Building design effect picture automatic generation method and system, medium and program product

Abstract

The invention discloses an automatic building design effect picture generation method and system, a medium and a program product, and the method comprises the steps: obtaining three-dimensional geometric data of a building group, and building a spatial visual importance network of the building group according to the three-dimensional geometric data; the building group is divided into a plurality of visual correlation sub-regions based on the spatial visual importance network, the spatial entropy value of each visual correlation sub-region is calculated, and the spatial entropy values are used for representing the spatial distribution complexity degree of building units in the visual correlation sub-regions; according to the spatial entropy values and the three-dimensional geometric data of the building units in the visual association sub-regions, generating a multi-level detail model corresponding to each visual association sub-region; constructing a rendering pipeline based on the multi-stage detail model and the connection weight; and performing visual consistency synthesis on the output result of the rendering pipeline to generate a building design effect picture of the building group. The real-time performance and practicability of building design effect picture generation are improved.

Description

Technical Field

[0001] The present application relates to the field of digital cultural product production, and in particular to a method, system, medium and program product for automatically generating architectural design renderings. Background Art

[0002] Architectural design renderings are an important part of the architectural design process. They play an important role in demonstrating architectural design concepts, expressing design intentions, and effectively communicating with customers. However, the production of traditional architectural renderings mainly relies on designers to manually model, adjust materials and lights, set rendering parameters, etc. This process is not only time-consuming and labor-intensive, but also requires high professional skills of designers. At the same time, it is easy to have problems such as inconsistent rendering styles and uneven quality.

[0003] With the development of artificial intelligence technology, deep learning-based architectural rendering generation technology has gradually emerged. This technology builds a generative adversarial network model and uses a large number of architectural rendering samples for training, so that the model can learn the style characteristics and composition rules of architectural renderings, thereby realizing the automatic generation of architectural renderings. This method improves the efficiency of rendering generation to a certain extent and reduces the dependence on the professional skills of designers.

[0004] However, the existing deep learning-based architectural rendering generation technology has the problem of excessive consumption of computing resources when processing large-scale building groups. This is because building groups usually contain a large amount of geometric details and material information. During the generation process, a large amount of feature data needs to be processed simultaneously, resulting in a large amount of computing resources being consumed in both the model training and inference stages, which reduces the real-time and practicality of rendering generation. Summary of the invention

[0005] The present application provides a method, system, medium and program product for automatically generating architectural design renderings, which are used to improve the real-time and practicality of generating architectural design renderings.

[0006] In a first aspect, the present application provides a method for automatically generating architectural design renderings, which obtains three-dimensional geometric data of a building group, and establishes a spatial visual importance network of the building group based on the three-dimensional geometric data, wherein the nodes of the spatial visual importance network represent building units, and the connection weights between the nodes are determined according to the visual occlusion relationship between the building units; Based on the spatial visual importance network, the building group is divided into several visually associated sub-regions, and the spatial entropy value of each visually associated sub-region is calculated. The spatial entropy value is used to characterize the spatial distribution complexity of the building units in the visually associated sub-region. Generate a multi-level detail model corresponding to each visual association sub-area according to the spatial entropy value and the three-dimensional geometric data of the building unit in the visual association sub-area; Building a rendering pipeline based on multi-level detail models and connection weights; The output results of the rendering pipeline are synthesized for visual consistency to generate architectural design renderings of building groups.

[0007] By adopting the above technical solution, the three-dimensional geometric data of the building group is obtained to establish a spatial visual importance network, which can accurately reflect the visual occlusion relationship between the building units. Then, based on the network, the building group is divided into visually related sub-areas and the spatial entropy value is calculated, which can quantitatively characterize the spatial distribution characteristics of the building units in each sub-area. The multi-level detail model generated according to the spatial entropy value and the three-dimensional geometric data can reduce the complexity of the model while maintaining the visual effect. Combined with the rendering pipeline constructed by the connection weights, efficient rendering of the building group can be achieved. Finally, the visual consistency synthesis of the rendering results can reduce the visual discontinuity generated during the rendering process, enhance the realism and coherence of the overall rendering, and improve the real-time and practicality of the generation of architectural design renderings.

[0008] In combination with some embodiments of the first aspect, in some embodiments, generating a multi-level detail model corresponding to each visual association sub-region according to the spatial entropy value and the three-dimensional geometric data of the building unit in the visual association sub-region specifically includes: A geometric feature matrix is ​​constructed for the building units in the visual association sub-area, and each row of the geometric feature matrix represents a geometric feature vector of a building unit; The dimension threshold of the feature decomposition is determined based on the spatial entropy value, and the singular value decomposition is performed on the geometric feature matrix to obtain the main feature vector set; The main feature vector set is sorted according to the size of the singular value, and the features are reconstructed according to the dimension threshold to obtain a multi-level detail model.

[0009] By adopting the above technical solution, a geometric feature matrix is ​​constructed for the building units in the visual association sub-area, and the geometric features of the building units are vectorized to make feature extraction more systematic and standardized. The spatial entropy value is used to determine the dimension threshold of feature decomposition, and the feature dimension can be adaptively adjusted according to the complexity of the spatial distribution. The geometric feature matrix is ​​subjected to singular value decomposition to obtain the main feature vector set, and the feature reconstruction is performed after sorting according to the size of the singular value. It can reduce the data dimension while retaining the key geometric features. The generated multi-level detail model not only maintains the main geometric features of the building unit, but also reduces the amount of data, and improves the computational efficiency of the subsequent rendering process.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, constructing a rendering pipeline based on a multi-level detail model and connection weights specifically includes: Build a rendering task priority queue, where the task priority in the rendering task priority queue is determined based on the connection weight; Determine a detail level model for each rendering task in the rendering task priority queue according to a distance from a preset viewpoint to a corresponding visually associated sub-region; Create a rendering calculation graph. The nodes of the rendering calculation graph represent rendering tasks, and the edges between nodes represent task dependencies. Construct a multi-level cache structure based on the topological structure of the rendering calculation graph; The rendering task priority queue, multi-level cache structure and rendering calculation graph are integrated into a rendering pipeline.

[0011] By adopting the above technical solution, constructing a rendering task priority queue and determining the task priority according to the connection weight, the rendering order can be arranged reasonably, giving priority to the building units with greater visual impact. The detail level model is determined according to the distance from the preset viewpoint to the visually associated sub-area, so that the nearby buildings use high-detail models and the distant buildings use low-detail models, which not only ensures the visual effect but also improves the rendering efficiency. Establishing a rendering calculation graph to represent the task dependency and constructing a multi-level cache structure can optimize the execution order of rendering tasks and the data access efficiency. Efficient adaptive rendering of building groups is achieved, reducing computing resource consumption while ensuring rendering quality.

[0012] In combination with some embodiments of the first aspect, in some embodiments, before acquiring the three-dimensional geometric data of the building group, the method further includes: Construct a hierarchical data structure of building groups, which includes the spatial topological relationship, functional attributes and material parameters of building units; Perform geometric constraint checking based on a hierarchical data structure and generate constraint violation metrics, which are used to characterize the degree of spatial conflict between building units; The building units whose constraint violation metric index is less than a preset first threshold are positioned so that the constraint violation metric index is not less than the preset first threshold, thereby obtaining three-dimensional geometric data that meets the spatial constraints.

[0013] By adopting the above technical solution, a hierarchical data structure including spatial topological relationships, functional attributes and material parameters is constructed, which can comprehensively describe the spatial organization characteristics of the building group. Based on the data structure, geometric constraint checks are performed and constraint violation metrics are generated, which can quantitatively evaluate the degree of spatial conflict between building units. The position of the building unit whose constraint violation metric is less than the preset first threshold is adjusted until the constraint violation metric is not less than the preset first threshold, which can eliminate the unreasonable spatial relationship between the building units and obtain the three-dimensional geometric data that meets the spatial constraint requirements.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, performing geometric constraint checking based on a hierarchical data structure and generating constraint violation metrics specifically include: Construct a spatial index and store the bounding box information of the building unit in the corresponding spatial partitioning node; The spatial index is used to detect the occlusion relationship between building units and obtain the occlusion detection result; Based on the occlusion detection results, a constraint graph is constructed. The nodes of the constraint graph represent building units, and the edges represent spatial constraint relationships. Perform connectivity analysis on the constraint graph and calculate constraint violation metrics.

[0015] By adopting the above technical solution, a spatial index is constructed to store the bounding box information of the building unit, which can quickly locate and access the building units in the space, reducing the computational complexity of occlusion relationship detection. Using the spatial index for occlusion detection can accurately identify the spatial position relationship and occlusion degree between building units, reducing the redundant calculation caused by comparing all building units one by one. The constraint graph constructed based on the occlusion detection results converts the spatial constraint relationship between building units into a graph structure representation, which is convenient for quantitative analysis. By performing connectivity analysis on the constraint graph and calculating the constraint violation metric, the spatial conflict between each building unit in the building group can be comprehensively evaluated, which improves the efficiency and accuracy of geometric constraint checking and makes the spatial layout of the building group more reasonable.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after generating the architectural design renderings of the building group, the method further includes: Collect several groups of expert scoring samples and establish a scoring standard set, which includes the rationality of spatial layout, visual coherence and material authenticity; Perform multi-dimensional quality assessment on architectural design renderings based on the scoring standard set to obtain a quality scoring vector; The rendering pipeline is adjusted according to the quality score vector until each dimensional component of the quality score vector is greater than a preset second threshold.

[0017] By adopting the above technical solution, introducing expert scoring samples to establish a scoring standard set, the evaluation indicators of multiple dimensions such as spatial layout rationality, visual coherence and material authenticity are quantified, so that the quality evaluation of the renderings has an objective basis. The quality scoring vector obtained through multi-dimensional quality evaluation can accurately reflect the performance level of the renderings in different aspects. According to the quality scoring vector, the rendering pipeline is adjusted in a targeted manner, and through iterative optimization, the components of each dimension meet the preset threshold requirements, which can continuously improve the overall quality of the renderings, make the rendering generation process have clear quality control standards, ensure the professional level of the final renderings, reduce the workload of manual repeated modifications, and improve the automation and quality controllability of rendering generation.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, a multi-dimensional quality assessment is performed on the architectural design renderings based on a scoring standard set to obtain a quality scoring vector, specifically including: Extract local image blocks of the architectural design renderings and calculate feature statistics of each local image block; Perform cluster analysis on the local image block features to obtain a feature distribution map; The spatial layout score, visual coherence score, and material authenticity score were calculated based on the feature distribution map; The spatial layout score, visual coherence score, and material authenticity score are normalized to form a quality score vector.

[0019] By adopting the above technical solution, extracting local image blocks and calculating feature statistics, the detailed features and local visual effects in the renderings can be captured. The feature distribution map obtained by clustering the features of local image blocks reflects the spatial distribution law and organization of visual elements in the renderings. Based on the feature distribution map, the spatial layout score, visual coherence score and material authenticity score are calculated respectively, realizing a multi-angle quantitative evaluation of the quality of the renderings. The scores are normalized to form a quality score vector, so that the evaluation indicators of different dimensions are comparable, and the quality of the renderings can be systematically measured from the local to the whole, which improves the interpretability and reliability of the evaluation results. Through feature statistics and cluster analysis, quality problems in the renderings can be accurately identified, providing specific improvement directions for the optimization of the rendering pipeline.

[0020] In a second aspect, an embodiment of the present application provides a system for automatically generating architectural design renderings, the system comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, and the one or more processors call the computer instructions so that the system executes the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a system, causes the system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer program product, characterized in that when the computer program product runs on a system, the system executes the method described in any possible implementation manner in the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The present application provides a method for automatically generating architectural design renderings, which obtains the three-dimensional geometric data of the building group to establish a spatial visual importance network, which can accurately reflect the visual occlusion relationship between the building units, and then divide the building group into visually related sub-areas based on the network and calculate the spatial entropy value, which can quantitatively characterize the spatial distribution characteristics of the building units in each sub-area. The multi-level detail model generated according to the spatial entropy value and the three-dimensional geometric data can reduce the complexity of the model while maintaining the visual effect, and the rendering pipeline constructed by the connection weights can realize efficient rendering of the building group. Finally, the rendering results are synthesized for visual consistency, which can reduce the visual discontinuity generated during the rendering process, enhance the realism and coherence of the overall rendering, and improve the real-time and practicality of the generation of architectural design renderings.

[0024] 2. The present application provides a method for automatically generating architectural design renderings, constructing a hierarchical data structure containing spatial topological relationships, functional attributes, and material parameters, which can comprehensively describe the spatial organizational characteristics of a building group. Based on the data structure, a geometric constraint check is performed and a constraint violation metric is generated, which can quantitatively evaluate the degree of spatial conflict between building units. The position of the building unit whose constraint violation metric is less than a preset first threshold is adjusted until the constraint violation metric is not less than the preset first threshold, so that the unreasonable spatial relationship between the building units can be eliminated, and the obtained three-dimensional geometric data meets the spatial constraint requirements.

[0025] 3. This application provides a method for automatically generating architectural design renderings, introduces expert scoring samples to establish a scoring standard set, and quantifies evaluation indicators in multiple dimensions such as spatial layout rationality, visual coherence, and material authenticity, so that the quality assessment of the renderings has an objective basis. The quality score vector obtained through multi-dimensional quality assessment can accurately reflect the performance level of the renderings in different aspects. Targeted adjustments are made to the rendering pipeline based on the quality score vector, and each dimensional component is optimized through iterative optimization to meet the preset threshold requirements, which can continuously improve the overall quality of the renderings, so that the generation process of the renderings has clear quality control standards, ensure the professional level of the final renderings, reduce the workload of repeated manual modifications, and improve the degree of automation and quality controllability of rendering generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of a method for automatically generating architectural design renderings in an embodiment of the present application.

[0027] Figure 2 It is a flow chart of a quality assessment and optimization method based on expert scoring in an embodiment of the present application.

[0028] Figure 3It is a schematic diagram of the physical device structure of a system for automatically generating architectural design renderings provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more listed items.

[0030] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0031] The following uses an embodiment and combines Figure 1 , a method for automatically generating a building design rendering in an embodiment of the present application is described: See also Figure 1 , which is a flow chart of a method for automatically generating architectural design renderings in an embodiment of the present application.

[0032] S101, constructing a hierarchical data structure of building groups; The system constructs a hierarchical data structure of building groups, which includes the spatial topological relationship, functional attributes and material parameters of building units.

[0033] The system constructs a hierarchical data structure of building groups that includes the spatial topological relationship, functional attributes and material parameters of building units. The hierarchical data structure is a way of organizing data that can clearly express the spatial position relationship between each building unit in the building group, the use function of the building unit, and the building materials. By constructing a hierarchical data structure, it provides the necessary data foundation for subsequent steps such as geometric constraint checking and 3D geometric data generation.

[0034] The system can use a tree structure to represent the hierarchical data structure of a building group. The root node of the tree structure represents the entire building group, the middle nodes represent different areas or building types, and the leaf nodes represent specific building units. The edges between nodes represent spatial topological relationships, such as adjacency, inclusion, etc. Each node can also carry the functional attributes (such as residential, shop, office, etc.) and material parameters (such as masonry, glass, metal, etc.) of the building unit. The system can automatically extract the spatial topological relationship, functional attributes and material parameters of the building unit by parsing data sources such as architectural design drawings and BIM models, and construct a hierarchical data structure.

[0035] S102, performing geometric constraint checking based on the hierarchical data structure and generating constraint violation measurement indicators; The system performs geometric constraint checking based on the hierarchical data structure and generates constraint violation metrics, which are used to characterize the degree of spatial conflict between building units. Specifically: construct a spatial index and store the bounding box information of the building unit in the corresponding spatial partitioning node; The spatial index is used to detect the occlusion relationship between building units and obtain the occlusion detection result; Based on the occlusion detection results, a constraint graph is constructed. The nodes of the constraint graph represent building units, and the edges represent spatial constraint relationships. Perform connectivity analysis on the constraint graph and calculate constraint violation metrics.

[0036] Based on the hierarchical data structure constructed in the previous step, the system performs geometric constraint checks and generates constraint violation metrics that characterize the degree of spatial conflict between building units. Geometric constraints refer to the spatial position relationships that should be satisfied between building units, such as certain building units cannot block each other, and certain building units need to maintain a certain distance from each other. Through geometric constraint checks, spatial conflict problems in building groups can be discovered, providing a basis for subsequent adjustments to the positions of building units.

[0037] The system first builds a spatial index and stores the bounding box (i.e., the smallest rectangle that can completely contain the building unit) information of each building unit in the corresponding space division node. Common spatial indexes include octrees, kd trees, R trees, etc. Then, the system uses the spatial index to quickly detect whether there is an occlusion relationship between building units and obtain the occlusion detection result. Next, the system builds a constraint graph based on the occlusion detection result. The constraint graph is a graph model in which nodes represent building units and edges represent spatial constraint relationships between building units (such as occlusion, being too close to each other, etc.). Finally, the system performs connectivity analysis on the constraint graph and counts the number of connected subgraphs in the graph. The more connected subgraphs there are, the more serious the spatial conflict is, and the constraint violation metric is calculated based on this.

[0038] When performing geometric constraint checking, if there are many types of constraint relationships that need to be checked, the constraint graph may be too dense and the computational complexity of connectivity analysis may be high. To address this problem, the system can adopt an incremental constraint checking strategy. That is, first check the more basic constraint relationships (such as occlusion) and generate an initial constraint graph; then gradually add other types of constraint relationships on this basis, update the constraint graph, and perform incremental connectivity analysis at the same time. Through incremental constraint checking, the system can reduce the computational complexity of the algorithm while meeting the requirements of geometric constraint checking.

[0039] S103, adjusting the position of the building unit whose constraint violation metric index is less than a preset first threshold value until the constraint violation metric index is not less than the preset first threshold value, thereby obtaining three-dimensional geometric data that satisfies the spatial constraint; The system adjusts the position of the building units whose constraint violation metric index is less than the preset first threshold until the constraint violation metric index is not less than the preset first threshold, and obtains the three-dimensional geometric data that meets the spatial constraints. After that, a spatial visual importance network of the building group is established based on the three-dimensional geometric data. The nodes of the spatial visual importance network represent the building units, and the connection weights between the nodes are determined according to the visual occlusion relationship between the building units.

[0040] The system adjusts the position of building units with smaller constraint violation metrics (less than the preset threshold) until the constraint violation metrics are no longer less than the threshold, thereby obtaining the three-dimensional geometric data of the building group that meets the spatial constraints. Position adjustment is to eliminate the spatial conflict between building units by changing the spatial coordinates of the building units. After position adjustment, the spatial layout of the building group can meet the design specifications and usage requirements.

[0041] The system can adopt a position adjustment algorithm based on physical simulation. Each building unit is regarded as a particle, and there is repulsion (from constraint violation) and attraction (from design intent) between the particles. The system determines the magnitude of the repulsion based on the constraint violation metric, and determines the magnitude of the attraction based on the spatial relationship between the building units (such as symmetry, parallelism, etc.). Then, the system uses Newton's second law to simulate the motion trajectory of the particle and obtain the new coordinates of the particle as the adjusted position of the building unit. During the iteration process, the system continuously detects changes in the constraint violation metric and stops the iteration when the metric is no longer less than the threshold. In addition, the system can also perform collision detection on the building units after position adjustment to avoid introducing new spatial conflicts.

[0042] S104, dividing the building group into a number of visually associated sub-regions based on the spatial visual importance network, and calculating the spatial entropy value of each visually associated sub-region; Based on the spatial visual importance network, the system divides the building group into several visually associated sub-areas and calculates the spatial entropy value of each visually associated sub-area. The spatial entropy value is used to characterize the complexity of the spatial distribution of building units in the visually associated sub-area.

[0043] The system first establishes a spatial visual importance network of building groups based on the three-dimensional geometric data obtained in the previous step. The network reflects the visibility relationship between building units. The nodes in the network represent building units, and the edge weights between nodes are determined by the visibility of the building units. Then, the system performs community discovery based on the spatial visual importance network and divides building units with strong visibility into the same visually associated sub-area. Finally, the system calculates the spatial entropy value of each sub-area. The spatial entropy value measures the complexity of the spatial distribution of building units and provides a basis for the subsequent generation of multi-level detail models.

[0044] The system can use the Graph Cut algorithm to divide the visually associated sub-regions. The spatial visual importance network is regarded as an undirected weighted graph. The goal of the Graph Cut algorithm is to find a set of edges so that after deleting these edges, the graph is divided into several connected sub-graphs, while minimizing the sum of the weights of the cut edges. Intuitively, the sum of the weights of the cut edges reflects the strength of the visual association between different sub-regions. The visual association within the sub-region is strong, while the visual association between sub-regions is weak. After obtaining the visually associated sub-regions, the system calculates the spatial entropy value of each sub-region. The calculation of spatial entropy can be based on the spatial distribution histogram of building units in the sub-region. The system evenly divides the three-dimensional space and counts the number of building units falling into each grid, thereby obtaining a spatial distribution histogram. Then, the system calculates the Shannon entropy as the spatial entropy value based on the histogram.

[0045] S105, generating a multi-level detail model corresponding to each visual association sub-region according to the spatial entropy value and the three-dimensional geometric data of the building unit in the visual association sub-region; The system generates a corresponding multi-level detail model (LOD) for each visually associated sub-area. A multi-level detail model refers to using geometric models of different degrees of refinement to represent the same object at different viewing distances. The farther the viewing distance, the simpler the geometric model, and the closer the viewing distance, the finer the geometric model. The purpose of generating a multi-level detail model is to reduce the amount of rendering calculations as much as possible while ensuring the visualization effect. The system determines the level of detail contained in the sub-area LOD model and the degree of simplification of each level based on the spatial entropy value of the sub-area and the three-dimensional geometric data of the building unit.

[0046] For each visually associated sub-area, the system first determines the LOD level that needs to be generated based on the spatial entropy value. Generally, the larger the spatial entropy value, the more complex the distribution of the building unit, and the more LOD levels need to be generated. Then, for each LOD level, the system uses a geometric simplification algorithm to simplify the three-dimensional geometric data of the building unit. Common geometric simplification algorithms include vertex clustering, edge collapse, and patch merging. The vertex clustering algorithm reduces the number of model faces by merging vertices with close distances into one vertex; the edge collapse algorithm merges two adjacent patches into one patch by deleting the edges in the model; the patch merging algorithm simplifies the model by merging coplanar or small angle patches into one patch. The system can combine algorithms such as vertex clustering, edge collapse, and patch merging to obtain LOD models of building units with different degrees of simplification.

[0047] S106, constructing a rendering pipeline based on the multi-level detail model and the connection weights; Based on the multi-level detail model generated in the previous step and the connection weights of the spatial visual importance network, the system constructs a rendering pipeline for rendering building groups. The rendering pipeline refers to a series of operation steps to convert a three-dimensional model into a two-dimensional image, usually including geometry processing, lighting calculation, texture mapping, shadow generation, post-processing, etc. The system determines the execution mode and order of each stage in the rendering pipeline based on the hierarchy of the LOD model and the connection weights between visually associated sub-areas, thereby achieving efficient and high-quality rendering of building groups.

[0048] The system can use an adaptive tile rendering strategy to build a rendering pipeline. The system divides the screen space into several rendering tiles, each of which corresponds to a visually associated sub-area. For each rendering tile, the system executes the rendering pipeline from coarse to fine according to the LOD level of the building units it contains. In the geometry processing stage, the system loads the geometric model of the building unit corresponding to the LOD level, and transforms and projects the vertex coordinates. In the lighting calculation stage, the system calculates the vertex color according to the light source position and the material properties of the building unit. For different LOD levels, the system can choose lighting models of different complexity, such as Phong model, PBR model, etc. In the texture mapping stage, the system maps the texture image to the surface of the building unit to provide surface details. Similarly, for different LOD levels, the system can use textures of different resolutions. In the shadow generation stage, the system calculates the shadow relationship between building units. Considering the overhead of shadow calculation, the system can generate dynamic shadows only for building units with high LOD levels, and use pre-baked light maps for building units with low LOD levels. After each stage is completed, the system obtains the rendering results of all building units in the sub-area. Finally, the system mixes the rendering results of the sub-regions based on the connection weights of the visually associated sub-regions to obtain the final rendered image of the building object.

[0049] S107, performing visual consistency synthesis on the output results of the rendering pipeline to generate an architectural design rendering of the building complex.

[0050] The system post-processes the initial rendering results output by the rendering pipeline to generate the final architectural design rendering. Since the rendering pipeline is executed in blocks, there may be inconsistencies between the initial rendering results in different visually related sub-areas, such as discontinuous lighting and abrupt shadow changes. In order to eliminate these inconsistencies and improve the rendering quality, the system needs to perform visual consistency synthesis on the initial rendering results. The goal of visual consistency synthesis is to adjust the visual attributes such as color, brightness, and contrast in the initial rendering results so that adjacent sub-areas transition visually smoothly and the rendered image of the entire building group has global consistency.

[0051] The system can use an optimization-based method to achieve visual consistency synthesis. The system constructs an energy function to measure the visual consistency of the rendered image. The energy function consists of two parts: data term and smoothing term. The data term measures the difference between each pixel color and the initial rendering result. The smaller the difference, the smaller the energy function value. The smoothing term measures the difference between adjacent pixel colors. The smaller the difference, the smaller the energy function value. The goal of the system is to find a set of pixel colors that minimizes the energy function value. This can be converted into a least squares optimization problem. Considering the high resolution of architectural design renderings, the scale of the optimization problem is large. Therefore, the system adopts a multi-resolution optimization strategy. First, the system downsamples the initial rendering result to obtain an image with lower resolution; then, the optimization is performed on the low-resolution image to obtain a result with better visual consistency; finally, the system uses bilinear interpolation to upsample the optimization result, and uses it as the initial value to optimize on the original resolution image to obtain the final visual consistency image. During the optimization process, the system can also adaptively adjust the weight of the smoothing term according to the connection weight of the visually associated sub-region. The larger the connection weight of adjacent regions, the higher the smoothness requirement between them, and the larger the weight of the smoothness term.

[0052] In the above embodiment, the three-dimensional geometric data of the building group is obtained to establish a spatial visual importance network, which can accurately reflect the visual occlusion relationship between the building units, and then the building group is divided into visually related sub-areas based on the network and the spatial entropy value is calculated, which can quantitatively characterize the spatial distribution characteristics of the building units in each sub-area. The multi-level detail model generated according to the spatial entropy value and the three-dimensional geometric data can reduce the complexity of the model while maintaining the visual effect, and the rendering pipeline constructed by the connection weight can realize the efficient rendering of the building group. Finally, the visual consistency synthesis of the rendering results can reduce the visual discontinuity generated in the rendering process, enhance the realism and coherence of the overall rendering, and improve the real-time and practicality of the generation of architectural design renderings.

[0053] On the basis of the first embodiment, in order to further improve the quality of architectural design renderings, the embodiment of the present application also provides a quality assessment and optimization method based on expert scoring. This method establishes a standardized scoring system, performs multi-dimensional quality assessment on the generated renderings, and dynamically adjusts the rendering pipeline according to the assessment results, thereby achieving closed-loop quality control of the rendering generation process. Figure 2 , a quality assessment and optimization based on expert scoring in an embodiment of the present application is described: See also Figure 2 , which is a flow chart of a quality assessment and optimization method based on expert scoring in an embodiment of the present application.

[0054] S201, collecting several groups of expert scoring samples and establishing a scoring standard set; The system collects several groups of expert scoring samples and establishes a scoring standard set, which includes the rationality of spatial layout, visual coherence and material authenticity.

[0055] The system collects the scoring data of different architectural design renderings from multiple experts as samples to establish a standardized scoring system. The so-called expert scoring refers to the subjective evaluation and scoring of various indicators of the renderings by senior designers or review experts with rich experience in the field of architectural design. These indicators can include the rationality of spatial layout, the coherence of visual elements, the authenticity of material mapping and other aspects.

[0056] For different architectural design styles and types, the system can select multiple experts in the corresponding professional fields to score them separately, and summarize a set of general scoring rules and standards through the collection and statistical analysis of a large number of samples. For example, cluster analysis can be performed on sample data to extract statistical features such as the score range, mean, and variance of each indicator, and the scoring weights of each indicator can be summarized based on expert experience, thereby forming a standardized multi-dimensional scoring matrix.

[0057] In practical applications, the collection of expert scoring samples may encounter problems such as insufficient sample quantity and inconsistent scoring standards of individual experts. To this end, the system can expand the scale and scope of sample collection through crowdsourcing and other methods, and perform necessary preprocessing on the original scoring data, such as outlier removal and normalization, to ensure the representativeness and robustness of the scoring standard set. In addition, the system can also regularly update the iterative scoring standard set to adapt to the dynamic changes in architectural design aesthetics.

[0058] S202, extracting local image blocks of the architectural design rendering, and calculating feature statistics of each local image block; The system first divides the entire rendering into several local sub-regions, each of which is called an image block. The division method can be fixed-size grid segmentation, adaptive block division based on saliency, etc. Then, the system extracts multiple visual features from each image block and calculates the corresponding statistics. Common features can include color histograms, gradient histograms, texture features, etc., which correspond to different statistics such as mean, variance, skewness, and kurtosis.

[0059] When extracting features, the system can fully consider the high-level semantic characteristics of architectural design images. For example, it can use deep learning-based object detection and segmentation algorithms to extract key structural elements such as building outlines, windows, and balconies, and then calculate their size ratios, alignment, and other layout features. For another example, for interior decoration renderings, it can focus on extracting the style features of detailed objects such as furniture, lighting, and ornaments.

[0060] S203, performing cluster analysis on the local image block features to obtain a feature distribution map; This step uses cluster analysis to study the intrinsic distribution pattern of local image block features, thereby characterizing the overall visual characteristics of the rendering. Cluster analysis is an unsupervised learning method that can automatically classify similar data samples into one category and reveal the implicit structure within the data. Taking the feature vectors of all image blocks as input samples, the system can use classic algorithms such as K-means and spectral clustering for clustering to obtain several cluster centers, each of which represents a typical feature pattern.

[0061] Furthermore, the system can calculate the similarity between the feature vector of each image block and each cluster center to generate a feature distribution map. The position of the image block corresponds to the spatial layout of the rendering, and the degree of similarity corresponds to the degree of fit between the image block feature and a certain clustering pattern. The similarity value can be encoded by color or brightness to form an intuitive visualization result. By observing the feature distribution map, the visual characteristics of the rendering in different areas can be judged, such as the harmony of color matching and the uniformity of texture arrangement.

[0062] S204, calculating a spatial layout score, a visual coherence score, and a material authenticity score according to the feature distribution map; This step is based on the feature distribution map obtained above, and quantitatively evaluates the architectural design renderings from three aspects: spatial layout, visual coherence, and material authenticity. The system first refers to the manually annotated scoring samples to establish a mapping relationship between the feature distribution map and each score. For example, machine learning methods such as regression analysis and support vector machines can be used to train a prediction function model with the statistical features of the feature distribution map (such as histograms, co-occurrence matrices, etc.) as input, thereby obtaining an automatic scoring mechanism for new renderings.

[0063] Among them, the spatial layout score reflects the rationality of the layout relationship of the scale, proportion, alignment and other structural elements in the rendering. The system can extract the geometric parameters of key parts such as building outlines, doors and windows, columns and beams from the feature distribution map, such as area ratio, golden ratio, etc., and calculate the layout score in combination with the architectural design specifications. The visual coherence score measures the harmony and unity of the entire rendering in terms of color matching, light and shadow transition, etc. The system can use color aesthetics theory to analyze factors such as subject and object color, cold and warm contrast, light and dark levels from the feature distribution map, and consider composition rules such as shape repetition and rhythm to give a comprehensive score. The material authenticity score focuses on evaluating the realism of the texture material. The system can start with the texture features of the feature distribution map, such as roughness, glossiness, transparency, etc., and give a score based on the visual properties of the real material sample.

[0064] It should be noted that since there may be certain correlations and constraints between different evaluation indicators, the system needs to weigh the combined impact of multiple indicators when scoring. For example, a too dense spatial layout may affect the realism of the material, and too rich and jumping color combinations may also destroy the overall visual coherence. Therefore, the system can establish a dependency model between scoring indicators, and conduct joint reasoning on each score through methods such as conditional random fields and Bayesian networks to obtain a more comprehensive and objective evaluation result. At the same time, for different design styles and application scenarios, the system can also flexibly adjust the weight ratio of each indicator to adapt to the user's personalized preferences.

[0065] S205, normalizing the spatial layout score, the visual coherence score, and the material authenticity score to form a quality score vector; This step integrates the three scoring indicators obtained in the previous step into a comprehensive quality scoring vector, so as to optimize the rendering pipeline as a whole. Since the value ranges and dimensions of different scoring indicators are different, the system first needs to normalize them so that they are mapped to a unified interval (such as [0, 1]). Common normalization methods include maximum and minimum value normalization, zero mean unit variance normalization, etc.

[0066] After normalization, the system combines the three scores into a three-dimensional column vector, and each component represents the score of the corresponding indicator. By introducing the concept of scoring vector, the overall quality of the rendering can be expressed as a mathematical quantitative indicator, which is convenient for tracking the optimization process. At the same time, the scoring vector can also be used as a visualization tool to intuitively display the advantages and disadvantages of different design schemes in various indicators.

[0067] When constructing the scoring vector, the system can also dynamically adjust the weight of each component based on the user's interactive feedback. For example, an interactive interface can be designed to allow users to manually drag and adjust the importance of each scoring indicator. The system will adaptively generate a weighted scoring vector based on this and update the optimization focus accordingly. This interactive mechanism makes the evaluation process more flexible and can fully respect the user's subjective preferences.

[0068] S206 , adjusting the rendering pipeline according to the quality score vector until each dimensional component of the quality score vector is greater than a preset second threshold.

[0069] This step uses the quality score vector obtained above to achieve closed-loop optimization of the rendering pipeline by setting the target threshold. The system first sets an expected quality threshold for each component of the score vector. When all components exceed this threshold, the current design effect is considered to have met the comprehensive quality requirements.

[0070] During the optimization process, the system adaptively adjusts the various parameters and sub-modules of the rendering pipeline based on the score vector. Specifically, gradient-based optimization algorithms such as SGD and Adam can be used to regard the score vector as a multi-objective function, and iteratively update the rendering parameters by solving the gradient direction. Common adjustable parameters include the softness of global illumination, the sharpness of shadows, the highlight intensity of materials, the saturation contrast of colors, etc. At the same time, for certain indicators with low scores, the system can also selectively enhance the corresponding processing modules of the rendering pipeline, such as adding additional post-processing effects such as stylized filters and cartographic optimization, to focus on improving the performance of weak links.

[0071] The optimization process usually requires multiple rounds of iterations. The system updates the scoring vector based on the renderings of each round, and dynamically adjusts the optimization step size and direction accordingly, until all components of the scoring vector exceed the preset threshold. The entire closed-loop optimization mechanism realizes automatic feedback adjustment of rendering parameters and quality evaluation, improving the efficiency of generating design effects.

[0072] In the above embodiment, expert scoring samples are introduced to establish a scoring standard set, and evaluation indicators of multiple dimensions such as spatial layout rationality, visual coherence and material authenticity are quantified, so that the quality evaluation of the rendering has an objective basis. The quality scoring vector obtained by multi-dimensional quality evaluation can accurately reflect the performance level of the rendering in different aspects. According to the quality scoring vector, the rendering pipeline is adjusted in a targeted manner, and each dimensional component is optimized through iterative optimization to meet the preset threshold requirements, which can continuously improve the overall quality of the rendering, make the generation process of the rendering have clear quality control standards, ensure the professional level of the final rendering, reduce the workload of manual repeated modification, and improve the automation and quality controllability of the rendering generation.

[0073] The following describes the system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of the physical device structure of a system for automatically generating architectural design renderings provided in an embodiment of the present application.

[0074] It should be noted that Figure 3 The structure of the system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0075] like Figure 3 As shown, the system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0076] The following components are connected to the I / O interface 305: an input section 306 including a camera, an infrared sensor, etc.; an output section 307 including a liquid crystal display (LCD) and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0077] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0078] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.

[0079] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0080] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the system described in the above embodiment; or may exist independently without being assembled into the system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of a system, the system implements the method provided in the above embodiment.

[0081] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0082] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0083] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0084] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A method for automatically generating architectural design renderings, characterized in that: include: Acquire three-dimensional geometric data of a building group, and establish a spatial visual importance network of the building group according to the three-dimensional geometric data, wherein nodes of the spatial visual importance network represent building units, and connection weights between the nodes are determined according to visual occlusion relationships between the building units; Dividing the building group into a number of visually associated sub-regions based on the spatial visual importance network, and calculating a spatial entropy value of each of the visually associated sub-regions, wherein the spatial entropy value is used to characterize the complexity of the spatial distribution of the building units in the visually associated sub-regions; Generate a multi-level detail model corresponding to each visual association sub-region according to the spatial entropy value and the three-dimensional geometric data of the building unit in the visual association sub-region; Constructing a rendering pipeline based on the multi-level detail model and the connection weights; Visually consistent synthesis is performed on the output results of the rendering pipeline to generate an architectural design rendering of the building complex.

2. The method according to claim 1, characterized in that Generating a multi-level detail model corresponding to each visual association sub-region according to the spatial entropy value and the three-dimensional geometric data of the building unit in the visual association sub-region specifically includes: Constructing a geometric feature matrix for the building units in the visual association sub-area, wherein each row of the geometric feature matrix represents a geometric feature vector of the building unit; Determining a dimension threshold of feature decomposition based on the spatial entropy value, and performing singular value decomposition on the geometric feature matrix to obtain a main feature vector set; The main feature vector set is sorted according to the size of the singular value, and the feature is reconstructed according to the dimension threshold to obtain a multi-level detail model.

3. The method according to claim 1, characterized in that The constructing a rendering pipeline based on the multi-level detail model and the connection weights specifically includes: Constructing a rendering task priority queue, wherein the task priority in the rendering task priority queue is determined according to the connection weight; Determining a detail level model for each rendering task in the rendering task priority queue according to a distance from a preset viewpoint to a corresponding visually associated sub-region; Establishing a rendering calculation graph, wherein nodes of the rendering calculation graph represent rendering tasks, and edges between the nodes represent task dependencies; Constructing a multi-level cache structure according to the topological structure of the rendering calculation graph; The rendering task priority queue, the multi-level cache structure and the rendering calculation graph are integrated into a rendering pipeline.

4. The method according to claim 1, characterized in that: Before acquiring the three-dimensional geometric data of the building complex, the method further includes: Constructing a hierarchical data structure of the building group, the hierarchical data structure including spatial topological relationships, functional attributes and material parameters of the building units; Performing geometric constraint checking based on the hierarchical data structure to generate a constraint violation metric indicator, wherein the constraint violation metric indicator is used to characterize the degree of spatial conflict between the building units; The building units whose constraint violation metric index is less than a preset first threshold are positioned so that the constraint violation metric index is not less than the preset first threshold, thereby obtaining three-dimensional geometric data that meets the spatial constraint.

5. The method according to claim 4, characterized in that The performing of geometric constraint checking based on the hierarchical data structure and generating constraint violation metric indicators specifically includes: Constructing a spatial index, and storing the bounding box information of the building unit in the corresponding spatial division node; Using the spatial index to detect the occlusion relationship between the building units, and obtain an occlusion detection result; Constructing a constraint graph based on the occlusion detection result, wherein the nodes of the constraint graph represent the building units and the edges represent the spatial constraint relationships; Connectivity analysis is performed on the constraint graph to calculate constraint violation metrics.

6. The method according to claim 1, characterized in that After generating the architectural design renderings of the building complex, the method further includes: Collecting several groups of expert scoring samples and establishing a scoring standard set, wherein the scoring standard set includes rationality of spatial layout, visual coherence and material authenticity; Performing a multi-dimensional quality assessment on the architectural design rendering based on the scoring standard set to obtain a quality scoring vector; The rendering pipeline is adjusted according to the quality score vector until each dimensional component of the quality score vector is greater than a preset second threshold.

7. The method according to claim 6, characterized in that The multi-dimensional quality assessment of the architectural design renderings based on the scoring standard set to obtain a quality scoring vector specifically includes: Extracting local image blocks of the architectural design rendering, and calculating feature statistics of each local image block; Performing cluster analysis on the local image block features to obtain a feature distribution map; Calculating a spatial layout score, a visual coherence score, and a material authenticity score according to the feature distribution map; The spatial layout score, the visual coherence score and the material authenticity score are normalized to form a quality score vector.

8. A system for automatically generating architectural design renderings, characterized in that: The system comprises: One or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the system to execute the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a system, the system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a system, the system is caused to execute the method according to any one of claims 1 to 7.

Images