Fabricated building construction progress assessment method and device and readable medium
Through drone photography and deep convolutional neural network model combined with YOLO model, a three-dimensional model is created to evaluate the construction progress of prefabricated buildings, solving the problem that the construction progress is difficult to be intuitive and accurate, and improving the accuracy and safety of construction management.
Patent Information
- Application Number
- CN202411083751.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-12
AI Technical Summary
In prefabricated building construction, it is difficult for the project management team to understand the construction progress intuitively and accurately, resulting in increased construction risks.
The drone camera device is used to obtain two-dimensional aerial images, and through the deep convolutional neural network model and the YOLO model, a sparse three-dimensional point cloud, dense three-dimensional point cloud and fine three-dimensional model are created, and the construction progress is then evaluated.
It realizes intuitive and accurate display of construction progress, reduces construction risks, and improves data clarity and accuracy.
Smart Images

Figure CN118941994A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of building construction information technology, and in particular to a method, device and readable medium for evaluating the construction progress of an assembled building. Background Art
[0002] Prefabricated buildings refer to buildings that transfer a large amount of on-site work in traditional construction methods to factories, where building components and accessories (such as floor slabs, wall panels, stairs, balconies, etc.) are processed and manufactured in the factory, transported to the construction site, and assembled and installed on site through reliable connection methods.
[0003] In the above-mentioned prefabricated building construction, it is crucial for the project management team to grasp the construction progress. The project management team needs to adjust the construction strategy in time according to the construction progress to ensure the quality of the project and ensure the smooth progress of the project. However, the current construction progress is often presented through flat data such as engineering documents. The project management team cannot have a good and intuitive understanding of the current construction progress, and paper data is prone to errors, resulting in construction risks. Summary of the invention
[0004] The purpose of the present invention is to solve the problem described in the above background technology that there is currently a lack of an intuitive and accurate method to obtain the current construction progress. A method, device and readable medium for evaluating the construction progress of prefabricated buildings are proposed herein. Through this evaluation method, the current construction progress can be understood more intuitively, and the data obtained by this evaluation method is more accurate and clear.
[0005] The technical solution adopted by the present invention to solve the technical problem is: a method for evaluating the construction progress of an assembled building, the method comprising the following steps:
[0006] S1, using a drone camera to obtain a two-dimensional aerial image of a target building;
[0007] S2, inputting the two-dimensional aerial image into a deep convolutional neural network model to obtain a processed two-dimensional aerial image;
[0008] S3, creating a sparse three-dimensional point cloud based on the processed two-dimensional aerial image;
[0009] S4, performing densification processing on the sparse three-dimensional point cloud to obtain a dense three-dimensional point cloud;
[0010] S5, reconstructing the dense 3D point cloud into a 3D surface model;
[0011] S6, mapping the building texture in the two-dimensional aerial image to the three-dimensional surface model to obtain a fine three-dimensional model;
[0012] S7. Input the detailed three-dimensional model and the planned installation components of the target building into the YOLO model, and output the construction progress of the target building.
[0013] Preferably, the drone camera device comprises a drone and a plurality of cameras with different viewing angles mounted on the drone, and obtaining a two-dimensional aerial image of a target building using the drone camera device comprises the following steps:
[0014] S11, the drone flies above the target building to obtain the top and partial side images of the target building;
[0015] S12, the UAV flies around the target building in the same horizontal plane and keeps the focal length of the camera unchanged to obtain the surrounding images of the target building, wherein each camera collects at least 24 surrounding images, and the shooting angle of each adjacent surrounding image does not exceed 15°;
[0016] S13, the camera is aimed at the facade of the target building, and the camera angle and the position of the drone are adjusted according to the specific features of the target building to obtain a high-resolution image of the facade of the target building;
[0017] S14, integrating and arranging the top and part of the side images, the circumferential side images and the high-resolution images to form a two-dimensional aerial image.
[0018] Preferably, the deep convolutional neural network model comprises an input layer, a hidden layer and an output layer, and the hidden layer comprises a convolutional layer, a pooling layer and a fully connected layer;
[0019] The input layer is used to preprocess the input two-dimensional aerial image;
[0020] The convolution layer is used to extract image features from the two-dimensional aerial image;
[0021] The pooling layer is used to adjust the spatial size of the two-dimensional aerial image and filter image features;
[0022] The fully connected layer is used to integrate and transform the filtered image features;
[0023] The output layer is used to convert the integrated and converted image features into the final classification results, and the output is a processed two-dimensional aerial image.
[0024] Preferably, creating a sparse three-dimensional point cloud based on the processed two-dimensional aerial image comprises the following steps:
[0025] S31, extracting image features from the processed two-dimensional aerial image by a size-invariant feature transformation method;
[0026] S32, performing pairwise matching of image features on all processed two-dimensional aerial images through an optimal search operator to determine matching points;
[0027] S33, screening and removing matching points that do not meet the requirements through a random sampling consistency algorithm;
[0028] S34. Generate a sparse three-dimensional point cloud through a motion structure recovery algorithm.
[0029] Preferably, the sparse three-dimensional point cloud is densified to obtain a dense three-dimensional point cloud, comprising the following steps:
[0030] S41, clustering the image using the CMVS algorithm;
[0031] S42. The sparse three-dimensional point cloud is densified by the PMVS algorithm to obtain a dense three-dimensional point cloud.
[0032] Preferably, the dense three-dimensional point cloud is reconstructed into a three-dimensional surface model using a Poisson surface reconstruction algorithm.
[0033] Preferably, mapping the building texture in the two-dimensional aerial image to the three-dimensional surface model to obtain a fine three-dimensional model comprises the following steps:
[0034] S61, establishing a mapping relationship between the two-dimensional aerial image and the three-dimensional surface model space coordinates;
[0035] S62, determining the texture contribution of each two-dimensional aerial image to the three-dimensional surface model according to the calibration data of the drone camera device;
[0036] S63, for each triangular facet of the three-dimensional surface model, extracting texture information from the corresponding two-dimensional aerial image by calculating visibility and projection position of the triangular facet in different two-dimensional aerial images;
[0037] S64, mapping the texture information to the three-dimensional surface model through a mapping relationship to obtain a fine three-dimensional model.
[0038] Preferably, the detailed three-dimensional model and the planned installation components of the target building are input into the YOLO model, and the construction progress of the target building is output, including the following steps:
[0039] S71, counting all prefabricated components of the target building, and inputting the statistical data into the YOLO model;
[0040] S72, inputting the refined three-dimensional model into the YOLO model, performing disassembly and classification, and obtaining the number of prefabricated components actually installed;
[0041] S73. Obtain the current progress percentage of the target building by calculating the ratio of the number of prefabricated components actually installed to the number of all prefabricated components of the target building.
[0042] An electronic device, comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;
[0043] The memory is used to store the evaluation program of the construction progress of the prefabricated building;
[0044] The processor is used to execute the evaluation program, and when the evaluation program is executed, the steps of the evaluation method described above are implemented.
[0045] A computer-readable storage medium stores an evaluation program for the construction progress of an assembled building. When the evaluation program is executed by a processor, the steps of the evaluation method described above are implemented.
[0046] The beneficial effects of the present invention are:
[0047] 1. The present invention obtains two-dimensional aerial images of target buildings through three photography methods: orthographic photography, oblique photography and close-up photography, which can achieve no-dead-angle photography of buildings and provide a basis for subsequent three-dimensional modeling. Moreover, by combining the above three photography methods, it can meet the photography requirements of complex and irregular buildings with higher accuracy.
[0048] 2. The present invention uses the YOLO model to decompose the constructed fine three-dimensional model, classify it into different prefabricated structures, and compare it with all the prefabricated structures used in the design to obtain the current construction progress, and its display method is more intuitive and accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 The present invention is a flowchart of the method for evaluating the construction progress of prefabricated buildings. DETAILED DESCRIPTION
[0050] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.
[0051] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.
[0052] The present invention is now further described in conjunction with specific examples. The following examples are only for explaining the present invention, but do not constitute a limitation of the present invention. The test samples and test processes used in the following examples include the following (if the specific experimental conditions are not indicated in the examples, they are usually in accordance with conventional conditions or the conditions recommended by the reagent company; the reagents, consumables, etc. used in the following examples, if not otherwise specified, can all be obtained from commercial channels).
[0053] A method for evaluating the construction progress of an assembled building, wherein the assembled building is completed by constructing a plurality of assembled components, and the method for evaluating the construction progress comprises the following steps:
[0054] S1. Use a drone camera to obtain a two-dimensional aerial image of a target building.
[0055] Specifically, the drone camera device is composed of a drone and five cameras with different viewing angles installed on the drone. The drone camera device uses three photography methods, namely, orthographic photography, oblique photography and close-up photography, to obtain a two-dimensional aerial image of the target building, specifically:
[0056] S11, Orthophotography: The drone flies above the target building and adjusts the camera to look vertically downward to obtain the top and part of the side images of the target building;
[0057] S12, oblique photography: The UAV flies around the target building in the same horizontal plane and keeps the camera focal length unchanged to obtain the surrounding images of the target building, wherein each camera collects at least 24 surrounding images, and the shooting angle of each adjacent surrounding image does not exceed 15°;
[0058] S13, close-up photography: the camera is aimed at the facade of the target building, and the camera angle and the position of the drone are adjusted according to the specific features of the target building to obtain a high-resolution image of the facade of the target building;
[0059] S14, integrating and arranging the acquired top and partial side images, circumferential side images and high-resolution images to form a two-dimensional aerial image.
[0060] It should be noted that the number of cameras is not limited to 5. In other embodiments, the number may be 3, 4, 6, etc., as long as the number is sufficient to capture the target building in detail without blind spots.
[0061] S2. Input the two-dimensional aerial image into a deep convolutional neural network model to obtain a processed two-dimensional aerial image.
[0062] Specifically, the two-dimensional aerial images need to be processed before they can be used as input sources to create sparse three-dimensional point clouds. The deep convolutional neural network model consists of an input layer, a hidden layer, and an output layer. The hidden layer consists of a convolutional layer, a pooling layer, and a fully connected layer. The input layer is used to preprocess the input two-dimensional aerial images. The preprocessing operations include removing the mean (centering each dimension of the data to 0), normalization, and possible PCA (principal component analysis, used for dimensionality reduction), whitening, etc.
[0063] The convolution layer performs convolution operation with the two-dimensional aerial image through the convolution kernel, slides on the two-dimensional aerial image and calculates the weighted sum of the local area to extract image features;
[0064] The pooling layer is located between the convolutional layers. It adjusts the spatial size of the two-dimensional aerial image and reduces the resolution of the two-dimensional aerial image through maximum pooling or average pooling. At the same time, it removes redundant information and retains the most recognizable image features, thereby streamlining the model and reducing the amount of calculation.
[0065] The fully connected layer is located at the back end of the deep convolutional neural network model. It integrates and transforms the image features extracted by the previous layers. The fully connected layer maps the multi-dimensional features to a one-dimensional feature vector of a fixed length for further processing. In a series of fully connected layers, the output length of the last fully connected layer is usually consistent with the number of categories that the model needs to classify.
[0066] The output layer converts the output of the fully connected layer into the final classification result. In the image classification task, the softmax function or the sigmoid function is used to calculate the output of the last fully connected layer. These functions convert the original output into a probability distribution, thereby generating a probability value for each category, representing the possibility that the output belongs to each category. The final output is a processed two-dimensional aerial image.
[0067] S3. Create a sparse three-dimensional point cloud based on the processed two-dimensional aerial image.
[0068] Specifically, create a sparse 3D point cloud by following the steps below:
[0069] S31, extracting image features from the processed two-dimensional aerial image by a size-invariant feature transformation method;
[0070] S32, performing pairwise matching of image features on all processed two-dimensional aerial images through an optimal search operator to determine matching points;
[0071] S33, screening and removing matching points that do not meet the requirements through a random sampling consistency algorithm;
[0072] S34. Generate a sparse three-dimensional point cloud through a motion structure recovery algorithm.
[0073] S4, performing densification processing on the sparse three-dimensional point cloud to obtain a dense three-dimensional point cloud;
[0074] Specifically, the feature point matching algorithm is used to determine the initial seed points, and then these seed points are expanded outward to obtain dense point clouds or surface patches with normal information. The specific operations are as follows:
[0075] S41, clustering the image using the CMVS algorithm;
[0076] S42. The sparse three-dimensional point cloud is densified by the PMVS algorithm to obtain a dense three-dimensional point cloud.
[0077] S5. Use Poisson surface reconstruction algorithm to reconstruct the dense 3D point cloud into a 3D surface model.
[0078] The Poisson surface reconstruction algorithm is a 3D surface reconstruction technology based on solving the Poisson equation. According to the intrinsic relationship between the directed point set collected on the model surface and the model indicator function, it is first assumed that the points in the dense 3D point cloud are located on or near the surface of the 3D surface model. Then, by summing the normal vectors of the point cloud, the gradient function is calculated using the integral relationship between the gradient of the model indicator function and the surface normal field. The gradient function is then converted into a Poisson equation problem to deduce the indicator function of the model. Finally, the reconstructed surface is obtained by extracting the isosurface. After the reconstruction process is completed, a closed surface is formed, which contains the 3D scene of the target object. At the same time, the redundant parts need to be removed to obtain the final 3D surface model.
[0079] S6. Mapping the building texture in the two-dimensional aerial image to the three-dimensional surface model to obtain a detailed three-dimensional model.
[0080] Specifically, this step consists of the following steps:
[0081] S61, establishing a mapping relationship between the two-dimensional aerial image and the three-dimensional surface model space coordinates;
[0082] S62, determining the texture contribution of each two-dimensional aerial image to the three-dimensional surface model according to the calibration data of the drone camera device;
[0083] S63, for each triangular facet of the three-dimensional surface model, extracting texture information from the corresponding two-dimensional aerial image by calculating visibility and projection position of the triangular facet in different two-dimensional aerial images;
[0084] S64, mapping the texture information to the three-dimensional surface model through a mapping relationship to obtain a fine three-dimensional model.
[0085] Through the above steps, the texture information of the two-dimensional aerial image can be applied to the fine three-dimensional model to enhance the visual effect and realism of the ink fragrance.
[0086] S7. Input the detailed three-dimensional model and the planned installation components of the target building into the YOLO model, and output the construction progress of the target building.
[0087] Specifically, the following steps are included:
[0088] S71, counting all prefabricated components of the target building, and inputting the statistical data into the YOLO model;
[0089] S72, inputting the refined three-dimensional model into the YOLO model, performing disassembly and classification, and obtaining the number of prefabricated components actually installed;
[0090] S73. Obtain the current progress percentage of the target building by calculating the ratio of the number of prefabricated components actually installed to the number of all prefabricated components of the target building.
[0091] All prefabricated components of the prefabricated building are provided by the design unit. In one embodiment, the information of the prefabricated components planned for construction can also be input into the YOLO model, and compared with the types and quantities of the prefabricated components obtained by disassembly and classification of the fine three-dimensional model into the YOLO model to obtain the current completion status of the construction plan.
[0092] The YOLO model adopts the structure of "backbone + neck + detection head". The backbone network is a recognized architecture or network for feature extraction. It adopts CSPDarknet53 and is composed of stacked CSPResNeXt modules. Several ResNext Blocks are stacked in each module. The neck network is to enhance certain properties in the model, such as expanding the receptive field, introducing attention mechanism or enhancing feature integration capability, etc., using SPP and PANet modules. The detection head adopts the YOLO series network.
[0093] The present invention adopts YOLOv4 as the object detection algorithm, which allows single GPU training on the one hand, reducing the training cost; on the other hand, it adopts more BoF (Bag of Freebies) and BoS (Bag of Specials), significantly improving the detection accuracy without affecting the inference speed.
[0094] The present invention also provides an electronic device, which includes but is not limited to terminal devices with computing functions such as tablet computers, mobile terminals, and computers.
[0095] The electronic device is composed of a processor, a memory, a communication interface and a communication bus. The processor, the memory and the communication interface communicate with each other through the communication bus. The memory is used to store an evaluation program for the construction progress of prefabricated buildings; the processor is used to execute the evaluation program, and when the evaluation program is executed, the steps of the evaluation method described above are implemented.
[0096] In one embodiment, the assessment program for the construction progress of prefabricated buildings may be stored in a computer-readable storage medium, which may be a storage device such as a memory card or a hard disk.
[0097] The processor may optionally be a data processing chip such as a CPU, which is used to run an assessment program for the construction progress of prefabricated buildings and output assessment results.
[0098] The network interface may be a wired interface, a wireless interface, etc., which is used for data transmission and interaction between electronic devices and other electronic devices.
[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0100] The above is a detailed introduction to the method, device and readable medium for evaluating the construction progress of an assembled building provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.
Claims
1. A method for evaluating the construction progress of an assembled building, characterized in that: The evaluation method comprises the following steps: S1, using a drone camera to obtain a two-dimensional aerial image of a target building; S2, inputting the two-dimensional aerial image into a deep convolutional neural network model to obtain a processed two-dimensional aerial image; S3, creating a sparse three-dimensional point cloud based on the processed two-dimensional aerial image; S4, performing densification processing on the sparse three-dimensional point cloud to obtain a dense three-dimensional point cloud; S5, reconstructing the dense 3D point cloud into a 3D surface model; S6, mapping the building texture in the two-dimensional aerial image to the three-dimensional surface model to obtain a fine three-dimensional model; S7. Input the detailed three-dimensional model and the planned installation components of the target building into the YOLO model, and output the construction progress of the target building.
2. The evaluation method according to claim 1, characterized in that: The drone camera device includes a drone and a plurality of cameras with different viewing angles installed on the drone. The process of obtaining a two-dimensional aerial image of a target building using the drone camera device includes the following steps: S11, the drone flies above the target building to obtain the top and partial side images of the target building; S12, the UAV flies around the target building in the same horizontal plane and keeps the focal length of the camera unchanged to obtain the surrounding images of the target building, wherein each camera collects at least 24 surrounding images, and the shooting angle of each adjacent surrounding image does not exceed 15°; S13, the camera is aimed at the facade of the target building, and the camera angle and the position of the drone are adjusted according to the specific features of the target building to obtain a high-resolution image of the facade of the target building; S14, integrating and arranging the top and part of the side images, the circumferential side images and the high-resolution images to form a two-dimensional aerial image.
3. The evaluation method according to claim 1, characterized in that: The deep convolutional neural network model includes an input layer, a hidden layer and an output layer, and the hidden layer includes a convolutional layer, a pooling layer and a fully connected layer; The input layer is used to preprocess the input two-dimensional aerial image; The convolution layer is used to extract image features from the two-dimensional aerial image; The pooling layer is used to adjust the spatial size of the two-dimensional aerial image and filter image features; The fully connected layer is used to integrate and transform the filtered image features; The output layer is used to convert the integrated and converted image features into the final classification results, and the output is a processed two-dimensional aerial image.
4. The evaluation method according to claim 1, characterized in that: The step of creating a sparse three-dimensional point cloud based on the processed two-dimensional aerial image comprises the following steps: S31, extracting image features from the processed two-dimensional aerial image by a size-invariant feature transformation method; S32, performing pairwise matching of image features on all processed two-dimensional aerial images through an optimal search operator to determine matching points; S33, screening and removing matching points that do not meet the requirements through a random sampling consistency algorithm; S34. Generate a sparse three-dimensional point cloud through a motion structure recovery algorithm.
5. The evaluation method according to claim 1, characterized in that: Densifying the sparse 3D point cloud to obtain a dense 3D point cloud includes the following steps: S41, clustering the image using the CMVS algorithm; S42. The sparse three-dimensional point cloud is densified by the PMVS algorithm to obtain a dense three-dimensional point cloud.
6. The evaluation method according to claim 1, characterized in that: The dense three-dimensional point cloud is reconstructed into a three-dimensional surface model using a Poisson surface reconstruction algorithm.
7. The evaluation method according to claim 1, characterized in that: Mapping the building texture in the two-dimensional aerial image to the three-dimensional surface model to obtain a fine three-dimensional model includes the following steps: S61, establishing a mapping relationship between the two-dimensional aerial image and the three-dimensional surface model space coordinates; S62, determining the texture contribution of each two-dimensional aerial image to the three-dimensional surface model according to the calibration data of the drone camera device; S63, for each triangular facet of the three-dimensional surface model, extracting texture information from the corresponding two-dimensional aerial image by calculating visibility and projection position of the triangular facet in different two-dimensional aerial images; S64, mapping the texture information to the three-dimensional surface model through a mapping relationship to obtain a fine three-dimensional model.
8. The evaluation method according to claim 1, characterized in that: Inputting the detailed 3D model and the planned installation components of the target building into the YOLO model, and outputting the construction progress of the target building includes the following steps: S71, counting all prefabricated components of the target building, and inputting the statistical data into the YOLO model; S72, inputting the refined three-dimensional model into the YOLO model, performing disassembly and classification, and obtaining the number of prefabricated components actually installed; S73. Obtain the current progress percentage of the target building by calculating the ratio of the number of prefabricated components actually installed to the number of all prefabricated components of the target building.
9. An electronic device, characterized in that: The electronic device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store the evaluation program of the construction progress of the prefabricated building; The processor is used to execute the evaluation program, and when the evaluation program is executed, the steps of the evaluation method according to any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an evaluation program for the construction progress of prefabricated buildings. When the evaluation program is executed by the processor, the steps of the evaluation method as described in any one of claims 1 to 8 are implemented.
Citation Information
Cited By
Construction progress management method and device based on aerial photography, medium and equipment
CN119809573A
Construction site progress monitoring method and system based on two-dimensional image and three-dimensional point cloud
CN121095886A
Construction completion rate determination method and device, electronic equipment and storage medium
CN121392521A
