CN-122022536-A - Quality management system and method for engineering construction

CN122022536ACN 122022536 ACN122022536 ACN 122022536ACN-122022536-A

Abstract

The application relates to the field of engineering management, in particular to a quality management system and method for engineering construction, firstly, preprocessing a BIM model to construct a reference frame for fusing visual characteristics and space geometric information. When a problem image shot by a mobile terminal on a construction site is received, introducing a computer vision and space calculation technology, analyzing the two-dimensional site image into the accurate pose of a shooting camera in a three-dimensional BIM space, automatically tracing the line of sight in a BIM digital model based on the accurate pose, and intelligently deducing and searching out candidate components highly matched with the problem view angle. And finally, firmly binding the on-site unstructured problem data with the unique identifier of the BIM component through light interactive confirmation with a user to obtain a standardized structured problem record. Therefore, the problems of low manual positioning efficiency and unstable association can be overcome, and the full-flow intelligent closed loop from the collection of quality problem data to the structured warehousing can be realized.

Inventors

LIU HONGWANG
GUO SHUANJUN
ZHANG JINGTIAN
LI MENGJIAO
BAI ZHIBIN

Assignees

国家能源集团雄安能源有限公司

Dates

Publication Date: 20260512
Application Date: 20250814

Claims (8)

1. A quality management method for engineering construction, comprising: constructing a visual feature map and a lightweight BIM database based on the project BIM model; acquiring an original problem data packet uploaded by a mobile terminal APP, wherein the original problem data packet comprises a problem component image and a problem text description; based on the visual feature map, performing visual repositioning and pose solving on the problem component image to obtain a six-degree-of-freedom pose of the camera; Performing component ray casting based on the six-degree-of-freedom pose of the camera to obtain virtual rays; Querying all components intersected with the virtual ray by using the spatial index of the lightweight BIM database to obtain a candidate component set; Extracting the best candidate component from the candidate component set and sending the best candidate component to a mobile terminal APP; After obtaining the user confirmation action, obtaining the confirmed component unique identifier; The confirmed component unique identifier, the original problem data package and the camera six-degree-of-freedom pose are structured and correlated to obtain a structured problem record.
2. The quality management method for engineering construction according to claim 1, wherein performing visual repositioning and pose solving on the problem component image based on the visual feature map to obtain a camera six-degree-of-freedom pose, comprises: Extracting 2D visual feature point encoding vectors from the problem component image; Based on the 2D visual feature point coding vector, carrying out quick matching with feature descriptors in the visual feature map to obtain 2D-3D matching point pairs; and carrying out pose solving based on the 2D-3D matching point pairs to obtain the pose with six degrees of freedom of the camera.
3. The quality management method for engineering construction according to claim 2, wherein extracting 2D visual feature point code vectors from the problem member image comprises: Carrying out semantic embedded coding on the problem text description to obtain a problem text description semantic embedded coding vector; performing component visual feature extraction based on convolutional coding on the problem component image to obtain a problem component visual feature coding diagram; based on the problem text description semantic embedded coding vector, performing visual feature enhancement based on problem text modal guidance on the problem component visual feature coding diagram to obtain a 2D visual feature cross-modal guidance enhancement implicit coding vector; And carrying out decoding regression on the 2D visual feature cross-modal guiding enhanced implicit coding vector to obtain the 2D visual feature point coding vector.
4. A quality management method for engineering construction according to claim 3, wherein, based on the problem text description semantic embedded coding vector, performing problem text modality guidance-based visual feature enhancement on the problem component visual feature coding map to obtain a 2D visual feature cross-modality guidance enhancement implicit coding vector, comprising: the problem component visual feature coding diagram is subjected to node segmentation along the channel dimension to obtain a set of problem component visual feature vectors; calculating a modal inter-node adjacency matrix of the set of the visual feature vectors of the problem member to obtain a problem member visual modal inter-node adjacency matrix; Convolving the set of the visual feature vectors of the problem member and the adjacency matrix input graph between the nodes of the visual mode of the problem member with a neural network model to obtain a topological association matrix between the nodes of the visual feature of the problem member; Embedding the coding vector by using the description semantics of the problem text, and performing global scanning interaction on the topological association matrix among the visual feature nodes of the problem member to obtain the 2D visual feature cross-modal guiding enhanced implicit coding vector.
5. The method of claim 1, wherein projecting component rays based on the six degree of freedom pose of the camera to obtain virtual rays comprises transmitting the virtual rays in a BIM world coordinate system along a camera orientation from a position of the six degree of freedom pose of the camera.
6. The quality management method for engineering construction according to claim 5, wherein querying all components intersected by the virtual ray to obtain a candidate component set using a spatial index of the lightweight BIM database, comprises: Constructing an axis alignment bounding box of each BIM component in the lightweight BIM database; performing rough query on the lightweight BIM database by utilizing an octree traversal algorithm based on the axis alignment bounding boxes of the BIM components to obtain a temporary selection set; traversing each temporary candidate component in the temporary candidate set, and performing intersection tests on the triangular network model of each temporary candidate component and the virtual rays to obtain the candidate component set.
7. The method for quality management of engineering construction according to claim 6, wherein extracting an optimal candidate member from the candidate member set and transmitting it to the mobile terminal APP, comprises selecting a candidate member closest to the start point of the virtual ray from the candidate member set as the optimal candidate member.
8. A quality management system for engineering construction, comprising: The feature map and database construction module is used for constructing a visual feature map and a lightweight BIM database based on the project BIM model; the original problem data acquisition module is used for acquiring an original problem data packet uploaded by the mobile terminal APP, wherein the original problem data packet comprises a problem component image and a problem text description; The visual repositioning and pose resolving module is used for performing visual repositioning and pose resolving on the problem component image based on the visual characteristic map so as to obtain a six-degree-of-freedom pose of the camera; the component ray projection module is used for performing component ray projection based on the six-degree-of-freedom pose of the camera so as to obtain virtual rays; A candidate component query module for querying all components intersected with the virtual ray by using the spatial index of the lightweight BIM database to obtain a candidate component set; The best candidate component extraction module is used for extracting the best candidate components from the candidate component set and sending the best candidate components to the mobile terminal APP; The user confirmation and identification extraction module is used for acquiring the confirmed component unique identifier after acquiring the user confirmation action; and the problem record structuring module is used for structuring and correlating the confirmed component unique identifier, the original problem data packet and the six-degree-of-freedom pose of the camera to obtain a structured problem record.

Description

Quality management system and method for engineering construction Technical Field The present application relates to the field of engineering management, and more particularly, to a quality management system and method for engineering construction. Background With the deep integration of information technology and the building industry, the engineering industry is undergoing a deep revolution from the traditional mode to digital and intelligent management. The Building Information Model (BIM) technology is used as a core driving force for the innovation, and an unprecedented technical support and realization path is provided for multiparty collaborative work, dynamic process management and refined management of the whole life cycle of engineering projects by constructing a digital three-dimensional model integrating building geometric parameters, physical properties and functional information. Especially in the key links of the quality and safety management of engineering construction, the BIM is utilized to carry out visual technical traffic, carry out space conflict inspection in advance, implement the dynamic simulation of construction progress and the like, and the BIM is an important technical means for remarkably improving the management efficiency, effectively preventing various quality common diseases and practically guaranteeing the construction safety. However, in the conventional quality management scheme, recording of field quality and safety problems is mostly dependent on field personnel taking field photos, manually entering text descriptions or voice recordings through mobile devices, and these data, although being capable of capturing and presenting the actual condition of the field more conveniently, are essentially in the category of unstructured data, and still present a technical barrier for the association with specific components (such as specific beams, floors or pipelines) in the BIM model. In addition, the prior art often requires on-site personnel to manually browse and click a complex BIM model at a mobile terminal to position corresponding components, the process not only causes complicated operation due to dense construction site components and complex space relationship, but also has the positioning precision highly dependent on the professional proficiency and responsibility center of operators, and mismatch risks are very easy to generate. More importantly, the multi-mode characteristics (image semantics, text description and the like) of unstructured data and the structured data (component geometric parameters and attribute information) of the BIM model lack a unified correlation paradigm, so that stable mapping relation which can be resolved by a machine is difficult to form, and the automatic tracing and closed-loop management efficiency of quality problems is severely restricted. Therefore, an optimized quality management scheme for engineering construction is desired. Disclosure of Invention The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides a quality management system and a quality management method for engineering construction, which are characterized in that a BIM model is preprocessed to construct a reference frame integrating visual characteristics and space geometric information. When a problem image shot by a mobile terminal on a construction site is received, introducing a computer vision and space calculation technology, analyzing the two-dimensional site image into the accurate pose of a shooting camera in a three-dimensional BIM space, automatically tracing the line of sight in a BIM digital model based on the accurate pose, and intelligently deducing and searching out candidate components highly matched with the problem view angle. And finally, firmly binding the on-site unstructured problem data with the unique identifier of the BIM component through light interactive confirmation with a user to obtain a standardized structured problem record. Therefore, the problems of low manual positioning efficiency and unstable association can be overcome, and the full-flow intelligent closed loop from the collection of quality problem data to the structured warehousing can be realized. According to an aspect of the present application, there is provided a quality management method for engineering construction, comprising: constructing a visual feature map and a lightweight BIM database based on the project BIM model; acquiring an original problem data packet uploaded by a mobile terminal APP, wherein the original problem data packet comprises a problem component image and a problem text description; based on the visual feature map, performing visual repositioning and pose solving on the problem component image to obtain a six-degree-of-freedom pose of the camera; Performing component ray casting based on the six-degree-of-freedom pose of the camera to obtain virtual rays; Querying all compo