Method for optimizing layout of indoor three-dimensional scene based on wall surface features

Abstract

The application discloses a wall feature-based indoor three-dimensional scene layout optimization method, which comprises the following steps: firstly, obtaining furniture attributes of a scene from a data set, and structurally representing wall information; secondly, based on the furniture attributes and the wall information, dividing the indoor three-dimensional scene into a plurality of local scene units with regular furniture as the center; then, using a diffusion model to optimize the layout attributes of the furniture in the local scene units; finally, converting the local three-dimensional scene information obtained after the optimization into a global three-dimensional scene, and performing adversarial optimization to complete the indoor three-dimensional scene layout optimization. The wall structure extraction module based on a two-dimensional overhead view retains the wall structure as a spatial boundary constraint, enhances the spatial structure perception ability of the model, can better capture the fine-grained spatial relationship of the furniture layout, and thus improves the accuracy of the layout optimization.

Description

Technical Field

[0001] This invention relates to the task of optimizing the layout of indoor 3D scenes, and in particular to a method for optimizing the layout of indoor 3D scenes based on wall features. Background Technology

[0002] In fields such as game development, film and television production, and interior design, the technology for generating 3D interior scene layouts has demonstrated significant application value. By automatically generating high-quality virtual 3D interior scenes, this technology not only significantly reduces the cost of scene layout construction in film, television, and gaming, but also provides strong support for interior design, promoting a more intelligent and efficient design process.

[0003] However, existing methods face two major challenges: existing generative models do not adequately consider room layout and are difficult to accurately model the spatial relationships between furniture and between furniture and walls; the number of existing indoor 3D scene datasets is limited, and the scarcity of training data often leads to model overfitting, making the scarcity of training data a key bottleneck restricting the improvement of model performance.

[0004] The 3D-FRONT dataset, as a large-scale indoor 3D scene dataset, is widely used due to its rich indoor 3D scene models and professional furniture layout schemes. However, even such a high-quality dataset has some errors that may cause biases in models trained directly on the dataset, thereby reducing the spatial rationality and visual realism of the generated indoor 3D scenes. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a wall feature-based indoor 3D scene layout optimization method (WOSG), generating scenes with good spatial consistency and overall expressiveness. Users only need to input an original scene to be optimized (including room structure and furniture layout) to obtain an optimized scene. Experimental results show that this invention maintains the global style properties of the source scene, significantly improves the modeling ability of fine-grained spatial relationships between furniture, and the generated indoor 3D scenes exhibit clear advantages in terms of layout rationality and visual quality.

[0006] To accurately model the spatial relationships between furniture and between furniture and walls, this invention considers the strong constraint of walls in the scene and designs a wall extraction module. By constructing wall objects in the input scene data, the placement of furniture can be optimized. To overcome the limitations of the scale and diversity of the dataset, this invention focuses on local scene optimization and proposes a local-global optimization strategy. This strategy localizes the complete scene, enriching the training data while improving the model's perception of details.

[0007] First, a wall structure extraction method based on a 2D top-view is used to represent scene wall constraints using 2D corner points. The obtained wall constraints are then combined with furniture information to construct local scenes and expand the dataset. A U-Net diffusion model is trained using the expanded dataset, and finally, the resulting locally optimized scenes are globally integrated.

[0008] A local-global indoor 3D scene layout optimization method based on wall features includes the following steps:

[0009] Step 1: Obtain the furniture attributes of the scene from the dataset and represent the wall information in a structured manner;

[0010] Step 2: Based on furniture attributes and wall information, divide the interior 3D scene into several local scene units centered around regular furniture;

[0011] Step 3: Use the diffusion model to optimize the layout attributes of furniture in local scene units;

[0012] Step 4: Convert the optimized local 3D scene information into a global 3D scene and perform adversarial optimization to complete the indoor 3D scene layout optimization.

[0013] Beneficial effects of this invention:

[0014] 1. A wall structure extraction module based on a two-dimensional top view. This method preserves the wall structure as a spatial boundary constraint, enhancing the model's ability to perceive the spatial structure.

[0015] 2. A dataset augmentation method based on scene localization significantly expands the training data scale, effectively improves the model's generalization ability, enables the model to better capture the fine-grained spatial relationships of furniture layout, thereby improving the accuracy of layout optimization and preventing overfitting.

[0016] 3. A local scene optimization module based on the U-Net diffusion model: This method's feature encoder abstracts the input features step-by-step through multiple convolutions and residual blocks, mapping them to a high-dimensional semantic space; the corresponding decoder reconstructs the furniture's attribute information gradually through upsampling operations and skip connections. The cross-attention mechanism significantly improves the accuracy of feature extraction and reconstruction by modeling the correlations between different features. The time embedding module dynamically provides temporal information to the network, enabling the model to adaptively adjust feature representations according to different time steps. Attached Figure Description

[0017] Figure 1 Here is a diagram of the algorithm structure;

[0018] Figure 2 A top-down mask of the scene;

[0019] Figure 3 This is a schematic diagram of scene localization;

[0020] Figure 4 This is a schematic diagram of the network structure. Detailed Implementation

[0021] like Figure 1 As shown, the indoor 3D scene layout optimization method based on wall features includes the following steps: 1. Obtain the furniture and wall attributes of the scene from the dataset; 2. Extract and structure the wall information from the input scene; 3. Divide the complete indoor 3D scene into several local scene units centered on regular furniture; 4. Optimize the layout attributes of furniture in the local indoor 3D scene using a diffusion model; 5. Convert the optimized local 3D scene information into a global 3D scene; 6. Perform network adversarial optimization.

[0022] Step 1, obtain the furniture attributes of the scene:

[0023] This step requires defining an indoor 3D scene data representation and obtaining the corresponding representation from the dataset.

[0024] The furniture's positional information is defined using a world coordinate system, with the geometric center of the furniture model's 3D bounding box serving as the positional reference. Therefore, this algorithm focuses on the horizontal plane when modeling furniture positional information. Coordinates, their vertical direction ( The axis height can be determined by the dimensions of the furniture model. Accurate calculation yields, i.e. .

[0025] The furniture dimensions are modeled using a 3D bounding box (length × width × height); the furniture orientation is represented parametrically using angles in the world coordinate system, specifically... Taking the positive axis as the reference, the rotation angle is determined by rotating counterclockwise. The orientation of furniture is defined using this method. Furniture category information is represented using one-hot encoding, with different categories corresponding to discrete vectors in the semantic space. Because different rooms have different functional attributes, the number and types of furniture categories also vary. For example, in a living room scenario, furniture categories include sofas, TV cabinets, etc., totaling 23 types, while in a bedroom scenario, there are 19 types, including beds, wardrobes, etc.

[0026] Step 2, Wall Information Extraction:

[0027] After obtaining the furniture information, the wall information needs to be extracted.

[0028] In the indoor 3D scene dataset, the wall model is controlled by several corner points. However, these corner points have the following problems in practical applications: (1) the corner point coordinates are incorrect, affecting direct use; (2) redundant corner points do not make a substantial contribution to scene understanding; (3) the distance between adjacent corner points is too small to form an effective spatial constraint.

[0029] However, the two-dimensional top view and corner data are highly correlated, so the three-dimensional problem can be reduced to a two-dimensional plane using top-view projection. First, the two-dimensional outline of the room is extracted, and then the corner points of the walls are precisely matched and their positions are adjusted through an iterative nearest-point algorithm, finally obtaining a wall structure that meets the requirements.

[0030] When processing scene corner points, the 3D scene corner points are first projected onto a 2D plane to form a set of points to be registered. ,in Refers to the first One registration point; simultaneously, a mask image is obtained based on the top-down planar view of the 3D scene, such as... Figure 2 As shown, the reference point set is extracted. ,in Refers to the first There are several reference points. In each iteration, the algorithm adjusts the value based on the currently estimated rigid transformation matrix. Pair set Perform the update. The update process includes the following steps:

[0031] First, calculate the least squares error; the goal is to find a rigid transformation matrix. (in It is a rotation matrix. (is a displacement vector), making the point set with point set The distance between them is minimized. The error function can be expressed as:

[0032]

[0033] exist middle, This indicates that the points to be registered are transformed by the transformation matrix. Position after rotation This represents the displacement vector.

[0034] Further optimize the transformation matrix by minimizing the error function. To obtain the optimal rotation matrix and displacement vector Then update the point set by applying the calculated transformation matrix to the point set. Update:

[0035]

[0036] Finally, iterate and repeat the above steps until the error function is reached. Converging to a threshold indicates that the matching between the two point sets has reached its optimal level. Through multiple iterations, the ICP algorithm can continuously adjust the position of the wall corner points to align them with the target wall point set, thereby achieving accurate wall extraction.

[0037] During the wall extraction process, a complete geometric representation system was established, including three core parameters: position, rotation, and size. The corner data obtained through the above steps allows for the creation of an accurate wall description model. For each wall, its shape can be represented by the following parameters:

[0038] First, there are the dimensional parameters; the dimensions of a wall are usually expressed in terms of its length. ,thickness and height To describe. For a wall, its length It can be calculated using the following formula:

[0039]

[0040] in and These are two adjacent corner points in space. The height and thickness of the walls are standardized to the same value in this model to simplify calculations. Secondly, there's the rotation angle; the rotation angle of the walls... This describes the rotation of the wall relative to the reference coordinate system. This angle can be calculated from the relative positions of the two endpoints of the wall.

[0041]

[0042] in and These are the coordinates of the two endpoints of the wall.

[0043] Finally, there are the location parameters. The position of the wall can be determined by comparing the coordinates of adjacent corner points. and OK. First, calculate the horizontal coordinates of the wall's center:

[0044]

[0045] Considering that the wall may rotate in the world coordinate system, the rotation angle of the wall needs to be taken into account in order to transform the center coordinates to the world coordinate system. Assume the bottom of the wall is in contact with the ground, and the height is... along If distributed along the axis, its position parameters in three-dimensional space can be expressed as:

[0046]

[0047] in The coordinates are taken as This indicates the center height of the wall. By determining the aforementioned dimensions, rotation angles, and positional parameters, the geometry of the wall is fully described, providing accurate data support for subsequent spatial analysis and model construction.

[0048] Step 3, Local Scene Construction:

[0049] Only after obtaining the two-dimensional wall information can the scene data be complete. This step requires localized processing of the scene data.

[0050] This step divides the complete 3D indoor scene into several local scene units centered around regular furniture, enabling localized processing of scene data.

[0051] First, a comprehensive analysis of the standard furniture in the interior scene is performed, establishing spatial relationships by calculating the two-dimensional Euclidean distance between the furniture. Specifically, assuming the currently selected standard furniture... The position is Then furniture and furniture Two-dimensional Euclidean distance between It can be represented as:

[0052]

[0053] Based on a preset distance threshold (Unit distance), the criteria for determining the local scene: if Then determine the furniture Belongs to furniture A localized scene centered on the subject. For example... Figure 3 As shown, the first column represents a two-dimensional global schematic diagram, while the second and third columns represent corresponding two-dimensional local schematic diagrams. This distance measurement method can effectively capture spatial combinations with functional relationships (such as the spatial configuration relationship between a bed and a bedside table), providing basic spatial units for subsequent layout optimization.

[0054] To improve the model's training performance and generalization ability, additional filtering conditions were added to the obtained local scenes. The total number of regular furniture and walls in each local scene was set. Must meet The conditional constraints of this threshold range optimize the model training effect by balancing the richness of scene information and computational complexity. This ensures that enough furniture interaction relationships are retained for feature learning, while avoiding feature dilution caused by overly complex scenes, thereby improving the model's generalization ability in real-world environments.

[0055] For local scenes that meet the constraints, coordinate transformation and rotation operations are performed. First, for each piece of furniture within the local scene, the difference in its planar coordinates relative to the central furniture within the local scene is calculated. Assume the furniture... The position is And central furniture The position is Then furniture The planar coordinates relative to the center furniture are:

[0056]

[0057] Meanwhile, in this algorithm, the coordinate rotation of all furniture is based on the world coordinate origin of the room. To unify the orientation of the furniture, the orientation of the central furniture is fixed at "up," and then the coordinates are rotated using a rotation matrix. Rotate, specifically, if the angle of the furniture is It can be done through rotation matrix Transform its coordinates to new coordinates :

[0058]

[0059] Where the rotation matrix Defined as:

[0060]

[0061] This transformation converts the furniture's coordinates to a unified direction, making the relative position of the furniture clearer in subsequent calculations.

[0062] Step 4, Local Diffusion Optimization:

[0063] After constructing the local scene, the model needs to be used to optimize the layout through diffusion.

[0064] This step proposes a diffusion model based on the U-Net architecture, which aims to optimize the layout properties of furniture in local indoor 3D scenes.

[0065] As a breakthrough technology in generative artificial intelligence, the diffusion model's iterative denoising mechanism effectively captures the characteristics of data distribution, generating visual content with high fidelity and rich detail. In 2D image generation tasks, the diffusion model demonstrates excellent generalization ability, capable of generating diverse image content. When transferred to the task of generating interior 3D scene layouts, the diffusion model can generate interior 3D scenes with diverse styles and reasonable layouts based on a given room structure. However, the diffusion model has limitations in its ability to control furniture pose parameters. The layout of interior 3D scenes is highly dependent on the position, orientation, and spatial relationship of furniture with other furniture. Therefore, compared to 2D image generation, the task of interior 3D scene layout generation has more stringent requirements for furniture spatial arrangement.

[0066] The scene attributes optimized by the model in this invention mainly include location. ,size Orientation and categories In order to effectively process input furniture features, such as Figure 4 As shown, the diffusion model employs several key modules: a feature encoder, a feature decoder, a cross-attention mechanism, a self-attention mechanism, and residual blocks. The feature encoder abstracts the input scene attributes step-by-step through multiple convolutional layers and residual blocks, mapping them to a high-dimensional semantic space. The corresponding decoder reconstructs the furniture's attribute information gradually through upsampling operations and skip connections. The cross-attention mechanism significantly improves the accuracy of feature extraction and reconstruction by modeling the correlations between different features. The time embedding module dynamically provides temporal information to the network, enabling the diffusion model to adaptively adjust its feature representation according to different time steps.

[0067] In summary, this module takes a 15-dimensional attribute matrix of furniture in a local scene as input (including regular furniture and walls as special furniture). After being mapped by the encoder, conditional control and loss constraints are applied during the denoising process to optimize the layout. The wall attributes remain unchanged as a fixed context, which maintains the integrity of the scene structure and avoids redundant parameter updates, thus ensuring training stability.

[0068] Step 5, Global Scene Integration:

[0069] After obtaining the optimized local scene, it is necessary to integrate the local scene.

[0070] This step will convert the optimized local 3D scene information into a global 3D scene to present the final result of the algorithm.

[0071] Let the position of the central furniture in the local coordinate system be... Its rotation angle in the local coordinate system is This converts the position and angle into a global coordinate system.

[0072] First, there's the transformation of position, specifically the position information of the central furniture in the local coordinate system. Requires rotation matrix and displacement vector Transform to the global coordinate system. Assume the furniture's position in the global coordinate system is... Therefore, the transformation formula is:

[0073]

[0074] in It is a rotation matrix describing the rotation from the current local coordinate system to the global coordinate system. It is the rotation angle of the local coordinate system relative to the global coordinate system.

[0075] Then, the angles of other furniture within the local scene are transformed. In the local coordinate system, the rotation angle of the furniture is... This is usually relative to the local coordinate system, but in the global coordinate system, the angle needs to be adjusted appropriately to make it consistent. Let the rotation angle of the furniture in the global coordinate system be... Then its transformation formula is:

[0076]

[0077] in It is the rotation angle of the local coordinate system relative to the global coordinate system.

[0078] To recover information from the local coordinate system to the global coordinate system, the algorithm first uses an inverse normalization operation to restore the local data to the range of the global coordinate system. Specifically, by recovering furniture attributes such as position and angle from the local data, this local information can be mapped back to the global coordinate system. The recovery process includes two main steps: displacement recovery and angle recovery.

[0079] The first step is displacement recovery, where the position in the local coordinate system is adjusted by offsetting it relative to the origin of the local coordinate system, and further compensated using the offset from the global coordinate system.

[0080] Secondly, angle restoration is performed. For rotation angles, the restoration process uses a rotation matrix to adjust the angle. Let the rotation angle of the furniture in the local coordinate system be... The angle in the global coordinate system is .

[0081] In practice, the process of restoring the rotation angle is closely related to the process of restoring the position, and usually requires the use of a reverse rotation matrix for adjustment.

[0082] To ultimately integrate all local results into a global scene, the algorithm first integrates all local scenes...

[0083] The algorithm merges the position, angle, size, and category labels of the furniture. Since each piece of furniture may generate multiple optimized results after local reconstruction, the algorithm averages these merged results to obtain the final position, angle, and size information. By integrating this information, a complete global scene containing information about all furniture is finally obtained. The core formula of this process is:

[0084]

[0085] For each piece of furniture Its global information Calculated in the following way:

[0086]

[0087] in This represents the global scene and contains information about all furniture (including location, angle, size, and category label). This indicates the number of results for each piece of furniture after being recreated in a local scene. Indicates the first The furniture is in the first Information from partial restoration.

[0088] Through this processing method, local scene information is effectively integrated to generate a global scene containing complete furniture information.

[0089] Step 6, Countermeasure Optimization:

[0090] To effectively achieve joint optimization of multiple furniture attributes, this paper combines various loss functions, the specific loss function settings are as follows:

[0091] (1) Mean square error loss:

[0092] The mean squared error (MSE) loss is applied in the noise recovery stage of this model. This loss ensures that the network can accurately reconstruct the original layout during the diffusion process by minimizing the difference between the predicted noise and the actual noise. Specifically, the MSE loss is calculated as follows:

[0093]

[0094] in Indicates the model at time step The noise in the prediction; This represents the actual noise added during the diffusion process; it represents the noise added at the time step. Noisy scene data; Indicates conditional information, such as furniture category;

[0095] (2) Displacement and orientation loss:

[0096] The displacement loss of furniture is predicted using the L2 norm. With real location Euclidean distance:

[0097]

[0098] Considering the periodicity of angles, the angle loss for furniture uses cosine similarity loss to avoid the 2π periodicity problem:

[0099]

[0100] (3) Loss on conventional furniture transactions:

[0101] To improve the accuracy of bounding box prediction for standard furniture, the cross-union loss for standard furniture is designed to optimize the overlapping areas of bounding boxes between standard furniture pieces. The specific formula is as follows:

[0102]

[0103] in and These are the bounding boxes for the predicted different types of conventional furniture. and Let be the areas of their intersection and union, respectively.

[0104] Wall intersection ratio:

[0105] To effectively model the spatial constraints between conventional furniture and walls in an interior 3D scene, this paper designs a wall intersection-union ratio (IUU) loss function. This loss function quantifies the spatial overlap between the bounding boxes of conventional furniture and the bounding boxes of walls, guiding the model to learn furniture layout patterns that conform to building codes. The specific formula is as follows:

[0106]

[0107] in Indicates the first Predicted bounding boxes for a typical piece of furniture. This represents the actual bounding box of the corresponding wall.

Claims

1. A method for optimizing the layout of an indoor 3D scene based on wall features, characterized in that, Includes the following steps: Step 1: Obtain the furniture attributes of the scene from the dataset and represent the wall information in a structured manner; Step 2: Based on furniture attributes and wall information, the interior 3D scene is divided into several local scene units centered around regular furniture. The specific implementation process is as follows: First, the furniture in the indoor scene is traversed and analyzed. Spatial relationships between furniture are established by calculating the two-dimensional Euclidean distance between them. Assuming the currently selected furniture... Location Then furniture and furniture The two-dimensional Euclidean distance between them is expressed as ; Based on a preset distance threshold Construct the judgment criteria for local scenes: if Then determine the furniture Belongs to furniture A local scene centered on the subject; For local scenes that meet the constraints, coordinate transformation and rotation operations are performed: First, for each piece of furniture within the local scene, the difference in its planar coordinates relative to the central furniture within the local scene is calculated; assuming the furniture... The position is And central furniture Location Then furniture The planar coordinates relative to the center furniture are and The corresponding coordinates are subtracted to obtain the result; Meanwhile, the coordinate rotation of all furniture is based on the world coordinate origin of the room. To unify the orientation of the furniture, the orientation of the central furniture is fixed upwards, and the coordinates are adjusted using a rotation matrix. Rotate the furniture if the angle is... Through rotation matrix Transform its coordinates to new coordinates ; Step 3: Use the diffusion model to optimize the layout attributes of furniture in local scene units; Step 4: Convert the optimized local 3D scene information into a global 3D scene and perform adversarial optimization to complete the indoor 3D scene layout optimization.

2. The indoor three-dimensional scene layout optimization method based on wall features according to claim 1, characterized in that, The specific implementation process for obtaining the furniture attributes of the scene is as follows: The furniture's positional information is defined using a world coordinate system, with the geometric center of the furniture model's 3D bounding box serving as the positional reference. The modeling of furniture positional information focuses on the horizontal plane. The coordinates, whose vertical height is determined by the height of the furniture model dimensions. Calculated; The furniture's dimensions are modeled using 3D bounding boxes; the furniture's orientation is represented parametrically using angles in a world coordinate system. Taking the positive axis as the reference, the rotation angle is determined by rotating counterclockwise. Define the orientation of the furniture; the category information of the furniture is represented by one-hot encoding, and different categories of furniture correspond to discrete vectors in the semantic space. The number and types of furniture categories are different.

3. The indoor three-dimensional scene layout optimization method based on wall features according to claim 2, characterized in that, The specific implementation process of the structured representation of wall information is as follows: After obtaining the furniture information, the wall model in the indoor 3D scene dataset is controlled by several corner points; First, the two-dimensional outline of the room is extracted. Then, the corner points of the walls are matched and their positions are adjusted by the iterative nearest point algorithm to finally obtain a wall structure that meets the requirements. When processing scene corner points, the 3D scene corner points are first projected onto a 2D plane to form a set of points to be registered. , of which elements Refers to the first One registration point is needed; simultaneously, a mask image is obtained based on the top-down planar view of the 3D scene, and a set of reference points is extracted. , of which elements Refers to the first There are 1 reference point; in each iteration, the rigid transformation matrix is ​​estimated based on the current estimate. Pair set Update; For each wall, its shape is represented by the following parameters: The dimensions of the wall are based on its length. ,thickness and height Description; For a wall, its length It is calculated from two adjacent corner points in space; Secondly, there's the rotation angle, the rotation angle of the wall. This describes the rotation of the wall relative to a reference coordinate system, calculated using the relative positions of the wall's two endpoints. The wall's position is determined by the coordinate pairs of adjacent corner points. The horizontal coordinates of the wall's center are calculated using these adjacent corner point coordinate pairs, taking into account the wall's rotation angle. Assuming the bottom of the wall is in contact with the ground, and the height is... along axial distribution, based on the central horizontal coordinate Combined with rotation angle The positional parameters of the wall in three-dimensional space are obtained. ,in The coordinates are taken as This indicates the center height of the wall.

4. The indoor three-dimensional scene layout optimization method based on wall features according to claim 3, characterized in that, The pair set The update process is as follows: First, calculate the least squares error to find a rigid transformation matrix. ,in It is a rotation matrix. It is a displacement vector that makes the point set with point set The distance between them is minimized, and the error function is expressed as: ; Further optimize the transformation matrix by minimizing the error function. Obtain the optimal rotation matrix and displacement vector ; The point set is then updated, using the points to be registered through a transformation matrix. Position after rotation Add displacement vector Point set Update; Finally, iterate and repeat the above steps until the error function is reached. It converges to a threshold.

5. The indoor three-dimensional scene layout optimization method based on wall features according to claim 4, characterized in that, The specific implementation process of step 3 is as follows: A diffusion model based on the U-Net architecture is proposed to optimize the layout attributes of furniture in a local scene unit of an indoor 3D scene. The scene attributes optimized by the diffusion model include location. ,size Orientation and categories ; The diffusion model includes a feature encoder, a feature decoder, a cross-attention mechanism, a self-attention mechanism, and residual blocks. The feature encoder abstracts the scene attributes of the input step by step through multiple convolutions and residual blocks, mapping them to the semantic space. The corresponding decoder reconstructs the attribute information of the furniture step by step through upsampling operations and skip connections. The cross-attention mechanism improves the accuracy of feature extraction and reconstruction by modeling the correlation between different features. The temporal embedding module dynamically provides temporal information, enabling the diffusion model to adaptively adjust the feature representation according to different time steps.

6. The indoor three-dimensional scene layout optimization method based on wall features according to claim 5, characterized in that, The specific implementation process of converting the optimized local 3D scene information into a global 3D scene is as follows: Let the position of the central furniture in the local coordinate system be... Its rotation angle in the local coordinate system is This converts it into position and angle in the global coordinate system; First, there's the position transformation, specifically the position information of the central furniture in the local coordinate system. By multiplying by the rotation matrix Add displacement vector Transform to the global coordinate system to obtain the position of the furniture in the global coordinate system. ; Then, the angles of other furniture within the local scene are transformed. In the local coordinate system, the rotation angle of the furniture is... It is relative to the local coordinate system, while in the global coordinate system, the angles are adjusted to make them consistent, based on the rotation angle of the furniture. Add the rotation angle of the local coordinate system relative to the global coordinate system This gives the rotation angle of the furniture in the global coordinate system. ; To restore information from the local coordinate system to the global coordinate system, the local data is first restored to the range of the global coordinate system through an inverse normalization operation: by restoring the furniture attributes in the local data, the local information is mapped back to the global coordinate system. The restoration process includes two steps: displacement restoration and angle restoration.

7. The indoor three-dimensional scene layout optimization method based on wall features according to claim 6, characterized in that, The displacement recovery and angle recovery processes specifically include: merging the position information, angle information, size information and category labels of all furniture in the local scene, and obtaining the final position, angle and size information by averaging the merged results, ultimately obtaining a complete global scene containing all furniture information.

8. The indoor three-dimensional scene layout optimization method based on wall features according to claim 7, characterized in that, In the adversarial optimization process, the loss function is set as follows: Mean squared error loss: Applied to the noise recovery stage of this model, this loss minimizes the difference between the predicted noise and the true noise; Displacement and orientation loss: The L2 norm is used to measure the Euclidean distance between the predicted position and the true position; the angle loss of the furniture is achieved using cosine similarity loss. Furniture crossover and union ratio loss: used to optimize the overlapping area of ​​the bounding boxes between regular furniture, and design the wall crossover and union ratio loss. By quantifying and predicting the spatial overlap between the bounding boxes of regular furniture and the bounding boxes of walls, the model is guided to learn furniture layout patterns that conform to building codes.

Images