Engineering calculation amount automatic extraction method based on point cloud data and YOLOv11

Abstract

The invention discloses an engineering calculation amount automatic extraction method based on point cloud data and YOLOv11, and the method comprises the following steps: S1, collecting and obtaining the point cloud data of a building, carrying out the preprocessing of the collected point cloud data, and generating a three-dimensional point cloud model; s2, constructing a BIM model, and carrying out integration registration on the point cloud data and BIM model data; s3, a YOLOv11 model is established, and the YOLOv11 model is trained; s4, carrying out target detection by adopting the trained YOLOv11 model; and S5, correcting the construction position and the dimensional precision, automatically calculating the engineering quantity, and outputting the engineering quantity. Through automatic point cloud data preprocessing and a YOLOv11 target detection model, manual intervention is reduced, efficiency is improved, a large amount of data is automatically processed, and manual operation is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of building construction, and in particular to an automatic extraction method for engineering quantity based on point cloud data and YOLOv11. Background Art

[0002] In the construction industry, the automated analysis and calculation of construction drawings has always been one of the key technologies to improve engineering efficiency. With the continuous development of Building Information Modeling (BIM) technology, the digital processing and data management of engineering drawings have gradually become an important trend in the construction industry. BIM models can provide data support for the entire life cycle of building design, construction and operation, but there are still many challenges in the automated calculation of the bill of quantities (QTO).

[0003] The traditional method of calculating construction quantities relies on manual measurement, which usually involves a lot of drawing analysis, component identification and numerical calculation, which is cumbersome and error-prone. With the continuous expansion of the scale of construction projects, relying solely on manual calculations not only increases the workload, but also easily leads to data errors, calculation omissions and other problems, which in turn affects the overall progress and cost control of the project. Therefore, how to improve the accuracy and efficiency of engineering quantity calculations through automation technology has become a technical problem that needs to be solved urgently in the current construction industry.

[0004] In recent years, the progress of deep learning technology, especially in the field of computer vision, has provided new ideas for the automatic analysis of architectural drawings. Object detection algorithms based on deep learning, such as YOLO (You Only Look Once), have shown excellent performance in the fields of object detection and image recognition. Combining the YOLO model with the automatic analysis of architectural drawings is expected to achieve automatic identification and measurement of building components and materials, and provide an intelligent solution for the compilation of bills of quantities.

[0005] Defects and shortcomings of the existing technology:

[0006] 1. Much manual intervention and low efficiency: Traditional methods of calculating quantities in architectural drawings usually rely on manual item-by-item calculations and manual identification of building components. This method is not only cumbersome and error-prone, but also requires a huge workload in the calculation process for large or complex construction projects, which seriously affects work efficiency and accuracy. Especially in complex architectural drawings, manual operations make it difficult to quickly and accurately extract quantities, making it more difficult to manage project costs and time.

[0007] 2. Poor drawing recognition capability and low accuracy: Although existing drawing parsing technology can recognize some architectural symbols through traditional image processing methods, the recognition accuracy of traditional methods is often insufficient in complex architectural drawings, especially drawings of different styles and symbols, especially when facing some hand-drawn drawings or scanned drawings of poor quality. The recognition accuracy is low. Existing computer vision technology has limited automatic recognition capabilities for architectural elements (such as steel bars, nodes, material identification, etc.) and drawing annotations (such as dimensions, symbols, etc.), and it is difficult to stably apply them in various types and formats of architectural drawings.

[0008] 3. Lack of intelligence and adaptability: Current architectural drawing processing methods generally lack the ability to adapt to different drawing specifications, symbol standards and annotation methods. In practical applications, there are many different specifications and expressions for drawings. The existing technology has great limitations in the automatic mapping and processing of different symbols in drawings, resulting in a low degree of intelligence and inability to make flexible adjustments and adaptations according to the actual needs of the project.

[0009] 4. Difficulty in integration and poor compatibility: Existing automatic calculation methods for architectural drawings usually combine and apply various technologies (such as OCR, image processing, deep learning, etc.), but these technologies are poorly integrated and it is difficult to form an efficient and stable unified workflow. Many existing solutions lack compatibility with BIM data formats, point cloud data, and deep learning models in actual applications, resulting in difficulties in integrating different data sources, affecting the ease of use and application efficiency of the overall system.

[0010] 5. Poor processing speed and real-time performance: Although existing technologies have guaranteed accuracy, they often require a long processing time, especially in large-scale construction projects, where the processing efficiency cannot meet the real-time requirements. In actual construction projects, especially for the demand for real-time data, the processing speed of existing technologies cannot keep pace with the project progress requirements, resulting in a lag in data update and optimization during the construction phase, affecting project management decisions and resource scheduling. Summary of the invention

[0011] In order to solve the above problems, the present invention provides an automatic extraction method for engineering quantity based on point cloud data and YOLOv11, comprising the following steps: S1. Acquire point cloud data of a building, pre-process the acquired point cloud data and generate a three-dimensional point cloud model; S2. Build a BIM model, and integrate and align the point cloud data with the BIM model data; S3. Establish a YOLOv11 model and train the YOLOv11 model; S4. Use the trained YOLOv11 model for target detection; S5. Correct the construction position and dimensional accuracy, and automatically calculate the engineering quantity and output it.

[0012] Furthermore, the point cloud data of the building collected in step S1 is obtained specifically by a laser scanner, a drone or other three-dimensional scanning equipment.

[0013] Furthermore, the preprocessing of the collected point cloud data in step S1 specifically includes: removing noise from the point cloud data using statistical outliers, and downsampling by random sampling.

[0014] Furthermore, the use of statistical outlier points to remove noise from point cloud data specifically includes: calculating a threshold Di by setting a standard deviation parameter and a radius value, and when the distance from a point in the neighborhood to point i is greater than Di, it is regarded as an abnormal point and removed; wherein the calculation formula of the threshold Di is: Di=SD+s×σ; wherein Di represents the noise point judgment threshold; SD represents the standard deviation parameter; s represents the standard deviation coefficient; and σ represents the standard deviation of the distances from all points in the neighborhood to point i.

[0015] Furthermore, the downsampling by random sampling specifically includes: the original point cloud data set D = {d1, d2, ..., d N}, random sampling sample S = {d i1 , d i2 , ..., d iM}, the probability of each sample being selected is P(d i ∈S) = M / N; where i1, i2, ..., i M are all different indices randomly selected from {1,2,...,N}.

[0016] Furthermore, the integration and registration of the point cloud data with the BIM model data in the step S2 specifically includes the following steps: S21. Extracting plane patches through the BIM model and the three-dimensional point cloud model respectively; S22. Extracting four plane sets in the BIM model, and substituting them into the four plane sets for congruence in the three-dimensional point cloud model; S23. Calculating the transformation matrix, centroid support, plane support and centroid root mean square error, and applying the transformation matrix; S24. Clustering transformations and sorting the transformations.

[0017] Furthermore, the calculation formula for calculating the centroid root mean square error is:

[0018] Where T is the translation matrix; R is the rotation matrix; qi and pi are the i-th points in the reference point cloud Q and the point cloud to be registered P respectively; N is the number of point clouds.

[0019] Furthermore, the engineering quantity data output in step S5 is structured data.

[0020] Furthermore, it also includes result verification and optimization steps: verifying the accuracy of the engineering quantity data and checking the consistency between the components identified by YOLOv11 and the actual components in the BIM model.

[0021] The present invention provides an automatic extraction method for engineering quantity based on point cloud data and YOLOv11, which has the following beneficial effects:

[0022] The present invention reduces manual intervention and improves efficiency through automated point cloud data preprocessing and YOLOv11 target detection model. The point cloud data preprocessing steps include using the SOR algorithm to remove noise, the random sampling algorithm to downsample, and the ICP algorithm and PCA algorithm to perform point cloud registration. These steps automatically process a large amount of data and reduce manual operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0024] Figure 1 A flow chart of the method provided by the present invention;

[0025] Figure 2 This is a flow chart of the point cloud + BIM data alignment provided by the present invention. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0027] The following is a detailed description of the implementation method of the present invention in conjunction with the accompanying drawings. Only some embodiments are described, not all embodiments. For the purpose of clarity, representations and descriptions that are not related to the present invention are omitted in the drawings and descriptions.

[0028] In order to have a clearer understanding of the technical features, purposes and beneficial effects of the present invention, the technical solution of the present invention is now described in detail below. Obviously, the implementation cases described are part of the embodiments of the present invention, not all of the embodiments, and cannot be understood as limiting the scope of the implementation of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0029] like Figure 1As shown, the present invention provides an automatic extraction method for engineering quantity based on point cloud data and YOLOv11, comprising the following steps: S1. acquiring point cloud data of a building, preprocessing the acquired point cloud data and generating a three-dimensional point cloud model; S2. constructing a BIM model, and integrating and aligning the point cloud data with the BIM model data; S3. establishing a YOLOv11 model and training the YOLOv11 model; S4. using the trained YOLOv11 model for target detection; S5. correcting the construction position and dimensional accuracy, and automatically calculating the engineering quantity and then outputting it.

[0030] The point cloud data of the building collected in step S1 is obtained specifically through laser scanners, drones or other 3D scanning equipment. Using a laser scanner, the Tianbao TX 3D laser scanner was used in the test project to obtain the laser point cloud of the steel structure. The measurement accuracy of this equipment is better than 1mm, the maximum scanning radius is 120m, and the measurement speed is 976,000 points / s. It can quickly and comprehensively obtain the point cloud data of the indoor and outdoor surfaces of the building. In order to obtain complete point cloud data and avoid obstruction by objects on the construction site, multiple scanning stations are set up, multiple targets are arranged, and three-dimensional laser scanning of components is performed from multiple angles. Each station is guaranteed to contain more than 5 targets. The point cloud data scanned from multiple angles are spliced.

[0031] The preprocessing of the collected point cloud data in step S1 specifically includes: removing noise from the point cloud data using statistical outliers and downsampling through random sampling.

[0032] Affected by factors such as light and emission intensity, the original point cloud data collected by the 3D laser scanner contains noise points. In order to ensure the accuracy of the point cloud data, the point cloud data needs to be denoised. The use of statistical outliers to remove noise from point cloud data specifically includes: by setting the standard deviation parameter and radius value, calculating the threshold Di, when the distance from a point in the neighborhood to point i is greater than Di, it is regarded as an abnormal point and removed; the calculation formula of the threshold Di is: Di = SD + s × σ; where Di represents the noise point judgment threshold; SD represents the standard deviation parameter; s represents the standard deviation coefficient; σ represents the standard deviation of the distance from all points in the neighborhood to point i.

[0033] The point cloud data collected by the scanner is very large, and it is necessary to streamline the sampling points to improve the overall efficiency and facilitate subsequent operations. The random sampling algorithm is used to complete the point cloud data filtering. Random sampling is a simple and effective downsampling method, especially suitable for large-scale data sets or point cloud processing. In the point cloud data downsampling, the random sampling method without replacement is selected. Downsampling by random sampling specifically includes: the original point cloud data set D = {d1, d2, ..., d N}, random sampling sample S = {d i1 , di2 , ..., d iM}, the probability of each sample being selected is P(d i ∈S) = M / N; where i1, i2, ..., i M are all different indices randomly selected from {1,2,...,N}.

[0034] In step S2, the integration and registration of point cloud data and BIM model data specifically includes the following steps: S21. Extracting plane patches through the BIM model and the three-dimensional point cloud model respectively; S22. Extracting four plane sets in the BIM model and substituting them into the four plane sets in the three-dimensional point cloud model for congruence; S23. Calculating the transformation matrix, centroid support, plane support and centroid root mean square error, and applying the transformation matrix; S24. Clustering transformations and sorting the transformations.

[0035] The formula for calculating the root mean square error of the centroid is: Where T is the translation matrix; R is the rotation matrix; qi and pi are the i-th points in the reference point cloud Q and the point cloud to be registered P respectively; N is the number of point clouds.

[0036] The 3D laser scanning point cloud data is located in the world coordinate system, and the parsed BIM point cloud data is located in the local coordinate system, so it is necessary to agree on the coordinate system and use the PCA algorithm and ICP algorithm for registration. Point cloud + BIM data registration algorithm is as follows: Figure 2 As shown in the figure, the ICP algorithm has high requirements for the initial position of the point cloud to be registered, and it is easy to have a local optimal solution. Therefore, the PCA algorithm is used to roughly register the original point cloud to obtain the advantageous initial position. On this basis, the ICP algorithm is used to achieve accurate point cloud registration, effectively solving the problem of local optimal solution for registration. Since the main axis direction of the point cloud to be registered calculated by the PCA algorithm may be reversed, the rotation correction matrix A is established to constrain the rough registration process. Among them, the rotation correction matrix Where: a1, a2, a3 correspond to the x, y, z axes of the coordinate system. Matrix A is only used for spatial rotation of point cloud data. Its non-diagonal elements are all 0, and the diagonal elements ai can only take values ​​of 1 and -1, where 1 indicates the same direction of the axis, and -1 indicates the opposite direction of the axis.

[0037] After the above rough registration is completed, the initial registration position of the point cloud is determined, and the ICP algorithm is further used to iteratively register the point cloud to be registered: E is the iterative calculation error; T is the translation matrix; R is the rotation matrix; qi, pi are the i-th points in the reference point cloud Q and the point cloud to be registered P; N is the number of point clouds.

[0038] The engineering quantity data output in step S5 is structured data.

[0039] It also includes result verification and optimization steps: verifying the accuracy of the engineering quantity data and checking the consistency between the components identified by YOLOv11 and the actual components in the BIM model.

[0040] YOLOv11 is the latest version of Ultralytics YOLO series of real-time object detectors, redefining what is possible with cutting-edge accuracy, speed and efficiency. Building on the impressive progress of previous YOLO versions, YOLO11 introduces significant improvements in architecture and training methods, making it a versatile choice for a variety of computer vision tasks.

[0041] YOLO11 uses an improved backbone and neck architecture and enhances feature extraction capabilities to achieve more accurate target detection and complex task performance. With the advancement of model design, YOLOv11 achieves higher average precision (mAP) on the COCO dataset while using 22% fewer parameters than YOLOv11, thereby improving computational efficiency without affecting accuracy.

[0042] Use the trained model to train the format-converted data. During the training process, improve the generalization ability of the model through data enhancement, hyperparameter adjustment and other means to ensure that YOLOv11 can accurately identify different types of building components.

[0043] Finally, the coordinates, size, category, confidence level and other information of each component are output, and an EXCEL file containing structured data is generated.

[0044] The present invention reduces manual intervention and improves efficiency through automated point cloud data preprocessing and YOLOv11 target detection model. The point cloud data preprocessing steps include using the SOR algorithm to remove noise, the random sampling algorithm to downsample, and the ICP algorithm and PCA algorithm to perform point cloud registration. These steps automatically process a large amount of data and reduce manual operations.

[0045] By utilizing the advanced feature extraction capability of the YOLOv11 model, the present invention improves the accuracy of automatic identification and measurement of building components and materials. YOLOv11 significantly improves the feature extraction capability through the improved trunk and neck architecture, thereby improving the recognition accuracy. The present invention integrates BIM model data with point cloud data, so that the system can adapt to different drawing specifications and symbol standards. The BIM model and point cloud data are docked in the same coordinate system, achieving seamless data fusion and enhancing the system's adaptive capability.

[0046] The present invention solves the integration difficulties between different data sources by integrating point cloud and BIM data. The component information identified by YOLOv11 automatically calculates the engineering quantity of each component, and outputs the result as structured data, which improves the compatibility and ease of use of the system. The YOLOv11 model is optimized for efficiency and speed, providing faster processing speed and meeting real-time requirements. This balance ensures that YOLOv11 is an ideal choice for real-time and large-scale applications, and improves processing speed.

[0047] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. The automatic extraction method of engineering quantity based on point cloud data and YOLOv11 is characterized by: The following steps are involved: S1. Collecting point cloud data of the building, preprocessing the collected point cloud data and generating a three-dimensional point cloud model; S2. Build a BIM model and integrate and align the point cloud data with the BIM model data; S3. Establish a YOLOv11 model and train the YOLOv11 model; S4. Use the trained YOLOv11 model for target detection; S5. Correct the construction position and dimensional accuracy, and automatically calculate the engineering quantity and output it.

2. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 1 is characterized in that: The point cloud data of the building collected in step S1 is specifically obtained by a laser scanner, a drone or other three-dimensional scanning equipment.

3. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 2 is characterized in that: The preprocessing of the collected point cloud data in step S1 specifically includes: removing noise from the point cloud data using statistical outliers and downsampling by random sampling.

4. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 3 is characterized in that: The method of removing noise from point cloud data by using statistical outliers specifically includes: calculating the threshold D by setting the standard deviation parameter and the radius value. i , when the distance from the point in the neighborhood to point i is greater than D i When , it is regarded as an abnormal point and removed; Among them, the threshold D i The calculation formula is: i =SD+s×σ; where D i represents the noise point judgment threshold; SD represents the standard deviation parameter; s represents the standard deviation coefficient; σ represents the standard deviation of the distance from all points in the neighborhood to point i.

5. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 3 is characterized in that: The downsampling by random sampling specifically includes: the original point cloud data set D = {d1, d2, ..., d N }, random sampling sample S = {d i1 , d i2 , ..., d iM }, the probability of each sample being selected is P(d i ∈S) = M / N; where i1, i2, ..., i M are all different indices randomly selected from {1,2,...,N}.

6. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 1, characterized in that: In step S2, the integration and registration of point cloud data and BIM model data specifically includes the following steps: S21. extracting plane patches through the BIM model and the three-dimensional point cloud model respectively; S22. extracting four plane sets from the BIM model and substituting them into the four plane sets in the three-dimensional point cloud model to find congruence; S23. Calculate the transformation matrix, centroid support, plane support, and centroid root mean square error, and apply the transformation matrix; S24. Cluster and sort the transformations.

7. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 6 is characterized in that: The calculation formula for calculating the centroid root mean square error is: Where T is the translation matrix; R is the rotation matrix; qi, pi are the i-th points in the reference point cloud Q and the point cloud to be registered P respectively; N is the number of point clouds.

8. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 1, characterized in that: The engineering quantity data output in step S5 is structured data.

9. The method for automatic extraction of engineering quantities based on point cloud data and YOLOv11 according to claim 1, characterized in that: It also includes result verification and optimization steps: verifying the accuracy of the engineering quantity data and checking the consistency between the components identified by YOLOv11 and the actual components in the BIM model.

Images