A remote acceptance data management system and method based on a visualized model

Abstract

This invention relates to the field of remote acceptance, specifically to a remote acceptance data management system and method based on a visualization model. The system includes: a panoramic modeling module, an interaction optimization module, a data processing module, a remote guidance module, and a rectification and acceptance module. The panoramic modeling module generates a digital panoramic model; the interaction optimization module manages the panoramic model by partition; the data processing module converts unstructured datasets into structured data; the remote guidance module binds structured data to locations; and the rectification and acceptance module generates acceptance files. This invention enables effective management and control of construction sites, improves the targeting of guidance, enhances traceability efficiency, standardizes construction behavior, improves the efficiency of problem rectification, enhances the reusability of experience, and improves acceptance efficiency and accuracy. It has broad industry applicability, while simultaneously improving project quality, reducing management costs, and strengthening collaborative efficiency.

Description

Technical Field

[0001] This invention relates to the field of remote acceptance, specifically to a remote acceptance data management system and method based on a visualization model. Background Technology

[0002] Remote acceptance is a technology for remotely verifying projects, products, and equipment. It uses video, images, and sensor data to enable the acceptance party to complete project assessments off-site, replacing traditional on-site acceptance methods. Current remote guidance and acceptance models in the construction engineering field typically rely on video calls, text instructions, or distributed data recording. While these methods achieve basic cross-space communication, they suffer from deficiencies in accuracy, relevance, visualization, and standardization.

[0003] Traditional remote guidance suffers from vague regional positioning and insufficient targeting, failing to accurately pinpoint specific work locations within the area. Furthermore, it requires manual data integration across platforms, which can easily lead to data omissions and mismatches. It also lacks the ability to reproduce processes effectively and the functionality to mark problems, push rectification tasks, and track rectification progress in a visual medium. In complex projects such as zero-carbon factories and photovoltaic power stations, where the work area is vast and the process requirements are high, the accuracy of acceptance results cannot be guaranteed when relying heavily on manual verification of scattered rectification data.

[0004] In addition, the existing remote acceptance procedures lack standardized remote archiving capabilities, fail to effectively utilize data based on factors such as construction area, work period, and work environment, have poor identification capabilities for construction data, cannot reproduce the on-site environment and work scenario, and are difficult to form a complete guidance trajectory. Summary of the Invention

[0005] The purpose of this invention is to provide a remote acceptance data management system and method based on a visualization model to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, the present invention provides the following technical solution: a remote acceptance data management system based on a visualization model, comprising: a panoramic modeling module, an interaction optimization module, a data processing module, a remote guidance module, and a rectification and acceptance module;

[0007] The panoramic modeling module is used to capture images of the construction site using drones and ground panoramic cameras, acquire image data from various perspectives, and after stitching and optimization, store the images in a lightweight format. After three-dimensional reconstruction processing, an interactive digital panoramic model of the construction site is generated.

[0008] The interactive optimization module is used to manage the panoramic model by functional area and work point, locate the coordinates of the points, list all primary areas and secondary points in the panoramic model, set virtual coordinate anchor points for rapid target location, and verify the construction accuracy.

[0009] The data processing module is used to collect images of the construction process through smart safety helmets, fixed monitoring or drones, extract the motion characteristics of the construction points, establish a behavior dataset, standardize the unstructured dataset, convert it into structured data, and output a remote guidance dataset in CSV format.

[0010] The remote guidance module is used to send the remote guidance dataset to the corresponding responsible entity, continuously collect and upload videos of construction sites, extract key frames from the videos, generate preview thumbnails, and use the site number as a unique primary key to strongly bind structured data with the site and write it into the site attribute library of the panoramic model.

[0011] The rectification and acceptance module is used to initiate an acceptance application after the completion of regional construction, associate data of each point in the region, locate problem points in the panoramic model, mark the problem and add a problem description, upload photos of the problem site, automatically associate the construction information of the point, push rectification work orders to the responsible personnel, and generate an acceptance file after all points are judged to be qualified according to the acceptance standards.

[0012] Furthermore, the panoramic modeling module includes: an image acquisition unit, an image processing unit, and a 3D reconstruction unit;

[0013] The image acquisition unit is used to guide the drone to plan the photography route based on the initial building model obtained by the drone, determine the shooting viewpoint position according to the flight altitude, overlap and drone field of view, plan the photography path according to the ordered viewpoint and building layout, connect the camera group, and complete the high-altitude panoramic acquisition of the construction area according to the preset route.

[0014] The image processing unit is used to uniformly segment the orthophoto of the UAV, expand the samples using a generative adversarial network, extract features from the normal illumination domain and the low illumination domain, transfer the features from the source domain to the target domain, and establish a feature point KD tree index using unlabeled data in the target domain.

[0015] The 3D reconstruction unit is used to search for matching points of the target structure at the work site using a bidirectional approximate nearest neighbor algorithm, match the orthophoto dataset, fuse the matched images, and output an interactive 3D model through model simplification, texture compression, and LOD reconstruction.

[0016] Furthermore, the interaction optimization module includes: a partition management unit and a point cloud processing unit;

[0017] The partition management unit is used to extract feature points in each region based on the target shape descriptor, register the preprocessed target points using an iterative feature point automatic registration algorithm, construct a triangulation model for the ground point data, and logically partition the panoramic model according to functional zones and secondary operation points.

[0018] The point cloud processing unit is used to identify construction targets based on normal vectors and curvature features, register scene point clouds and construction target point clouds through an iterative nearest neighbor algorithm with normal vector constraints, acquire square wave images at different object distances, calculate optical point spread functions, determine the distance between two points in the panoramic model, and verify the construction accuracy of components.

[0019] Furthermore, the data processing module includes: a behavioral data unit and a structured data unit;

[0020] The behavior data unit is used to adaptively adjust the brightness of the original image by iteratively training the generator and discriminator, generate monitoring images, construct a video frame encoder, guide the perception of spatial information through pixel segmentation, capture the spatiotemporal dependencies of each frame in the image, learn global spatiotemporal feature representation, and learn inter-frame dynamic change information with a differential frame encoder.

[0021] The structured unit is used to extract the position of moving joints using the HRNet model, collect the operating specifications, technical instructions and process requirements issued by the management personnel, build a dynamic behavior detection model, and generate a remote guidance dataset. The remote guidance dataset contains text specifications, video links and key frame screenshots of the construction process.

[0022] Furthermore, the remote guidance module includes: a video restoration unit and a guidance binding unit;

[0023] The video restoration unit is used to automatically trigger a reminder when the video lacks standard references or is blurry. It estimates the blur parameters based on the initial image spectrum, compares the artificially blurred image with the actual image for similarity, selects the blur kernel of the image with the highest similarity, and iteratively updates the blur kernel and the high-frequency part of the blurred image alternately to perform video restoration processing.

[0024] The guidance binding unit is used to generate a preview thumbnail for each video segment, and mark the acquisition time and acquisition terminal type. It generates a playback link, uses the point number as a unique primary key, binds the guidance data, video playback link, keyframe preview and accuracy verification report to the corresponding point in the 3D model, records the binding time, binding person and data source, and generates a binding log.

[0025] Furthermore, the rectification and acceptance module includes: a problem marking unit, a rectification tracking unit, and an acceptance archiving unit;

[0026] The problem labeling unit is used to continuously collect target point images by tracing back point by point through the panoramic model, and transmit them to the host. The host generates a target point cloud, classifies the point cloud according to the curvature value of each point, and determines the model boundary points according to the directional Hausdorff distance.

[0027] The rectification tracking unit is used to construct a local coordinate system based on the query point and the set of neighboring points using the eigenvalue decomposition method, calculate the spatial transformation matrix, calculate the deviation values ​​of key dimensions, flatness and verticality indicators, compare them with the acceptance standard values, judge the construction effect, and for points that fail the inspection, locate the problem location in the panoramic model and generate guidance cards.

[0028] The acceptance and archiving unit is used to obtain the video of the rectification after the rectification is completed, verify whether the rectification meets the standards, close the work order if it meets the standards, and re-rectify if it does not meet the standards, and generate a rectification progress dashboard. Based on the guidance records, process videos and rectification status, a digital acceptance file is generated.

[0029] A remote acceptance data management method based on a visualization model includes the following steps:

[0030] Step S1. Capture images of the construction site, obtain image data from various perspectives, and after stitching and optimization, store the images in a lightweight format. After 3D reconstruction processing, generate an interactive digital panoramic model of the construction site.

[0031] Step S2. Divide the panoramic model into functional zones and work points for zone management, locate the coordinates of the points, set virtual coordinate anchor points for rapid target positioning, and verify the construction accuracy.

[0032] Step S3. Collect images of the construction process, extract the motion features of the construction points, establish a behavior dataset, standardize the unstructured dataset, convert it into structured data, and output a remote guidance dataset.

[0033] Step S4. Send the remote guidance dataset, collect and upload videos of the construction points, generate preview thumbnails, use the point number as the unique primary key, strongly bind the structured data with the point, and write it into the point attribute library of the panoramic model.

[0034] Step S5. Link the data of each point, locate the problem point, mark the problem and add a problem description, link the construction information of the point, push the rectification work order, and after the regional construction is completed, initiate the acceptance application. After all points are judged to be qualified according to the acceptance standards, the acceptance file is generated.

[0035] Furthermore, step S1 includes:

[0036] Step S11. Guide the drone to plan the photography route based on the initial building model obtained. Determine the shooting viewpoint position according to the flight altitude, overlap and drone field of view. Plan the photography path according to the ordered viewpoint and building layout. Connect the camera group and complete the high-altitude panoramic acquisition of the construction area according to the preset route.

[0037] Step S12. The UAV orthophoto is uniformly segmented, and the sample is expanded using a generative adversarial network. Features in the normal illumination domain and low illumination domain are extracted, and the features in the source domain are transferred to the target domain. A feature point KD tree index is established using unlabeled data in the target domain. Matching points of the target structure at the operation point are searched using a bidirectional approximate nearest neighbor algorithm. The orthophoto dataset is matched, and the matched images are fused. Through model simplification, texture compression and LOD reconstruction, an interactive 3D model is output.

[0038] Furthermore, step S2 includes:

[0039] Step S21. Extract feature points in each region based on the target shape descriptor, use the automatic registration algorithm of iterative feature points to register the preprocessed target points, construct a triangulation model for the ground point data, and logically partition the panoramic model according to functional zones and secondary operation points.

[0040] Step S22. Identify the construction target based on the normal vector and curvature features, register the scene point cloud and the construction target point cloud using the iterative nearest neighbor algorithm with normal vector constraints, collect square wave images at different object distances, calculate the optical point spread function, determine the distance between two points in the panoramic model, and verify the construction accuracy of the components.

[0041] Furthermore, step S3 includes:

[0042] Step S31. By iteratively training the generator and discriminator, the brightness of the original image is adaptively adjusted to generate a monitoring image, a video frame encoder is constructed, and spatial information is guided by pixel segmentation to capture the spatiotemporal dependencies of each frame in the image. The global spatiotemporal feature representation is learned, and the inter-frame dynamic change information is learned by the differential frame encoder.

[0043] Step S32. Use the HRNet model to extract the positions of moving joints, collect the operating specifications, technical instructions and process requirements issued by the management personnel, construct a dynamic behavior detection model, and generate a remote guidance dataset. The remote guidance dataset contains text specifications, video links and key frame screenshots of the construction process.

[0044] Furthermore, step S4 includes:

[0045] Step S41. When the video lacks standardized basis or is blurry, an alert is automatically triggered. The blur parameters are estimated based on the initial image spectrum. The artificially blurred image is compared with the actual image in terms of similarity. The blur kernel of the image with the highest similarity is selected. The blur kernel and the high-frequency part of the blurred image are alternately and iteratively updated to restore the video.

[0046] Step S42. Generate a preview thumbnail for each video segment and label it with the acquisition time and acquisition terminal type. Generate a playback link, using the point number as the unique primary key. Bind the guidance data, video playback link, keyframe preview, and accuracy verification report to the corresponding point in the 3D model, record the binding time, binding person, and data source, and generate a binding log.

[0047] Furthermore, step S5 includes:

[0048] Step S51. Using the panoramic model, backtrack point by point, continuously collect images of target points, transmit them to the host, the host generates target point cloud, classifies the point cloud according to the curvature value of each point, determines the model boundary points according to the directional Hausdorff distance, and constructs a local coordinate system using the eigenvalue decomposition method based on the query point and the set of neighboring points, and calculates the spatial transformation matrix.

[0049] Step S52. Calculate the deviation values ​​of key dimensions, flatness, and verticality indicators, compare them with the acceptance standard values, judge the construction effect, locate the problem location in the panoramic model for unqualified points, generate guidance cards, obtain the rectification video after rectification, verify whether the rectification meets the standards, close the work order if it meets the standards, and re-rectify if it does not meet the standards, and generate a rectification progress board. Based on the guidance records, process videos and rectification status, generate a digital acceptance file.

[0050] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0051] 1. Based on the topography and building layout of the construction site, this invention acquires multi-view image data through drone aerial photography and ground panoramic cameras, generates an interactive digital panoramic model, and manages the panoramic model by partition, accurately binding all guidance data to specific work points in the panoramic model, enabling effective management and control of the construction site, improving the targeting of guidance, increasing traceability efficiency, and avoiding traceability failures caused by manual recording.

[0052] 2. This invention comprehensively collects remote guidance data and video images during the construction process, standardizes the collected unstructured data, automatically generates standardized guidance cards, and strongly binds them to corresponding locations. It eliminates the need for cross-platform data queries, solves the problems of limited perspective and incomplete scenes in traditional videos, realizes real-time monitoring and dynamic guidance of the construction process, standardizes construction behavior, improves the efficiency of problem rectification, enhances the reusability of experience, and controls the range of quality fluctuations in different projects.

[0053] 3. This invention accurately marks problem locations in a panoramic model, records problem details, pushes problem work orders to the corresponding responsible parties, tracks rectification progress in real time, verifies rectification effects, forms a rectification closed loop, and thus conducts standardized acceptance, improving acceptance efficiency and accuracy. It has wide industry applicability, improves project quality, reduces management costs, and enhances collaborative efficiency, demonstrating significant practical value. Attached Figure Description

[0054] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0055] Figure 1 This is a schematic diagram of the structure of a remote acceptance data management system based on a visualization model according to the present invention;

[0056] Figure 2 This is a schematic diagram illustrating the steps of a remote acceptance data management method based on a visualization model according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Please see Figures 1 to 2 The present invention provides a technical solution: a remote acceptance data management system based on a visualization model, comprising: a panoramic modeling module, an interaction optimization module, a data processing module, a remote guidance module, and a rectification and acceptance module;

[0059] The panoramic modeling module is used to capture images of the construction site using drones and ground panoramic cameras, acquire image data from various perspectives, and after stitching and optimization, store the images in a lightweight format. After three-dimensional reconstruction processing, an interactive digital panoramic model of the construction site is generated.

[0060] The panoramic modeling module includes: an image acquisition unit, an image processing unit, and a three-dimensional reconstruction unit;

[0061] The image acquisition unit is used to guide the drone to plan the photography route based on the initial building model obtained by the drone, determine the shooting viewpoint position according to the flight altitude, overlap and drone field of view, plan the photography path according to the ordered viewpoint and building layout, connect the camera group, and complete the high-altitude panoramic acquisition of the construction area according to the preset route.

[0062] The image processing unit is used to uniformly segment the orthophoto of the UAV, expand the samples using a generative adversarial network, extract features from the normal illumination domain and the low illumination domain, transfer the features from the source domain to the target domain, and establish a feature point KD tree index using unlabeled data in the target domain.

[0063] The 3D reconstruction unit is used to search for matching points of the target structure at the work site using a bidirectional approximate nearest neighbor algorithm, match the orthophoto dataset, fuse the matched images, and output an interactive 3D model through model simplification, texture compression, and LOD reconstruction.

[0064] The interactive optimization module is used to manage the panoramic model by functional area and work point, locate the coordinates of the points, list all primary areas and secondary points in the panoramic model, set virtual coordinate anchor points for rapid target location, and verify the construction accuracy.

[0065] The interaction optimization module includes: a partition management unit and a point cloud processing unit;

[0066] The partition management unit is used to extract feature points in each region based on the target shape descriptor, register the preprocessed target points using an iterative feature point automatic registration algorithm, construct a triangulation model for the ground point data, and logically partition the panoramic model according to functional zones and secondary operation points.

[0067] The point cloud processing unit is used to identify construction targets based on normal vectors and curvature features, register scene point clouds and construction target point clouds through an iterative nearest neighbor algorithm with normal vector constraints, acquire square wave images at different object distances, calculate optical point spread functions, determine the distance between two points in the panoramic model, and verify the construction accuracy of components.

[0068] The data processing module is used to collect images of the construction process through smart safety helmets, fixed monitoring or drones, extract the motion characteristics of the construction points, establish a behavior dataset, standardize the unstructured dataset, convert it into structured data, and output a remote guidance dataset in CSV format.

[0069] The data processing module includes: a behavioral data unit and a structured data unit;

[0070] The behavior data unit is used to adaptively adjust the brightness of the original image by iteratively training the generator and discriminator, generate monitoring images, construct a video frame encoder, guide the perception of spatial information through pixel segmentation, capture the spatiotemporal dependencies of each frame in the image, learn global spatiotemporal feature representation, and learn inter-frame dynamic change information with a differential frame encoder.

[0071] The structured unit is used to extract the position of moving joints using the HRNet model, collect the operating specifications, technical instructions and process requirements issued by the management personnel, build a dynamic behavior detection model, and generate a remote guidance dataset. The remote guidance dataset contains text specifications, video links and key frame screenshots of the construction process.

[0072] The remote guidance module is used to send the remote guidance dataset to the corresponding responsible entity, continuously collect and upload videos of construction sites, extract key frames from the videos, generate preview thumbnails, and use the site number as a unique primary key to strongly bind structured data with the site and write it into the site attribute library of the panoramic model.

[0073] The remote guidance module includes: a video restoration unit and a guidance binding unit;

[0074] The video restoration unit is used to automatically trigger a reminder when the video lacks standard references or is blurry. It estimates the blur parameters based on the initial image spectrum, compares the artificially blurred image with the actual image for similarity, selects the blur kernel of the image with the highest similarity, and iteratively updates the blur kernel and the high-frequency part of the blurred image alternately to perform video restoration processing.

[0075] The guidance binding unit is used to generate a preview thumbnail for each video segment, and mark the acquisition time and acquisition terminal type. It generates a playback link, uses the point number as a unique primary key, binds the guidance data, video playback link, keyframe preview and accuracy verification report to the corresponding point in the 3D model, records the binding time, binding person and data source, and generates a binding log.

[0076] The rectification and acceptance module is used to initiate an acceptance application after the completion of regional construction, associate data of each point in the region, locate problem points in the panoramic model, mark the problem and add a problem description, upload photos of the problem site, automatically associate the construction information of the point, push rectification work orders to the responsible personnel, and generate an acceptance file after all points are judged to be qualified according to the acceptance standards.

[0077] The rectification and acceptance module includes: a problem marking unit, a rectification tracking unit, and an acceptance archiving unit;

[0078] The problem labeling unit is used to continuously collect target point images by tracing back point by point through the panoramic model, and transmit them to the host. The host generates a target point cloud, classifies the point cloud according to the curvature value of each point, and determines the model boundary points according to the directional Hausdorff distance.

[0079] The rectification tracking unit is used to construct a local coordinate system based on the query point and the set of neighboring points using the eigenvalue decomposition method, calculate the spatial transformation matrix, calculate the deviation values ​​of key dimensions, flatness and verticality indicators, compare them with the acceptance standard values, judge the construction effect, and for points that fail the inspection, locate the problem location in the panoramic model and generate guidance cards.

[0080] The acceptance and archiving unit is used to obtain the video of the rectification after the rectification is completed, verify whether the rectification meets the standards, close the work order if it meets the standards, and re-rectify if it does not meet the standards, and generate a rectification progress dashboard. Based on the guidance records, process videos and rectification status, a digital acceptance file is generated.

[0081] A remote acceptance data management method based on a visualization model includes the following steps:

[0082] Step S1. Capture images of the construction site, obtain image data from various perspectives, and after stitching and optimization, store the images in a lightweight format. After 3D reconstruction processing, generate an interactive digital panoramic model of the construction site.

[0083] Step S1 includes:

[0084] Step S11. Guide the drone to plan the photography route based on the initial building model obtained. Determine the shooting viewpoint position according to the flight altitude, overlap and drone field of view. Plan the photography path according to the ordered viewpoint and building layout. Connect the camera group and complete the high-altitude panoramic acquisition of the construction area according to the preset route.

[0085] Step S12. The UAV orthophoto is uniformly segmented, and the sample is expanded using a generative adversarial network. Features in the normal illumination domain and low illumination domain are extracted, and the features in the source domain are transferred to the target domain. A feature point KD tree index is established using unlabeled data in the target domain. Matching points of the target structure at the operation point are searched using a bidirectional approximate nearest neighbor algorithm. The orthophoto dataset is matched, and the matched images are fused. Through model simplification, texture compression and LOD reconstruction, an interactive 3D model is output.

[0086] Step S2. Divide the panoramic model into functional zones and work points for zone management, locate the coordinates of the points, set virtual coordinate anchor points for rapid target positioning, and verify the construction accuracy.

[0087] Step S2 includes:

[0088] Step S21. Extract feature points in each region based on the target shape descriptor, use the automatic registration algorithm of iterative feature points to register the preprocessed target points, construct a triangulation model for the ground point data, and logically partition the panoramic model according to functional zones and secondary operation points.

[0089] Step S22. Identify the construction target based on the normal vector and curvature features, register the scene point cloud and the construction target point cloud using the iterative nearest neighbor algorithm with normal vector constraints, collect square wave images at different object distances, calculate the optical point spread function, determine the distance between two points in the panoramic model, and verify the construction accuracy of the components.

[0090] Step S3. Collect images of the construction process, extract the motion features of the construction points, establish a behavior dataset, standardize the unstructured dataset, convert it into structured data, and output a remote guidance dataset.

[0091] Step S3 includes:

[0092] Step S31. By iteratively training the generator and discriminator, the brightness of the original image is adaptively adjusted to generate a monitoring image, a video frame encoder is constructed, and spatial information is guided by pixel segmentation to capture the spatiotemporal dependencies of each frame in the image. The global spatiotemporal feature representation is learned, and the inter-frame dynamic change information is learned by the differential frame encoder.

[0093] Step S32. Use the HRNet model to extract the positions of moving joints, collect the operating specifications, technical instructions and process requirements issued by the management personnel, construct a dynamic behavior detection model, and generate a remote guidance dataset. The remote guidance dataset contains text specifications, video links and key frame screenshots of the construction process.

[0094] Step S4. Send the remote guidance dataset, collect and upload videos of the construction points, generate preview thumbnails, use the point number as the unique primary key, strongly bind the structured data with the point, and write it into the point attribute library of the panoramic model.

[0095] Step S4 includes:

[0096] Step S41. When the video lacks standardized basis or is blurry, an alert is automatically triggered. The blur parameters are estimated based on the initial image spectrum. The artificially blurred image is compared with the actual image in terms of similarity. The blur kernel of the image with the highest similarity is selected. The blur kernel and the high-frequency part of the blurred image are alternately and iteratively updated to restore the video.

[0097] Step S42. Generate a preview thumbnail for each video segment and label it with the acquisition time and acquisition terminal type. Generate a playback link, using the point number as the unique primary key. Bind the guidance data, video playback link, keyframe preview, and accuracy verification report to the corresponding point in the 3D model, record the binding time, binding person, and data source, and generate a binding log.

[0098] Step S5. Link the data of each point, locate the problem point, mark the problem and add a problem description, link the construction information of the point, push the rectification work order, and after the regional construction is completed, initiate the acceptance application. After all points are judged to be qualified according to the acceptance standards, the acceptance file is generated.

[0099] Step S5 includes:

[0100] Step S51. Using the panoramic model, backtrack point by point, continuously collect images of target points, transmit them to the host, the host generates target point cloud, classifies the point cloud according to the curvature value of each point, determines the model boundary points according to the directional Hausdorff distance, and constructs a local coordinate system using the eigenvalue decomposition method based on the query point and the set of neighboring points, and calculates the spatial transformation matrix.

[0101] Step S52. Calculate the deviation values ​​of key dimensions, flatness, and verticality indicators, compare them with the acceptance standard values, judge the construction effect, locate the problem location in the panoramic model for unqualified points, generate guidance cards, obtain the rectification video after rectification, verify whether the rectification meets the standards, close the work order if it meets the standards, and re-rectify if it does not meet the standards, and generate a rectification progress board. Based on the guidance records, process videos and rectification status, generate a digital acceptance file.

[0102] Example: Taking the installation of steel column 3B-GZ-15 as an example, the panoramic modeling module collects images of the area north of column #3 and constructs a panoramic model. The area division and numbering module assigns the number 3B-GZ-15 to the 15th steel column in this area. The guidance card automatic generation module generates guidance cards containing "anchor bolt type M42, verticality deviation ≤3mm / m" based on the steel column installation specifications. The data-area binding module binds the guidance card, the steel column installation technical requirements, and the 3B-GZ-15 point.

[0103] On the day of construction: Workers wearing smart safety helmets filmed first-person perspective videos of the hoisting of the 3B-GZ-15 steel column, which were automatically uploaded to the platform. Managers issued real-time guidance to the workers through the remote visualization platform, stating that the hoisting speed should be ≤0.5m / s. The data was automatically linked to the 3B-GZ-15 point, and the data structuring module organized the video and guidance data and updated the attribute library of that point.

[0104] Problem rectification during construction: When the supervisor inspected the panoramic model, he found that the verticality deviation of the 3B-GZ-15 steel column was 5mm. The problem was marked by the problem marking module and a rectification work order was generated. After receiving the work order, the responsible personnel adjusted the position of the steel column and uploaded a video of the rectification (showing a deviation of 2mm). The supervisor verified that the rectification met the standard through the rectification tracking module and closed the work order.

[0105] Acceptance and Archiving: The construction team initiates an acceptance application for the North Side Area of ​​No. 3. The acceptance personnel click on point 3B-GZ-15 in the panoramic model, review the guidance card, construction video, and rectification records of the steel column, and after confirming that it meets the acceptance standards, they sign the acceptance opinion. The system automatically generates the "Installation Acceptance File of Steel Column on the North Side of No. 3" and archives it to the project management system.

[0106] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0107] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A remote acceptance data management method based on a visualization model, characterized in that, The method includes the following steps: Step S1. Capture images of the construction site, obtain image data from various perspectives, and after stitching and optimization, store the images in a lightweight format. After 3D reconstruction processing, generate an interactive digital panoramic model of the construction site. Step S2. Divide the panoramic model into functional zones and work points for zone management, locate the coordinates of the points, set virtual coordinate anchor points for rapid target positioning, and verify the construction accuracy. Step S3. Collect images of the construction process, extract the motion features of the construction points, establish a behavior dataset, standardize the unstructured dataset, convert it into structured data, and output a remote guidance dataset. Step S4. Send the remote guidance dataset, collect and upload videos of the construction points, generate preview thumbnails, use the point number as the unique primary key, strongly bind the structured data with the point, and write it into the point attribute library of the panoramic model. Step S5. Link the data of each point, locate the problem point, mark the problem and add a problem description, link the construction information of the point, push the rectification work order, and after the regional construction is completed, initiate the acceptance application. After all points are judged to be qualified according to the acceptance standards, the acceptance file is generated.

2. The remote acceptance data management method based on a visualization model according to claim 1, characterized in that: Step S1 includes: Step S11. Guide the drone to plan the photography route based on the initial building model obtained. Determine the shooting viewpoint position according to the flight altitude, overlap and drone field of view. Plan the photography path according to the ordered viewpoint and building layout. Connect the camera group and complete the high-altitude panoramic acquisition of the construction area according to the preset route. Step S12. The UAV orthophoto is uniformly segmented, and the sample is expanded using a generative adversarial network. Features in the normal illumination domain and low illumination domain are extracted, and the features in the source domain are transferred to the target domain. A feature point KD tree index is established using unlabeled data in the target domain. Matching points of the target structure at the operation point are searched using a bidirectional approximate nearest neighbor algorithm. The orthophoto dataset is matched, and the matched images are fused. Through model simplification, texture compression and LOD reconstruction, an interactive 3D model is output.

3. The remote acceptance data management method based on a visualization model according to claim 2, characterized in that: Step S2 includes: Step S21. Extract feature points in each region based on the target shape descriptor, use the automatic registration algorithm of iterative feature points to register the preprocessed target points, construct a triangulation model for the ground point data, and logically partition the panoramic model according to functional zones and secondary operation points. Step S22. Identify the construction target based on the normal vector and curvature features, register the scene point cloud and the construction target point cloud using the iterative nearest neighbor algorithm with normal vector constraints, collect square wave images at different object distances, calculate the optical point spread function, determine the distance between two points in the panoramic model, and verify the construction accuracy of the components. Step S3 includes: Step S31. By iteratively training the generator and discriminator, the brightness of the original image is adaptively adjusted to generate a monitoring image, a video frame encoder is constructed, and spatial information is guided by pixel segmentation to capture the spatiotemporal dependencies of each frame in the image. The global spatiotemporal feature representation is learned, and the inter-frame dynamic change information is learned by the differential frame encoder. Step S32. Use the HRNet model to extract the positions of moving joints, collect the operating specifications, technical instructions and process requirements issued by the management personnel, construct a dynamic behavior detection model, and generate a remote guidance dataset. The remote guidance dataset contains text specifications, video links and key frame screenshots of the construction process.

4. The remote acceptance data management method based on a visualization model according to claim 3, characterized in that: Step S4 includes: Step S41. When the video lacks standardized basis or is blurry, an alert is automatically triggered. The blur parameters are estimated based on the initial image spectrum. The artificially blurred image is compared with the actual image in terms of similarity. The blur kernel of the image with the highest similarity is selected. The blur kernel and the high-frequency part of the blurred image are alternately and iteratively updated to restore the video. Step S42. Generate a preview thumbnail for each video segment and label it with the acquisition time and acquisition terminal type. Generate a playback link, using the point number as the unique primary key. Bind the guidance data, video playback link, keyframe preview, and accuracy verification report to the corresponding point in the 3D model, record the binding time, binding person, and data source, and generate a binding log.

5. The remote acceptance data management method based on a visualization model according to claim 4, characterized in that: Step S5 includes: Step S51. Using the panoramic model, backtrack point by point, continuously collect images of target points, transmit them to the host, the host generates target point cloud, classifies the point cloud according to the curvature value of each point, determines the model boundary points according to the directional Hausdorff distance, and constructs a local coordinate system using the eigenvalue decomposition method based on the query point and the set of neighboring points, and calculates the spatial transformation matrix. Step S52. Calculate the deviation values ​​of key dimensions, flatness, and verticality indicators, compare them with the acceptance standard values, judge the construction effect, locate the problem location in the panoramic model for unqualified points, generate guidance cards, obtain the rectification video after rectification, verify whether the rectification meets the standards, close the work order if it meets the standards, and re-rectify if it does not meet the standards, and generate a rectification progress board. Based on the guidance records, process videos and rectification status, generate a digital acceptance file.

6. A remote acceptance data management system based on a visualization model, characterized in that, The system includes the following modules: panoramic modeling module, interaction optimization module, data processing module, remote guidance module, and rectification and acceptance module; The panoramic modeling module is used to capture images of the construction site using drones and ground panoramic cameras, acquire image data from various perspectives, and after stitching and optimization, store the images in a lightweight format. After three-dimensional reconstruction processing, an interactive digital panoramic model of the construction site is generated. The interactive optimization module is used to manage the panoramic model by functional area and work point, locate the coordinates of the points, list all primary areas and secondary points in the panoramic model, set virtual coordinate anchor points for rapid target positioning, and verify the construction accuracy. The data processing module is used to collect images of the construction process through smart safety helmets, fixed monitoring or drones, extract the motion characteristics of the construction points, establish a behavior dataset, standardize the unstructured dataset, convert it into structured data, and output a remote guidance dataset in CSV format. The remote guidance module is used to send the remote guidance dataset to the corresponding responsible entity, continuously collect and upload videos of construction sites, extract key frames from the videos, generate preview thumbnails, and use the site number as a unique primary key to strongly bind structured data with the site and write it into the site attribute library of the panoramic model. The rectification and acceptance module is used to initiate an acceptance application after the completion of regional construction, associate data of each point in the region, locate problem points in the panoramic model, mark the problem and add a problem description, upload photos of the problem site, automatically associate the construction information of the point, push rectification work orders to the responsible personnel, and generate an acceptance file after all points are judged to be qualified according to the acceptance standards.

7. A remote acceptance data management system based on a visualization model according to claim 6, characterized in that: The panoramic modeling module includes: an image acquisition unit, an image processing unit, and a three-dimensional reconstruction unit; The image acquisition unit is used to guide the drone to plan the photography route based on the initial building model obtained by the drone, determine the shooting viewpoint position according to the flight altitude, overlap and drone field of view, plan the photography path according to the ordered viewpoint and building layout, connect the camera group, and complete the high-altitude panoramic acquisition of the construction area according to the preset route. The image processing unit is used to uniformly segment the orthophoto of the UAV, expand the samples using a generative adversarial network, extract features from the normal illumination domain and the low illumination domain, transfer the features from the source domain to the target domain, and establish a feature point KD tree index using unlabeled data in the target domain. The 3D reconstruction unit is used to search for matching points of the target structure at the work site using a bidirectional approximate nearest neighbor algorithm, match the orthophoto dataset, fuse the matched images, and output an interactive 3D model through model simplification, texture compression, and LOD reconstruction.

8. The remote acceptance data management system based on a visualization model according to claim 7, characterized in that: The interaction optimization module includes: a partition management unit and a point cloud processing unit; The partition management unit is used to extract feature points in each region based on the target shape descriptor, register the preprocessed target points using an iterative feature point automatic registration algorithm, construct a triangulation model for the ground point data, and logically partition the panoramic model according to functional zones and secondary operation points. The point cloud processing unit is used to identify construction targets based on normal vectors and curvature features, register scene point clouds and construction target point clouds through an iterative nearest neighbor algorithm with normal vector constraints, acquire square wave images at different object distances, calculate optical point spread functions, determine the distance between two points in the panoramic model, and verify the construction accuracy of components. The data processing module includes: a behavioral data unit and a structured data unit; The behavior data unit is used to adaptively adjust the brightness of the original image by iteratively training the generator and discriminator, generate monitoring images, construct a video frame encoder, guide the perception of spatial information through pixel segmentation, capture the spatiotemporal dependencies of each frame in the image, learn global spatiotemporal feature representation, and learn inter-frame dynamic change information with a differential frame encoder. The structured unit is used to extract the position of moving joints using the HRNet model, collect the operating specifications, technical instructions and process requirements issued by the management personnel, build a dynamic behavior detection model, and generate a remote guidance dataset. The remote guidance dataset contains text specifications, video links and key frame screenshots of the construction process.

9. A remote acceptance data management system based on a visualization model according to claim 8, characterized in that: The remote guidance module includes: a video restoration unit and a guidance binding unit; The video restoration unit is used to automatically trigger a reminder when the video lacks standard references or is blurry. It estimates the blur parameters based on the initial image spectrum, compares the artificially blurred image with the actual image for similarity, selects the blur kernel of the image with the highest similarity, and iteratively updates the blur kernel and the high-frequency part of the blurred image alternately to perform video restoration processing. The guidance binding unit is used to generate a preview thumbnail for each video segment, and mark the acquisition time and acquisition terminal type. It generates a playback link, uses the point number as a unique primary key, binds the guidance data, video playback link, keyframe preview and accuracy verification report to the corresponding point in the 3D model, records the binding time, binding person and data source, and generates a binding log.

10. A remote acceptance data management system based on a visualization model according to claim 9, characterized in that: The rectification and acceptance module includes: a problem marking unit, a rectification tracking unit, and an acceptance archiving unit; The problem labeling unit is used to continuously collect target point images by tracing back point by point through the panoramic model, and transmit them to the host. The host generates a target point cloud, classifies the point cloud according to the curvature value of each point, and determines the model boundary points according to the directional Hausdorff distance. The rectification tracking unit is used to construct a local coordinate system based on the query point and the set of neighboring points using the eigenvalue decomposition method, calculate the spatial transformation matrix, calculate the deviation values ​​of key dimensions, flatness and verticality indicators, compare them with the acceptance standard values, judge the construction effect, and for points that fail the inspection, locate the problem location in the panoramic model and generate guidance cards. The acceptance and archiving unit is used to obtain the video of the rectification after the rectification is completed, verify whether the rectification meets the standards, close the work order if it meets the standards, and re-rectify if it does not meet the standards, and generate a rectification progress dashboard. Based on the guidance records, process videos and rectification status, a digital acceptance file is generated.

Images