant data of the current project at least includes the construction plan data, construction environment data and construction personnel data of the current project.

[0065] Step 150 : determining a similar historical project class to the current project from the multiple historical project classes based on the second relevant data of the current project and the first relevant data of the multiple historical projects.

[0066] Figure 2 FIG. 1 is a flowchart of determining a similar historical engineering class of a current engineering from multiple historical engineering classes in one embodiment of the present application, such as Figure 2 As shown, step 150 includes:

[0067] S11, for each historical project class, a plurality of historical projects are sampled from the historical project class as sampled historical projects, and the construction plan similarity between the sampled historical project and the current project is calculated based on the construction plan data of the sampled historical project and the construction plan data of the current project; the construction environment similarity between the sampled historical project and the current project is calculated based on the construction environment data of the sampled historical project and the construction environment data of the current project; the construction personnel similarity between the sampled historical project and the current project is calculated based on the construction personnel data of the sampled historical project and the construction personnel data of the current project; the comprehensive similarity between the sampled historical project and the current project is calculated based on the construction plan similarity, construction environment similarity and construction personnel similarity between the sampled historical project and the current project; the project similarity between the current project and each sampled historical project is calculated based on the comprehensive similarity, wherein the method parameters for calculating the construction plan similarity, construction environment similarity and construction personnel similarity between the sampled historical project and the current project are as described in step 120 for calculating the construction plan similarity, construction environment similarity and construction personnel similarity of two historical projects, and are not repeated here;

[0068] S12, determine whether the current sampling round is greater than the sampling round threshold, if not, execute S11, if yes, execute S13;

[0069] S13, calculating a global fluctuation parameter sequence based on the similarity between the current project and each historical project type in each round of sampling, wherein the global fluctuation parameter sequence includes global fluctuation parameters corresponding to any two adjacent rounds of sampling;

[0070] S14, based on the global fluctuation parameter sequence, determine whether to perform the next round of sampling, if yes, execute S11, if not, execute S15;

[0071] S15. Determine the target sampling round based on the global fluctuation parameter sequence;

[0072] S16. Determine a similar historical engineering class of the current project from multiple historical engineering classes based on the engineering similarity between the current project and each historical engineering class in the target sampling round.

[0073] Specifically, for each sampling round, the average of the comprehensive similarities between the current project and each sampled historical project corresponding to the sampling round can be calculated as the project similarity between the current project corresponding to the sampling round and the historical project class.

[0074] Preferably, the global fluctuation parameter corresponding to two adjacent sampling rounds can be calculated based on the following formula:

[0075]

[0076] Among them, η(t,t+1) is the global fluctuation parameter corresponding to the adjacent t-th sampling round and the t+1-th sampling round, S (t,k) is the engineering similarity between the current engineering project corresponding to the tth sampling round and the kth historical engineering class, S (t+1,k) is the engineering similarity between the current engineering corresponding to the t+1th sampling round and the kth historical engineering class, and K is the total number of historical engineering classes.

[0077] The variance of the global fluctuation parameter sequence can be calculated. When the variance of the global fluctuation parameter sequence is greater than the variance threshold, it is determined to proceed to the next round of sampling.

[0078] For each sampling round, calculate the absolute value of the difference between the global fluctuation parameter of the previous sampling round and the global fluctuation parameter of this sampling round, and the absolute value of the difference between the global fluctuation parameter of the next sampling round and the global fluctuation parameter of this sampling round. Sum the absolute values ​​of the two differences and take the reciprocal to obtain the screening value corresponding to the sampling round, and take the sampling round with the largest screening value as the target sampling round.

[0079] Based on the engineering similarity between the current project and each historical engineering class in the target sampling round, the historical engineering class with the greatest engineering similarity is selected as the similar historical engineering class of the current project.

[0080] Step 160 , based on the engineering safety items corresponding to similar historical engineering classes of the current project and the dependency coefficient between any two engineering safety items, determine the engineering safety items of multiple construction nodes of the current project.

[0081] Specifically include:

[0082] Establishing a safety project prediction model corresponding to each historical engineering class, wherein the safety project prediction model may be a convolutional neural network model;

[0083] Based on the similar historical engineering class of the current engineering, the target safety project prediction model is retrieved, wherein the target safety project prediction model is the safety project prediction model corresponding to the similar historical engineering class of the current engineering;

[0084] Extracting characteristic data of the current project from the second related data of the current project;

[0085] Determine the first engineering safety project of multiple construction nodes of the current project based on the characteristic data of the current project through the target safety project prediction model;

[0086] Based on the dependency coefficient between any two engineering safety projects and the first engineering safety projects of multiple construction nodes of the current project, the second engineering safety projects of multiple construction nodes of the current project are determined. For example, at a certain construction node, the engineering safety project corresponding to the construction node and having a dependency coefficient greater than a dependency coefficient threshold with the first engineering safety project of the construction node is used as the second engineering safety project of the construction node.

[0087] As a preferred method, a safety project prediction model corresponding to each historical engineering category is established, including:

[0088] According to the engineering safety items corresponding to each historical engineering class and the dependency coefficient between any two engineering safety items, the clustering parameters of any two historical engineering classes are calculated. For example, the overlap of engineering safety items corresponding to the two historical engineering classes can be calculated, and according to the dependency coefficient between any two engineering safety items corresponding to each of the two historical engineering classes, the similarity of the dependency coefficients of the two historical engineering classes is calculated. The overlap of engineering safety items and the similarity of the dependency coefficients of the two historical engineering classes are weightedly summed to obtain the clustering parameters of the two historical engineering classes.

[0089] Using a hierarchical clustering algorithm, multiple historical engineering classes are clustered according to the clustering parameters of any two historical engineering classes to determine multiple historical engineering clusters;

[0090] For each historical engineering cluster, an initial prediction model corresponding to the cluster center of the historical engineering cluster is established, a training sample corresponding to the cluster center of the historical engineering cluster is generated, and the initial prediction model is trained using the training sample corresponding to the cluster center of the historical engineering cluster to generate a trained initial prediction model corresponding to the cluster center of the historical engineering cluster;

[0091] For each historical engineering class included in the historical engineering class cluster, the trained initial prediction model corresponding to the cluster center of the historical engineering class cluster is adjusted, a safety project prediction model corresponding to the historical engineering class is established, and training samples corresponding to the historical engineering class are generated. The training samples corresponding to the historical engineering class are used to train the safety project prediction model corresponding to the historical engineering class to generate a safety project prediction model corresponding to the historical engineering class.

[0092] Step 170, based on the engineering safety items of the multiple construction nodes of the current project, determine the image acquisition scheme of the multiple construction nodes corresponding to the current project.

[0093] Specifically include:

[0094] For each construction node of the current project, multiple image monitoring positions are determined based on the engineering safety items of the construction node. For example, for each engineering safety item, the position of the construction site where the safety accident corresponding to the engineering safety item is likely to occur is used as the image monitoring position;

[0095] Based on multiple image monitoring positions of each construction node of the current project, multiple candidate image acquisition schemes for multiple construction nodes corresponding to the current project are generated;

[0096] Establish a scheme evaluation system, which may include multiple scheme evaluation indicators, such as image monitoring position coverage indicator, image acquisition field overlap indicator, image acquisition quantity indicator, image acquisition difference indicator between two adjacent construction nodes, etc.;

[0097] According to the scheme evaluation system, multiple candidate image acquisition schemes for multiple construction nodes corresponding to the current project are screened to determine the image acquisition schemes for multiple construction nodes corresponding to the current project.

[0098] Specifically, the score of each candidate image acquisition scheme in multiple scheme evaluation indicators may be calculated, and the candidate image acquisition scheme with the highest comprehensive score may be selected as the image acquisition scheme for multiple construction nodes corresponding to the current project.

[0099] Step 180 , based on the image acquisition schemes of multiple construction nodes corresponding to the current project, multiple image acquisition devices are set at the construction site of the current project.

[0100] Specifically, parameters such as the number, installation positions, and image acquisition angles of the multiple image acquisition devices change at different construction nodes corresponding to the current project according to the image acquisition schemes for the multiple construction nodes corresponding to the current project.

[0101] Step 190 , analyzing the construction site images collected by multiple image acquisition devices based on the engineering safety items of multiple construction nodes of the current project, and generating safety prompt information.

[0102] Specifically include:

[0103] For each engineering safety project, an image analysis model corresponding to the engineering safety project is established, wherein the image analysis model may be a convolutional neural network model;

[0104] Based on the engineering safety projects of multiple construction nodes of the current project, multiple target image analysis models are retrieved;

[0105] For each target image analysis model, a construction site image corresponding to the target image analysis model is extracted from construction site images acquired by multiple image acquisition devices, and the construction site image corresponding to the target image analysis model is analyzed by the target image analysis model to generate safety prompt information.

[0106] For example, when a target image analysis model detects that the stacked construction materials exceed the safe height or are in an improper position, which may cause the risk of collapse, it will immediately issue an early warning and notify the management staff to conduct an inspection.

[0107] For example, if a target image analysis model detects abnormal movement or cessation of mechanical equipment, it will immediately issue a warning and notify management personnel to conduct inspections to prevent safety accidents caused by equipment failure.

[0108] For example, a target image analysis model calculates the number of people in a specific area and compares it with a preset safety threshold. When the crowd density exceeds the safety limit, an early warning is issued immediately.

[0109] Preferably, for each engineering safety project, an image analysis model corresponding to the engineering safety project is established, including:

[0110] For any two engineering safety items, determine the image features corresponding to the two engineering safety items (e.g., detection targets (e.g., people, equipment, etc.), whether there is a time sequence, etc.), and determine the similarity of the image features of the two engineering safety items;

[0111] A hierarchical clustering algorithm is used to cluster multiple engineering safety projects according to the image feature similarity of any two engineering safety projects, and multiple engineering safety project clusters are determined;

[0112] For each engineering safety project cluster, an initial image analysis model corresponding to the cluster center of the engineering safety project cluster is established, a training sample corresponding to the cluster center of the engineering safety project cluster is generated, and the initial image analysis model corresponding to the cluster center of the engineering safety project cluster is trained using the initial image analysis model corresponding to the cluster center of the engineering safety project cluster to generate a trained initial image analysis model corresponding to the cluster center of the engineering safety project cluster;

[0113] For each engineering safety project included in the engineering safety project cluster, the trained initial image analysis model corresponding to the cluster center of the engineering safety project cluster is adjusted, an image analysis model corresponding to the engineering safety project is established, and training samples corresponding to the engineering safety project are generated. The image analysis model corresponding to the engineering safety project is trained using the training samples corresponding to the engineering safety project to generate an image analysis model corresponding to the engineering safety project.

[0114] Figure 3 FIG. 1 is a block diagram of a system for monitoring engineering safety projects using intelligent image analysis, as shown in an embodiment of the present application. Figure 3 As shown, a system for monitoring engineering safety projects using intelligent image analysis may include a data analysis module, a data acquisition module, a project determination module, a solution determination module and an image analysis module.

[0115] The data analysis module can be used to obtain the first relevant data of multiple historical projects, wherein the first relevant data of historical projects at least include the construction plan data, construction environment data, construction personnel data and engineering safety risk data of multiple construction nodes of the historical projects. It is also used to classify multiple historical projects based on the first relevant data of multiple historical projects and determine multiple historical project classes. It is also used to determine, for each historical project class, the engineering safety items corresponding to the historical project class and the dependency coefficient between any two engineering safety items based on the first relevant data of the historical projects included in the historical project class.

[0116] The data acquisition module can be used to acquire second relevant data of the current project, wherein the second relevant data of the current project at least includes construction plan data, construction environment data and construction personnel data of the current project.

[0117] The project determination module can be used to determine similar historical engineering classes of the current project from multiple historical engineering classes based on the second relevant data of the current project and the first relevant data of multiple historical projects. It can also be used to determine engineering safety projects of multiple construction nodes of the current project based on the engineering safety projects corresponding to the similar historical engineering classes of the current project and the dependency coefficient between any two engineering safety projects.

[0118] The scheme determination module can be used for engineering safety projects based on multiple construction nodes of the current project to determine the image acquisition scheme for multiple construction nodes corresponding to the current project. It can also be used to set up multiple image acquisition devices at the construction site of the current project based on the image acquisition scheme for multiple construction nodes corresponding to the current project.

[0119] The image analysis module can be used to analyze the construction site images collected by multiple image acquisition devices based on the engineering safety items of multiple construction nodes of the current project, and generate safety prompt information.

[0120] A system for monitoring engineering safety projects using intelligent image analysis can be used to execute a method for monitoring engineering safety projects using intelligent image analysis. For more descriptions of a system for monitoring engineering safety projects using intelligent image analysis, please refer to the relevant descriptions of a method for monitoring engineering safety projects using intelligent image analysis, which will not be repeated here.

[0121] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A method for monitoring engineering safety projects using intelligent image analysis, characterized in that: include: Acquire first relevant data of a plurality of historical projects, wherein the first relevant data of the historical projects at least includes construction plan data, construction environment data, construction personnel data, and engineering safety risk data of a plurality of construction nodes of the historical projects; Based on the first relevant data of the plurality of historical projects, classifying the plurality of historical projects to determine a plurality of historical project categories; For each of the historical engineering categories, determining the engineering safety items corresponding to the historical engineering category and the dependency coefficient between any two engineering safety items based on the first relevant data of the historical engineering included in the historical engineering category; Acquire second relevant data of the current project, wherein the second relevant data of the current project at least includes construction plan data, construction environment data and construction personnel data of the current project; Determining a similar historical project class to the current project from the multiple historical project classes based on the second relevant data of the current project and the first relevant data of the multiple historical projects; Determine the engineering safety items of multiple construction nodes of the current project based on the engineering safety items corresponding to the similar historical engineering classes of the current project and the dependency coefficient between any two engineering safety items; Determine, based on the engineering safety items of the multiple construction nodes of the current project, an image acquisition scheme for the multiple construction nodes corresponding to the current project; Based on the image acquisition schemes of multiple construction nodes corresponding to the current project, multiple image acquisition devices are set up at the construction site of the current project; The construction site images collected by the multiple image acquisition devices are analyzed based on the engineering safety items of the multiple construction nodes of the current project to generate safety prompt information.

2. A method for monitoring engineering safety projects using intelligent image analysis according to claim 1, characterized in that: Based on the first relevant data of the plurality of historical projects, the plurality of historical projects are classified to determine a plurality of historical project categories, including: numbering each of said historical projects; For any two historical projects, based on the construction plan data of the two historical projects, the similarity of the construction plans of the two historical projects is calculated; based on the construction environment data of the two historical projects, the similarity of the construction environments of the two historical projects is calculated; based on the construction personnel data of the two historical projects, the similarity of the construction personnel of the two historical projects is calculated; based on the similarity of the construction plans of the two historical projects, the similarity of the construction environments and the similarity of the construction personnel, the comprehensive similarity of the two historical projects is calculated; For each historical project, based on the comprehensive similarity between the historical project and other historical projects and the serial number of each historical project, a similarity sequence corresponding to the historical project is generated; For any two historical projects, based on the similarity sequences corresponding to each of the historical projects, the sequence similarity of the two historical projects is calculated; The plurality of historical projects are classified according to the sequence similarity of any two historical projects by using a hierarchical clustering algorithm to determine a plurality of historical project classes.

3. The method for monitoring engineering safety projects using intelligent image analysis according to claim 1, characterized in that: Determining, based on first relevant data of historical projects included in the historical project category, engineering safety items corresponding to the historical project category and a dependency coefficient between any two engineering safety items, includes: For each historical project included in the historical project category, determining the engineering safety items of the multiple construction nodes of the historical project based on the engineering safety risk data of the multiple construction nodes of the historical project; Determining the engineering safety project corresponding to the historical engineering class based on the engineering safety projects of multiple construction nodes of each historical engineering included in the historical engineering class; For each engineering safety project corresponding to the historical engineering category, based on the engineering safety risk data of each historical engineering project included in the historical engineering category, determine the project risk sequence of the engineering safety project; For any two engineering safety projects corresponding to the historical engineering class, the dependency coefficient between the two engineering safety projects is calculated based on the project risk sequences of the two engineering safety projects.

4. The method for monitoring engineering safety projects using intelligent image analysis according to claim 1, characterized in that: Determining a similar historical project class of the current project from the multiple historical project classes based on the second relevant data of the current project and the first relevant data of the multiple historical projects includes: S11. For each of the historical project classes, a plurality of historical projects are sampled from the historical project class as sampled historical projects, and the construction plan similarity between the sampled historical project and the current project is calculated based on the construction plan data of the sampled historical project and the construction plan data of the current project; the construction environment similarity between the sampled historical project and the current project is calculated based on the construction environment data of the sampled historical project and the construction environment data of the current project; the construction personnel similarity between the sampled historical project and the current project is calculated based on the construction personnel data of the sampled historical project and the construction personnel data of the current project; the comprehensive similarity between the sampled historical project and the current project is calculated based on the construction plan similarity, construction environment similarity and construction personnel similarity between the sampled historical project and the current project; and the engineering similarity between the current project and the historical project class is calculated based on the comprehensive similarity between the current project and each of the sampled historical projects; S12, determine whether the current sampling round is greater than the sampling round threshold, if not, execute S11, if yes, execute S13; S13, calculating a global fluctuation parameter sequence based on the similarity between the current project and each of the historical project types in each round of sampling, wherein the global fluctuation parameter sequence includes global fluctuation parameters corresponding to any two adjacent sampling rounds; S14, based on the global fluctuation parameter sequence, determine whether to perform the next round of sampling, if yes, execute S11, if not, execute S15; S15, determining a target sampling round based on the global fluctuation parameter sequence; S16. Determine a similar historical engineering class to the current project from the multiple historical engineering classes based on the engineering similarity between the current project and each of the historical engineering classes in the target sampling round.

5. A method for monitoring engineering safety projects using intelligent image analysis according to any one of claims 1 to 4, characterized in that: Based on the engineering safety items corresponding to the similar historical engineering classes of the current project and the dependency coefficient between any two engineering safety items, the engineering safety items of multiple construction nodes of the current project are determined, including: Establishing a safety project prediction model corresponding to each of the historical engineering categories; Based on similar historical engineering classes of the current engineering, a target safety project prediction model is retrieved; Extracting characteristic data of the current project from second related data of the current project; Determining a first engineering safety project of a plurality of construction nodes of the current project based on the characteristic data of the current project by using the target safety project prediction model; According to the dependency coefficient between any two engineering safety items and the first engineering safety items of the multiple construction nodes of the current project, the second engineering safety items of the multiple construction nodes of the current project are determined.

6. A method for monitoring engineering safety projects using intelligent image analysis according to claim 5, characterized in that: Establishing a safety project prediction model corresponding to each of the historical engineering categories, including: Calculate the clustering parameters of any two historical engineering classes according to the engineering safety items corresponding to each of the historical engineering classes and the dependency coefficient between any two engineering safety items; Clustering the plurality of historical engineering classes according to clustering parameters of any two of the historical engineering classes by a hierarchical clustering algorithm to determine a plurality of historical engineering class clusters; For each of the historical engineering clusters, an initial prediction model corresponding to the cluster center of the historical engineering cluster is established, a training sample corresponding to the cluster center of the historical engineering cluster is generated, and the initial prediction model is trained using the training sample corresponding to the cluster center of the historical engineering cluster to generate a trained initial prediction model corresponding to the cluster center of the historical engineering cluster; For each historical engineering class included in the historical engineering class cluster, the trained initial prediction model corresponding to the cluster center of the historical engineering class cluster is adjusted, a safety project prediction model corresponding to the historical engineering class is established, and a training sample corresponding to the historical engineering class is generated. The training sample corresponding to the historical engineering class is used to train the safety project prediction model corresponding to the historical engineering class to generate a safety project prediction model corresponding to the historical engineering class.

7. A method for monitoring engineering safety projects using intelligent image analysis according to any one of claims 1 to 4, characterized in that: Based on the engineering safety items of the multiple construction nodes of the current project, determining the image acquisition scheme of the multiple construction nodes corresponding to the current project includes: For each construction node of the current project, determining multiple image monitoring locations based on the engineering safety items of the construction node; Based on multiple image monitoring positions of each construction node of the current project, generating multiple candidate image acquisition schemes for multiple construction nodes corresponding to the current project; Establish a program evaluation system; A plurality of candidate image acquisition schemes for a plurality of construction nodes corresponding to the current project are screened according to the scheme evaluation system to determine an image acquisition scheme for a plurality of construction nodes corresponding to the current project.

8. A method for monitoring engineering safety projects using intelligent image analysis according to any one of claims 1 to 4, characterized in that: The construction site images collected by the multiple image acquisition devices are analyzed based on the engineering safety items of the multiple construction nodes of the current project to generate safety prompt information, including: For each of the engineering safety items, an image analysis model corresponding to the engineering safety item is established; Based on the engineering safety items of multiple construction nodes of the current project, multiple target image analysis models are retrieved; For each of the target image analysis models, the construction site image corresponding to the target image analysis model is extracted from the construction site images acquired by the multiple image acquisition devices, and the construction site image corresponding to the target image analysis model is analyzed by the target image analysis model to generate safety prompt information.

9. A method for monitoring engineering safety projects using intelligent image analysis according to claim 8, characterized in that: For each of the engineering safety items, an image analysis model corresponding to the engineering safety item is established, including: For any two of the engineering safety items, determine the image features corresponding to the two engineering safety items, and determine the similarity of the image features of the two engineering safety items; A hierarchical clustering algorithm is used to cluster multiple engineering safety projects according to the image feature similarity of any two engineering safety projects, and multiple engineering safety project clusters are determined; For each of the engineering safety project clusters, an initial image analysis model corresponding to the cluster center of the engineering safety project cluster is established, a training sample corresponding to the cluster center of the engineering safety project cluster is generated, and the initial image analysis model corresponding to the cluster center of the engineering safety project cluster is trained using the training sample corresponding to the cluster center of the engineering safety project cluster to generate a trained initial image analysis model corresponding to the cluster center of the engineering safety project cluster; For each engineering safety project included in the engineering safety project cluster, the trained initial image analysis model corresponding to the cluster center of the engineering safety project cluster is adjusted, an image analysis model corresponding to the engineering safety project is established, and a training sample corresponding to the engineering safety project is generated. The image analysis model corresponding to the engineering safety project is trained using the training sample corresponding to the engineering safety project to generate an image analysis model corresponding to the engineering safety project.

10. A system for monitoring engineering safety projects using intelligent image analysis, characterized in that: A method for monitoring engineering safety projects using intelligent image analysis according to any one of claims 1 to 9, comprising: A data analysis module, for obtaining first relevant data of a plurality of historical projects, wherein the first relevant data of the historical projects at least include construction plan data, construction environment data, construction personnel data and engineering safety risk data of a plurality of construction nodes of the historical projects, and for classifying the plurality of historical projects based on the first relevant data of the plurality of historical projects to determine a plurality of historical project classes, and for determining, for each of the historical project classes, engineering safety items corresponding to the historical project class and a dependency coefficient between any two engineering safety items based on the first relevant data of the historical projects included in the historical project class; A data acquisition module, used to acquire second relevant data of the current project, wherein the second relevant data of the current project at least includes construction plan data, construction environment data and construction personnel data of the current project; A project determination module, configured to determine a similar historical project class of the current project from the multiple historical project classes based on the second relevant data of the current project and the first relevant data of the multiple historical projects, and further configured to determine the engineering safety projects of multiple construction nodes of the current project based on the engineering safety projects corresponding to the similar historical project classes of the current project and the dependency coefficient between any two engineering safety projects; A scheme determination module, used to determine the image acquisition schemes for the multiple construction nodes corresponding to the current project based on the engineering safety items of the multiple construction nodes of the current project, and also used to set up multiple image acquisition devices at the construction site of the current project based on the image acquisition schemes for the multiple construction nodes corresponding to the current project; The image analysis module is used to analyze the construction site images collected by the multiple image acquisition devices based on the engineering safety items of multiple construction nodes of the current project to generate safety prompt information.

Images