Building prefabricated part hoisting guiding method and system based on multi-modal perception and dynamic knowledge graph
By using a multimodal perception and dynamic knowledge graph-based method for guiding the hoisting of prefabricated building components, multimodal data is acquired and fused in real time. The hoisting dynamic knowledge graph is used for risk assessment and multi-level machine instructions are generated. This solves the problems of data fusion difficulties and risk response lag in the hoisting process in existing technologies, and achieves efficient and safe hoisting control.
Patent Information
- Application Number
- CN202511260290.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-04
- Publication Date
- 2026-01-20
AI Technical Summary
In the current process of hoisting prefabricated building components, multimodal data fusion is difficult, risk identification and response are delayed, hoisting guidance decisions are inaccurate and have low precision, rely on human experience, have limited safety assurance, and have poor machine equipment instruction executability and poor human information readability.
A method for guiding the hoisting of prefabricated building components using multimodal perception and dynamic knowledge graphs is adopted. Multimodal data is acquired in real time, dynamic feature weighting and fusion are performed, and similarity retrieval is carried out using a dynamic knowledge graph library for prefabricated building hoisting. Machine-executable instructions are generated to achieve multi-level risk response and collaborative control.
It improves the accuracy and efficiency of the hoisting process, reduces risk response delays, generates professional and executable machine instructions, enhances human-machine collaboration efficiency and hoisting precision, and ensures construction safety.
Smart Images
Figure CN121365795A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the technical field of intelligent construction, and more particularly relates to a building prefabricated component hoisting guiding method and system based on multi-modal perception and dynamic knowledge graph. BACKGROUND
[0002] Building prefabricated component hoisting guidance is a key link to ensure safe, efficient and precise construction of fabricated buildings. It realizes the installation of components by guiding and controlling the hoisting of components. The current guidance of building prefabricated component hoisting is basically manually operated by tower crane drivers, ground command personnel and other relevant construction personnel.
[0003] In the process of building prefabricated component hoisting guidance, various hoisting state information such as the image of the component, the distance between the component and the surrounding obstacles, the rotation angle and angular velocity and angular acceleration of the hoisting equipment, the swing amplitude of the component, and the environmental wind speed are involved. It is difficult for construction personnel to accurately and efficiently use these multi-modal data to judge the current hoisting situation at the same time. There are problems such as difficulty in multi-source heterogeneous sensor data fusion, significant time delay in risk identification and response, etc., which further lead to inaccurate hoisting guidance decision, low hoisting efficiency and limited precision, and difficulty in real-time judgment of the current risk situation and timely response according to the risk situation. In addition, the executability of the mechanical equipment execution instructions generated by the existing method is poor, and the human readability of the hoisting guidance information is poor. Therefore, the current hoisting of building prefabricated components has core problems such as risk response lag, hoisting precision depending on manual experience, limited safety protection, and low hoisting quality and efficiency. SUMMARY
[0004] In view of the above defects or improvement needs of the prior art, the present application provides a building prefabricated component hoisting guiding method and system based on multi-modal perception and dynamic knowledge graph, which aims to realize multi-modal data collaboration and rapid dynamic risk response in building prefabricated component hoisting guidance, and improve the accuracy of decision-making, hoisting efficiency and precision.
[0005] To achieve the above purpose, the present application provides a building prefabricated component hoisting guiding method based on multi-modal perception and dynamic knowledge graph, comprising: S1, acquiring multi-modal data of the current hoisting process in real time; S2, dynamically weighting and fusing the multi-modal data to generate a fusion feature vector; S3, similarity retrieval is performed on the fusion feature vector and the prefabricated building hoisting dynamic knowledge graph library, if a risk level under the hoisting condition represented by the multi-modal data is retrieved, step S4 is entered, otherwise, step S1 is returned; wherein the prefabricated building hoisting dynamic knowledge graph library stores data including historical risk cases, and the historical risk cases are used to represent historical prefabricated component hoisting cases and corresponding risk levels in the prefabricated building hoisting field; S4, machine executable instructions of corresponding levels are generated according to different risk levels to control corresponding execution devices to perform corresponding operations, and step S1 is returned until the current component hoisting is completed.
[0006] Further, in S4, the machine executable instructions of corresponding levels are generated according to different risk levels to control corresponding devices to perform corresponding operations, including: When the risk level is low, hoisting guidance instructions are generated to control the hoisting guidance terminal to perform hoisting guidance; When the risk level is medium, control instructions of the hoisting equipment control interface are generated to control the hoisting equipment to adjust the rotation angular velocity of the tower crane jib and the height of the current component to reduce the collision risk of the current component and the obstacle; When the risk level is high, emergency brake triggering instructions are generated to trigger the hydraulic locking device to make the tower crane emergency brake.
[0007] Further, when the risk level is low, the hoisting guidance instructions are visual hoisting guidance instructions, and the guidance information in the visual hoisting guidance instructions includes: the current risk level, the adjustment direction, the adjustment distance and the adjustment angle of the current component hoisting guidance; Wherein, the adjustment direction, the adjustment distance and the adjustment angle are determined based on the multi-modal data.
[0008] Further, the prefabricated building hoisting dynamic knowledge graph library stores triple data of attributes of each component in the prefabricated building hoisting field, hoisting process specifications and the historical risk cases; The fusion feature vector is subjected to similarity retrieval with the prefabricated building hoisting dynamic knowledge graph library, and the risk level under the hoisting condition represented by the multi-modal data is retrieved, including: According to the multi-modal data, the triple data is filtered to obtain a historical risk case candidate set consistent with the attributes of the current component and the hoisting process specification; convert each historical risk case in the candidate set of historical risk cases into a feature vector respectively; similarity calculation is carried out between the fusion feature vector and each feature vector respectively; if there is at least one similarity exceeding a preset risk threshold, then in the similarities exceeding the preset risk threshold, the risk level of the historical risk case with the highest similarity is selected as the risk level under the current hoisting condition.
[0009] Further, the hoisting dynamic knowledge graph database of the fabricated building is constructed, comprising: An entity definition layer is constructed, comprising component entities, process entities and case entities; The component entity is a physical attribute comprising component model, size, weight, center of gravity position, lifting point distribution and material strength, and is used to represent the attribute of the component; The process entity represents the hoisting process specification of the component; The case entity is used to represent the historical risk case; A dynamic updating layer is constructed, which is used to automatically retrieve and dynamically update the hoisting process specification and historical risk case at a set time interval.
[0010] Further, the multi-modal data comprises image data of the component, pose data of the component, environmental wind speed data, and rotation angle, angular velocity and angular acceleration data of the tower crane boom; In S2, when the environmental wind speed data is greater than a preset wind speed threshold, the weight of the component pose data is automatically increased and the weight of the component image data is reduced during dynamic feature weighting of the multi-modal data.
[0011] The application also provides a building prefabricated component hoisting guidance system based on multi-modal perception and dynamic knowledge graph, which is used to execute any of the building prefabricated component hoisting guidance methods described above, and comprises a multi-modal data perception subsystem, a dynamic feature weighting module, a risk level judgment module, a multi-level machine executable instruction generation module and a multi-level execution subsystem. The multi-modal data perception subsystem is used to acquire multi-modal data of the current hoisting process in real time. The dynamic feature weighting module is used to dynamically weight and fuse the multi-modal data to generate a fusion feature vector. The risk level judgment module is used to perform similarity retrieval between the fusion feature vector and the hoisting dynamic knowledge graph database of the fabricated building, if the risk level under the hoisting condition represented by the multi-modal data is retrieved, then the multi-level machine executable instruction generation module is entered, otherwise, the multi-modal data perception subsystem is returned; The hoisting dynamic knowledge graph database of the fabricated building stores data including historical risk cases, and the historical risk cases are used to represent historical prefabricated component hoisting cases and their corresponding risk levels in the field of hoisting of fabricated buildings. The multi-level machine executable instruction generation module is used to generate machine executable instructions of corresponding levels according to different risk levels to control the multi-level execution subsystem to execute corresponding operations, and returns to the multi-modal data perception subsystem until the current component hoisting is completed.
[0012] The application also provides a building prefabricated component hoisting guiding device based on multi-modal perception and dynamic knowledge graph, comprising a computer readable storage medium and a processor. The computer readable storage medium is used to store executable instructions. The processor is used to read the executable instructions stored in the computer readable storage medium to execute the building prefabricated component hoisting guiding method.
[0013] The application also provides a computer readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the building prefabricated component hoisting guiding method.
[0014] The application also provides a computer program product, characterized in that comprising a computer program, when the computer program runs on a computer, so that the computer executes the building prefabricated component hoisting guiding method.
[0015] Overall, the above technical solutions conceived by the application can achieve the following beneficial effects: (1) The application adopts a "perception-decision-execution" closed-loop architecture, based on real-time multi-modal data perception, cooperates with the prefabricated building hoisting dynamic knowledge graph database to make real-time risk level judgment, adopts a multi-level instruction coordination mechanism for different levels of risk, realizes multi-modal data coordination and dynamic risk response, improves the dynamic adaptability of the system, avoids risk response lag, and also improves the accuracy of decision-making, hoisting efficiency and precision.
[0016] (2) Further, the machine device instruction executable instructions and human information (hoisting guiding information) generated according to different risk levels contain professional and rich content, the instruction executability is strong, and the human information readability is good.
[0017] (3) As a preferred, considering that when the environmental wind speed is large, the swing amplitude of the component is large, and the visual picture quality is reduced, when the environmental wind speed data is greater than the preset wind speed threshold, the weight of the component pose data is automatically increased, and the weight of the component image data is reduced, which can further improve the accuracy of the multi-modal fusion data for representing the current hoisting condition.
[0018] Overall, the application is based on a building prefabricated component hoisting guidance method based on multi-modal perception and dynamic knowledge graph library, which can be applied to the hoisting scene of modular building prefabricated components, can improve the man-machine cooperation efficiency and hoisting accuracy of the building prefabricated component hoisting process, enable multi-modal data to be synergistically analyzed and comprehensively utilized, and timely respond to risks, solving the problems of existing technology, such as hoisting accuracy relying on manual experience, multi-modal data synergy, low man-machine cooperation efficiency, poor machine equipment instruction executability, poor human information readability, and risk response lag. It has important significance for the transformation and upgrading of the construction industry, ensuring the construction safety of building prefabricated component hoisting projects, and improving construction production efficiency and quality. BRIEF DESCRIPTION OF DRAWINGS
[0019] Figure 1 A building prefabricated component hoisting guidance method based on multi-modal perception and dynamic knowledge graph in the embodiments of the application. DETAILED DESCRIPTION
[0020] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.
[0021] Embodiment 1 The embodiment of the present application provides a building prefabricated component hoisting guidance method based on multi-modal perception and dynamic knowledge graph, comprising the following steps: S1, real-time acquisition of multi-modal data of the current component hoisting process; in the embodiment of the present application, the multi-modal data of the component hoisting process includes: image data of the component, tower crane boom rotation angle and angular velocity and angular acceleration data, pose data of the component, and environmental wind speed data.
[0022] S2, dynamic feature weighted fusion of multi-modal data to generate a fusion feature vector. In the embodiment of the present application, the Attention-LSTM model is used for weight distribution and feature fusion of multi-modal data.
[0023] Considering that when the environmental wind speed is large, the swing amplitude of the component is large, and the visual picture quality is reduced, in order to further improve the accuracy of the multi-modal fusion data in representing the current hoisting situation, as a preferred embodiment, when the environmental wind speed data is greater than a preset wind speed threshold, the weight of the component pose data is automatically increased, and the weight of the component image data is reduced. In the embodiment of the present application, the preset wind speed threshold is 8 m / s, the weight of the improved component pose data is 0.7, and the weight of the reduced low component image data is 0.2.
[0024] S3, the fusion feature vector is searched for similarity with the prefabricated building hoisting dynamic knowledge graph library, if the risk level under the hoisting condition represented by the current multi-modal data is searched, step S4 is entered, otherwise, step S1 is returned, and the risk level at the next moment is judged; As a specific implementation, in S3, the prefabricated building hoisting dynamic knowledge graph library stores the triples of the attributes of each component in the prefabricated building hoisting field, the hoisting process specification and the historical risk case, the historical risk case is used to represent the historical prefabricated component hoisting case in the prefabricated building hoisting field and the risk level corresponding thereto, and different risk levels include high, medium and low risks. In the embodiment of the application, the Neo4j is used to build the hoisting dynamic knowledge graph library.
[0025] As a specific implementation, in S3, the prefabricated building hoisting dynamic knowledge graph library is built, including: The entity definition layer includes: The component entity contains physical properties such as component model, size, weight, center of gravity position, hoisting point distribution and material strength, and is used to represent the attributes of the component; The process entity is associated with hoisting specifications, wind speed thresholds, component pose swing amplitude thresholds and the like, and is used to represent the hoisting process specification of the component; The case entity includes accident types such as overturning, collision and rope rupture and the corresponding risk level, and is associated with environmental parameters of historical cases, such as swing amplitude exceeding limit cases when the wind speed is greater than 10 m / s.
[0026] The dynamic updating layer automatically retrieves and dynamically updates the hoisting process specification and historical risk case data in the hoisting dynamic knowledge graph library through network crawlers and the like every certain period of time; and after the component size and other attributes are modified, the center of gravity of the component can be dynamically calculated and the hoisting point data of the component can be updated.
[0027] In S3, the fusion feature vector is searched for similarity with the prefabricated building hoisting dynamic knowledge graph library, and the risk level under the hoisting condition represented by the current multi-modal data is obtained, including: According to the current multi-modal data, the historical risk case candidate set consistent with the current to-be-hoisted component attributes and hoisting process specification is obtained by rapid screening in the prefabricated building hoisting dynamic knowledge graph library; After each historical risk case in the historical risk case candidate set is respectively converted into a feature vector, the feature vector is compared with the fusion feature vector in similarity; if at least one similarity exceeds a preset risk threshold, the risk level of the historical risk case with the highest similarity in the similarities exceeding the preset risk threshold is selected as the risk level under the current hoisting condition. In the embodiment of the application, each historical risk case in the historical risk case candidate set is converted into a feature vector by using a Transformer-CRF model.
[0028] If all similarities are less than the preset risk threshold, return to re-execute S1-S3 to acquire the multi-modal data at the next moment, calculate the fusion feature vector, and judge the risk level.
[0029] In the embodiment of the application, the cosine similarity is used to calculate the similarity between the feature vector corresponding to each historical risk case and the fusion feature vector; in other embodiments, other similarity calculation methods can also be selected. In the embodiment of the application, the preset risk threshold is 0.85.
[0030] S4, generate machine executable instructions (hoisting PLC controller executable instructions) and human readable information (hoisting guide information) of corresponding levels according to different risk levels, the corresponding equipment executes corresponding operations based on the machine executable instructions of the corresponding levels, and returns to step S1, and repeats steps S1-S4 until the component hoisting is completed; wherein the multi-level machine executable instructions include: When the risk level is low, according to the current risk level, the Transformer-CRF model is used to generate visual hoisting guide instructions for controlling the hoisting guide terminal, and the current hoisting guide is performed; wherein the visual hoisting guide information in the hoisting guide instructions includes: the current risk level, the adjustment direction, the adjustment distance, the adjustment angle of the current hoisting moment, the center of gravity and the hoisting point position of the next component, etc.; the adjustment direction, the adjustment distance and the adjustment angle of the current hoisting moment are determined based on the multi-modal data at the current moment; the center of gravity and the hoisting point position of the next component are determined according to the hoisting dynamic knowledge graph library.
[0031] When the risk level is medium, according to the current risk level, the Transformer-CRF model is used to generate control instructions for controlling the hoisting equipment control interface, to adjust the swing angle velocity of the boom and the height of the component, so as to reduce the collision risk of the component and the obstacle.
[0032] When the risk level is high, according to the current risk level, the Transformer-CRF model is used to generate an emergency brake trigger instruction and send the trigger instruction to the emergency brake module, to trigger the hydraulic locking device and make the tower crane emergency brake.
[0033] The hoisting guide terminal, the equipment control interface or the emergency braking module receives and executes instructions and performs corresponding operations, specifically including: After the hoisting guide terminal receives the visual hoisting guide instruction, a visual adjustment guide picture is displayed to guide the worker to finely adjust the current component hoisting; wherein, the human-readable information of the current hoisting adjustment direction, adjustment distance, adjustment angle, risk level, reasoning basis, etc. are read and referenced by the worker.
[0034] After the equipment control interface receives the regulation and control instruction, the PLC controller is called to control the large arm of the hoisting equipment, including reducing the rotation angle speed of the large arm and adjusting the height of the component to reduce the collision risk of the component and the obstacle, and as an optimization, the sound and light alarm is controlled. Alarm.
[0035] After the emergency braking module receives the trigger instruction, the hydraulic locking device is triggered to make the tower crane emergency brake; as an optimization, the sound and light alarm is controlled. Alarm.
[0036] Embodiment 2 The embodiment of the application provides a building prefabricated component hoisting guide system based on multi-modal perception and dynamic knowledge graph, comprising: A multi-modal data perception subsystem is used to acquire multi-modal data of the current component hoisting process in real time; in the embodiment of the application, the multi-modal data perception subsystem comprises: A binocular vision sensor is used to acquire image data of the component; An absolute value encoder is used to acquire tower crane large arm rotation angle and angular velocity and angular acceleration data; A UWB positioning module is used to acquire pose data of the component; A wind speed sensor is used to acquire environmental wind speed data.
[0037] A cross-modal decision engine comprises: A dynamic feature weighting module is used to perform weight distribution and feature fusion on the multi-modal data based on an Attention-LSTM model.
[0038] A risk level judgment module is used to perform similarity retrieval on the fusion feature vector and the prefabricated building hoisting dynamic knowledge graph database; if the risk level under the hoisting condition represented by the current multi-modal data is retrieved, the multi-level machine executable instruction generation module is entered, otherwise, the multi-modal data perception subsystem is returned.
[0039] A multi-level machine executable instruction generation module is used to generate machine executable instructions of corresponding levels according to different risk levels to control the multi-level execution subsystem to perform corresponding operations.
[0040] The multi-level execution subsystem comprises: The lifting guidance terminal is used to display human-readable information such as visual adjustment guidance pictures, adjustment directions, adjustment distances, adjustment angles, risk levels, reasoning bases, etc., and generate visual lifting guidance instructions based on the human-readable information when the risk is low, and perform current lifting guidance.
[0041] The device control interface is used to send machine-readable instructions to the lifting PLC controller according to the generated control instructions when the risk is medium, and call the PLC controller to control the large arm of the lifting device to adjust the rotation angular velocity of the large arm and the height of the component to reduce the collision risk of the component and the obstacle, or control the sound and light alarm to sound and light alarm.
[0042] The emergency braking module is used to trigger the hydraulic locking device according to the generated emergency braking trigger instruction when the risk is high.
[0043] The specific role of each module is described in the description of the corresponding steps in Embodiment 1, which will not be repeated here.
[0044] Embodiment 3 The embodiment of the application provides a building prefabricated component lifting guidance device based on multi-modal perception and a dynamic knowledge graph, which comprises a memory and a processor, the memory stores a computer program, and the processor realizes the steps of the method in Embodiment 1 above when executing the computer program.
[0045] The related technical solutions are the same as above, and will not be repeated here.
[0046] Embodiment 4 The embodiment of the application provides a computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the method in Embodiment 1 above.
[0047] Specifically, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory devices.
[0048] The related technical solutions are the same as above, and will not be repeated here.
[0049] Embodiment 5 The embodiment of the application provides a computer program product, which comprises a computer program, and when the computer program runs on a computer, the computer executes the steps of the method in Embodiment 1 above.
[0050] The related technical solutions are the same as above, and will not be repeated here.
[0051] It is to be understood that the above description is intended to be illustrative and not restrictive. Many other embodiments will be apparent to those of skill in the art upon reading and understanding the above description. The scope of the application should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.
Claims
1. A method for guiding the hoisting of a building prefabricated component based on multi-modal perception and dynamic knowledge graph, characterized in that, The method comprises the following steps: S1, acquiring multi-modal data of the current hoisting process in real time; S2, performing dynamic feature weighted fusion on the multi-modal data to generate a fusion feature vector; S3, performing similarity retrieval on the fusion feature vector and a prefabricated building hoisting dynamic knowledge graph library, if a risk level under the hoisting condition represented by the multi-modal data is retrieved, then step S4 is entered, otherwise, step S1 is returned; the prefabricated building hoisting dynamic knowledge graph library stores data including historical risk cases, and the historical risk cases are used to represent historical prefabricated component hoisting cases and the corresponding risk levels in the prefabricated building hoisting field; S4, generating machine executable instructions of corresponding levels according to different risk levels to control corresponding execution devices to perform corresponding operations, and returning to step S1 until the current component hoisting is completed.
2. The building precast member hoisting guide method according to claim 1, characterized by, In S4, generating machine executable instructions of corresponding levels according to different risk levels to control corresponding devices to perform corresponding operations, comprising: When the risk level is low, generating hoisting guidance instructions to control the hoisting guidance terminal to guide hoisting; When the risk level is medium, generating control instructions of the hoisting equipment control interface to control the hoisting equipment to adjust the rotation angular velocity of the tower crane jib and the height of the current component to reduce the collision risk between the current component and the obstacle; When the risk level is high, generating emergency brake triggering instructions to trigger the hydraulic locking device to make the tower crane brake urgently.
3. The building precast member hoisting guide method according to claim 2, characterized by, When the risk level is low, the hoisting guidance instructions are visual hoisting guidance instructions, and the guidance information in the visual hoisting guidance instructions includes: the current risk level, the adjustment direction, the adjustment distance and the adjustment angle of the current component hoisting guidance; The adjustment direction, the adjustment distance and the adjustment angle are determined based on the multi-modal data.
4. The building precast member hoisting guide method according to any one of claims 1 to 3, characterized by, The prefabricated building hoisting dynamic knowledge graph library stores triple data of attributes of each component in the prefabricated building hoisting field, hoisting process specifications and the historical risk cases; The similarity retrieval of the fusion feature vector and the prefabricated building hoisting dynamic knowledge graph library retrieves the risk level under the hoisting condition represented by the multi-modal data, comprising: Filtering in the triple data according to the multi-modal data to obtain a historical risk case candidate set consistent with the attributes of the current component and the hoisting process specification; Converting each historical risk case in the historical risk case candidate set into a feature vector, performing similarity calculation on the fusion feature vector and each feature vector, and if at least one similarity exceeds a preset risk threshold, selecting the risk level of the historical risk case with the highest similarity in the similarity exceeding the preset risk threshold as the risk level under the current hoisting condition.
5. The building precast member hoisting guide method according to claim 4, characterized by, The prefabricated building hoisting dynamic knowledge graph library is constructed, comprising: The entity definition layer includes a component entity, a process entity, and a case entity; the component entity is a physical attribute including a component model, a size, a weight, a gravity center position, a lifting point distribution, and a material strength, and is used to represent the attribute of the component; the process entity represents the lifting process specification of the component; and the case entity is used to represent the historical risk case; The dynamic updating layer is used to automatically retrieve and dynamically update the lifting process specification and the historical risk case at a set time interval.
6. The building precast member hoisting guide method according to any one of claims 1 to 3, characterized by, The multi-modal data includes image data of the component, pose data of the component, environmental wind speed data, and rotation angle, angular velocity, and angular acceleration data of the tower crane boom; In S2, when the environmental wind speed data is greater than a preset wind speed threshold, the weight of the component pose data is automatically increased and the weight of the component image data is reduced during dynamic feature weighting of the multi-modal data.
7. A multi-modal perception and dynamic knowledge graph based building precast component hoisting guidance system, characterized in that, The building prefabricated component lifting guiding system comprises a multi-modal data perception subsystem, a dynamic feature weighting module, a risk level judgment module, a multi-level machine executable instruction generation module, and a multi-level execution subsystem. The multi-modal data perception subsystem is used to acquire multi-modal data of a current lifting process in real time. The dynamic feature weighting module is used to dynamically weight and fuse the multi-modal data to generate a fused feature vector. The risk level judgment module is used to perform similarity retrieval of the fused feature vector and a prefabricated building lifting dynamic knowledge graph database, and if a risk level under a lifting condition represented by the multi-modal data is retrieved, the multi-level machine executable instruction generation module is entered, otherwise, the multi-modal data perception subsystem is returned; the prefabricated building lifting dynamic knowledge graph database stores data including historical risk cases, and the historical risk cases are used to represent historical prefabricated component lifting cases and corresponding risk levels in the prefabricated building lifting field. The multi-level machine executable instruction generation module is used to generate machine executable instructions of corresponding levels according to different risk levels to control the multi-level execution subsystem to perform corresponding operations, and returns to the multi-modal data perception subsystem until the current component lifting is completed.
8. A multi-modal perception and dynamic knowledge graph based building precast component hoisting guidance device, characterized in that, The computer readable storage medium is used to store executable instructions. The processor is used to read the executable instructions stored in the computer readable storage medium to execute the building prefabricated component lifting guiding method. The program is executed by the processor to implement the building prefabricated component lifting guiding method.
9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program makes the computer execute the building prefabricated component lifting guiding method when the computer program runs on the computer.
10. A computer program product, characterised in that,