CN-121982581-A - Automatic quality inspection system and method for power transmission engineering quality based on Building Information Model (BIM) guidance and unmanned aerial vehicle multi-mode data fusion

CN121982581ACN 121982581 ACN121982581 ACN 121982581ACN-121982581-A

Abstract

The invention discloses a BIM model-based automatic quality inspection system and method for power transmission engineering quality based on multi-mode data fusion of unmanned aerial vehicle, and aims to solve the problems of low quality inspection efficiency, high risk and insufficient precision in the prior art. The method comprises the steps of firstly controlling an unmanned aerial vehicle carrying a laser radar and a camera to collect high-precision point cloud and images of a built line, fusing the high-precision point cloud and images into a built real-scene digital model, secondly carrying out millimeter-level accurate registration on the real-scene model and a BIM model serving as a design reference, carrying out BIM guided targeting analysis, namely automatically defining a local three-dimensional region of interest (ROI) in the real-scene model based on the design position and the size of a component in the BIM, calling an AI model to identify the component instance only in the ROI, judging the installation state of the component instance, and finally quantitatively calculating the deviation between the identified component instance and BIM design parameters and generating a visual quality inspection report containing multi-view evidence. According to the invention, through a 'according to the drawing' target analysis strategy, the efficiency, stability and precision of automatic quality inspection are greatly improved, objective, quantitative and full coverage inspection of the quality of the power transmission engineering is realized, and a foundation is laid for digital delivery and operation and maintenance.

Inventors

LI TIANTIAN
ZHANG QINGFENG
SONG XIAOKAI
JING CHUAN
YU BO
YUAN ZITAO
LIANG FEIHONG
ZHANG CHENG
CHEN WEISHEN

Assignees

国网河南省电力公司经济技术研究院
河南省电力建设工程质量监督中心站

Dates

Publication Date: 20260505
Application Date: 20251231

Claims (10)

1. The automatic quality inspection method for the power transmission engineering quality based on Building Information Model (BIM) guidance and unmanned aerial vehicle multi-mode data fusion is characterized by comprising the following steps: a. Controlling an unmanned aerial vehicle carrying a laser radar and a high-resolution camera, collecting high-precision three-dimensional point cloud and high-definition images of a built transmission line, and fusing the three-dimensional point cloud and the high-definition images to generate a built live-action digital model under a unified coordinate system; b. A model accurate registration step, namely carrying out space coordinate alignment on a preset BIM model of the power transmission line serving as a design reference and the completion live-action digital model to realize millimeter-level accurate registration; c. BIM guided targeting analysis step: i. traversing each component to be inspected in the BIM in sequence based on a preset quality inspection list; Aiming at the currently traversed component to be detected, automatically defining a corresponding local three-dimensional region of interest (ROI) in the registered completion live-action digital model according to the design three-dimensional position and size information recorded in the BIM model; calling a pre-trained three-dimensional point cloud segmentation network or a two-dimensional image analysis network only in the three-dimensional region of interest (ROI), and analyzing the as-built digital model to identify and segment an instance of a target component and judge the installation state of the target component; d. and the deviation quantification and report generation step is that the actual space parameters of the component instance identified in the three-dimensional region of interest (ROI) are compared with the design parameters of the component in the BIM model in three dimensions, the deviation value is quantitatively calculated, and a visual quality inspection report containing multi-view evidence is automatically generated.
2. The method of claim 1, further comprising, prior to the data acquisition and modeling step, automatically planning generation of a three-dimensional surrounding flight path for the unmanned aerial vehicle based on three-dimensional bounding box information of a tower in the BIM model to ensure data coverage integrity for all components under inspection.
3. The method of claim 1, wherein the model accurate registration step specifically comprises the steps of firstly performing rough registration on the BIM model and the as-built live-action digital model by using geographic coordinate information, then respectively extracting large structures such as a main pole leg and a cross arm from the two models as characteristic primitives, and performing iterative optimization on the rough registered models by adopting an iterative closest point ICP algorithm to realize fine registration.
4. The method of claim 1, wherein the BIM guided target analysis step further comprises the steps of back projecting the designed three-dimensional coordinates of the detail features in the BIM model onto a specific high-definition image covering the coordinate points shot by the unmanned aerial vehicle to determine a two-dimensional analysis area on the high-definition image, and calling a two-dimensional image recognition model to analyze the two-dimensional analysis area to determine whether the installation of the detail features is tight or not.
5. The method of claim 1, wherein the BIM guided targeting analysis step further comprises defining one or more blank regions ROI within a pre-defined area of the BIM model that is to be blank in design, and performing a target detection algorithm within the blank regions ROI for detecting whether off-design or construction remains are present.
6. An automatic quality inspection system for power transmission engineering based on Building Information Model (BIM) guidance and unmanned aerial vehicle multi-modal data fusion, comprising: the data acquisition device is an unmanned aerial vehicle carrying a laser radar and a high-resolution camera and is used for acquiring high-precision three-dimensional point clouds and high-definition images of finished transmission lines; a memory for storing a transmission line BIM model as a design reference and a computer program; and the processor is connected with the memory and is used for realizing the following functions when executing the computer program: The data acquisition device is controlled to acquire data, and the acquired data are fused to generate a completion live-action digital model under a unified coordinate system; performing millimeter-level accurate registration on the BIM model and the completion live-action digital model; traversing components to be detected in a BIM model based on a quality inspection list, automatically defining a corresponding three-dimensional region of interest (ROI) in a registered finished live-action digital model according to the design position and size of each component to be detected, and calling an analysis network to identify an instance of a target component only in the ROI and judge the state of the instance; comparing the actual parameters of the component examples identified in the ROI with BIM design parameters, quantitatively calculating deviation values, and generating a visual quality inspection report.
7. The system of claim 6, wherein the processor, when executing the computer program, is further configured to automatically plan for the unmanned aerial vehicle to generate a three-dimensional surrounding flight path for refined data acquisition based on three-dimensional bounding box information of a tower in the BIM model prior to data acquisition.
8. The system of claim 6, wherein the processor, when performing the precise registration function, is configured to first perform a coarse registration using geographic coordinates, then extract tower main legs, cross arms as features, and invoke an iterative closest point ICP algorithm to iteratively optimize the BIM model and the as built digital model to complete the fine registration.
9. The system of claim 6, wherein the processor, when performing a BIM guided targeting analysis function, is further configured for a bolt, cotter pin, or like detail feature to backproject its BIM design three-dimensional coordinates onto a corresponding high definition image to determine a two-dimensional analysis area and invoke a two-dimensional image recognition model to analyze the two-dimensional analysis area to determine the installation status of the detail feature.
10. The system of claim 6, wherein the visual quality inspection report generated by the processor automatically correlates its unique component ID, quantified deviation data in the BIM model for each problem item in the project, with BIM design views, as-built live-action model snapshots, and high-definition raw images as multi-view evidence materials.

Description

Automatic quality inspection system and method for power transmission engineering quality based on Building Information Model (BIM) guidance and unmanned aerial vehicle multi-mode data fusion Technical Field The invention relates to the technical fields of electric power engineering construction quality detection, digital twin technology and automatic detection, in particular to a system and a method for realizing automatic, high-precision and component level checking of electric power engineering construction quality, which are used for integrating a Building Information Model (BIM), unmanned plane multi-mode data acquisition, three-dimensional point cloud processing and artificial intelligent visual analysis technology. Background The power transmission line engineering is used as a key infrastructure for national energy transmission, and the construction quality of the power transmission line engineering directly influences the safety and stable operation of a power grid system. Currently, the quality supervision work of power transmission engineering still depends on the traditional manual field inspection and sampling detection mode to a great extent. In this mode, the supervision engineer needs to carry data such as design drawings, and attends to the construction site, and checks the installation conditions of thousands of components such as transmission towers, hardware fittings, insulators one by means of visual inspection, measuring tools (such as tape measures and angle meters), tower climbing operation and the like. This traditional quality supervision mode has a series of inherent technical drawbacks. First, the contradiction between efficiency and safety is very prominent. The construction path of the transmission line engineering often passes through complicated terrains such as mountain areas, rivers and the like, so that the manual inspection is inconvenient, the operation range is limited, and the time cost is high. Meanwhile, the tower climbing operation belongs to high-risk high-altitude operation, and the personal safety of operators is seriously threatened. These factors together lead to difficulties in achieving full coverage and high frequency verification of engineering quality for traditional quality monitoring work. Secondly, the accuracy and consistency of the detection result are difficult to ensure. The accuracy of manual interpretation is highly dependent on the personal experience and responsibility of the supervision engineer. For small deviations of the mounting positions and angles of the components and the small mounting states of the fasteners (such as bolts and cotters), the problems of measurement errors or standard deviation are easily caused by subjective judgment, and thus quality hidden troubles may be left. Furthermore, data management presents serious disjointing problems. The field detection result mostly exists in the form of a paper form or an isolated digital photo, and the data lacks a direct and systematic mapping relation with a three-dimensional design model, so that a digital island is formed. Not only does this make the traceability and responsibility definition of quality problems difficult, but it is not possible to form systematic quality big data assets, and it is also not possible to provide accurate and reliable completion data (As-Built) for digital delivery of projects and full life cycle intelligent operation and maintenance. In recent years, although unmanned aerial vehicle aerial photographing measurement and three-dimensional reconstruction technology are applied in certain fields, the existing technical scheme still has obvious defects when being applied to the construction quality inspection of power transmission engineering. On the one hand, the application of the techniques is mainly focused on the operation and maintenance stage of the power transmission line, and is mainly used for inspection of macroscopic defects, such as bird nest searching, component rust, tree obstacle searching and the like, but cannot be effectively applied to the construction stage, and accurate quantitative verification is carried out on whether component installation strictly accords with design specifications. On the other hand, even if a three-dimensional model of a finished scene is obtained by an unmanned aerial vehicle, its application often remains on the level for manual visual comparison. How to realize automatic, high-precision and component-level deviation quantitative analysis between the As-Built model and the designed BIM model (As-Designed) is still a pending technical problem. Particularly, in the real-scene data of a large scene and a complex structure, how to stably and efficiently finish the accurate registration of the real-scene model and the BIM model and automatically identify and check each specific component is a technical bottleneck to be broken through in the current field. Disclosure of Invention The invention aims to solve the tech