A digital installation method and system for steel anchor beams

Abstract

This invention relates to the field of bridge construction technology, and in particular to a digital installation method and system for steel anchor beams. The method includes: deploying positioning targets on the steel bracket; installing several binocular camera modules on the anchor beam; constructing a three-dimensional model of the steel bracket and anchor beam, clarifying the fixed spatial relationship between the positioning targets, binocular camera modules, and anchor beam, and calibrating and associating the three-dimensional model with the on-site construction coordinate system; deploying several BeiDou measurement reference stations to provide a basis for real-time position adjustment of the steel bracket during installation by measuring the coordinates of the positioning targets in real time; and activating each binocular camera module, which acquires images of the positioning targets in real time to provide a basis for real-time position adjustment of the anchor beam during installation. This application can effectively improve installation accuracy, overcome measurement fluctuations caused by environmental interference, meet the high-precision requirements of modern engineering, reduce the probability of safety accidents, solve the problem of work stoppages due to environmental constraints in traditional methods, and ensure the controllability of the construction period.

Description

Technical Field

[0001] This invention relates to the field of bridge construction technology, and in particular to a digital installation method and system for steel anchor beams. Background Technology

[0002] In large-scale steel structure projects, the installation of steel anchor beams is one of the core construction steps. Steel anchor beams consist of anchor beams and steel brackets. The main function of the anchor beam is to transfer the tension of the bridge cables or the load of the building structure to the steel bracket, and then to the main bridge structure. Traditional steel anchor beam installation methods mainly rely on manual operation and measurement. The specific process is as follows: the anchor beam is hoisted to the vicinity of the steel bracket by a crane, and construction workers use measuring equipment such as total stations and levels to manually measure the positional deviation of the anchor beam. Then, the crane operator is directed via walkie-talkie to adjust the anchor beam's posture, repeating the adjustment until the deviation meets the requirements, and finally fixing it in place. However, this traditional method has the following drawbacks: Manual measurement is significantly affected by environmental factors, such as changes in lighting, wind interference, and temperature deformation. A single measurement error typically ranges from 5 to 10 mm, which is insufficient to meet the modern engineering requirements for installation accuracy within ±2 mm. Furthermore, anchor beams sway due to wind during hoisting, making it difficult for manual adjustments to track deviations in real time, leading to unstable final installation accuracy. In addition, the manual measurement and adjustment process is cumbersome, requiring multiple measurements and adjustments for each installation, which is time-consuming. In complex environments, such as high-altitude or nighttime construction, the time required will be further extended, severely impacting the overall project progress.

[0003] Furthermore, traditional installation methods require multiple workers to collaborate, including surveyors, crane operators, and workers at height. This not only increases labor costs but also poses significant safety risks. Workers at height face the risk of falls, and improper crane operation during adjustment could cause the anchor beam to collide with the steel bracket, leading to component damage or accidents. Traditional methods are also highly dependent on the construction environment; inclement weather makes normal construction difficult, necessitating work stoppages and impacting project schedules. Additionally, manual adjustment of large and heavy steel anchor beams is even more challenging and requires greater precision control.

[0004] As steel structure projects become larger and more complex, traditional installation methods can no longer meet the project requirements. A digital and automated installation technology is needed to achieve high-precision, high-efficiency, and high-safety installation of steel anchor beams. Summary of the Invention

[0005] This invention provides a digital installation method and system for steel anchor beams, which can effectively solve the problems pointed out in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A digital installation method for steel anchor beams includes: Based on the designed installation position of the anchor beam relative to the steel bracket, several positioning targets are set on the steel bracket. Several binocular camera modules are installed on the anchor beam, such that the shooting range of any one of the binocular camera modules covers at least two of the positioning targets. Construct a three-dimensional model of the steel bracket and anchor beam, clarify the fixed spatial relationship of the positioning target, binocular camera module and anchor beam in the three-dimensional model, and calibrate and associate the three-dimensional model with the on-site construction coordinate system; Several Beidou measurement reference stations are set up around the steel bracket installation area. The Beidou measurement reference stations measure the coordinates of the positioning target in real time to provide a basis for the real-time position adjustment of the steel bracket during the installation process, so that the installation deviation of the steel bracket relative to the on-site construction coordinate system is within a first set range. Each of the binocular camera modules is activated. The binocular camera modules acquire images of the positioning target in real time, providing a basis for real-time position adjustment of the anchor beam during installation, so that the installation deviation of the anchor beam relative to the steel bracket is within a second set range.

[0007] Furthermore, the images captured by the binocular camera module are transmitted wirelessly to the hoisting control system in real time, including: High-quality images are obtained by identifying and discarding blurry frames using a pre-trained model at the binocular camera module. Feature points of the positioning target are extracted from the high-quality image, and the local three-dimensional coordinate data of the anchor beam are calculated based on the principle of binocular vision. Assign a high-priority transmission marker to the local three-dimensional coordinate data, and assign a low-priority transmission marker to the high-quality image; The hoisting control system prioritizes receiving and processing the local three-dimensional coordinate data, and fuses the local three-dimensional coordinate data of each perspective corresponding to several binocular camera modules. The fusion is performed using the weighted least squares method, and the anchor beam installation deviation is calculated.

[0008] Furthermore, when the hoisting control system detects that the rate of change of the local three-dimensional coordinate data exceeds a set threshold, it retrieves the high-quality image at the corresponding time. The pre-trained model is reused to identify the occlusion type. If it is determined to be dust or fog occlusion, the weight of the current binocular camera module is reduced, and the weight of other binocular camera modules is increased. If it is determined to be mechanical interference occlusion, the corresponding positioning target is marked as invalid.

[0009] Furthermore, the pre-trained model is constructed using a multi-stage transfer learning framework, the construction process of which includes: In the basic training phase, a network model is trained using a public dataset. The network model must at least recognize motion blur, defocus blur, and environmental occlusion features. In the domain adaptation phase, the parameters of the network model are fine-tuned based on historical images of steel structure engineering to adapt to the construction scenario of steel anchor beams. During the edge deployment phase, the fine-tuned network model is quantized into a set format and embedded into the chip of the binocular camera module.

[0010] Furthermore, the positioning target accounts for more than 15% of the image in the field of view of the binocular camera module.

[0011] Furthermore, the positioning target is a barcode target, the barcode adopts a black and white stripe structure of equal width, the stripe width is 1 to 2 millimeters, the target base is made of metal material, and the surface is covered with a reflective coating.

[0012] Furthermore, the binocular camera module has a resolution higher than 5 million pixels, a frame rate higher than 30fps, and a lens focal length of 8 to 12 millimeters.

[0013] Furthermore, it also includes a system calibration procedure before construction, including: A calibration platform is set up near the steel bracket, and a standard calibration plate of known size is placed there. The hoisting equipment is controlled to move the anchor beam to different preset positions; The binocular camera module acquires images of the standard calibration board, calculates the measurement coordinates of the feature points on the standard calibration board, and compares them with standard coordinates. The intrinsic and extrinsic parameters of the binocular camera module were corrected based on the comparison results.

[0014] Furthermore, the image acquisition frequency of the binocular camera module during system calibration is 0.4 to 0.6 Hz, and the image acquisition frequency during construction is 1.8 to 2.2 Hz.

[0015] A digital installation and construction system for steel anchor beams, comprising: Several positioning targets are arranged on the steel bracket, with the designed installation position of the anchor beam relative to the steel bracket as the reference. Several binocular camera modules are installed on the anchor beam, and the shooting range of any one of the binocular camera modules covers at least two of the positioning targets. The 3D modeling module constructs a 3D model of the steel bracket and anchor beam, clarifies the fixed spatial relationship between the positioning target, the binocular camera module and the anchor beam in the 3D model, and calibrates and associates the 3D model with the on-site construction coordinate system. The Beidou positioning module includes several Beidou measurement reference stations deployed around the steel bracket installation area, used to measure the coordinates of the positioning target in real time and output the position adjustment basis of the steel bracket so that the installation deviation of the steel bracket relative to the on-site construction coordinate system is within a first set range. The dynamic control module activates each of the binocular camera modules to acquire images of the positioning target in real time, and generates a basis for adjusting the position of the anchor beam so that the installation deviation of the anchor beam relative to the steel bracket is within a second set range.

[0016] The technical solution of this invention can achieve the following technical effects: This application uses the BeiDou measurement reference station to correct the steel bracket's pose in real time, and combines binocular vision with multi-angle fusion calculation to solve the anchor beam's spatial position, effectively improving installation accuracy and overcoming measurement fluctuations caused by environmental interference, meeting the high-precision requirements of modern engineering. In addition, it can shorten the installation time per operation, reduce the number of personnel working at heights, eliminate the risk of collisions caused by human error, and reduce the probability of safety accidents. The anti-interference design of the positioning target and the protection capabilities of the binocular camera module can ensure stable operation under complex working conditions such as strong winds and nighttime, solving the problem of work stoppages due to environmental constraints caused by traditional methods, and ensuring the controllability of the project schedule. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart for the digital installation construction method of steel anchor beams; Figure 2 This is a front view of the steel anchor beam; Figure 3 for Figure 2 Sectional view at point AA; Figure 4 A flowchart illustrating how images captured by a binocular camera module are transmitted wirelessly to a hoisting control system in real time. Figure 5 A flowchart illustrating the construction of a pre-trained model using a multi-stage transfer learning framework; Reference numerals: 01, Anchor beam; 02, Steel bracket; 03, First installation area; 04, Second installation area. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Example 1 A digital installation construction method for steel anchor beams, such as Figures 1 to 3 As shown, it includes: A1: Using the designed installation position of anchor beam 01 relative to steel bracket 02 as a reference, several positioning targets are set on steel bracket 02; taking the example that the supporting surface of steel bracket 02 extends beyond the bottom surface of anchor beam 01 on both sides, the positioning targets in this step can be specifically set at the extension position of the supporting surface relative to anchor beam 01, such as... Figure 3 The first installation area 03 shown in the figure; as a further specific implementation, two positioning targets are respectively arranged on each side of the steel bracket 02, so that a total of 8 positioning targets are arranged on the two steel brackets 02 that support the anchor beam 01; by using the positioning targets as spatial reference points, the problem of reference positioning during the installation of the anchor beam 01 can be solved, ensuring that there are stable targets for subsequent visual measurement; A2: Install several binocular camera modules on anchor beam 01, ensuring that the shooting range of any binocular camera module covers at least two positioning targets; this step constructs a multi-view observation system to provide hardware support for real-time capture of the spatial position of the positioning targets, avoiding the risk of single-point failure; as a further specific implementation method, see also... Figure 3 Mounting brackets are installed in the second installation area 04 at the four corners of the anchor beam 01, and the four binocular camera modules are fixed separately through the mounting brackets. In order to make the use of the binocular camera modules more flexible, the mounting brackets can be further optimized into an angle and length adjustable structure, which facilitates the adjustment of the binocular camera modules. A3: Construct 3D models of steel bracket 02 and anchor beam 01. Define the fixed spatial relationships of the positioning targets, binocular camera modules, and anchor beam 01 within the 3D model. Also, calibrate and associate the 3D model with the on-site construction coordinate system. The 3D model in this step can specifically be a LOD4.0 level BIM model. Taking the placement of 8 positioning targets in step A1 and the installation of 4 binocular camera modules in step A2 as examples, extract the design 3D coordinates of feature points for each positioning target, feature points for each binocular camera module, and feature points for anchor beam 01 from the model. This yields a specific implementation method for defining fixed spatial relationships through these design 3D coordinates. It should be noted that each feature point can be specifically determined through the intersection of the symmetry axes of the component's outer contour, centroid calculation, or explicit marking on design drawings, and has unique coordinates in 3D space. This step calibrates the virtual model with the construction coordinate system, resolving the issue of coordinate consistency between the model and the physical site. A4: Several BeiDou measurement reference stations are deployed around the installation area of ​​the steel bracket 02. These stations measure the coordinates of the positioning targets in real time, providing a basis for real-time position adjustment of the steel bracket 02 during installation, ensuring that the installation deviation of the steel bracket 02 relative to the on-site construction coordinate system is within a first set range. Preferably, this first set range is ±1 mm. During implementation, two BeiDou measurement reference stations are preferred. After the steel bracket 02 is initially fixed, as a specific implementation method, a BeiDou mobile measurement terminal, such as a high-precision GNSS receiver, is used to measure the actual coordinates of eight positioning targets point by point at a frequency of 1 Hz. The deviation from the designed three-dimensional coordinates is calculated, and a least-squares method is used to fit the adjustment scheme, thereby adjusting the position of the steel bracket 02. Each positioning target can be measured at least three times consecutively, and the average value is taken. The working method of the BeiDou measurement reference stations in this step is existing technology and will not be described in detail here. A5: Activate each binocular camera module. The binocular camera modules acquire images of the positioning target in real time, providing a basis for real-time position adjustment of the anchor beam 01 during installation, ensuring that the installation deviation of the anchor beam 01 relative to the steel bracket 02 is within a second set range. Preferably, the second set range can be ±1 mm. During implementation, the real-time pose of the anchor beam 01 is calculated through multi-view data fusion, which can then be linked with the hoisting control system to dynamically adjust the position of the anchor beam 01, achieving high-precision docking with the steel bracket 02. In this specific implementation, images of the positioning target are acquired at a frequency greater than 30 fps, corresponding to a specific implementation using four binocular camera modules. For the four sets of camera data acquired in real time, the spatial position parameters of the anchor beam 01 relative to the steel bracket 02 can be obtained by fusing the four sets of camera data using a weighted least squares method.

[0021] This application uses the BeiDou measurement reference station to correct the pose of the steel bracket 02 in real time, and combines binocular vision multi-angle fusion to calculate the spatial position of the anchor beam 01, effectively improving installation accuracy and overcoming measurement fluctuations caused by environmental interference, meeting the high-precision requirements of modern engineering. In addition, it can shorten the installation time per operation, reduce the number of personnel working at heights, eliminate the risk of collisions caused by human error, and reduce the probability of safety accidents. The anti-interference design of the positioning target and the protection capabilities of the binocular camera module can ensure stable operation under complex working conditions such as strong winds and nighttime, solving the problem of work stoppage due to environmental constraints in traditional methods, and ensuring the controllability of the project schedule.

[0022] As a preferred embodiment of the above, such as Figure 4 As shown, the images captured by the binocular camera module are transmitted wirelessly to the hoisting control system in real time, including: B1: At the binocular camera module end, a pre-trained model is used to identify and discard blurry frames to obtain high-quality images, thereby filtering out a certain proportion of invalid images, avoiding the input of erroneous feature points into subsequent stages, reducing invalid data transmission, and alleviating the pressure on the wireless channel. B2: Extract feature points of the positioning target from high-quality images and calculate the local three-dimensional coordinate data of the anchor beam 01 based on the principle of binocular vision. Specifically, by combining the pre-calibrated intrinsic and extrinsic parameters of the binocular camera module, the local three-dimensional coordinates of the positioning target in the coordinate system of the anchor beam 01 can be calculated by the principle of triangulation. The working method of the binocular camera module in this step is the existing technology and will not be described in detail here. B3: Assign high-priority transmission markers to local 3D coordinate data and low-priority transmission markers to high-quality images, thereby achieving differential quality of service through wireless protocols. In practice, this can specifically meet the requirement that high-priority data transmission latency is less than or equal to 50ms, satisfying the dynamic adjustment requirements of hoisting, while low-priority high-quality images can be transmitted when the channel is idle, avoiding blocking control commands. B4: The hoisting control system prioritizes receiving and processing local three-dimensional coordinate data, and fuses the local three-dimensional coordinate data of each perspective corresponding to several binocular camera modules. The fusion is performed using the weighted least squares method, and the installation deviation of anchor beam 01 is calculated. Specifically, the installation deviation of anchor beam 01 can be obtained by constructing and solving the objective function that minimizes the sum of squared deviations, and this is used as the optimal adjustment amount.

[0023] Similarly, taking the above embodiment with four binocular camera modules as an example, the hoisting control system receives the local three-dimensional coordinate data of the four binocular camera modules; as a further preferred embodiment of this application, for the weighted least squares method, the initial weights can be allocated according to the camera calibration accuracy.

[0024] The above-mentioned preferred scheme in this application can improve adaptability to complex environments. During implementation, for extreme working conditions such as strong winds at high altitudes, tower crane vibrations, and welding fumes, the dual anti-interference mechanism of front-end fuzzy frame recognition and multi-view data fusion can improve the effective data acquisition rate. The three-dimensional coordinate calculation is brought down to the camera end, and combined with the high-priority transmission mechanism, the delay in the generation of control commands is compressed, and dynamic correction is performed at the moment of the anchor beam 01 swing, reducing the number of high-altitude docking adjustments.

[0025] After hoisting is completed, the anchor beam 01 and steel bracket 02 can be pre-fixed with high-strength bolts. The binocular camera module can be removed, and the installation accuracy can be checked with a total station. After passing the inspection, the final welding or bolt tightening can be carried out.

[0026] As a preferred embodiment of the above, when the hoisting control system detects that the rate of change of local three-dimensional coordinate data exceeds the set threshold, the set threshold here is specifically set according to the occlusion of the binocular camera module in the actual working conditions, such as when strong winds, dust, fog or mechanical interference occlusion occurs, and high-quality images at the corresponding time are retrieved. The pre-trained model is reused to identify the type of occlusion. If it is determined to be dust or fog occlusion, the weight of the current binocular camera module is reduced and the weight of other binocular camera modules is increased. If it is determined to be mechanical interference occlusion, the corresponding positioning target is marked as invalid.

[0027] In the implementation of this preferred solution, the judgment of dust and fog obstruction can be triggered by the identification of diffuse noise features, which can then trigger dynamic weight adjustment. For example, the weight of the current binocular camera module can be reduced to 0.3, while the weights of others can be increased to 1.2, thereby maintaining the continuous operation of the system. For mechanical interference, the judgment can be made based on the hard edge obstruction features, and corresponding measures can be taken. As a further optimization of this application, specific measures include, but are not limited to: interrupting the current data input of the binocular camera module to prevent erroneous coordinates from contaminating the fusion system; sending an alarm to the tower crane control panel and simultaneously activating the hydraulic brake preparation; and realizing backup positioning switching, such as activating the Beidou positioning module to replace the failed positioning target, etc.

[0028] This optimized solution reuses the same pre-trained model to perform fuzzy frame filtering and occlusion analysis, reducing the memory usage of edge devices and enabling a single device to simultaneously run positioning and diagnostic tasks. High-altitude strong winds cause frequent dynamic occlusions such as dust and welding fumes. Switching between multiple dedicated models would result in significant delays and would fail to match the swing speed of anchor beam 01. This solution reuses the same model to complete fuzzy frame filtering and occlusion type determination, reducing model switching latency and improving anomaly response speed, thus gaining a critical time window for tower crane correction.

[0029] As a preferred embodiment of the above, such as Figure 5 As shown, the pre-trained model is built using a multi-stage transfer learning framework, and the construction process includes: In the basic training phase, the network model is trained using public datasets. The network model can at least identify motion blur, defocus blur and environmental occlusion features. The use of public data can solve the problem of scarce annotation data in the field of steel structures. In the domain adaptation phase, the network model parameters are fine-tuned based on historical images of steel structure engineering to adapt to the construction scenario of steel anchor beam 01. This phase specifically strengthens scene feature extraction, which can effectively improve the accuracy of occlusion recognition. During the edge deployment phase, the fine-tuned network model is quantized into a set format and embedded into the chip of the binocular camera module. In the implementation process, INT8 quantization technology can be used to compress the model to more than half of its original volume. After the camera NPU chip is implanted, it successfully matches the power supply and heat dissipation limitations of high-altitude equipment.

[0030] For this preferred solution, model fine-tuning can be achieved through the following techniques to achieve accurate scenario adaptation, including: Interference features are enhanced in depth, including but not limited to injecting high-frequency vibration fuzzy datasets for high-altitude strong wind interference, such as the vibration trajectory of tower cranes from 5 to 12 Hz, and constructing a metal glare sample library for steel structure-specific interference such as metal reflection, such as the 01 reflection mode of anchor beams under different sunlight angles.

[0031] Data augmentation strategies are optimized, including but not limited to implementing dynamic occlusion simulation, specifically by manually adding moving occluders such as slings and tools to historical images, and implementing environmental degradation generation, specifically by using GAN networks to synthesize extreme weather images such as salt spray corrosion and rain fog.

[0032] In practical implementation, the network model can specifically adopt residual networks. The residual structure can extract general features, such as motion blur and defocus blur, through the bottom convolutional layers, while the high-level network adapts to scene features and naturally supports multi-stage transfer learning. The lightweight fine-tuning capability of the residual structure can support freezing the bottom parameters, retaining the pre-trained generalization ability, and only fine-tuning the top fully connected layers and classifiers, significantly reducing the amount of computation, which is suitable for the data-scarce scenario in the steel structure field.

[0033] As a preferred embodiment of the above, the imaging proportion of the positioning target in the field of view of the binocular camera module is higher than 15%. This preferred solution can ensure the accuracy of feature point extraction, solve the problem of blurred recognition caused by the small size of the positioning target when shooting from high altitude, reduce the pixel coordinate error of feature points, and support high-precision positioning accuracy; in strong wind environment, it can prevent measurement failure caused by insufficient imaging proportion due to shaking of the positioning target.

[0034] As a preferred embodiment of the above, the positioning target is a barcode target. The barcode uses a black and white stripe structure of equal width, with a stripe width of 1 to 2 millimeters. The target substrate is made of metal and has a reflective coating on its surface. In this preferred embodiment, the black and white stripe target of equal width combined with the reflective metal substrate can significantly improve the recognition robustness under complex lighting conditions; the metal material adapts to temperature difference deformation in steel structure environments, and the reflective coating enhances the imaging contrast in low-light environments; the sharp edges of the barcode target improve the efficiency of coordinate calculation, meeting the requirements of real-time processing.

[0035] As a preferred embodiment of the above, the binocular camera module has a resolution higher than 5 megapixels, a frame rate higher than 30fps, and a lens focal length of 8 to 12 mm. In this preferred solution, the combination of 5 megapixel resolution and 8 to 12 mm focal length ensures clear imaging of the target stripes within a distance of 50 meters, eliminating the loss of details in long-distance shooting; the frame rate of 30fps or higher accurately captures the swing trajectory of the anchor beam 01, which can reduce the lag time of hoisting adjustment during implementation and avoid the cumulative positioning deviation caused by insufficient frame rate in traditional solutions.

[0036] As a preferred embodiment of the above, the digital installation method for steel anchor beams further includes a system calibration step before construction, including: C1: Set up a calibration platform near the steel bracket 02 and place a standard calibration plate of known size thereon; C2: Control the hoisting equipment to move the anchor beam 01 to different preset positions; C3: Acquire images of the standard calibration board using a binocular camera module, calculate the measurement coordinates of feature points on the standard calibration board, and compare them with the standard coordinates; C4: Based on the comparison results, correct the internal and external parameters of the binocular camera module to ultimately control system error.

[0037] In this preferred solution, the camera's intrinsic and extrinsic parameters are corrected before hoisting using a standard calibration plate of known size. The intrinsic parameters include lens distortion and focal length error, while the extrinsic parameters include spatial pose deviation. This can specifically compress the system's measurement error and avoid the cumulative error caused by the lack of on-site calibration in traditional methods.

[0038] As a preferred embodiment of the above, the image acquisition frequency of the binocular camera module during system calibration is 0.4 to 0.6 Hz, preferably 0.5 Hz; the image acquisition frequency during construction is 1.8 to 2.2 Hz, preferably 2 Hz; the differentiated frequency design can simultaneously achieve different levels of response speed on resource-constrained edge devices.

[0039] Example 2 A digital installation and construction system for steel anchor beams, comprising: Several positioning targets are arranged on the steel bracket 02, with the designed installation position of the anchor beam 01 relative to the steel bracket 02 as the reference. Several binocular camera modules are installed on the anchor beam 01, and the shooting range of any binocular camera module covers at least two positioning targets. The 3D modeling module constructs 3D models of steel bracket 02 and anchor beam 01, clarifies the fixed spatial relationship between the positioning target, binocular camera module and anchor beam 01 in the 3D model, and calibrates and associates the 3D model with the on-site construction coordinate system. The Beidou positioning module includes several Beidou measurement reference stations deployed around the installation area of ​​the steel bracket 02. These stations are used to measure the coordinates of the positioning target in real time and output the position adjustment basis for the steel bracket 02 so that the installation deviation of the steel bracket 02 relative to the on-site construction coordinate system is within a first set range. The dynamic control module activates each binocular camera module to collect images of the positioning target in real time, generating a basis for adjusting the position of the anchor beam 01, so that the installation deviation of the anchor beam 01 relative to the steel bracket 02 is within the second set range.

[0040] The technical effects achieved in this embodiment are as described in Embodiment 1 above, and will not be repeated here.

[0041] Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A digital installation construction method for steel anchor beams, characterized in that, include: Based on the designed installation position of the anchor beam relative to the steel bracket, several positioning targets are set on the steel bracket. Several binocular camera modules are installed on the anchor beam, such that the shooting range of any one of the binocular camera modules covers at least two of the positioning targets. Construct a three-dimensional model of the steel bracket and anchor beam, clarify the fixed spatial relationship of the positioning target, binocular camera module and anchor beam in the three-dimensional model, and calibrate and associate the three-dimensional model with the on-site construction coordinate system; Several Beidou measurement reference stations are set up around the steel bracket installation area. The Beidou measurement reference stations measure the coordinates of the positioning target in real time to provide a basis for the real-time position adjustment of the steel bracket during the installation process, so that the installation deviation of the steel bracket relative to the on-site construction coordinate system is within a first set range. Each of the binocular camera modules is activated. The binocular camera modules acquire images of the positioning target in real time, providing a basis for real-time position adjustment of the anchor beam during installation, so that the installation deviation of the anchor beam relative to the steel bracket is within a second set range.

2. The digital installation construction method for steel anchor beams according to claim 1, characterized in that, The images captured by the binocular camera module are transmitted wirelessly to the hoisting control system in real time, including: High-quality images are obtained by identifying and discarding blurry frames using a pre-trained model at the binocular camera module. Feature points of the positioning target are extracted from the high-quality image, and the local three-dimensional coordinate data of the anchor beam are calculated based on the principle of binocular vision. Assign a high-priority transmission marker to the local three-dimensional coordinate data, and assign a low-priority transmission marker to the high-quality image; The hoisting control system prioritizes receiving and processing the local three-dimensional coordinate data, and fuses the local three-dimensional coordinate data of each perspective corresponding to several binocular camera modules. The fusion is performed using the weighted least squares method, and the anchor beam installation deviation is calculated.

3. The digital installation construction method for steel anchor beams according to claim 2, characterized in that, When the hoisting control system detects that the rate of change of the local three-dimensional coordinate data exceeds a set threshold, it retrieves the high-quality image at the corresponding moment. The pre-trained model is reused to identify the occlusion type. If it is determined to be dust or fog occlusion, the weight of the current binocular camera module is reduced, and the weight of other binocular camera modules is increased. If the problem is determined to be mechanical interference or obstruction, the corresponding positioning target is marked as invalid.

4. The digital installation construction method for steel anchor beams according to claim 3, characterized in that, The pre-trained model is constructed using a multi-stage transfer learning framework, and the construction process includes: In the basic training phase, a network model is trained using a public dataset. The network model must at least recognize motion blur, defocus blur, and environmental occlusion features. In the domain adaptation phase, the parameters of the network model are fine-tuned based on historical images of steel structure engineering to adapt to the construction scenario of steel anchor beams. During the edge deployment phase, the fine-tuned network model is quantized into a set format and embedded into the chip of the binocular camera module.

5. The digital installation construction method for steel anchor beams according to claim 1, characterized in that, The positioning target accounts for more than 15% of the image in the field of view of the binocular camera module.

6. The digital installation construction method for steel anchor beams according to claim 1, characterized in that, The positioning target is a barcode target. The barcode adopts a black and white stripe structure with equal width. The stripe width is 1 to 2 millimeters. The target base is made of metal and the surface is covered with a reflective coating.

7. The digital installation construction method for steel anchor beams according to claim 1, characterized in that, The binocular camera module has a resolution of over 5 million pixels, a frame rate of over 30fps, and a lens focal length of 8 to 12 millimeters.

8. The digital installation construction method for steel anchor beams according to claim 1, characterized in that, It also includes a system calibration procedure before construction, including: A calibration platform is set up near the steel bracket, and a standard calibration plate of known size is placed there. The hoisting equipment is controlled to move the anchor beam to different preset positions; The binocular camera module acquires images of the standard calibration board, calculates the measurement coordinates of the feature points on the standard calibration board, and compares them with standard coordinates. The intrinsic and extrinsic parameters of the binocular camera module were corrected based on the comparison results.

9. The digital installation construction method for steel anchor beams according to claim 8, characterized in that, The image acquisition frequency of the binocular camera module during system calibration is 0.4 to 0.6 Hz, and the image acquisition frequency during construction is 1.8 to 2.2 Hz.

10. A digital installation and construction system for steel anchor beams, characterized in that, include: Several positioning targets are arranged on the steel bracket, with the designed installation position of the anchor beam relative to the steel bracket as the reference. Several binocular camera modules are installed on the anchor beam, and the shooting range of any one of the binocular camera modules covers at least two of the positioning targets. The 3D modeling module constructs a 3D model of the steel bracket and anchor beam, clarifies the fixed spatial relationship between the positioning target, the binocular camera module and the anchor beam in the 3D model, and calibrates and associates the 3D model with the on-site construction coordinate system. The Beidou positioning module includes several Beidou measurement reference stations deployed around the steel bracket installation area, used to measure the coordinates of the positioning target in real time and output the position adjustment basis of the steel bracket so that the installation deviation of the steel bracket relative to the on-site construction coordinate system is within a first set range. The dynamic control module activates each of the binocular camera modules to acquire images of the positioning target in real time, and generates a basis for adjusting the position of the anchor beam so that the installation deviation of the anchor beam relative to the steel bracket is within a second set range.

Images