Intelligent building refined three-dimensional modeling method

Abstract

The invention provides an intelligent building refined three-dimensional modeling method. The method comprises the steps of image data acquisition, data preprocessing, three-dimensional model construction, hole filling, light and shadow simulation, model precision optimization and the like. According to the method, the unmanned aerial vehicle is used for collecting multi-source data from multiple angles, the coordinate reference established by the total station is matched, the data accuracy is greatly improved, the accuracy of the model is improved, and compared with a traditional three-dimensional modeling method, the method is more efficient and more accurate. Through a plurality of data processing modes, noise in the data is accurately removed, and a hole detection step is provided, so that the three-dimensional model is more complete and higher in accuracy. The method further comprises a shadow simulation step, dynamic shadow simulation can be achieved, and data support is provided for building lighting analysis, energy-saving transformation and other green building applications. Whether the current model needs to be iterated or not can be judged according to the evaluation result, closed-loop quality control and iteration are achieved, and the refinement degree of the model is improved.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, and in particular to a method for refined 3D modeling of smart buildings based on multi-source image enhancement and geometric constraint optimization. Background Technology

[0002] As a crucial vehicle for urban low-carbon development, smart buildings rely on sophisticated 3D modeling as a key technology for promoting energy conservation, emission reduction, and green performance optimization. The construction sector urgently needs to leverage digital means to improve energy efficiency and precisely control carbon emissions. High-precision 3D models can fully recreate building structures, spatial layouts, and external environmental characteristics, providing a data foundation for green scenarios such as natural lighting potential analysis, intelligent photovoltaic equipment deployment, and thermal performance evaluation of building envelopes, guiding passive energy-saving building design. Furthermore, the intelligent operation and maintenance of smart buildings depends on 3D models to achieve real-time energy consumption monitoring and equipment linkage control, thereby dynamically optimizing energy allocation and reducing resource waste. With the deep integration of the Internet of Things, artificial intelligence, and Building Information Modeling (BIM) technologies, 3D modeling is becoming a core link connecting physical buildings and digital twin systems, providing technical support for urban sustainable development goals.

[0003] Smart buildings, through refined 3D modeling technology, upgrade buildings from static physical entities to dynamic digital twins, empowering low-carbon management throughout the building's entire lifecycle. Its role is reflected in three aspects: first, accurately analyzing building space utilization to assist in functional zoning adjustments and facility layout improvements; second, combining models with video surveillance systems to locate safety hazards and simulate emergency response paths; and third, relying on model-driven intelligent operation and maintenance to achieve precise matching of energy supply and demand, reduce carbon emission intensity, and contribute to the green transformation of cities. Unfortunately, while current mainstream photogrammetry technology can acquire building facade information from images, it is limited by single image sources, uneven point cloud density, and noise interference, often resulting in surface damage and geometric distortion in the models. Furthermore, traditional modeling processes lack integration with dynamic lighting characteristics, limiting their application in areas such as the deployment of light-driven equipment and energy efficiency optimization. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a refined 3D modeling method for smart buildings, which solves the problems of model damage, geometric distortion, and limited application scope due to the lack of integration of dynamic lighting characteristics.

[0005] According to an embodiment of the present invention, a method for refined 3D modeling of smart buildings includes the following steps:

[0006] S1. Image Data Acquisition: Design a composite flight path combining vertical and oblique views. The vertical flight path acquires orthophotos of the top, while the oblique flight path surrounds the area to collect details of the facade. Ground control points are set up in the survey area, and coordinate references are obtained using a total station. The UAV simultaneously records imagery, positioning, and attitude data, forming a multi-source dataset with spatiotemporal labels.

[0007] S2. Data preprocessing: The acquired images are processed to correct lens distortion, adjust image grayscale distribution, remove salt-and-pepper noise and unify resolution, providing a high-quality data source for subsequent matching and modeling.

[0008] S3. Constructing a 3D model: Extract feature points from the image and generate descriptors, filter corresponding points, convert 2D image coordinates into 3D coordinates, construct 3D point cloud data, and establish an error objective function.

[0009] S4. Hole Filling: Detects holes and fills them to make the repaired area consistent with the geometry of the original model. The Laplacian smoothing algorithm is used to optimize vertex positions, eliminate surface noise, and improve the overall quality of the model.

[0010] S5. Light and Shadow Simulation: A solar light source simulation module is introduced into the 3D model. Based on the building's geographical location (latitude and longitude) and seasonal parameters, the solar light path in different seasons is simulated through the light analysis algorithm, generating dynamic light and shadow effects on the model surface to intuitively demonstrate the building's lighting performance.

[0011] S6. Model Accuracy Optimization: Evaluate and output the model accuracy. Set up checkpoints on the building surface, compare the total station measured coordinates with the model coordinates, calculate the plane and height errors, and iteratively optimize based on the evaluation results.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] 1. The intelligent building refined 3D modeling method provided by this invention is more efficient and accurate than traditional 3D modeling methods. This method uses drones to collect multi-source data from multiple angles and, together with the coordinate reference established by the total station, greatly enhances the accuracy of the data and improves the accuracy of the model.

[0014] 2. This invention uses multiple data processing methods to accurately remove noise from the acquired image data and provides a hole detection step to repair model holes caused by acquisition gaps, making the three-dimensional model more complete and more accurate.

[0015] 3. The present invention also includes a light and shadow simulation step, which calculates the sun's position based on geographical information and simulates the light source illumination in different seasons, thereby simulating the changes of light on the smart building model, realizing dynamic light and shadow simulation, and providing data support for green building applications such as building lighting analysis and energy-saving renovation.

[0016] 4. This invention can determine whether the current model needs to be iterated based on the evaluation results, thereby achieving closed-loop quality control and iteration and improving the refinement of the model. Attached Figure Description

[0017] Figure 1 This is a flowchart of the three-dimensional modeling method according to an embodiment of the present invention.

[0018] Figure 2 This is a flowchart of the Delaunay triangulation algorithm in step S42 of this embodiment of the invention; Detailed Implementation

[0019] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] like Figure 1 As shown in the figure, this embodiment of the invention provides a method for refined 3D modeling of smart buildings, including the following steps:

[0021] S1. Image Data Acquisition: Design a composite flight path combining vertical and oblique views. The vertical flight path acquires orthophotos of the top, while the oblique flight path surrounds the area to collect details of the facade. Ground control points are set up in the survey area, and coordinate references are obtained using a total station. The UAV simultaneously records imagery, positioning, and attitude data, forming a multi-source dataset with spatiotemporal labels.

[0022] S1.1 Prepare the data acquisition equipment. The data acquisition is based on a low-altitude UAV, equipped with a multi-lens photography device (1 vertical lens + 4 tilt lenses), RTK positioning data and IMU attitude data to achieve spatiotemporal synchronous acquisition of multi-view images. The RTK positioning data includes longitude L, latitude B, and elevation H, and the IMU attitude data includes heading angle φ, pitch angle ω, and roll angle κ.

[0023] S1.2. Ground control points (GCPs) are evenly distributed in the target survey area. Their three-dimensional coordinates (planar position and elevation) are measured using a total station to ensure that the distribution of points is reasonable and covers the key areas of the survey area. At the same time, several check points (ICPs) are set up for accuracy verification. The geometric accuracy of the three-dimensional modeling is evaluated by comparing the measured coordinates with the inverse coordinates of the model.

[0024] S1.3 Calculate key aerial survey parameters. The formula for calculating flight altitude H is:

[0025]

[0026] Where f is the lens focal length, GSD is the ground sampling distance (set to 2-5cm), and α is the sensor pixel size;

[0027] The scale 1 / M is determined by flight altitude and focal length:

[0028]

[0029] S1.4. Using the camera platform already mounted on the UAV, acquire high-precision images of ground targets from five angles (vertical, front, back, left, and right) representing different directions of ground movement, and store them as a timestamped dataset D = {I} t ,(L t B t H t ,φ t ,ω t ,κ t )}.

[0030] S2. Data Preprocessing: The acquired images are processed to correct lens distortion, adjust image grayscale distribution, remove salt-and-pepper noise, and unify resolution, providing a high-quality data source for subsequent matching and modeling.

[0031] S2.1. The Brown-Conrady model is used to correct radial (barrel / pincushion) and tangential distortions. The pixel coordinate correction formula is as follows:

[0032]

[0033] Where r 2 =x 2 +y 2 k1 and k2 are radial distortion coefficients, and p1 and p2 are tangential distortion coefficients, which are obtained through calibration using a checkerboard calibration plate;

[0034] S2.2 Utilize the Wallis filter to adjust the smart building images with significant differences in pixel grayscale value distribution, thereby promoting a linear distribution of grayscale values ​​on the target image:

[0035] h0(x',y')=αh1(x',y')+β

[0036] In the formula, h0(x',y') represents the gray value at pixel (x',y') of the smart building image after color and light equalization processing; h1(x',y') represents the original gray value; α represents the linear transformation addition coefficient; β represents the linear transformation multiplication coefficient;

[0037] S2.3. Median filtering algorithm (3×3 window) is used to remove salt-and-pepper noise. Combined with bicubic interpolation, the image resolution is unified to 200dpi to ensure the scale consistency of subsequent matching and modeling and avoid feature extraction deviation caused by resolution differences.

[0038] S3. Constructing a 3D model: Extract feature points from the image and generate descriptors, filter corresponding points, convert 2D image coordinates into 3D coordinates, construct 3D point cloud data, and establish an error objective function.

[0039] S3.1 Use the Harris corner detection algorithm to extract image feature points, and combine it with the SIFT algorithm to generate feature descriptors. The formula is D=SIFT(I,σ,θ), where σ is the scale space factor and θ is the gradient direction, to ensure that the feature points have scale and rotation invariance.

[0040] S3.2. A bidirectional matching strategy (A→B and B→A) is used to filter points with the same name. The matching cost function is:

[0041]

[0042] Where p and q are feature points in images I1 and I2, respectively; d(p) and d(q) are SIFT descriptors of the feature points, retaining point pairs with C < 0.6;

[0043] S3.3. Convert two-dimensional coordinates to three-dimensional coordinates using collinearity equations. Assume that in the image space coordinate system, there exists an image with coordinates (x, y, -f), and in the auxiliary image space coordinate system, the coordinates are (x, y, -f). δ ,Y δ Z δ If the two are denoted by , then the relationship between them can be expressed by the following formula:

[0044]

[0045] Where R is the rotation matrix, expressed as follows:

[0046]

[0047] The equations for collinearity are:

[0048]

[0049] Where (x, y) represents the image plane coordinate system coordinates with the image principal point as the origin, (X, Y, Z) represents the object space coordinates corresponding to the ground target point, and f is the image principal distance parameter in the photogrammetric interior orientation elements;

[0050] S3.4 Establish the error objective function:

[0051]

[0052] Among them, (υ x ,υ y ) represents the residual error of the measured image points on each image, ΔX S ΔY S ΔZ S , Δω and Δκ are the correction values ​​for the elements, a ij It is a function of the three attitude angles of the image's exterior orientation element.

[0053] S4. Hole Filling: Detects holes and fills them to ensure the repaired area is consistent with the original model geometry. The Laplacian smoothing algorithm is used to optimize vertex positions, eliminate surface noise, and improve the overall quality of the model.

[0054] S4.1. Holes are detected by performing a comprehensive visualization check on the model and by analyzing point cloud density. In the point cloud density analysis, a reasonable point cloud density threshold is set. When the point cloud density of a certain area is lower than the threshold, it is marked as an area that may contain holes. At the same time, the geometric structure information of the model is combined to determine whether these areas are indeed holes, thereby ensuring the accuracy of hole detection.

[0055] S4.2 For detected voids, the following method shall be used: Figure 2 The Delaunay triangulation algorithm shown is used for filling. This algorithm fills the hole area by constructing triangles between the boundary points of the hole. In the process of constructing triangles, the Delaunay triangulation principle is followed, that is, the circumcircle of any triangle does not contain other points, which ensures that the quality of the generated triangle mesh is good and that the repaired area is consistent with the surrounding model in terms of geometry.

[0056] S4.3. Optimize the entire model using the Laplace smoothing algorithm, the formula is as follows:

[0057]

[0058] Among them, υ i It is the original vertex position, υ i ' is the optimized vertex position, λ takes a value of 0.2, is a smoothing factor used to control the degree of smoothing, and N(i) represents the set of adjacent vertices of vertex i;

[0059] By iterating multiple times (5-10 times), noise on the model surface is eliminated, making the model smoother and improving its quality.

[0060] S5. Light and Shadow Simulation: A solar light source simulation module is introduced into the 3D model. Based on the building's geographical location (latitude and longitude) and seasonal parameters, the solar path in different seasons is simulated through a light analysis algorithm, generating dynamic light and shadow effects on the model surface to intuitively demonstrate the building's lighting performance.

[0061] S5.1 Obtain the latitude and longitude information of the building based on its geographical location, and determine the seasonal parameters required for the current modeling; different seasons correspond to different solar declination angles. Using these latitude, longitude, and seasonal parameters, combine them with astronomical formulas for calculating the sun's position, such as those in spherical trigonometry:

[0062]

[0063] Where h is the solar altitude angle, δ is the local latitude, ω is the solar declination angle, and ω is the hour angle.

[0064] By calculating the solar altitude angle and azimuth angle, we can determine the changes in the sun's position in the sky during different seasons, providing basic data for subsequent simulations of the sun's path.

[0065] S5.2 Add a light source that changes with the seasons. Determine the direction of the light based on the sun's position (elevation angle, azimuth angle), project parallel light rays onto the surface of the building model, calculate the intersection point of the light rays with the 3D model, and lay the light rays at the intersection point. If there is an obstruction between a point and the sun (other buildings or structures), it is marked as a shadow area and the light intensity is reduced to zero. This can simulate the changes of light on smart buildings and provide a basis for the installation of light-driven equipment.

[0066] S6. Model Accuracy Optimization: Evaluate and output the model accuracy. Set up checkpoints on the building surface, compare the total station measured coordinates with the model coordinates, calculate the plane and height errors, and iteratively optimize based on the evaluation results.

[0067] S6.1 Set up checkpoints on the outer surface of the building, use a total station to measure their true three-dimensional coordinates, compare them with the coordinates of the corresponding points on the model, and calculate the plane error and elevation error.

[0068] S6.2 Optimize the model based on the evaluation results. If the error exceeds the set threshold (plane ≤ 8cm, elevation ≤ 10cm), return to the model optimization step for iterative processing until the accuracy meets the standard, ensuring that the model meets the accuracy requirements of intelligent building fine modeling.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for refined 3D modeling of smart buildings, characterized in that, Includes the following steps: S1. Image Data Acquisition: Set up ground control points in the area where the target building is located, obtain coordinate references through a total station, and collect and record image, positioning and attitude data simultaneously through a drone to form a multi-source dataset with spatiotemporal labels. S2. Data preprocessing: The acquired image data is processed to correct lens distortion, adjust image grayscale distribution, remove salt-and-pepper noise and unify resolution, providing a high-quality data source for subsequent matching and modeling. S3. Constructing a 3D model: Extract feature points from the processed image data and generate descriptors, filter corresponding points, convert 2D image coordinates into 3D coordinates, construct a 3D point cloud data model, and establish an error objective function. S4. Hole Filling: Detect holes in the 3D model and fill the detected holes to make the repaired area consistent with the geometry of the original model. The Laplacian smoothing algorithm is used to optimize vertex positions, eliminate surface noise, and improve the overall quality of the model. S5. Light and Shadow Simulation: A solar light source simulation module is introduced into the 3D model. Based on the building's geographical location and seasonal parameters, the solar path in different seasons is simulated through a light analysis algorithm, generating dynamic light and shadow effects on the model surface to intuitively demonstrate the building's lighting performance. S6. Model accuracy optimization: Set up checkpoints on the surface of the target building, compare the total station measured coordinates with the model coordinates, calculate the plane and height errors, and perform iterative optimization based on the evaluation results.

2. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that: In step S1, when the UAV collects data, it adopts a combination of vertical and oblique flight paths. The vertical flight path acquires an orthophoto of the top of the target, while the oblique flight path surrounds the target to collect facade details.

3. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that, Step S1 specifically includes the following sub-steps: S1.1 Prepare the data acquisition equipment. The data acquisition is based on a low-altitude UAV, equipped with multi-lens photography equipment, RTK positioning data and IMU attitude data to achieve spatiotemporal synchronous acquisition of multi-view images. The RTK positioning data includes longitude L, latitude B and elevation H, and the IMU attitude data includes heading angle φ, pitch angle ω and roll angle κ. S1.

2. Ground control points are evenly distributed in the target survey area, and their three-dimensional coordinates are measured using a total station to ensure that the distribution of points is reasonable and covers the key areas of the survey area. At the same time, several checkpoints are set up for accuracy verification. The geometric accuracy of the three-dimensional modeling is evaluated by comparing the measured coordinates with the inverse coordinates of the model. S1.3 Calculate key aerial survey parameters. The formula for calculating flight altitude H is: Where f is the lens focal length, GSD is the ground sampling distance, and α is the sensor pixel size; The scale 1 / M is determined by flight altitude and focal length: S1.

4. Using the camera platform already mounted on the UAV, acquire high-precision images of ground targets from five angles (vertical, front, back, left, and right) representing different directions of ground movement, and store them as a timestamped dataset D = {I} t ,(L t B t H t ,φ t ,ω t ,κ t )}.

4. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that, Step S2 includes the following sub-steps: S2.

1. The Brown-Conrady model is used to correct radial and tangential distortions. The pixel coordinate correction formula is as follows: Where r 2 =x 2 +y 2 k1 and k2 are radial distortion coefficients, and p1 and p2 are tangential distortion coefficients, which are obtained through calibration using a checkerboard calibration plate; S2.2 Utilize the Wallis filter to adjust the smart building images with significant differences in pixel grayscale value distribution, thereby promoting a linear distribution of grayscale values ​​on the target image: h0(x',y')=αh1(x',y')+β In the formula, h0(x',y') represents the gray value at pixel (x',y') of the smart building image after color and light equalization processing; h1(x',y') represents the original gray value; α represents the linear transformation addition coefficient; β represents the linear transformation multiplication coefficient; S2.

3. Median filtering algorithm is used to remove salt-and-pepper noise, and bicubic interpolation is used to unify the image resolution to 200 dpi to ensure the scale consistency of subsequent matching and modeling, and to avoid feature extraction deviation caused by resolution differences.

5. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that, Step S3 includes the following sub-steps: S3.1 Use the Harris corner detection algorithm to extract image feature points, and combine it with the SIFT algorithm to generate feature descriptors. The formula is D=SIFT(I,σ,θ), where σ is the scale space factor and θ is the gradient direction, to ensure that the feature points have scale and rotation invariance. S3.

2. A bidirectional matching strategy (A→B and B→A) is used to filter points with the same name. The matching cost function is: Where p and q are feature points in images I1 and I2, respectively; d(p) and d(q) are SIFT descriptors of the feature points, retaining point pairs with C < 0.6; S3.

3. Convert two-dimensional coordinates to three-dimensional coordinates using collinearity equations. Assume that in the image space coordinate system, there exists an image with coordinates (x, y, -f), and in the auxiliary image space coordinate system, the coordinates are (x, y, -f). δ ,Y δ Z δ If the two are denoted by , then the relationship between them can be expressed by the following formula: Where R is the rotation matrix, expressed as follows: The equations for collinearity are: Where (x, y) represents the image plane coordinate system coordinates with the image principal point as the origin, (X, Y, Z) represents the object space coordinates corresponding to the ground target point, and f is the image principal distance parameter in the photogrammetric interior orientation elements; S3.4 Establish the error objective function: Among them, (υ x ,υ y ) represents the residual error of the measured image points on each image, ΔX S ΔY S ΔZ S , Δω and Δκ are the correction values ​​for the elements, a ij It is a function of the three attitude angles of the image's exterior orientation element.

6. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that, Step S4 includes the following sub-steps: S4.

1. Holes are detected by performing a comprehensive visualization check on the model and by analyzing point cloud density. In the point cloud density analysis, a reasonable point cloud density threshold is set. When the point cloud density of a certain area is lower than the threshold, it is marked as an area that may contain holes. At the same time, the geometric structure information of the model is combined to determine whether these areas are indeed holes, thereby ensuring the accuracy of hole detection. S4.2 For the detected holes, the Delaunay triangulation algorithm is used for filling. This algorithm fills the hole area by constructing triangles between the boundary points of the hole. In the process of constructing triangles, the principle of Delaunay triangulation is followed, that is, the circumcircle of any triangle does not contain other points, ensuring that the quality of the generated triangular mesh is good, so that the repaired area is consistent with the surrounding model in terms of geometry. S4.

3. Optimize the entire model using the Laplace smoothing algorithm, the formula is as follows: Among them, υ i It is the original vertex position, υ i ' is the optimized vertex position, λ takes a value of 0.2, is a smoothing factor used to control the degree of smoothing, and N(i) represents the set of adjacent vertices of vertex i; Through multiple iterations, noise on the model surface is eliminated, making the model smoother and improving its quality.

7. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that, Step S5 has the following sub-steps: S5.1 Obtain the latitude and longitude information of the building based on its geographical location, and determine the seasonal parameters required for the current modeling; different seasons correspond to different solar declination angles. Using these latitude, longitude, and seasonal parameters, combine them with astronomical formulas for calculating the sun's position, such as those in spherical trigonometry: Where h is the solar altitude angle, δ is the local latitude, ω is the solar declination angle, and ω is the hour angle. By calculating the solar altitude angle and azimuth angle, we can determine the changes in the sun's position in the sky during different seasons, providing basic data for subsequent simulations of the sun's path. S5.2 Add a light source that changes with the seasons. Determine the direction of the light according to the position of the sun, project parallel light rays onto the surface of the building model, calculate the intersection point of the light rays with the 3D model, and lay the light rays at the intersection point. If there is a blockage between a point and the sun, it is marked as a shadow area and the light intensity is reduced to zero. This can simulate the changes of light on the smart building and provide a basis for the installation of light-driven equipment.

8. The method for refined 3D modeling of smart buildings as described in claim 1, characterized in that, Step S6 includes the following sub-steps: S6.1 Set up checkpoints on the outer surface of the building, use a total station to measure their true three-dimensional coordinates, compare them with the coordinates of the corresponding points on the model, and calculate the plane error and elevation error. S6.2 Optimize the model based on the evaluation results. If the error exceeds the set threshold, return to the model optimization step for iterative processing until the accuracy meets the standard, ensuring that the model meets the accuracy requirements of intelligent building fine modeling.

Images