A construction risk assessment method and device based on multi-source data fusion

Abstract

This invention discloses a construction risk assessment method and apparatus based on multi-source data fusion. The method includes: acquiring a construction scene image; and based on the construction scene image, determining the objects corresponding to the construction scene image and the category information corresponding to the objects, wherein the objects include workers and objects on the construction site; based on the objects and the category information, determining worker behavior information and the correlation between the workers and the objects on the construction site, wherein the correlation reflects the protective relationship or potential harm relationship between the objects on the construction site and the workers; and based on the worker behavior information and the correlation, determining construction risk level information. This invention enables refined assessment of construction risks, determines the construction risk level, and fully considers the characteristics of the construction scene, which is beneficial for accurately assessing construction risks.

Description

Technical Field

[0001] This invention relates to the field of construction risk assessment technology, and in particular to a construction risk assessment method and apparatus based on multi-source data fusion. Background Technology

[0002] The construction industry, as a vital force driving my country's economic and social development, employs over 70 million people. Due to the complex construction environment and sites, safety risks for construction workers are easily created.

[0003] Current image-based construction risk behavior recognition methods only focus on individual worker behaviors (such as smoking, not wearing a safety helmet, not wearing a safety belt, etc.) but fail to fully consider scene characteristics and cannot accurately assess construction risks.

[0004] Therefore, existing technologies still need to be improved and enhanced. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a construction risk assessment method and device based on multi-source data fusion, which addresses the above-mentioned deficiencies of the prior art. The aim is to solve the problem that the prior art cannot accurately assess construction risks by fully considering the characteristics of the scenario.

[0006] In a first aspect, the present invention provides a construction risk assessment method based on multi-source data fusion, wherein the method includes:

[0007] Acquire construction scene images, and based on the construction scene images, determine the objects corresponding to the construction scene images and the category information corresponding to the objects, wherein the objects include workers and objects on the construction site;

[0008] Based on the object and category information, worker behavior information and the association between the worker and the objects at the construction site are determined, wherein the association is used to reflect the protective relationship or potential harm relationship between the objects at the construction site and the worker;

[0009] Based on the worker behavior information and the correlation, the construction risk level information is determined.

[0010] In one implementation, determining the object corresponding to the construction scene image and the category information corresponding to the object based on the construction scene image includes:

[0011] A pre-trained image segmentation model is obtained, and the construction scene image is input into the image segmentation model;

[0012] The construction scene image is segmented based on the image segmentation model, and the objects and category information in the construction scene image are identified. The image segmentation model is trained based on the YOLOv5 model.

[0013] In one implementation, the training process of the image segmentation model includes:

[0014] Collect scene sample images, which include the positional and distance relationships between workers and objects at the construction site in different scenes;

[0015] The workers and construction site objects in the scene sample image are marked, and the category information of the construction site objects is marked;

[0016] The scene sample images of the labeled workers, construction site objects, and the corresponding category information of the construction site objects are input into the YOLOv5 model for training to obtain the image segmentation model.

[0017] In one implementation, determining worker behavior information and the association between the worker and the objects at the construction site based on the object and the category information includes:

[0018] The positional and distance relationships between the worker and the objects at the construction site are obtained, and the worker's behavior information is determined based on the positional and distance relationships.

[0019] Based on the category information and in combination with the location and distance relationships, the impact information of the objects at the construction site on the workers is determined. The impact information includes the positive and negative impacts of the objects at the construction site on the workers.

[0020] Based on the influence information, the relationship between the worker and the objects at the construction site is determined.

[0021] In one implementation, determining worker behavior information and the association between the worker and the objects at the construction site based on the object and the category information further includes:

[0022] A pre-trained behavior recognition model is obtained, and the construction scene image after determining the object and the category information corresponding to the object is input into the behavior recognition model;

[0023] Based on the behavior recognition model, the worker behavior information is output.

[0024] In one implementation, the training process of the behavior recognition model includes:

[0025] Workers and construction site objects in several scene sample images are labeled in advance, and the category information of the construction site objects is determined.

[0026] Based on image recognition analysis, the behavioral information of workers in each scene sample image is obtained, and the behavioral information, workers and category information are bound together to obtain a mapping relationship;

[0027] The behavior recognition model is obtained by training the mapping relationship based on the residual convolutional neural network.

[0028] In one implementation, determining the construction risk level information based on the worker behavior information and the correlation includes:

[0029] If the worker behavior information is a safe behavior and the association relationship is a protective relationship, then the construction risk level information is determined to be a low-risk level.

[0030] If the worker behavior information is a dangerous behavior and the association relationship is a protective relationship, then the construction risk level information is determined to be the second lowest risk level.

[0031] If the worker behavior information is a safe behavior and the correlation is a potential harm relationship, then the construction risk level information is determined to be a high risk level.

[0032] If the worker behavior information is a dangerous behavior and the correlation is a potential harm relationship, then the construction risk level information is determined to be a high-risk level.

[0033] Secondly, embodiments of the present invention also provide a construction risk assessment device based on multi-source data fusion, characterized in that the device comprises:

[0034] The scene image analysis module is used to acquire construction scene images and, based on the construction scene images, determine the objects corresponding to the construction scene images and the category information corresponding to the objects, wherein the objects include workers and objects on the construction site;

[0035] The association analysis module is used to determine worker behavior information and the association between the worker and the objects at the construction site based on the object and the category information. The association is used to reflect the protective relationship or potential harm relationship between the objects at the construction site and the worker.

[0036] The construction risk assessment module is used to determine the construction risk level information based on the worker behavior information and the correlation.

[0037] Thirdly, embodiments of the present invention also provide a terminal device, wherein the terminal device is a commercial display terminal or a projection terminal, the terminal device includes a memory, a processor, and a construction risk assessment program based on multi-source data fusion stored in the memory and executable on the processor, wherein when the processor executes the construction risk assessment program based on multi-source data fusion, it implements the steps of the construction risk assessment method based on multi-source data fusion in any of the above schemes.

[0038] Fourthly, embodiments of the present invention also provide a computer-readable storage medium, wherein the computer-readable storage medium stores a construction risk assessment program based on multi-source data fusion, and when the construction risk assessment program based on multi-source data fusion is executed by a processor, it implements the steps of the construction risk assessment method based on multi-source data fusion as described in any of the above schemes.

[0039] Beneficial Effects: Compared with existing technologies, this invention provides a construction risk assessment method based on multi-source data fusion. First, this invention acquires construction scene images and, based on these images, determines the objects corresponding to the images and their corresponding category information. The objects include workers and other objects on the construction site. Based on the objects and category information, it determines worker behavior information and the correlation between the workers and the objects on the construction site. This correlation reflects the protective or potential harm relationship between the objects on the construction site and the workers. Based on the worker behavior information and the correlation, it determines the construction risk level. This invention enables refined assessment of construction risks, determines the construction risk level, and fully considers the characteristics of the construction scene, thus facilitating accurate risk assessment. Attached Figure Description

[0040] Figure 1 A flowchart illustrating a specific implementation of the construction risk assessment method based on multi-source data fusion provided in this embodiment of the invention.

[0041] Figure 2 A functional principle diagram of a construction risk assessment device based on multi-source data fusion provided in an embodiment of the present invention.

[0042] Figure 3 A schematic diagram of a terminal device provided in an embodiment of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0044] This embodiment provides a construction risk assessment method based on multi-source data fusion. Based on this method, a refined assessment of construction risks can be achieved, determining the risk level and fully considering the characteristics of the construction scenario, which is beneficial for accurate risk assessment. In specific implementation, this embodiment first acquires images of the construction scene and, based on these images, determines the objects corresponding to the images and their corresponding category information. The objects include workers and other objects on the construction site. Then, based on the objects and category information, worker behavior information and the correlation between the workers and the objects on the construction site are determined. This correlation reflects the protective or potential harm relationship between the objects on the construction site and the workers. Finally, based on the worker behavior information and the correlation, the construction risk level information is determined. Therefore, this embodiment determines the relationship between workers and objects on the construction site based on construction scene images. This relationship reflects the protective relationship or potential harm relationship between the objects on the construction site and the workers. Furthermore, this embodiment can also identify workers' behavior information. By comprehensively considering workers' behavior information and the relationship, and realizing the fusion analysis of multi-source data, the construction risk level information can be determined, thereby achieving a refined assessment of construction risks.

[0045] Exemplary methods

[0046] The construction risk assessment method based on multi-source data fusion in this embodiment is applied to terminal equipment, which includes intelligent product terminals such as computers. Specifically, as shown... Figure 1 As shown in the figure, the construction risk assessment method based on multi-source data fusion in this embodiment includes the following steps:

[0047] Step S100: Obtain a construction scene image, and based on the construction scene image, determine the object corresponding to the construction scene image and the category information corresponding to the object, wherein the object includes workers and objects on the construction site.

[0048] This embodiment first acquires a construction scene image, which reflects people or objects at the construction site. The construction scene image is a direct photograph of the construction site; therefore, it includes workers and other objects. Different objects have different functions and impacts on workers. Therefore, this embodiment needs to determine the category information of these objects (including workers and construction site objects) to determine their impact on workers in subsequent steps. This facilitates analysis of whether the objects provide protection or harm, thus helping to identify construction risks.

[0049] In one implementation, this embodiment includes the following steps when determining category information:

[0050] Step S101: Obtain a pre-trained image segmentation model and input the construction scene image into the image segmentation model;

[0051] Step S102: Segment the construction scene image based on the image segmentation model, and identify the objects and category information in the construction scene image, wherein the image segmentation model is trained based on the YOLOv5 model.

[0052] Specifically, in this embodiment, an image segmentation model is pre-set to segment construction scene images and identify workers and construction site objects within them. First, this embodiment acquires scene sample images, which include the positional and distance relationships between workers and construction site objects in different scenarios. Therefore, these scene sample images include both workers and construction site objects. Next, this embodiment uses image recognition to identify the workers and construction site objects in each scene sample image and labels them, along with the category information of the construction site objects. At this point, the category information of the construction site objects in each scene sample image has been labeled. Therefore, this embodiment inputs the labeled scene sample images containing workers, construction site objects, and their corresponding category information into the YOLOv5 model for training to obtain the image segmentation model. When a construction scene image is input into the image segmentation model, the model can automatically identify the workers and construction site objects in the image and automatically output the category information corresponding to the construction site objects. For example, if the object at the construction site is a safety helmet, the corresponding category information is protective equipment; if the object at the construction site is an excavator, the corresponding category information is ground construction equipment.

[0053] In another implementation, this embodiment can also identify workers and construction site objects from construction scene images based on image recognition. Then, the identified construction site objects are directly compared with a pre-set construction site image library. This library stores images of various equipment at the construction site from different angles, and the images in the library are categorized according to the type of construction site object. Therefore, by comparing the construction site objects with the pre-set construction site image library, it is possible to determine what the specific construction site object is and its category information.

[0054] Step S200: Based on the object and the category information, determine the worker behavior information and the association between the worker and the objects at the construction site, wherein the association is used to reflect the protective relationship or potential harm relationship between the objects at the construction site and the worker.

[0055] Once the workers / construction site objects in the construction scene image and their corresponding category information are identified, this embodiment can determine worker behavior information based on the objects (workers / construction site objects) and the category information. This worker behavior information reflects what work the worker is performing in the construction scene. Furthermore, this embodiment can also determine the association between the worker and the construction site objects, whereby the association reflects the protective or potential harm relationship that the construction site objects provide to the worker.

[0056] In one implementation, step S200 of this embodiment specifically includes the following steps:

[0057] Step S201: Obtain the positional relationship and distance relationship between the worker and the objects at the construction site, and determine the worker's behavior information based on the positional relationship and the distance relationship;

[0058] Step S202: Based on the category information and in combination with the positional and distance relationships, determine the impact information of the objects at the construction site on the workers. The impact information includes the positive and negative impacts of the objects at the construction site on the workers.

[0059] Step S203: Based on the influence information, determine the relationship between the worker and the objects at the construction site.

[0060] Specifically, this embodiment first obtains the positional and distance relationships between the worker and objects at the construction site. These relationships reflect whether the worker is operating the objects (e.g., operating an excavator) or wearing appropriate protective gear (e.g., a safety helmet). Therefore, this embodiment can determine worker behavior information based on these relationships, identifying the work the worker is performing. In determining the positional and distance relationships between the worker and objects at the construction site, image recognition technology can be used to identify the workers and objects in the construction scene image, determining their positional relationships. Then, based on the distance scale in the construction scene image, the distance relationship between the worker and objects is obtained. Finally, based on these positional and distance relationships, worker behavior information can be determined. This worker behavior information includes both safe and dangerous behaviors.

[0061] In another implementation, when recognizing worker behavior information in this embodiment, the construction scene image after determining the object and its corresponding category information can be input into a pre-set behavior recognition model. Then, based on the behavior recognition model, the worker behavior information is output. Specifically, in this embodiment, workers and construction site objects in several scene sample images are pre-labeled, and the category information of the construction site objects is determined. Then, based on image recognition analysis, the worker behavior information in each scene sample image is obtained. This behavior information can be determined by recognizing the movement posture of the worker's limbs in each scene sample image. Then, the behavior information, worker, and category information are bound together to obtain a mapping relationship. Next, the mapping relationship is trained based on a residual convolutional neural network to obtain the behavior recognition model. This behavior recognition model can then determine the category information of workers and construction site objects, as well as the corresponding behavior information of workers, from the construction scene image. Therefore, when the construction scene image after determining the object and its corresponding category information is input into the behavior recognition model, the worker behavior information can be output based on the behavior recognition model.

[0062] Next, this embodiment can determine the impact information of the construction site objects on the workers based on the category information of the objects, combined with the positional and distance relationships. This impact information includes both positive and negative impacts of the construction site objects on the workers. For example, based on positional and distance relationships, it can be determined that the safety helmet (i.e., the construction site object) is located above the worker's head, indicating the worker is wearing a safety helmet. Since the safety helmet's category information is protective equipment, its impact on the worker can be determined to be positive. As another example, based on positional and distance relationships, it can be determined that the worker is located on a high-rise building (i.e., the construction site object). The worker's behavior information is high-altitude work, and the high-rise building's category information is dangerous construction. Therefore, its impact on the worker can be determined to be negative. Once the impact information is determined, this embodiment can determine the association between the worker and the construction site objects based on this information. In this embodiment, when the impact relationship is positive, it can be determined that the construction site objects provide protection to the worker (e.g., the safety helmet provides protection), therefore, the association between the construction site objects and the worker is a protective relationship. When the influence relationship is negative, it can be determined that the objects at the construction site pose a potential harm to the workers (such as the potential harm to workers from working on high-rise buildings). Therefore, the relationship between the objects at the construction site and the workers is a potential harm relationship.

[0063] Step S300: Based on the worker behavior information and the correlation, determine the construction risk level information.

[0064] In this embodiment, worker behavior information includes safe behaviors and dangerous behaviors, and the relationships between them include protective relationships and potential harm relationships. Therefore, after determining the worker behavior information and the relationships between them, this embodiment can comprehensively consider the worker behavior information and the relationships between them to determine the construction risk level information.

[0065] Specifically, if the worker's behavior information is a safe behavior and the association is a protective relationship, then the construction risk level information is determined to be low-risk. If the worker's behavior information is a dangerous behavior and the association is a protective relationship, then the construction risk level information is determined to be the second lowest risk level. If the worker's behavior information is a safe behavior and the association is a potential harm relationship, then the construction risk level information is determined to be high-risk. If the worker's behavior information is a dangerous behavior and the association is a potential harm relationship, then the construction risk level information is determined to be high-risk. In this embodiment, the risk index is: low-risk level < second lowest risk level < higher risk level < high-risk level.

[0066] Furthermore, this embodiment can also train a risk assessment model. This embodiment can construct the correspondence between worker, individual behavior, construction site object category information, and construction risk level information. The worker, individual behavior, and construction site object category information are used as independent variables, and the construction risk level information is used as the dependent variable for training, resulting in the risk assessment model. This risk assessment model can automatically output construction risk level information directly after identifying workers and construction site objects from construction scene images, thereby achieving automatic assessment of construction risks.

[0067] In summary, this embodiment first acquires a construction scene image, and based on the image, determines the objects corresponding to the image and their corresponding category information. The objects include workers and other objects on the construction site. Based on the objects and category information, it determines worker behavior information and the relationship between the worker and the objects on the construction site. This relationship reflects the protective or potential harm relationship between the objects and the worker. Based on the worker behavior information and the relationship, it determines the construction risk level. This embodiment enables a refined assessment of construction risks, determines the risk level, and fully considers the characteristics of the construction scene, thus facilitating accurate risk assessment.

[0068] Exemplary device

[0069] Based on the above embodiments, the present invention also provides a construction risk assessment device based on multi-source data fusion, such as... Figure 2As shown in the diagram, the apparatus of this embodiment includes: a scene image analysis module 10, a correlation analysis module 20, and a construction risk assessment module 30. Specifically, the scene image analysis module 10 in this embodiment is used to acquire construction scene images and, based on the construction scene images, determine the objects corresponding to the construction scene images and the category information corresponding to the objects, wherein the objects include workers and objects on the construction site. The correlation analysis module 20 is used to determine worker behavior information and the correlation between the workers and the objects on the construction site based on the objects and the category information, wherein the correlation is used to reflect the protective relationship or potential harm relationship caused by the objects on the construction site to the workers. The construction risk assessment module 30 is used to determine construction risk level information based on the worker behavior information and the correlation.

[0070] In one implementation, the scene image analysis module 10 includes:

[0071] An image input unit is used to acquire a pre-trained image segmentation model and input the construction scene image into the image segmentation model;

[0072] An image processing unit is used to segment the construction scene image based on the image segmentation model and identify objects and category information in the construction scene image, wherein the image segmentation model is trained based on the YOLOv5 model.

[0073] In one implementation, the apparatus includes an image segmentation model training module, the image segmentation model training module comprising:

[0074] An image acquisition unit is used to acquire scene sample images, which include the positional and distance relationships between workers and objects at the construction site in different scenes;

[0075] An information labeling unit is used to label the workers and construction site objects in the scene sample image, and to label the category information of the construction site objects;

[0076] The model training unit is used to input scene sample images of labeled workers, construction site objects, and the corresponding category information of the construction site objects into the YOLOv5 model for training, so as to obtain the image segmentation model.

[0077] In one implementation, the association analysis module 20 includes:

[0078] The behavior analysis unit is used to obtain the positional and distance relationships between the worker and the objects at the construction site, and to determine the worker's behavior information based on the positional and distance relationships.

[0079] An impact analysis unit is used to determine the impact information of the objects at the construction site on the workers based on the category information and in combination with the location and distance relationships. The impact information includes the positive and negative impacts of the objects at the construction site on the workers.

[0080] The relationship determination unit is used to determine the association between the worker and the objects at the construction site based on the influence information.

[0081] In one implementation, the association analysis module 20 further includes:

[0082] The model acquisition unit is used to acquire a pre-trained behavior recognition model and input the construction scene image after determining the object and the category information corresponding to the object into the behavior recognition model;

[0083] The behavior recognition unit is used to output the worker behavior information based on the behavior recognition model.

[0084] In one implementation, the device further includes a behavior recognition model training module, wherein the behavior recognition model training model includes:

[0085] The information annotation processing unit is used to pre-annotate workers and construction site objects in several scene sample images and determine the category information of the construction site objects;

[0086] The mapping relationship establishment unit is used to analyze the worker's behavior information in each scene sample image based on image recognition, and bind the behavior information, worker, and category information to obtain the mapping relationship;

[0087] The behavior recognition model training unit trains the mapping relationship based on a residual convolutional neural network to obtain the behavior recognition model.

[0088] In one implementation, the construction risk assessment module 30 includes:

[0089] The first risk level determination unit is used to determine the construction risk level information as low risk level if the worker behavior information is a safe behavior and the association relationship is a protective relationship.

[0090] The second risk level determination unit is used to determine the construction risk level information as the second lowest risk level if the worker behavior information is a dangerous behavior and the association relationship is a protective relationship.

[0091] The third risk level determination unit is used to determine the construction risk level information as a higher risk level if the worker behavior information is a safe behavior and the correlation is a potential injury relationship.

[0092] The fourth risk level determination unit is used to determine the construction risk level information as high-risk if the worker behavior information is dangerous behavior and the correlation is a potential injury relationship.

[0093] The working principle of each template in the construction risk assessment device based on multi-source data fusion in this embodiment is the same as the principle of each step in the above method embodiment, and will not be repeated here.

[0094] Based on the above embodiments, the present invention also provides a terminal device, the principle block diagram of which is shown in Figure 3. The terminal device is the host computer in the above embodiments, such as a computer device. The terminal device may include one or more processors 100. Figure 3 (Only one is shown in the image), a memory 101, and a computer program 102 stored in the memory 101 and executable on one or more processors 100, such as a program for construction risk assessment based on multi-source data fusion. When one or more processors 100 execute the computer program 102, they can implement the various steps in the method embodiment of construction risk assessment based on multi-source data fusion. Alternatively, when one or more processors 100 execute the computer program 102, they can implement the functions of each template / unit in the apparatus embodiment of construction risk assessment based on multi-source data fusion, which is not limited here.

[0095] In one embodiment, the processor 100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0096] In one embodiment, memory 101 may be an internal storage unit of an electronic device, such as a hard drive or RAM. Memory 101 may also be an external storage device of the electronic device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, memory 101 may include both internal and external storage units. Memory 101 is used to store computer programs and other programs and data required by the terminal device. Memory 101 can also be used to temporarily store data that has been output or will be output.

[0097] Those skilled in the art will understand that Figure 3 The block diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal device to which the present invention is applied. The specific terminal device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0098] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the methods described above. Any references to memory, storage, operational databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual operating data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0099] In summary, this invention discloses a construction risk assessment method and apparatus based on multi-source data fusion. The method includes: acquiring a construction scene image; determining, based on the construction scene image, the objects corresponding to the construction scene image and the category information corresponding to the objects, wherein the objects include workers and objects on the construction site; determining worker behavior information and the correlation between the workers and the objects on the construction site based on the objects and the category information, wherein the correlation reflects the protective relationship or potential harm relationship between the objects on the construction site and the workers; and determining construction risk level information based on the worker behavior information and the correlation. This invention enables refined assessment of construction risks, determines the construction risk level, and fully considers the characteristics of the construction scene, which is beneficial for accurately assessing construction risks.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A construction risk assessment method based on multi-source data fusion, characterized in that, The method comprises: acquiring a construction scene image, and determining an object corresponding to the construction scene image and category information corresponding to the object based on the construction scene image, wherein the object comprises a worker and a construction site object; determining worker behavior information and an association relationship between the worker and the construction site object based on the object and the category information, wherein the association relationship is used to reflect a protection relationship or a potential harm relationship caused by the construction site object to the worker; determining construction risk level information based on the worker behavior information and the association relationship; the determination of the object corresponding to the construction scene image and the category information corresponding to the object based on the construction scene image comprises: acquiring a pre-trained image segmentation model, and inputting the construction scene image into the image segmentation model; performing segmentation processing on the construction scene image based on the image segmentation model, and identifying the object and the category information in the construction scene image, wherein the image segmentation model is trained based on a YOLOv5 model; alternatively, the determination of the object corresponding to the construction scene image and the category information corresponding to the object based on the construction scene image comprises: identifying the worker and the construction site object from the construction scene image based on an image recognition manner, comparing the identified construction site object with a preset construction site library, the construction site library stores images of various equipment in the construction site at various angles, and the images in the construction site library are classified and set according to the category information of the construction site object; after comparing the construction site object with the preset construction site library, the category information of the construction site object can be determined; the determination of the worker behavior information and the association relationship between the worker and the construction site object based on the object and the category information comprises: acquiring a position relationship and a distance relationship between the worker and the construction site object, and determining the worker behavior information according to the position relationship and the distance relationship, the position relationship and the distance relationship can reflect whether the worker is operating the construction site object or whether the worker is wearing the construction site object; determining influence information of the construction site object on the worker according to the category information and in combination with the position relationship and the distance relationship, the influence information comprises positive influence and negative influence of the construction site object on the worker; determining the association relationship between the worker and the construction site object based on the influence information; the method further comprises: building a corresponding relationship among the worker, personal behavior, category information of the construction site object and construction risk level information, and training the worker, personal behavior, category information of the construction site object as independent variables and the construction risk level information as a dependent variable to obtain a risk assessment model, the risk assessment model can automatically output the construction risk level information after identifying the worker and the construction site object from the construction scene image, and realize automatic assessment of the construction risk. 2.The construction risk assessment method based on multi-source data fusion according to claim 1, characterized in that, The training process of the image segmentation model comprises: collecting scene sample images, the scene sample images comprising position relationships and distance relationships between workers and construction site objects in different scenes; labeling the workers and the construction site objects in the scene sample images and labeling category information of the construction site objects; inputting the labeled workers, construction site objects and corresponding category information of the construction site objects in the scene sample images into the YOLOv5 model for training to obtain the image segmentation model. 3.The construction risk assessment method based on multi-source data fusion according to claim 1, characterized in that, The determining of worker behavior information and the association relationship between the worker and the construction site object based on the object and the category information further comprises: obtaining a pre-trained behavior recognition model and inputting a construction scene image after determining the object and the category information corresponding to the object into the behavior recognition model; outputting the worker behavior information based on the behavior recognition model. 4.The construction risk assessment method based on multi-source data fusion according to claim 3, characterized in that, The training process of the behavior recognition model comprises: pre-labeling workers and construction site objects in a plurality of scene sample images and determining category information of the construction site objects; analyzing worker behavior information in each scene sample image based on image recognition and binding the behavior information, the worker and the category information to obtain a mapping relationship; training the mapping relationship based on a residual convolutional neural network to obtain the behavior recognition model. 5.The construction risk assessment method based on multi-source data fusion according to claim 1, characterized in that, The determining of construction risk level information based on the worker behavior information and the association relationship comprises: if the worker behavior information is safe behavior and the association relationship is a protection relationship, determining that the construction risk level information is a low risk level; if the worker behavior information is dangerous behavior and the association relationship is a protection relationship, determining that the construction risk level information is a second-low risk level; if the worker behavior information is safe behavior and the association relationship is a potential harm relationship, determining that the construction risk level information is a higher risk level; if the worker behavior information is dangerous behavior and the association relationship is a potential harm relationship, determining that the construction risk level information is a high risk level.

6. A construction risk assessment device based on multi-source data fusion, characterized by, The device is used to implement steps of the construction risk assessment method based on multi-source data fusion according to any one of claims 1-5, and the device comprises: a scene image analysis module configured to obtain a construction scene image and determine an object corresponding to the construction scene image and category information corresponding to the object based on the construction scene image, wherein the object comprises a worker and a construction site object; an association relationship analysis module configured to determine worker behavior information and an association relationship between the worker and the construction site object based on the object and the category information, wherein the association relationship is used to reflect a protection relationship or a potential harm relationship caused by the construction site object to the worker; a construction risk assessment module configured to determine construction risk level information based on the worker behavior information and the association relationship.

7. A terminal device, characterized by, The terminal device comprises a memory, a processor, and a construction risk assessment program based on multi-source data fusion stored in the memory and executable on the processor; when the processor executes the construction risk assessment program based on multi-source data fusion, the steps of the construction risk assessment method based on multi-source data fusion according to any one of claims 1-5 are implemented.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a construction risk assessment program based on multi-source data fusion; when the processor executes the construction risk assessment program based on multi-source data fusion, the steps of the construction risk assessment method based on multi-source data fusion according to any one of claims 1-5 are implemented.

Images