An AI visual recognition method and system for T-beam assembly gap identification and correction

Through AI visual recognition technology, efficient identification and correction of T-beam assembly gaps are achieved, solving the problem of low detection efficiency in existing technologies and ensuring the accuracy and stability of the bridge.

CN120580527BActive Publication Date: 2025-09-26SICHUAN ROAD & BRIDGE CONSTRUCTION GROUP CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202511094216.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-06
Publication Date
2025-09-26
Estimated Expiration
2045-08-06

AI Technical Summary

Technical Problem

The existing technology has low efficiency in detecting gaps in T-beam assembly in bridge projects, making it difficult to achieve efficient and accurate gap identification and correction.

Method used

Using AI visual recognition methods, image recognition calculations and AI simulation predictions are used to obtain the spatial distance and posture data between T-beams, diaphragms and bridge decks, perform assembly simulation and correction, and combine AI learning methods to achieve pre-correction and real-time monitoring of gaps.

Benefits of technology

It improves the efficiency and accuracy of T-beam assembly, ensures that the gap size is within the tolerance range, reduces the need for multiple adjustments, and improves the stability and safety of the bridge.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120580527B_ABST
    Figure CN120580527B_ABST
Patent Text Reader

Abstract

The present invention discloses an AI visual recognition method and system for identifying and correcting T-beam assembly gaps, and relates to the field of AI visual recognition technology. The AI ​​visual recognition method includes: S01, setting an assembly area and a pre-assembly area in a spatial area according to the T-beam assembly process, and obtaining basic dimension data and real-time image data of the T-beam, diaphragm and bridge deck. The present invention uses the AI ​​visual recognition method, based on the basic dimension data combined with the tracking of boundary feature points on the T-beam, diaphragm and bridge deck, so as to understand the spatial distance data of the T-beam, diaphragm and bridge deck. The spatial distance data can be obtained through the boundary feature points, without the need to process a large amount of point cloud data, and the three-dimensional size of the gap can be recognized through the spatial distance data. The processing process is simple, and the correction of the T-beam is achieved by controlling the movement of the boundary feature points and the boundary feature lines, thereby ensuring the convenience of calculation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of AI visual recognition technology, and in particular to an AI visual recognition method and system for identifying and correcting T-beam assembly gaps. Background Art

[0002] In the field of bridge engineering, T-beams are key load-bearing components of precast concrete structures. Their assembly accuracy directly affects the overall mechanical properties and service life of the bridge. Therefore, it is necessary to measure the T-beam assembly gap to ensure that the gap size is within the gap tolerance range. The T-beam gap mainly refers to the gap between the T-beam and the bridge deck, diaphragm and other adjacent structures.

[0003] For example, patent publication number "CN116958259A", named "Displacement deviation detection method, system, terminal and medium for bridge widening and splicing", performs point cloud model coordinate positioning according to the corresponding engineering anchor points, reverse models the original steel box girder bridge at the specified location, and obtains a reverse modeling model; the reverse modeled original bridge model and the BIM widened bridge model are spliced ​​and verified, and then imported into data comparison software for splicing deviation analysis, generating a corresponding splicing deviation report, and providing reliable and accurate data support for displacement deviation detection at the splicing of steel box girder bridges.

[0004] The above patent collects original point cloud data to detect the displacement deviation of splicing. Although the displacement deviation of the bridge width joint can be detected by comparing and analyzing the point cloud data with the BIM model, the above method involves point cloud data processing and model comparison and analysis, resulting in low detection efficiency. For this reason, an AI visual recognition method and system for T-beam assembly gap identification and correction was invented. Summary of the Invention

[0005] The purpose of the present invention is to provide an AI visual recognition method and system for T-beam assembly gap identification and correction to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an AI visual recognition method for identifying and correcting T-beam assembly gaps, the AI ​​visual recognition method comprising:

[0007] S01, setting an assembly area and a pre-assembly area in a spatial area according to the T-beam assembly process, and obtaining basic dimension data and real-time image data of the T-beam, diaphragm and bridge deck;

[0008] S02, the processor receives real-time image data and obtains spatial distance data between the T-beam, the diaphragm and the bridge deck and the spatial posture of the T-beam through image recognition calculation method;

[0009] S03, performing assembly simulation on the T-beam in the pre-assembly area using an AI simulation prediction method, and correcting the spatial posture of the T-beam in the pre-assembly area based on the assembly simulation results to obtain calculation correction data;

[0010] S04, after the T-beams are assembled, the spatial posture of the T-beams in the assembly area is corrected according to the assembly results to obtain actual correction data and actual gap three-dimensional dimensions;

[0011] S05: An AI learning method for assembling T-beams on the same floor is provided. The AI ​​learning method includes:

[0012] Step 1: Assemble the first T-beam on the same floor according to S01, S02, S03, and S04;

[0013] Step 2: Combine the calculated correction data with the actual correction data to obtain reference correction data. The actual spatial posture of the first T-beam is obtained through the actual three-dimensional size of the gap. The standard virtual spatial posture in the pre-assembly area is obtained through simulation calculation. Based on S02, the No. 1 spatial posture of subsequent T-beams on the same layer is obtained. The standard virtual spatial posture and the No. 1 spatial posture are overlapped through the spatial correction method.

[0014] Step 3: According to S04, the subsequent T-beams on the same layer that have completed the overlap are moved to the assembly area for assembly, and the actual correction data No. 1 is obtained;

[0015] Step 4: The actual correction data No. 1 is combined with the reference correction data according to the process in step 2 to obtain new reference correction data. Steps 2, 3 and 4 are repeated based on the new reference correction data until the complete assembly of the T-beams on the same floor is completed.

[0016] Furthermore, the assembly results include obtaining the real spatial posture of the T-beam and the real three-dimensional size of the gap after the T-beam is assembled through image recognition calculation methods;

[0017] Correcting the spatial posture of the T-beam in the assembly area includes: determining whether the three-dimensional size of the actual gap is within the gap tolerance range through the spatial posture correction method, correcting the actual spatial posture of the T-beam in the assembly area through the spatial posture correction method, and obtaining actual correction data generated by the spatial posture correction method.

[0018] Furthermore, the AI ​​simulation prediction method includes calculating the gap size after the T-beam is moved and spliced ​​according to the current posture of the T-beam in the pre-assembly area by combining the spatial distance data and the spatial posture of the T-beam to obtain the calculated three-dimensional size of the gap;

[0019] The assembly simulation results include determining whether the calculated three-dimensional size of the gap is within the gap tolerance range. If it is not within the gap tolerance range, the spatial posture of the T-beam in the pre-assembly area is corrected using the spatial posture correction method. If it is within the gap tolerance range, no posture correction is required.

[0020] Obtain calculation correction data during the operation of the spatial attitude correction method.

[0021] Furthermore, the image recognition calculation method includes:

[0022] An AI automatic recognition system is provided for identifying the edge positions of T-beams, diaphragms and bridge decks. The AI ​​automatic recognition system performs pixel-level segmentation on the edges of T-beams, diaphragms and bridge decks in real-time images, extracts target boundaries, and realizes edge detection between T-beams, diaphragms and bridge decks. A virtual edge line is generated based on the detected edge, boundary feature points are set on the virtual edge line, and based on the basic dimension data, the image is perspective-processed to generate a virtual edge line that is not displayed in the image data. A point tracking system is provided for tracking the boundary feature points. The spatial posture of the T-beam is obtained based on the boundary feature points of the T-beam, and the spatial distance data between the T-beam, diaphragm and bridge deck are obtained based on the basic dimension data, the boundary feature points on the T-beam, the boundary feature points on the diaphragm and the boundary feature points on the bridge deck.

[0023] Furthermore, the spatial correction method includes analyzing difference data between the standard virtual space posture and the first spatial posture using a spatial comparison method, generating standard correction data based on the difference data by a processor, and adjusting the first spatial posture based on the standard correction data so that the standard virtual space posture and the first spatial posture coincide with each other;

[0024] Generate a standard virtual spatial posture of the T-beam in the pre-assembly area, perform image detection on subsequent T-beams on the same layer, calculate the No. 1 spatial posture of subsequent T-beams on the same layer through image calculation methods, analyze the difference data between the standard virtual spatial posture and the No. 1 spatial posture through spatial comparison methods, and the processor generates standard correction data based on the difference data, and adjusts the No. 1 spatial posture based on the standard correction data, so that the standard virtual spatial posture and the No. 1 spatial posture coincide with each other.

[0025] Furthermore, the simulation calculation to obtain the standard virtual space posture includes: the actual space posture is matched in the pre-assembly area by referring to the correction data, and a standard virtual space posture identical to the actual space posture is established in the pre-assembly area.

[0026] Furthermore, the spatial contrast method includes:

[0027] A virtual boundary feature line is set on the standard virtual space posture, and a boundary feature edge line No. 1 corresponding to the virtual boundary feature line is set on the No. 1 space posture according to the spatial position of the virtual boundary feature line on the T-beam. The spatial difference between the virtual boundary feature points with the corresponding relationship and the boundary feature edge line No. 1 is analyzed to obtain difference data.

[0028] Furthermore, the spatial posture correction method includes:

[0029] The calculated three-dimensional size of the gap is input into the processing terminal, which determines whether the calculated three-dimensional size of the gap is within the gap tolerance range. If not, a correction instruction for adjusting the posture of the T-beam in the pre-assembly area is generated, and the spatial posture of the T-beam is adjusted according to the correction instruction. Based on the change in the spatial posture of the T-beam, the new calculated three-dimensional size of the gap is repeatedly calculated and input into the processing terminal until the calculated three-dimensional size of the gap is within the gap tolerance range, and the calculated correction data is recorded.

[0030] An AI visual recognition system for identifying and correcting gaps in T-beam assembly is provided, which adopts the aforementioned AI visual recognition method for identifying and correcting gaps in T-beam assembly; the recognition system comprises:

[0031] Image acquisition module: obtains real-time image data of the assembly area and pre-assembly area;

[0032] Processing module: Receives real-time image data and obtains the spatial distance data between the T-beam, diaphragm and bridge deck, as well as the spatial posture of the T-beam through image recognition and calculation methods;

[0033] The T-beam in the pre-assembly area is assembled by using an AI simulation prediction method. Based on the assembly simulation results, the spatial posture of the T-beam in the pre-assembly area is corrected to obtain calculation correction data;

[0034] For the T-beams being assembled, based on the assembly results, the spatial posture of the T-beams in the assembly area is corrected to obtain the actual correction data and the actual three-dimensional size of the gap;

[0035] AI learning module: The processor is equipped with an AI learning method for assembling T-beams on the same floor. The AI ​​learning method includes:

[0036] Step 1: Assemble the first T-beam on the same floor according to S01, S02, S03, and S04;

[0037] Step 2: Combine the calculated correction data with the actual correction data to obtain reference correction data. The actual spatial posture of the first T-beam is obtained through the actual three-dimensional size of the gap. Simulation calculation is performed with reference to the correction data and the actual spatial posture to obtain the standard virtual spatial posture in the pre-assembly area. Based on S02, the No. 1 spatial posture of the subsequent T-beams on the same layer is obtained. The spatial correction method is used to achieve the coincidence of the standard virtual spatial posture and the No. 1 spatial posture.

[0038] Step 3: According to S04, the subsequent T-beams on the same layer that have completed the overlap are moved to the assembly area for assembly, and the actual correction data No. 1 is obtained;

[0039] Step 4: The actual correction data No. 1 is combined with the reference correction data according to the process in step 2 to obtain new reference correction data. Steps 2, 3 and 4 are repeated based on the new reference correction data until the complete assembly of the T-beams on the same floor is completed.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The AI ​​visual recognition method and system for identifying and correcting T-beam assembly gaps use the AI ​​visual recognition method, based on basic dimension data combined with tracking of boundary feature points on the T-beam, cross beam and bridge deck, so as to understand the spatial distance data of the T-beam, cross beam and bridge deck. The spatial distance data can be obtained through the boundary feature points without processing a large amount of point cloud data. The three-dimensional size of the gap can be identified through the spatial distance data. The processing process is simple, and the correction of the T-beam is achieved by controlling the movement of the boundary feature points and boundary feature lines, thereby ensuring the convenience of calculation.

[0042] At the same time, through the AI ​​simulation prediction method, the T-beams in the pre-assembly area are simulated, and the spatial posture of the T-beams is pre-corrected based on the simulation results to obtain calculation correction data. This allows most of the correction work of the T-beams to be completed before the formal assembly, thereby ensuring that the T-beams can be assembled quickly and accurately after moving to the assembly area. This pre-correction method shortens the assembly cycle. At the same time, in the AI ​​learning method, based on the assembly results of the first T-beam, the spatial posture of subsequent T-beams in the pre-assembly area can be automatically adjusted to coincide with the standard virtual spatial posture, thereby reducing the need for multiple adjustments to the T-beams in the assembly area and ensuring the uniformity of the installation of T-beams on the same floor.

[0043] The AI ​​visual recognition method ensures that the size of the T-beam assembly gap is strictly within the gap tolerance range by measuring and correcting the spatial posture and spatial distance data of the T-beam, thereby ensuring the accuracy and stability of the T-beam assembly. It also effectively avoids bridge structure problems that may be caused by assembly gaps that are too large or too small. The real-time monitoring and correction mechanism not only ensures the accuracy and quality of T-beam assembly, but also significantly improves the stability and safety of the overall bridge structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Schematic diagram of the AI ​​visual recognition method of the present invention;

[0045] Figure 2 Schematic diagram of the AI ​​learning method of the present invention;

[0046] Figure 3 This is a diagram combining the AI ​​visual recognition method and the AI ​​learning method of the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] like Figure 1-Figure 3 As shown, the present invention provides a technical solution: an AI visual recognition method for identifying and correcting T-beam assembly gaps, the AI ​​visual recognition method comprising:

[0049] S01, setting an assembly area and a pre-assembly area in a spatial area according to the T-beam assembly process, and obtaining basic dimension data and real-time image data of the T-beam, diaphragm and bridge deck;

[0050] S02, the processor receives real-time image data and obtains spatial distance data between the T-beam, the diaphragm and the bridge deck and the spatial posture of the T-beam through image recognition calculation method;

[0051] S03, performing assembly simulation on the T-beam in the pre-assembly area using an AI simulation prediction method, and correcting the spatial posture of the T-beam in the pre-assembly area based on the assembly simulation results to obtain calculation correction data;

[0052] S04, after the T-beams are assembled, the spatial posture of the T-beams in the assembly area is corrected according to the assembly results to obtain actual correction data and actual gap three-dimensional dimensions;

[0053] S05: An AI learning method for assembling T-beams on the same floor is provided. The AI ​​learning method includes:

[0054] Step 1: Assemble the first T-beam on the same floor according to S01, S02, S03, and S04;

[0055] Step 2: Combine the calculated correction data with the actual correction data to obtain reference correction data. The actual spatial posture of the first T-beam is obtained through the actual three-dimensional size of the gap. The standard virtual spatial posture in the pre-assembly area is obtained through simulation calculation. Based on S02, the No. 1 spatial posture of subsequent T-beams on the same layer is obtained. The standard virtual spatial posture and the No. 1 spatial posture are overlapped through the spatial correction method.

[0056] Step 3: According to S04, move the subsequent T-beams on the same layer that have completed the overlap to the assembly area for assembly, and obtain the actual correction data No. 1.

[0057] Step 4: The actual correction data No. 1 is combined with the reference correction data according to the process in step 2 to obtain new reference correction data. Steps 2, 3 and 4 are repeated based on the new reference correction data until the complete assembly of the T-beams on the same floor is completed.

[0058] The assembly results include the real spatial posture of the T-beam and the real three-dimensional size of the gap after T-beam assembly obtained through image recognition calculation method;

[0059] Correcting the spatial posture of the T-beam in the assembly area includes: determining whether the three-dimensional size of the actual gap is within the gap tolerance range through the spatial posture correction method, correcting the actual spatial posture of the T-beam in the assembly area through the spatial posture correction method, and obtaining actual correction data generated by the spatial posture correction method.

[0060] The AI ​​simulation prediction method includes combining the spatial distance data and the spatial posture of the T-beam in the pre-assembly area to calculate the gap size after the T-beam is moved and spliced ​​according to the current posture, and obtain the calculated three-dimensional size of the gap;

[0061] The assembly simulation results include determining whether the calculated three-dimensional size of the gap is within the gap tolerance range. If it is not within the gap tolerance range, the spatial posture of the T-beam in the pre-assembly area is corrected using the spatial posture correction method. If it is within the gap tolerance range, no posture correction is required.

[0062] Obtain calculation correction data during the operation of the spatial attitude correction method.

[0063] Image recognition computational methods include:

[0064] An AI automatic recognition system is provided for identifying the edge positions of T-beams, diaphragms and bridge decks. The AI ​​automatic recognition system performs pixel-level segmentation on the edges of T-beams, diaphragms and bridge decks in real-time images, extracts target boundaries, and realizes edge detection between T-beams, diaphragms and bridge decks. A virtual edge line is generated based on the detected edge, boundary feature points are set on the virtual edge line, and based on the basic dimension data, the image is perspective-processed to generate a virtual edge line that is not displayed in the image data. A point tracking system is provided for tracking the boundary feature points. The spatial posture of the T-beam is obtained based on the boundary feature points of the T-beam, and the spatial distance data between the T-beam, diaphragm and bridge deck are obtained based on the basic dimension data, the boundary feature points on the T-beam, the boundary feature points on the diaphragm and the boundary feature points on the bridge deck.

[0065] The spatial correction method includes analyzing difference data between the standard virtual space posture and the No. 1 spatial posture through a spatial comparison method, generating standard correction data based on the difference data by a processor, and adjusting the No. 1 spatial posture based on the standard correction data so that the standard virtual space posture and the No. 1 spatial posture coincide with each other;

[0066] Generate a standard virtual spatial posture of the T-beam in the pre-assembly area, perform image detection on subsequent T-beams on the same layer, calculate the No. 1 spatial posture of subsequent T-beams on the same layer through image calculation methods, analyze the difference data between the standard virtual spatial posture and the No. 1 spatial posture through spatial comparison methods, and the processor generates standard correction data based on the difference data, and adjusts the No. 1 spatial posture based on the standard correction data, so that the standard virtual spatial posture and the No. 1 spatial posture coincide with each other.

[0067] The simulation calculation to obtain the standard virtual space posture includes: the actual space posture is matched in the pre-assembly area by referring to the correction data, and a standard virtual space posture identical to the actual space posture is established in the pre-assembly area.

[0068] Spatial contrast methods include:

[0069] A virtual boundary feature line is set on the standard virtual space posture, and a boundary feature edge line No. 1 corresponding to the virtual boundary feature line is set on the No. 1 space posture according to the spatial position of the virtual boundary feature line on the T-beam. The spatial difference between the virtual boundary feature points with the corresponding relationship and the boundary feature edge line No. 1 is analyzed to obtain difference data.

[0070] The spatial posture correction method includes:

[0071] The calculated three-dimensional size of the gap is input into the processing terminal, which determines whether the calculated three-dimensional size of the gap is within the gap tolerance range. If not, a correction instruction for adjusting the posture of the T-beam in the pre-assembly area is generated, and the spatial posture of the T-beam is adjusted according to the correction instruction. Based on the change in the spatial posture of the T-beam, the new calculated three-dimensional size of the gap is repeatedly calculated and input into the processing terminal until the calculated three-dimensional size of the gap is within the gap tolerance range, and the calculated correction data is recorded.

[0072] An AI visual recognition system for identifying and correcting gaps in T-beam assembly is provided, which employs the aforementioned AI visual recognition method for identifying and correcting gaps in T-beam assembly; the recognition system comprises:

[0073] Image acquisition module: obtains real-time image data of the assembly area and pre-assembly area;

[0074] Processing module: Receives real-time image data and obtains the spatial distance data between the T-beam, diaphragm and bridge deck, as well as the spatial posture of the T-beam through image recognition and calculation methods;

[0075] The T-beam in the pre-assembly area is assembled by using an AI simulation prediction method. Based on the assembly simulation results, the spatial posture of the T-beam in the pre-assembly area is corrected to obtain calculation correction data;

[0076] For the T-beams being assembled, based on the assembly results, the spatial posture of the T-beams in the assembly area is corrected to obtain the actual correction data and the actual three-dimensional size of the gap;

[0077] AI learning module: The processor is equipped with an AI learning method for assembling T-beams on the same floor. The AI ​​learning method includes:

[0078] Step 1: Assemble the first T-beam on the same floor according to S01, S02, S03, and S04;

[0079] Step 2: Combine the calculated correction data with the actual correction data to obtain reference correction data. The actual spatial posture of the first T-beam is obtained through the actual three-dimensional size of the gap. Simulation calculation is performed with reference to the correction data and the actual spatial posture to obtain the standard virtual spatial posture in the pre-assembly area. Based on S02, the No. 1 spatial posture of the subsequent T-beams on the same layer is obtained. The spatial correction method is used to achieve the coincidence of the standard virtual spatial posture and the No. 1 spatial posture.

[0080] Step 3: According to S04, the subsequent T-beams on the same layer that have completed the overlap are moved to the assembly area for assembly, and the actual correction data No. 1 is obtained;

[0081] Step 4: The actual correction data No. 1 is combined with the reference correction data according to the process in step 2 to obtain new reference correction data. Steps 2, 3 and 4 are repeated based on the new reference correction data until the complete assembly of the T-beams on the same floor is completed.

[0082] The T-beam is moved by lifting equipment to the top of the assembly position, and then the T-beam is moved down and corrected to realize the assembly of the T-beam. Based on the assembly process of the T-beam, an assembly area and a pre-assembly area are set in space. The pre-assembly area is at the top of the assembly area. The T-beam can be assembled with the cross beam and the bridge deck in the assembly area, and the spatial posture of the T-beam can be adjusted in the pre-assembly area. Ideally, after the spatial posture is adjusted in the pre-assembly area, during assembly, it is only necessary to move the T-beam to realize assembly, without the need for adjustments such as angle axis deflection, thereby reducing the correction process in the assembly area.

[0083] The correction data includes adjustments to the position and angle of the T-beam. After assembly, the real spatial posture and the real three-dimensional size of the gap can be obtained through the spatial distance data between the T-beam, diaphragm and bridge deck and the spatial posture of the T-beam. After assembly, the spatial distance data is replaced by the real three-dimensional size of the gap to distinguish them. In essence, the three-dimensional size of the gap also indicates the spatial distance data between the T-beam, diaphragm and bridge deck. The spatial posture refers to the angle and axis deflection of the T-beam, while the spatial distance data represents the spatial distance parameters between the boundary feature points on the T-beam, diaphragm and bridge deck.

[0084] When acquiring real-time image data, areas can be divided in the real-time image data to obtain assembly areas and pre-assembly areas. An image monitoring unit is provided for photographing the assembly areas and the pre-assembly areas. The gap tolerance range is established based on the basic dimension data of the current T-beam, cross beam and bridge deck. Through the gap tolerance range, it is ensured that the actual gap tolerance range can be within the gap tolerance range, thereby ensuring the use effect of the T-beam. The establishment of the gap tolerance range is an existing technology.

[0085] The actual three-dimensional size of the gap is obtained through the image recognition calculation method, and the spatial posture correction method is used to make the actual three-dimensional size of the gap fall within the gap tolerance range, and the actual correction data generated by the spatial posture correction method is obtained. That is, under the correction of the actual correction data, the actual three-dimensional size of the gap is changed so that it can fall within the gap tolerance range, and the actual three-dimensional size of the gap at this time and within the gap tolerance range is obtained simultaneously. In the AI ​​learning method, the actual three-dimensional size of the gap will be used as a standard reference to adjust the subsequent T-beams of the same layer in the pre-assembly area.

[0086] The processor determines the difference between the three-dimensional size of the gap and the gap tolerance range, and determines whether the three-dimensional size of the gap is within the gap tolerance range. When the processing terminal determines that the three-dimensional size data of the gap is not within the gap tolerance range, it controls the lifting equipment to move the T-beam, thereby controlling the spatial posture of the T-beam to change, obtaining a new three-dimensional size, and determining whether the new three-dimensional size is qualified. The above process is repeated until it is judged to be qualified. At the same time, the movement data of the lifting equipment moving the T-beam is obtained in the process. Due to the different areas, in the pre-assembly area, only the spatial posture of the T-beam needs to be controlled and adjusted. In the assembly area, if necessary, the spatial posture and spatial distance data of the T-beam are adjusted. The AI ​​visual recognition method ensures that the size of the T-beam assembly gap can be strictly within the gap tolerance range by measuring and correcting the spatial posture and spatial distance data of the T-beam, thereby ensuring the accuracy and stability of the T-beam assembly, and effectively avoiding bridge structure problems that may be caused by an assembly gap that is too large or too small. The real-time monitoring and correction mechanism not only ensures the accuracy and quality of the T-beam assembly, but also significantly improves the stability and safety of the overall bridge structure.

[0087] Whether in the pre-assembly area or in the assembly area, the spatial posture of the T-beam will be adjusted many times, which will cause duplicate data in the movement data. Duplicate data refers to the process of moving data. For example, after moving 3mm to the left and then moving 2mm to the right, the two will be combined, and duplicate data will exist. Therefore, when the calculated correction data and the actual correction data are combined, deduplication processing is required. The calculated correction data and the actual correction data and other correction data are essentially the movement and adjustment of boundary feature points and boundary feature lines. After deduplication processing, the effective data obtained is a 1mm left shift. Combined with the overall assembly of the T-beam, it is not only limited to movement, but also includes the adjustment of angles and axis deflections.

[0088] When analyzing the difference data between the subsequent standard virtual space posture and the No. 1 space posture, boundary feature lines are set on the boundary of the T-beam to achieve the overlap between the standard virtual space posture and the No. 1 space posture. When the boundary feature lines on the T-beam and the boundary feature lines on the standard virtual space posture overlap, the overlap between the standard virtual space posture and the No. 1 space posture can be achieved. The boundary feature lines are extracted from the key parts of the geometric contour of the T-beam (such as the boundary line between the web and the flange). These lines have clear geometric invariance and can fully characterize the three-dimensional spatial posture of the T-beam. Therefore, the three-dimensional spatial posture of the T-beam can be known only by the edge feature lines. Therefore, it is only necessary to achieve the overlap of the boundary feature lines on the T-beam and the boundary feature lines on the standard virtual space posture to achieve the overlap of the standard virtual space posture and the No. 1 space posture in the pre-assembly area, so as to reduce the number of times the T-beam is corrected in the assembly area. After the spatial postures overlap, the remaining parts can be moved to achieve assembly.

[0089] By using AI visual recognition methods, based on basic dimensional data combined with tracking of boundary feature points on T-beams, cross beams and bridge decks, the spatial distance data of T-beams, cross beams and bridge decks can be understood. Spatial distance data can be obtained through boundary feature points without processing large amounts of point cloud data. The three-dimensional size of the gap can be identified through spatial distance data. The processing process is simple, and the correction of the T-beam is achieved by controlling the movement of boundary feature points and boundary feature lines, ensuring the convenience of calculation.

[0090] In the longitudinal direction of the bridge, a cross beam or beam is set at a certain interval to connect multiple T beams into a whole. The T beams between every two adjacent cross beams constitute a "layer". A bridge may be divided into several layers of T beams by multiple cross beams. Each layer contains several T beams. When assembling T beams on the same layer, the T beams on the same layer have highly consistent geometric parameters and mechanical environments, and even the ambient temperature may be similar. By recording the assembly data of the first T beam, a standardized assembly parameter template can be generated to directly guide the positioning of subsequent beam sections and reduce the amount of repeated calculations. At the same time, through the installation of the first T beam, the subsequent T beams on the same layer can be installed in the pre-assembly area to realize early adjustment of the T beam, reducing the situation of multiple adjustments to the T beam in the assembly area. The T beams between adjacent cross beams constitute a layer. The T beams between two adjacent cross beams constitute a layer. Multiple T-beams will be spliced ​​during installation. Since the installation positions of T-beams on the same layer are similar and they will be installed within the same period of time, the first T-beam on the same layer will be assembled when the T-beams on the same layer are installed, and the data during the assembly process of the first T-beam will be used as a reference. In this way, the assembly of T-beams on the same layer in the AI ​​learning method is realized. The image of the T-beam in the area to be assembled is processed based on real-time image data, and the movement data of the first T-beam on the same layer is obtained. Based on the movement data of the first T-beam on the same layer, the subsequent T-beams on the same layer are used as a reference to obtain the calculated correction data, actual correction data and actual gap three-dimensional size after the first T-beam is assembled. The calculated correction data and the actual correction data are combined to obtain the reference correction data. The actual spatial posture of the first T-beam after assembly can be obtained through the actual gap three-dimensional size.

[0091] Image recognition and calculation methods are used to accurately capture the spatial distance data between T-beams, diaphragms, and the bridge deck, as well as the spatial orientation of the T-beams. This precise acquisition relies on AI automatic recognition technology, which performs pixel-level segmentation of the T-beams, diaphragms, and bridge deck edges in real-time images, extracting target boundaries, generating virtual edges, and setting boundary feature points. Compared to traditional manual measurement and calibration methods, this method not only significantly improves measurement accuracy and efficiency, but also effectively reduces errors that can be caused by human factors.

[0092] The image monitoring unit can be fixedly installed on the diaphragm, lifting equipment and bridge deck. Through real-time detection by multiple image monitoring units, real-time shooting of the T-beam movement is guaranteed. In addition, through real-time detection by multiple image monitoring units, boundary feature points or virtual boundary lines may be blocked. However, combined with basic dimension data and the mutual correspondence between boundary feature points on the same object, the blocked boundary feature points or virtual boundary lines can also be calculated and processed, which facilitates subsequent AI calculations. In the captured image, due to the light blocking of the T-beam itself, some virtual boundary lines cannot be obtained by the image monitoring unit, but the virtual boundary lines can be calculated through the boundary feature points. Therefore, the virtual boundary lines that are not visible in the image data are calculated through the visible virtual boundary lines. This method is not only applicable to T-beams, but can also be implemented on diaphragms and bridge decks. By calculating the blocked boundaries, the undisplayed virtual boundary lines are obtained for subsequent AI simulation calculations and AI recognition of spatial posture and spatial distance data in the image.

[0093] The application of the point tracking system enables the boundary points on the T-beam to be tracked during its movement. By connecting two boundary points, the boundary line can be obtained. At the same time, when obtaining the boundary points, the spatial position of the boundary points on the T-beam is also obtained. The boundary points and spatial positions are combined to obtain boundary feature points. The two boundary feature points are connected to obtain the boundary feature line. At the same time, since the image is shot from multiple angles, the boundary feature line can be tracked by tracking the boundary feature points.

[0094] The captured image is uploaded to the processor, and AI automatically recognizes the real-time image data. AI automatically recognizes the pixel-level segmentation of the T-beams, cross beams and bridge deck edges in the real-time image, extracts the target boundary, and generates a virtual edge line by identifying the boundary position of the T-beams, cross beams and bridge deck. Boundary feature points are set on the virtual edge line, and the boundary feature points of the T-beams, cross beams and bridge deck are tracked based on the real-time image. The boundary feature line is obtained by connecting the two points into a line.

[0095] The processor generates standard correction data based on the difference data. Since the difference data is obtained by analyzing the spatial difference between the virtual boundary feature line and the No. 1 boundary feature edge line with a corresponding relationship, the standard correction data is obtained by adjusting the No. 1 spatial posture of the subsequent T-beam on the same layer so that the No. 1 boundary feature edge line and the virtual boundary feature line coincide with each other.

[0096] Afterwards, the corrected subsequent T-beams on the same floor are moved to the assembly area for assembly, and the actual correction data No. 1 during the assembly process is obtained. The actual correction data No. 1 is combined with the reference correction data, and this cycle is repeated to achieve the installation of the T-beams on the same floor. Through the AI ​​simulation prediction method, the assembly simulation of the T-beams in the pre-assembly area is performed, and the spatial posture of the T-beams is pre-corrected based on the simulation results to obtain the calculated correction data. This allows most of the correction work of the T-beams to be completed before the formal assembly, thereby ensuring that the T-beams can be quickly and accurately assembled after being moved to the assembly area. This pre-correction method shortens the assembly cycle. At the same time, in the AI ​​learning method, based on the assembly results of the first T-beam, the spatial posture of the subsequent T-beams in the pre-assembly area can be automatically adjusted to coincide with the standard virtual spatial posture, thereby reducing the need for multiple adjustments to the T-beams in the assembly area and ensuring the uniformity of the installation of the T-beams on the same floor.

[0097] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is limited by the accompanying embodiments and their equivalents.

Claims

1. An AI visual recognition method for T-beam assembly gap identification and correction, characterized in that: The AI ​​visual recognition method includes: S01, setting an assembly area and a pre-assembly area in a spatial area according to the T-beam assembly process, and obtaining basic dimension data and real-time image data of the T-beam, diaphragm and bridge deck; S02, the processor receives real-time image data and obtains spatial distance data between the T-beam, the diaphragm and the bridge deck and the spatial posture of the T-beam through an image recognition calculation method; S03, performing assembly simulation on the T-beam in the pre-assembly area using an AI simulation prediction method, and correcting the spatial posture of the T-beam in the pre-assembly area based on the assembly simulation results to obtain calculation correction data; S04, after the T-beams are assembled, the spatial posture of the T-beams in the assembly area is corrected according to the assembly results to obtain actual correction data and actual gap three-dimensional dimensions; S05: An AI learning method for assembling T-beams on the same floor is provided. The AI ​​learning method includes: Step 1: Assemble the first T-beam on the same floor according to S01, S02, S03, and S04; Step 2: Combine the calculated correction data with the actual correction data to obtain reference correction data. The actual spatial posture of the first T-beam is obtained through the actual three-dimensional size of the gap. The standard virtual spatial posture in the pre-assembly area is obtained through simulation calculation. Based on S02, the No. 1 spatial posture of subsequent T-beams on the same layer is obtained. The standard virtual spatial posture and the No. 1 spatial posture are overlapped through the spatial correction method. Step 3: According to S04, the subsequent T-beams on the same layer that have completed the overlap are moved to the assembly area for assembly, and the actual correction data No. 1 is obtained; Step 4: The actual correction data No. 1 is combined with the reference correction data according to the process in step 2 to obtain new reference correction data. Steps 2, 3 and 4 are repeated based on the new reference correction data until the complete assembly of the T-beams on the same floor is completed.

2. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 1, characterized in that: The assembly results include the real spatial posture of the T-beam and the real three-dimensional size of the gap after T-beam assembly obtained through image recognition calculation method; Correcting the spatial posture of the T-beam in the assembly area includes: determining whether the three-dimensional size of the actual gap is within the gap tolerance range through the spatial posture correction method, correcting the actual spatial posture of the T-beam in the assembly area through the spatial posture correction method, and obtaining actual correction data generated by the spatial posture correction method.

3. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 1, characterized in that: The AI ​​simulation prediction method includes calculating the gap size after the T-beam is moved and spliced ​​according to the current posture of the T-beam in the pre-assembly area by combining the spatial distance data and the spatial posture of the T-beam to obtain the calculated three-dimensional size of the gap; The assembly simulation results include determining whether the calculated three-dimensional size of the gap is within the gap tolerance range. If it is not within the gap tolerance range, the spatial posture of the T-beam in the pre-assembly area is corrected using the spatial posture correction method. If it is within the gap tolerance range, no posture correction is required. Obtain calculation correction data during the operation of the spatial attitude correction method.

4. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 1, characterized in that: The image recognition calculation method includes: An AI automatic recognition system is provided for identifying the edge positions of T-beams, diaphragms and bridge decks. The AI ​​automatic recognition system performs pixel-level segmentation on the edges of T-beams, diaphragms and bridge decks in real-time images, extracts target boundaries, and realizes edge detection between T-beams, diaphragms and bridge decks. A virtual edge line is generated based on the detected edge, boundary feature points are set on the virtual edge line, and based on the basic dimension data, the image is perspective-processed to generate a virtual edge line that is not displayed in the image data. A point tracking system is provided for tracking the boundary feature points. The spatial posture of the T-beam is obtained based on the boundary feature points of the T-beam, and the spatial distance data between the T-beam, diaphragm and bridge deck are obtained based on the basic dimension data, the boundary feature points on the T-beam, the boundary feature points on the diaphragm and the boundary feature points on the bridge deck.

5. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 1, characterized in that: The spatial correction method includes analyzing difference data between the standard virtual space posture and the first spatial posture using a spatial comparison method, generating standard correction data based on the difference data by a processor, and adjusting the first spatial posture based on the standard correction data so that the standard virtual space posture and the first spatial posture coincide with each other; Generate a standard virtual spatial posture of the T-beam in the pre-assembly area, perform image detection on subsequent T-beams on the same layer, calculate the No. 1 spatial posture of subsequent T-beams on the same layer through image calculation methods, analyze the difference data between the standard virtual spatial posture and the No. 1 spatial posture through spatial comparison methods, and the processor generates standard correction data based on the difference data, and adjusts the No. 1 spatial posture based on the standard correction data, so that the standard virtual spatial posture and the No. 1 spatial posture coincide with each other.

6. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 1, characterized in that: The simulation calculation to obtain the standard virtual space posture includes: the actual space posture is matched in the pre-assembly area by referring to the correction data, and a standard virtual space posture identical to the actual space posture is established in the pre-assembly area.

7. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 5, characterized in that: The spatial contrast method includes: A virtual boundary feature line is set on the standard virtual space posture, and a boundary feature edge line No. 1 corresponding to the virtual boundary feature line is set on the No. 1 space posture according to the spatial position of the virtual boundary feature line on the T-beam. The spatial difference between the virtual boundary feature points with the corresponding relationship and the boundary feature edge line No. 1 is analyzed to obtain difference data.

8. The AI ​​visual recognition method for T-beam assembly gap identification and correction according to claim 2, characterized in that: The space posture correction method comprises: The calculated three-dimensional size of the gap is input into the processing terminal, which determines whether the calculated three-dimensional size of the gap is within the gap tolerance range. If not, a correction instruction for adjusting the posture of the T-beam in the pre-assembly area is generated, and the spatial posture of the T-beam is adjusted according to the correction instruction. Based on the change in the spatial posture of the T-beam, the new calculated three-dimensional size of the gap is repeatedly calculated and input into the processing terminal until the calculated three-dimensional size of the gap is within the gap tolerance range, and the calculated correction data is recorded.

9. An AI visual recognition system for identifying and correcting gaps in T-beam assembly, employing the AI ​​visual recognition method for identifying and correcting gaps in T-beam assembly according to any one of claims 1 to 8; characterized in that: The identification system comprises: Image acquisition module: obtains real-time image data of the assembly area and pre-assembly area; Processing module: Receives real-time image data and obtains the spatial distance data between the T-beam, diaphragm and bridge deck, as well as the spatial posture of the T-beam through image recognition and calculation methods; The T-beam in the pre-assembly area is assembled by using an AI simulation prediction method. Based on the assembly simulation results, the spatial posture of the T-beam in the pre-assembly area is corrected to obtain calculation correction data; For the T-beams being assembled, based on the assembly results, the spatial posture of the T-beams in the assembly area is corrected to obtain the actual correction data and the actual three-dimensional size of the gap; AI learning module: The processor is equipped with an AI learning method for assembling T-beams on the same floor. The AI ​​learning method includes: Step 1: Assemble the first T-beam on the same floor according to S01, S02, S03, and S04; Step 2: Combine the calculated correction data with the actual correction data to obtain reference correction data. The actual spatial posture of the first T-beam is obtained through the actual three-dimensional size of the gap. Simulation calculation is performed with reference to the correction data and the actual spatial posture to obtain the standard virtual spatial posture in the pre-assembly area. Based on S02, the No. 1 spatial posture of the subsequent T-beams on the same layer is obtained. The spatial correction method is used to achieve the coincidence of the standard virtual spatial posture and the No. 1 spatial posture. Step 3: According to S04, the subsequent T-beams on the same layer that have completed the overlap are moved to the assembly area for assembly, and the actual correction data No. 1 is obtained; Step 4: The actual correction data No. 1 is combined with the reference correction data according to the process in step 2 to obtain new reference correction data. Steps 2, 3 and 4 are repeated based on the new reference correction data until the complete assembly of the T-beams on the same floor is completed.

Citation Information

Patent Citations

  • Displacement deviation value detection method and system for bridge width splicing, terminal and medium

    CN116958259A

  • Steel bridge simulation pre-assembly method integrating ground type laser scanning and handheld type laser scanning

    CN114332402A

  • Algorithm for predicting operation time of ship t-beam welding robot on basis of machine learning

    WO2023221476A1