Completed model project acceptance method, device and equipment based on AI image recognition and storage medium

Abstract

The invention discloses a completion model project acceptance method, device and equipment based on AI image recognition and a storage medium, and relates to the technical field of project acceptance. According to the method, a current scene image is segmented based on an image semantic segmentation model, a segmentation mask is obtained and mapped to a three-dimensional point cloud space, and a target point cloud comprising a plurality of point cloud clusters with semantic tags is obtained, so that point cloud processing is simplified; and performing geometric feature registration on the target point cloud and the preset BIM model to obtain a space alignment relationship between the target point cloud and the preset BIM model, thereby generating a current two-dimensional drawing according to the space alignment relationship, and performing difference comparison on the current two-dimensional drawing and the initial two-dimensional drawing to generate an acceptance report capable of objectively and accurately reflecting the completion condition. And the completion acceptance efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of engineering acceptance technology, and in particular to a method, device, equipment and storage medium for acceptance of completed model engineering projects based on AI image recognition. Background Art

[0002] Final acceptance is a critical step in construction projects before project delivery. Its core purpose is to verify that the building entity strictly complies with design drawings, construction specifications, and safety standards. While the development of Building Information Modeling (BIM) technology has provided a foundation for digital management of the entire design-construction-acceptance cycle, the acceptance stage still faces the challenge of efficiently and accurately comparing the building entity with the BIM model.

[0003] In the existing method, building entities are usually restored by relying on point cloud data collected on-site. However, the lack of semantic information in point cloud data leads to a heavy reliance on manual labeling for its automated comparison with the BIM model. In addition, the processing efficiency of massive point clouds is low, making it difficult to meet real-time requirements, which ultimately restricts the reliability of acceptance results and the efficiency of project delivery. Summary of the Invention

[0004] The main purpose of this application is to provide a method, device, equipment and storage medium for completion model engineering acceptance based on AI image recognition, aiming to solve the technical problems of the existing method based on manual comparison.

[0005] To achieve the above objectives, this application proposes a method for acceptance of completed model projects based on AI image recognition, which includes:

[0006] Acquire a current scene image, and segment the current scene image based on an image semantic segmentation model to obtain a segmentation mask;

[0007] Mapping the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud, wherein the target point cloud includes a plurality of point cloud clusters with semantic labels;

[0008] Performing geometric feature registration on the target point cloud and the preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model;

[0009] A current two-dimensional drawing is generated according to the spatial alignment relationship, and the current two-dimensional drawing is compared with the initial two-dimensional drawing to generate an acceptance report.

[0010] In one embodiment, the step of mapping the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud includes:

[0011] Obtain the initial point cloud of the current scene through the scanner;

[0012] Aligning the segmentation mask with the initial point cloud using camera intrinsic and extrinsic parameters corresponding to the current scene image, and transferring the semantic label corresponding to the segmentation mask to the initial point cloud;

[0013] Dividing the initial point cloud into point cloud clusters according to the semantic labels, and extracting the point cloud clusters corresponding to the building components;

[0014] Each of the point cloud clusters is labeled to obtain a target point cloud.

[0015] In one embodiment, the step of performing geometric feature registration on the target point cloud and the preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model includes:

[0016] Extracting building components with geometric features from the target point cloud, and extracting corresponding point cloud geometric features;

[0017] Extracting corresponding model geometric features from the preset BIM model;

[0018] Matching the point cloud geometric features and the model geometric features using a preset point matching algorithm to determine an initial registration point pair;

[0019] The random sampling consistency algorithm is used to perform coarse registration on the initial registration point pairs, and the coarse registration results are optimized by an iterative closest point algorithm to obtain the spatial alignment relationship between the target point cloud and the preset BIM model.

[0020] In one embodiment, the step of generating the current two-dimensional drawing according to the spatial alignment relationship includes:

[0021] Based on the spatial alignment relationship, the target point cloud and the preset BIM model are converted into the same coordinate system to obtain three-dimensional scene data with unified coordinates;

[0022] Comparing the geometric differences between each point cloud cluster and each building component in the preset BIM model using the three-dimensional scene data with unified coordinates;

[0023] The preset BIM model is updated according to the geometric difference comparison result, and the current two-dimensional drawing is generated based on the updated preset BIM model.

[0024] In one embodiment, after the step of converting the target point cloud and the preset BIM model to the same coordinate system based on the spatial alignment relationship to obtain three-dimensional scene data with unified coordinates, the method further includes:

[0025] Constructing a three-dimensional scene model according to the three-dimensional scene data, wherein the three-dimensional scene model is formed by superimposing the target point cloud and the preset BIM model in the same coordinate system;

[0026] The current scaling ratio is determined according to the three-dimensional scene model, and the initial two-dimensional drawing is scaled according to the current scaling ratio, so that the scaled initial two-dimensional drawing is updated to the bottom surface of the three-dimensional scene model, and the three-dimensional scene model with the updated bottom surface is displayed through an interactive page.

[0027] In one embodiment, after the step of constructing a three-dimensional scene model according to the three-dimensional scene data, the method further includes:

[0028] Obtaining a correspondence between the preset BIM model and the building components of the initial two-dimensional drawing;

[0029] Determining a simultaneous correspondence between the three-dimensional scene model and the initial two-dimensional drawing according to the spatial alignment relationship and the building component correspondence relationship;

[0030] Displaying the three-dimensional scene model and the initial two-dimensional drawing in sections on the interactive interface;

[0031] In response to a user selecting a target building component in the three-dimensional scene model through the interactive interface, locating and marking the target building component in the initial two-dimensional drawing according to the simultaneous correspondence relationship;

[0032] Alternatively, in response to a user selecting a target building component in the initial two-dimensional drawing through the interactive interface, the target building component is located and marked in the three-dimensional scene model according to the simultaneous correspondence relationship.

[0033] In one embodiment, the step of comparing the current two-dimensional drawing with the initial two-dimensional drawing to generate an acceptance report includes:

[0034] Obtaining a preset key component list, and marking key components in the current two-dimensional drawing and the initial two-dimensional drawing according to the preset key list, and obtaining the current drawing position and the standard drawing position of each key component;

[0035] When the current drawing position and the standard drawing position do not satisfy a preset position error condition, the corresponding key components are determined as components to be concerned, and an acceptance report is generated based on the components to be concerned.

[0036] In addition, to achieve the above objectives, the present application also proposes a device for acceptance of completed model projects based on AI image recognition, the device comprising:

[0037] A semantic segmentation module is used to obtain a current scene image and segment the current scene image based on an image semantic segmentation model to obtain a segmentation mask;

[0038] A point cloud mapping module is used to map the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud, wherein the target point cloud includes a plurality of point cloud clusters with semantic labels;

[0039] A geometric registration module is used to perform geometric feature registration on each of the point cloud clusters and a preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model;

[0040] The result generation module is used to generate a current two-dimensional drawing according to the spatial alignment relationship, compare the current two-dimensional drawing with the initial two-dimensional drawing, and generate an acceptance report.

[0041] In addition, to achieve the above-mentioned purpose, the present application also proposes a completion model engineering acceptance device based on AI image recognition, the device including: a memory, a processor, and a completion model engineering acceptance program based on AI image recognition stored on the memory and runnable on the processor, the computer program being configured to implement the steps of the completion model engineering acceptance method based on AI image recognition as described above.

[0042] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and the storage medium stores a completion model engineering acceptance program based on AI image recognition. When the completion model engineering acceptance program based on AI image recognition is executed by the processor, the steps of the completion model engineering acceptance method based on AI image recognition as described above are implemented.

[0043] The present application discloses a method for acceptance of completed model projects based on AI image recognition. The method obtains a current scene image and segments the current scene image based on an image semantic segmentation model to obtain a segmentation mask; maps the segmentation mask to a three-dimensional point cloud space through camera parameters to obtain a target point cloud, wherein the target point cloud includes a number of point cloud clusters with semantic labels; performs geometric feature registration on the target point cloud and a preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model; generates a current two-dimensional drawing based on the spatial alignment relationship, and compares the current two-dimensional drawing with the initial two-dimensional drawing to generate an acceptance report. Since the present application quickly generates a segmentation mask for the current scene image through an image semantic segmentation model, and then maps it to a three-dimensional point cloud space to add semantic labels to the point cloud data, simplified processing of the point cloud data is achieved. Then, the point cloud clusters with semantic labels are spatially aligned with the preset BIM model, and the current two-dimensional drawing is generated and compared with the initial two-dimensional drawing to obtain an acceptance report. This can objectively and accurately reflect the project acceptance situation and help improve acceptance efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0046] Figure 1 This is a flow chart of the first embodiment of the completion model project acceptance method based on AI image recognition in this application;

[0047] Figure 2 This is a flow chart of the second embodiment of the completion model project acceptance method based on AI image recognition of this application;

[0048] Figure 3 This is a flow chart of the third embodiment of the completion model project acceptance method based on AI image recognition of this application;

[0049] Figure 4 This is a schematic diagram of the module structure of the first embodiment of the completion model engineering acceptance device based on AI image recognition in this application;

[0050] Figure 5 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the completion model engineering acceptance method based on AI image recognition in the embodiment of the present application. DETAILED DESCRIPTION

[0051] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0052] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0053] The present application embodiment provides a method for acceptance of completed model engineering based on AI image recognition, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the completion model project acceptance method based on AI image recognition in this application. In this embodiment, the method includes: steps S10 to S40:

[0054] Step S10: Acquire a current scene image, and segment the current scene image based on an image semantic segmentation model to obtain a segmentation mask.

[0055] It should be noted that the execution subject of the method of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, personal computer, mobile phone, etc. Or other electronic devices that can be connected to the project completion acceptance system. This embodiment takes the completion model project acceptance device based on AI image recognition (hereinafter referred to as "acceptance device") as an example to illustrate this embodiment and the following embodiments.

[0056] It is understood that the current scene image can be a series of images captured in real time by the acceptance personnel using the laser scanner's built-in high dynamic range camera in the current as-built scene. For each image captured by the camera, the extrinsic parameters used to describe the matrix transformation are recorded as camera parameters for subsequent mapping.

[0057] It should be understood that the image semantic segmentation model can be a Segment Anything Model (SAM) or other large visual models such as DINOV2. These large visual models can use AI technology to semantically segment images, that is, classify each pixel in the image into different semantic categories, such as walls, doors and windows, fire hydrants, pipes and other building components, and generate pixel-level masks. The image semantic segmentation model can be pre-trained using a large number of labeled building scene images and combined with data enhancement technology to improve its adaptability to complex scenes.

[0058] In the specific implementation, the current scene image is input into a pre-trained image semantic segmentation model, which performs semantic segmentation on the input image and generates a pixel-level mask for identifying various categories of building components in the current scene image.

[0059] Step S20: Mapping the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud, wherein the target point cloud includes a plurality of point cloud clusters with semantic labels.

[0060] It should be understood that the camera parameters may include the intrinsic parameters (such as focal length, principal point coordinates, etc.) and extrinsic parameters (such as rotation matrix, translation vector, etc.) of the aforementioned camera, which are used to convert pixel points in the image coordinate system into the three-dimensional world coordinate system and ensure the correspondence between the current scene image and the target point cloud in three-dimensional space.

[0061] It is understood that the target point cloud can be obtained by annotating the initial point cloud with semantic labels. The target point cloud includes several semantically labeled point cloud clusters, each semantic label used to identify a different type of building component, that is, each point cloud cluster corresponds to a building component. The initial point cloud can be scanned by a laser scanner and acquired synchronously with the current scene image.

[0062] In the specific implementation, the initial point cloud of the current scene can be first obtained through a scanner; the segmentation mask is aligned with the initial point cloud through the camera internal and external parameters corresponding to the current scene image, and the semantic labels corresponding to the segmentation mask are transferred to the initial point cloud; the initial point cloud is divided into point cloud clusters according to each semantic label, and the point cloud clusters corresponding to each building component are extracted; each point cloud cluster is labeled to obtain the target point cloud.

[0063] Step S30: performing geometric feature registration on the target point cloud and the preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model.

[0064] It should be understood that the preset BIM model can be a design model of the current as-built scene generated based on the initial two-dimensional drawing, which can include the original design information of the building entity in the as-built scene, such as the installation position, shape, size, etc. of each building component.

[0065] It should be noted that geometric feature registration can be the process of aligning a target point cloud with a pre-set BIM model by matching geometric features in the target point cloud with those in the BIM model. The geometric features can be planes, edges, and edge corners.

[0066] In the specific implementation, geometric features are extracted from the target point cloud and the preset BIM model respectively to obtain the point cloud geometric features and the model geometric features; then the feature point matching algorithm can be used to perform feature matching on the point cloud geometric features and the model geometric features to find the corresponding feature point pairs, thereby obtaining the spatial alignment relationship between the target point cloud and the preset BIM model.

[0067] Furthermore, in order to improve the alignment accuracy between the target point cloud and the preset BIM model, step S30 specifically includes: steps S301 to S304:

[0068] Step S301: extracting building components with geometric features from the target point cloud, and extracting corresponding point cloud geometric features.

[0069] Step S302: extracting corresponding model geometric features from the preset BIM model.

[0070] It should be noted that since the point cloud clusters with semantic labels in the target point cloud can be some specific building components, such as sprinklers, fire escape indicator lights, smoke detectors, etc., these specific building components may not have geometric features such as planes, edges, and edge corners. Therefore, it is possible to first determine the building components with geometric features from the target point cloud, such as wall columns, and then extract the point cloud geometric features corresponding to the building components with geometric features.

[0071] It should be understood that by extracting geometric features from the target point cloud and the preset BIM model respectively, representative features in the point cloud and the BIM model are obtained, providing a basis for subsequent feature matching.

[0072] Step S303: Matching the point cloud geometric features and the model geometric features using a preset point matching algorithm to determine an initial registration point pair.

[0073] It should be understood that the preset feature point matching algorithm can be a feature descriptor algorithm such as Fast Point Feature Histogram (FPFH), Signature of Histograms of Orientations (SHOT), etc., or a deep learning matching (3DMatch) algorithm, which is not limited in this embodiment.

[0074] It is understandable that obtaining the initial registration point pair establishes the correspondence between the target point cloud and the preset BIM model, thereby providing a basis for subsequent registration.

[0075] Step S304: A random sampling consensus algorithm is used to perform a coarse registration on the initial registration point pairs, and the coarse registration result is optimized by an iterative closest point algorithm to obtain a spatial alignment relationship between the target point cloud and the preset BIM model.

[0076] It should be noted that the registration algorithm can be divided into two stages: coarse registration and fine registration. In the coarse registration stage, a random sample consensus (RANSAC) algorithm or a feature-based global registration method can be used to initially align the target point cloud and the preset BIM model, eliminating large-scale position and posture differences. The goal of coarse registration is to quickly bring the target point cloud and the preset BIM model to a similar position and posture, providing initial conditions for subsequent fine registration.

[0077] During the fine registration phase, an iterative closest point (ICP) algorithm or an optimization-based registration method (such as the Gauss-Newton method) can be used to further optimize the alignment accuracy between the target point cloud and the pre-set BIM model, ensuring that the registration error is within the millimeter level. The goal of fine registration is to achieve high-precision alignment between the target point cloud and the pre-set BIM model through iterative optimization based on the coarse registration.

[0078] It should be noted that the above-mentioned multi-stage registration algorithm ensures that the target point cloud and the preset BIM model are aligned with millimeter-level accuracy, which is conducive to meeting the high-precision requirements of final acceptance.

[0079] Step S40: generating a current two-dimensional drawing according to the spatial alignment relationship, performing a difference comparison between the current two-dimensional drawing and the initial two-dimensional drawing, and generating an acceptance report.

[0080] It should be understood that based on the spatial alignment relationship, the registered target point cloud and the preset BIM model can be projected onto a two-dimensional plane to generate a two-dimensional drawing with a coordinate system, namely the current two-dimensional drawing. The current two-dimensional drawing can reflect the status of the building entity.

[0081] It should be noted that the initial two-dimensional drawings are the original two-dimensional drawings in the design stage, which can be used as the basis for final acceptance. By comparing the current two-dimensional drawings with the initial two-dimensional drawings, it is possible to check whether the actual construction status of the building is consistent with the design.

[0082] In specific implementation, the current two-dimensional drawing and the initial two-dimensional drawing can be compared through an image difference algorithm, so that the areas where the actual construction conditions differ from the design can be marked on the initial two-dimensional drawing and output as an acceptance report, so that the acceptance personnel can intuitively obtain the actual construction conditions of the current completion scene and ensure the objectivity and traceability of the acceptance conclusion.

[0083] In addition, to speed up the generation of acceptance reports, after obtaining several target point clouds with semantically labeled point cloud clusters, the target point clouds can be directly projected onto the initial two-dimensional drawing. The point positions of the corresponding point cloud clusters on the initial two-dimensional drawing can be obtained to determine the projected point positions of each building component in the target point cloud. The difference is then compared with the original point positions of each building component on the initial two-dimensional drawing to generate an acceptance report.

[0084] This embodiment uses an image semantic segmentation model to quickly generate a segmentation mask for the current scene image, and then maps it to the three-dimensional point cloud space to add semantic labels to the point cloud data, thereby simplifying the processing of the point cloud data. The target point cloud is then spatially aligned with the preset BIM model, and the current two-dimensional drawing is generated and compared with the initial two-dimensional drawing to obtain an acceptance report. This can objectively and accurately reflect the project acceptance status and help improve acceptance efficiency.

[0085] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 2 , Figure 2 This is a flow chart of the second embodiment of the completion model engineering acceptance method based on AI image recognition in this application.

[0086] In this embodiment, to specifically illustrate the process of generating the acceptance report, step S40 includes steps S401 to S405:

[0087] Step S401: Based on the spatial alignment relationship, the target point cloud and the preset BIM model are converted into the same coordinate system to obtain three-dimensional scene data with unified coordinates.

[0088] It should be understood that since the target point cloud and the preset BIM model come from different sources, in order to facilitate the subsequent update of the preset BIM model based on the building entity, the target point cloud and the preset BIM model can be first converted into the same coordinate system.

[0089] In a specific implementation, the transformation matrix (such as the rotation matrix and translation vector) in the spatial alignment relationship obtained above can be used to transform all points in the target point cloud and the preset BIM model into the same coordinate system as a prerequisite for subsequent geometric difference comparison.

[0090] Step S402: using the three-dimensional scene data with unified coordinates, comparing the geometric differences between each point cloud cluster and each building component in the preset BIM model.

[0091] It should be understood that in the three-dimensional scene data with unified coordinates, it is necessary to compare the geometric differences of each building component between the building entity (point cloud cluster) and the design model (preset BIM model) to determine whether the construction meets the design requirements.

[0092] In a specific implementation, for each point cloud cluster in the target point cloud, the corresponding building component in the preset BIM model is found, and then the corresponding geometric difference is calculated. The geometric difference may include: size difference, shape difference, and position deviation, etc.

[0093] Step S403: updating the preset BIM model according to the geometric difference comparison result, and generating a current two-dimensional drawing based on the updated preset BIM model.

[0094] It should be understood that based on the results of the geometric difference comparison, the pre-set BIM model needs to be updated to reflect the latest status of the actual building. The updated BIM model can be used to generate the current 2D drawing as the basis for acceptance. The current 2D drawing can be set to the same scale as the initial 2D drawing.

[0095] In practice, based on the geometric difference comparison results, the geometric parameters of the corresponding building components in the pre-set BIM model can be adjusted to align with the building entity. The updated pre-set BIM model can then be projected onto a 2D plane to generate the current 2D drawing.

[0096] Step S404: obtaining a preset key component list, and marking key components in the current two-dimensional drawing and the initial two-dimensional drawing according to the preset key list, and obtaining the current drawing position and the standard drawing position of each key component.

[0097] It's important to note that during final acceptance, the construction quality of some building components may be critical to the safety and functionality of the entire building, or the inspector may need to focus on certain building components based on specific acceptance objectives. For example, in a fire inspection scenario, these building components may include sprinklers, fire hydrants, pipes, and so on. Therefore, a list of one or more critical components can be pre-generated to remind inspectors or to guide the generation of a targeted acceptance report.

[0098] In practice, you first obtain a preset list of key components. Then, locate and mark these key components in both the current and initial 2D drawings, recording the location of each key component in both the current and standard drawings. By marking key components, you can ensure that these important parts are inspected and evaluated during the acceptance process, improving the pertinence and effectiveness of the acceptance process.

[0099] Step S405: when the current drawing position and the standard drawing position do not satisfy a preset position error condition, the corresponding key components are determined as components to be concerned, and an acceptance report is generated based on the components to be concerned.

[0100] It should be noted that for the marked key components, it is necessary to check whether their positions in the current drawing and the standard drawing meet the preset position error conditions. If not, it means that there are deviations in the construction of these building components, and the acceptance personnel need to further confirm them.

[0101] In practice, the position of key components in the current drawing is compared with the standard drawing, and the position deviation is calculated. The calculated position deviation is then determined to be within a pre-set error range. If the deviation exceeds the preset error range, these key components are identified as requiring further attention, and detailed error information for these components is recorded to generate an acceptance report.

[0102] In this example, the target point cloud and the pre-set BIM model are converted to the same coordinate system, geometric differences are compared, the BIM model is updated, and a current 2D drawing is generated. Key components are marked and their positional deviations are checked, and finally an acceptance report is generated. This ensures the accuracy and pertinence of project acceptance, helping to further improve acceptance efficiency and quality.

[0103] Based on the first and second embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the first and second embodiments can be referred to the above introduction and will not be described in detail later. Figure 3 , Figure 3 This is a flow chart of the third embodiment of the completion model engineering acceptance method based on AI image recognition in this application.

[0104] In this embodiment, in order to facilitate the acceptance personnel to intuitively obtain the differences between the building entity and the design model in the current as-built scene and improve the interactivity and flexibility of the acceptance process, after step S401, the following steps are further included:

[0105] Step S501: constructing a three-dimensional scene model according to the three-dimensional scene data, wherein the three-dimensional scene model is formed by superimposing the target point cloud and the preset BIM model in the same coordinate system.

[0106] It should be understood that since the target point cloud and the preset BIM model in the 3D scene data are in the same coordinate system, they can be superimposed on each other to construct a 3D scene model. This can intuitively display the overall appearance and details of the building, allowing non-professional acceptance personnel (such as project managers and owners) to easily understand the building's current status and design intent.

[0107] Step S502: Determine the current scaling ratio based on the three-dimensional scene model, and scale the initial two-dimensional drawing based on the current scaling ratio to update the scaled initial two-dimensional drawing to the bottom surface of the three-dimensional scene model, and display the three-dimensional scene model after the bottom surface is updated through an interactive page.

[0108] It should be understood that the current scaling ratio may be a ratio that makes the 3D scene model consistent with the size of the initial 2D drawing. Considering that 2D scaling consumes less computing power than 3D scaling, the current scaling ratio may be the ratio used to scale the initial 2D drawing. By scaling the initial 2D drawing, the initial 2D drawing can be scaled to the size of the bottom surface of the 3D scene model, and the scaled initial 2D drawing can then be directly used as the bottom surface of the 3D scene model.

[0109] It should be noted that, since the preset BIM model in the three-dimensional scene model is a design model generated based on the initial two-dimensional drawing, the three-dimensional building components and the building components in the initial two-dimensional drawing after the bottom surface is scaled can be directly connected through connecting lines in the three-dimensional scene model.

[0110] It should also be noted that when the acceptance personnel browse the completion acceptance model on the interactive page, they can enter the model by switching the model observation perspective, so as to intuitively know the correspondence between the same building components in the three-dimensional model and the two-dimensional drawing through the connecting lines, reducing the user's time cost in data understanding and model operation, and improving the overall efficiency of the acceptance work.

[0111] Furthermore, in order to meet the acceptance personnel's diverse display requirements for the acceptance results, after step S501, the following steps are further included: steps S601 to S605:

[0112] Step S601: Obtaining the correspondence between the preset BIM model and the building components in the initial two-dimensional drawing.

[0113] Step S602: determining a simultaneous correspondence relationship between the three-dimensional scene model and the initial two-dimensional drawing according to the spatial alignment relationship and the building component correspondence relationship.

[0114] It should be understood that since the preset BIM model is a design model generated based on the initial two-dimensional drawings, the mapping relationship between each building component in the preset BIM model and each building component in the initial two-dimensional drawings can be directly obtained.

[0115] It should be noted that, since the aforementioned process determines the spatial alignment relationship between the target point cloud and the preset BIM model, combined with the correspondence between the preset BIM model and the building components of the initial two-dimensional drawing, the joint correspondence between the three-dimensional scene model obtained by superimposing the target point cloud and the preset BIM model and the initial two-dimensional drawing can be determined.

[0116] Step S603: displaying the three-dimensional scene model and the initial two-dimensional drawing in sections on the interactive interface.

[0117] It should be understood that the interactive interface can display the 3D scene model and the initial 2D drawings in different areas, allowing acceptance personnel to jointly review and analyze the 3D model and 2D drawings without switching between different software or tools, saving time and effort. Furthermore, acceptance personnel with different professional backgrounds (such as architects and structural engineers) can jointly review and discuss the model through the interactive interface, promoting collaborative work among multiple disciplines.

[0118] Step S604: In response to the user selecting a target building component in the three-dimensional scene model through the interactive interface, the target building component is located and marked in the initial two-dimensional drawing according to the simultaneous correspondence relationship.

[0119] Step S605: In response to the user selecting a target building component in the initial two-dimensional drawing through the interactive interface, the target building component is located and marked in the three-dimensional scene model according to the simultaneous correspondence relationship.

[0120] It should be understood that the aforementioned joint correspondence ensures an accurate match between the three-dimensional model and the two-dimensional drawing, reducing possible errors that may occur when manually searching and matching components. Through the partition display and two-way selection functions of the interactive interface, acceptance personnel can more conveniently switch and view between the three-dimensional model and the two-dimensional drawing, improving the convenience and intuitiveness of the operation.

[0121] In the specific implementation, the interactive interface displays the 3D scene model and the initial 2D drawings in separate sections. Inspection personnel can use the interactive interface to select target building components in the 3D scene model and then locate and mark them in the initial 2D drawings based on the joint correspondence. Conversely, inspection personnel can also select target building components in the initial 2D drawings and then locate and mark them in the 3D scene model.

[0122] Furthermore, on the interactive interface, another area can be divided to display a 360-degree panoramic video of the current scene. The panoramic video can also be captured in real time by the acceptance personnel using the built-in high dynamic range camera of the laser scanner in the current completed scene, and spatially and temporally aligned with the current scene image and the initial point cloud (or target point cloud) to ensure the spatiotemporal consistency of the three-dimensional scene model and the 360-degree panoramic video displayed on the interactive interface.

[0123] Specifically, after the acceptance personnel select the target building component in the three-dimensional scene model in the interactive interface, they can switch the observation perspective based on the target building component, thereby displaying real-life videos under different observation perspectives based on the target building component in the 360-degree panoramic video, and marking the observation perspective based on the target building component on the initial two-dimensional drawing.

[0124] In this embodiment, a three-dimensional scene model is constructed by superimposing the target point cloud and the preset BIM model, and the initial two-dimensional drawing is scaled to replace the bottom surface of the three-dimensional scene model for display, so that users can intuitively view and understand the corresponding position relationship between building components in two dimensions and three dimensions; and the spatial alignment relationship between the target point cloud and the preset BIM model and the corresponding relationship between the preset BIM model and the building components of the initial two-dimensional drawing are combined to obtain the joint corresponding relationship between the three-dimensional scene model and the initial two-dimensional drawing, so that through the partition display and two-way selection function of the interactive interface, users can more conveniently switch and view between the three-dimensional model and the two-dimensional drawing, further improving the interactivity and flexibility of the acceptance process.

[0125] It should be noted that the above examples are only used to understand this application and do not constitute a limitation on the completion model engineering acceptance method based on AI image recognition in this application. More simple transformations based on this technical concept are all within the scope of protection of this application.

[0126] In addition, this application also provides a completion model engineering acceptance device based on AI image recognition, referring to Figure 4 , Figure 4 This is a structural block diagram of the first embodiment of the completion model engineering acceptance device based on AI image recognition in this application; Figure 4 As shown, the device includes:

[0127] The semantic segmentation module 401 is used to obtain a current scene image and segment the current scene image based on an image semantic segmentation model to obtain a segmentation mask;

[0128] A point cloud mapping module 402 is configured to map the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud, wherein the target point cloud includes a plurality of point cloud clusters with semantic labels;

[0129] A geometric registration module 403 is configured to perform geometric feature registration on the target point cloud and a preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model;

[0130] The result generation module 404 is used to generate a current two-dimensional drawing according to the spatial alignment relationship, and compare the current two-dimensional drawing with the initial two-dimensional drawing to generate an acceptance report.

[0131] Furthermore, the point cloud mapping module 402 is also used to obtain an initial point cloud of the current scene through a scanner; align the segmentation mask with the initial point cloud through the camera internal and external parameters corresponding to the current scene image, and transfer the semantic labels corresponding to the segmentation mask to the initial point cloud; divide the initial point cloud into point cloud clusters according to each semantic label, and extract the point cloud clusters corresponding to each building component; and label each point cloud cluster to obtain a target point cloud.

[0132] Furthermore, the geometric registration module 403 is also used to extract building components with geometric features from the target point cloud and extract corresponding point cloud geometric features; extract corresponding model geometric features from the preset BIM model; match the point cloud geometric features and the model geometric features through a preset point matching algorithm to determine initial registration point pairs; use a random sampling consistency algorithm to coarsely align the initial registration point pairs, and optimize the coarse registration results through an iterative nearest point algorithm to obtain the spatial alignment relationship between the target point cloud and the preset BIM model.

[0133] Furthermore, the result generation module 404 is also used to convert the target point cloud and the preset BIM model into the same coordinate system based on the spatial alignment relationship to obtain three-dimensional scene data with unified coordinates; compare the geometric differences between each point cloud cluster and each building component in the preset BIM model through the three-dimensional scene data with unified coordinates; update the preset BIM model according to the geometric difference comparison results, and generate the current two-dimensional drawing based on the updated preset BIM model.

[0134] Furthermore, the result generation module 404 is also used to construct a three-dimensional scene model based on the three-dimensional scene data, wherein the three-dimensional scene model is composed of the target point cloud and the preset BIM model superimposed in the same coordinate system; determine the current scaling ratio based on the three-dimensional scene model, and scale the initial two-dimensional drawing according to the current scaling ratio to update the scaled initial two-dimensional drawing to the bottom surface of the three-dimensional scene model, and display the three-dimensional scene model after the bottom surface is updated through an interactive page.

[0135] Furthermore, the result generation module 404 is also used to obtain the correspondence between the preset BIM model and the building components of the initial two-dimensional drawing; determine the joint correspondence between the three-dimensional scene model and the initial two-dimensional drawing based on the spatial alignment relationship and the building component correspondence; display the three-dimensional scene model and the initial two-dimensional drawing in partitions on the interactive interface; in response to the user's selection operation of the target building component in the three-dimensional scene model through the interactive interface, locate and mark the target building component in the initial two-dimensional drawing according to the joint correspondence; in response to the user's selection operation of the target building component in the initial two-dimensional drawing through the interactive interface, locate and mark the target building component in the three-dimensional scene model according to the joint correspondence.

[0136] Furthermore, the result generation module 404 is also used to obtain a preset key component list, and mark the key components in the current two-dimensional drawing and the initial two-dimensional drawing according to the preset key list, and obtain the current drawing position and standard drawing position of each key component; when the current drawing position and the standard drawing position do not meet the preset position error conditions, the corresponding key components are determined as components to be concerned, and an acceptance report is generated based on each component to be concerned.

[0137] This embodiment uses an image semantic segmentation model to quickly generate a segmentation mask for the current scene image, and then maps it to the three-dimensional point cloud space to add semantic labels to the point cloud data, thereby simplifying the processing of the point cloud data. The target point cloud is then spatially aligned with the preset BIM model, and the current two-dimensional drawing is generated and compared with the initial two-dimensional drawing to obtain an acceptance report. This can objectively and accurately reflect the project acceptance status and help improve acceptance efficiency.

[0138] In addition, the present application also provides a completion model engineering acceptance device based on AI image recognition. The completion model engineering acceptance device based on AI image recognition includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can execute the completion model engineering acceptance method based on AI image recognition in the above-mentioned embodiment one.

[0139] Reference below Figure 5 , Figure 5This is a schematic diagram of the structure of the completed model engineering acceptance device based on AI image recognition in this application. The completed model engineering acceptance device based on AI image recognition in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The completion model engineering acceptance equipment based on AI image recognition shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0140] like Figure 5 As shown, the completion model engineering acceptance equipment based on AI image recognition may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the completion model engineering acceptance equipment based on AI image recognition are also stored in RAM1004. The processing device 1001, ROM1002 and RAM1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the completion model engineering acceptance device based on AI image recognition to communicate wirelessly or wired with other devices to exchange data. Although the figure shows the completion model engineering acceptance device based on AI image recognition with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have instead.

[0141] In addition, the present application also provides a computer-readable storage medium having computer-readable program instructions (i.e., computer program) stored thereon, and the computer-readable program instructions are used to execute the completion model engineering acceptance method based on AI image recognition in the above-mentioned embodiment.

[0142] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0143] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional elements in the process, method, article, or system comprising the element.

[0144] The serial numbers of the above-mentioned embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments. Moreover, they are only some embodiments of the present application and do not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the description and drawings of the present application under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for acceptance of completed model projects based on AI image recognition, characterized in that: The method comprises: Acquire a current scene image, and segment the current scene image based on an image semantic segmentation model to obtain a segmentation mask; Mapping the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud, wherein the target point cloud includes a plurality of point cloud clusters with semantic labels; Performing geometric feature registration on the target point cloud and the preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model; A current two-dimensional drawing is generated according to the spatial alignment relationship, and the current two-dimensional drawing is compared with the initial two-dimensional drawing to generate an acceptance report.

2. The method according to claim 1, wherein The step of mapping the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud comprises: Obtain the initial point cloud of the current scene through the scanner; Aligning the segmentation mask with the initial point cloud using camera intrinsic and extrinsic parameters corresponding to the current scene image, and transferring the semantic label corresponding to the segmentation mask to the initial point cloud; Dividing the initial point cloud into point cloud clusters according to the semantic labels, and extracting the point cloud clusters corresponding to the building components; Each of the point cloud clusters is labeled to obtain a target point cloud.

3. The method according to claim 1, wherein The step of performing geometric feature registration on the target point cloud and the preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model includes: Extracting building components with geometric features from the target point cloud, and extracting corresponding point cloud geometric features; Extracting corresponding model geometric features from the preset BIM model; Matching the point cloud geometric features and the model geometric features using a preset point matching algorithm to determine an initial registration point pair; The random sampling consistency algorithm is used to perform coarse registration on the initial registration point pairs, and the coarse registration results are optimized by an iterative closest point algorithm to obtain the spatial alignment relationship between the target point cloud and the preset BIM model.

4. The method according to claim 1, wherein The step of generating the current two-dimensional drawing according to the spatial alignment relationship includes: Based on the spatial alignment relationship, the target point cloud and the preset BIM model are converted into the same coordinate system to obtain three-dimensional scene data with unified coordinates; Comparing the geometric differences between each of the point cloud clusters and each building component in the preset BIM model using the three-dimensional scene data with unified coordinates; The preset BIM model is updated according to the geometric difference comparison result, and the current two-dimensional drawing is generated based on the updated preset BIM model.

5. The method according to claim 4, wherein After the step of converting the target point cloud and the preset BIM model into the same coordinate system based on the spatial alignment relationship to obtain three-dimensional scene data with unified coordinates, the method further includes: Constructing a three-dimensional scene model according to the three-dimensional scene data, wherein the three-dimensional scene model is formed by superimposing the target point cloud and the preset BIM model in the same coordinate system; The current scaling ratio is determined according to the three-dimensional scene model, and the initial two-dimensional drawing is scaled according to the current scaling ratio, so that the scaled initial two-dimensional drawing is updated to the bottom surface of the three-dimensional scene model, and the three-dimensional scene model with the updated bottom surface is displayed through an interactive page.

6. The method according to claim 5, wherein After the step of constructing a three-dimensional scene model according to the three-dimensional scene data, the method further includes: Obtaining a correspondence between the preset BIM model and the building components of the initial two-dimensional drawing; Determining a simultaneous correspondence between the three-dimensional scene model and the initial two-dimensional drawing according to the spatial alignment relationship and the building component correspondence relationship; Displaying the three-dimensional scene model and the initial two-dimensional drawing in sections on the interactive interface; In response to a user selecting a target building component in the three-dimensional scene model through the interactive interface, locating and marking the target building component in the initial two-dimensional drawing according to the simultaneous correspondence relationship; Alternatively, in response to a user selecting a target building component in the initial two-dimensional drawing through the interactive interface, the target building component is located and marked in the three-dimensional scene model according to the simultaneous correspondence relationship.

7. The method according to claim 4, wherein The step of comparing the current two-dimensional drawing with the initial two-dimensional drawing to generate an acceptance report includes: Obtaining a preset key component list, and marking key components in the current two-dimensional drawing and the initial two-dimensional drawing according to the preset key list, and obtaining the current drawing position and the standard drawing position of each key component; When the current drawing position and the standard drawing position do not satisfy a preset position error condition, the corresponding key components are determined as components to be concerned, and an acceptance report is generated based on the components to be concerned.

8. A device for acceptance of completed model projects based on AI image recognition, characterized in that: The device comprises: A semantic segmentation module is used to obtain a current scene image and segment the current scene image based on an image semantic segmentation model to obtain a segmentation mask; A point cloud mapping module is used to map the segmentation mask to a three-dimensional point cloud space using camera parameters to obtain a target point cloud, wherein the target point cloud includes a plurality of point cloud clusters with semantic labels; A geometric registration module is used to perform geometric feature registration on each of the point cloud clusters and a preset BIM model to obtain a spatial alignment relationship between the target point cloud and the preset BIM model; The result generation module is used to generate a current two-dimensional drawing according to the spatial alignment relationship, compare the current two-dimensional drawing with the initial two-dimensional drawing, and generate an acceptance report.

9. A completion model engineering acceptance device based on AI image recognition, characterized in that: The device includes: a memory, a processor, and a completion model engineering acceptance program based on AI image recognition stored in the memory and runnable on the processor. The completion model engineering acceptance program based on AI image recognition is configured to implement the steps of the completion model engineering acceptance method based on AI image recognition as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that The computer storage medium stores a completion model engineering acceptance program based on AI image recognition. When the completion model engineering acceptance program based on AI image recognition is executed by the processor, the steps of the completion model engineering acceptance method based on AI image recognition as described in any one of claims 1 to 7 are implemented.

Images