Data processing method based on building structure model design

Abstract

The present application relates to the field of data processing, especially to a kind of data processing method based on building structure model design, the present application extracts visual feature to the three-dimensional space in design sample library, respectively in each dimension selects feature constraint range to construct the corresponding visual constraint range combination image of three-dimensional space, and constructs the exclusive associated database for different visual constraint range combination image, the associated database has the associated parameter between different model bodies, subsequent in the process of creating three-dimensional space in user end, after user end executes predetermined instruction, according to current three-dimensional space, select associated database, and verify the visual association fuzzy category of three-dimensional space, subsequent adaptively push model body, the present application constructs visual feature understanding mechanism in the process of creating three-dimensional space in user end, adaptively push the model body that conforms to the visual feature of current three-dimensional space, reduce the browsing and screening burden when user end calls model body, improve design efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a data processing method based on building structure model design. Background Technology

[0002] With the rapid development of internet and computer vision technologies, 3D modeling technology has been widely applied in industrial and architectural design. Various 3D design software and building information modeling platforms provide designers with an intuitive and visual design environment, enabling the digital expression and spatial organization of design elements such as building structures and interior furnishings in the form of 3D models, significantly improving the intuitiveness and readability of design results.

[0003] For example, Chinese Patent Publication No. CN106485784A discloses an interior design imaging processing method based on multi-terminal device data synchronization. This method constructs an imaging model database adapted to various imaging media terminal system platforms to store three-dimensional imaging models of various interior design elements. It also establishes data conversion relationships between three-dimensional apartment models and interior design model calling parameters under different imaging media terminal system platforms. When performing three-dimensional imaging display of interior design, it reads a set of three-dimensional apartment models and interior design model calling parameters adapted to the system platform of the corresponding imaging media terminal from the interior design model dataset of the corresponding apartment interior design scheme. It then retrieves the corresponding three-dimensional imaging model resources of the interior design elements from the imaging model database to perform visualization processing and achieve imaging. This solves problems such as rapid data growth in three-dimensional imaging models of interior design across various imaging media terminal system platforms, large modeling workload, and cumbersome model data modification workflows.

[0004] However, the following problems still exist in the existing technology. Existing 3D design software typically has built-in or external model libraries. During the design process, users can select 3D models of the corresponding category from the model library and load them into a specified location in the current 3D space. However, due to the large number of similar models in the model library, users often have to manually filter through a massive number of candidate instances when selecting a certain type of model. The filtering process lacks effective reference and intelligent guidance. Furthermore, existing technologies generally lack a mechanism for understanding the overall visual characteristics of 3D space, and this lack of integration with model filtering and recommendation leads to low design efficiency on the user end. Summary of the Invention

[0005] To address this, the present invention provides a data processing method based on building structure model design, which overcomes the problems in the prior art where, due to the large amount of data, users usually have to select from a large number of models when calling a single type of model, and the design software lacks an understanding mechanism of the overall visual characteristics of three-dimensional space and is not combined with model selection and recommendation, resulting in low design efficiency on the user side.

[0006] To achieve the above objectives, the present invention provides a data processing method based on building structure model design, comprising: Several three-dimensional spaces are obtained from the design sample library, and scene visual features are extracted for the three-dimensional spaces, including obtaining two-dimensional images from different perspectives based on the viewfinder, and obtaining several dimensions of visual features based on each of the two-dimensional images. Feature constraint ranges are selected in each dimension to construct a visual constraint range combination portrait in three-dimensional space. The three-dimensional space corresponding to the visual constraint range combination portrait is clustered to identify the model bodies in the three-dimensional space, so as to determine the correlation parameters between the model bodies and construct the correlation database of the visual constraint range combination portrait mapping. In response to the detection of a user executing a predetermined command, scene features are extracted based on the current three-dimensional space to match the corresponding visual constraint range combined image and select the associated database; Verify the visual association fuzzy category in 3D space based on the association database, and push model volumes based on the visual association fuzzy category, including, The model body type is determined according to the predetermined instructions, the field of view is adjusted to extract scene visual features, and the visual features of each dimension are reconstructed to correct the selected association database. The model body to be pushed is determined based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the association database. Alternatively, the model body to be pushed can be determined directly based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding model bodies in the associated database; The predetermined instruction includes retrieving the model body.

[0007] Furthermore, the process of acquiring visual features in several dimensions includes, Create a viewfinder at the center of a three-dimensional space to obtain two-dimensional images from different perspectives; Identify several object contours in each of the two-dimensional images, extract the texture density of local images within the object contours, and set the mean texture density as a texture dimension feature. Extract the gradient direction histogram of the local image within the contour of each object to determine the kurtosis of the gradient direction histogram, and determine the mean kurtosis as the line dimension feature; Line detection is performed on the object's outline, and the ratio of the straight line length to the total length of the object's outline is determined as the geometric dimension feature. Extract the joint distribution histogram of saturation and brightness of the two-dimensional image, calculate the color style shift based on the distance between the joint distribution peak point and the preset neutral reference point, and set the mean of the color style shift as the color dimension feature. The visual features include texture distribution dimension features, line dimension features, geometric dimension features, and color dimension features.

[0008] Furthermore, feature constraint ranges are selected in each dimension to construct a combined image of visual constraint ranges in three-dimensional space. Determine the distribution range of visual features for each dimension, and divide the distribution range into several sub-intervals to obtain several feature constraint ranges corresponding to each dimension; Select individual feature constraint ranges in each dimension to form a combined image of several visual constraint ranges; The combination of feature constraint ranges contained in each of the aforementioned visual constraint range combination images is different.

[0009] Furthermore, the process of identifying model objects in three-dimensional space, determining the correlation parameters between each model object, and constructing the correlation database of the visual constraint range combined image mapping includes, Determine the probability of simultaneous occurrence of model volumes in three-dimensional space; Extract the visual features of each dimension corresponding to the model body, and determine the visual feature compatibility parameters between model bodies; The correlation parameter is obtained by normalizing and weighting the probability and visual feature compatibility parameters; Record the corresponding correlation parameters between each model body, construct a correlation database, and establish a mapping relationship between the correlation database and the combined image of visual constraint range.

[0010] Furthermore, the process of determining the visual feature compatibility parameters between model bodies includes, The mean ratio of visual feature differences among several model objects in each three-dimensional space within the statistical design sample library is used as the calibration quantile. For two model bodies, determine the offset ratio of the visual feature difference ratio between the two model bodies in each three-dimensional space relative to the calibration value, and solve for the mean of the offset ratio as the visual feature compatibility parameter between the two model bodies.

[0011] The offset ratio is the ratio of the absolute difference between two values ​​to the mean of the two values.

[0012] Furthermore, the process of extracting scene features based on the current three-dimensional space to match the corresponding visual constraint range and combine the images includes, By extracting scene visual features, the feature constraint range of each dimension of visual features is determined, and the visual constraint range combination image corresponding to the current 3D scene is determined. Determine the associated database for the visual constraint range combination image mapping, and select the associated database.

[0013] Furthermore, the process of verifying the visual association fuzzy category in 3D space based on the association database includes, Determine the correlation parameters between model bodies in 3D space based on the associated database; The variance of the correlation parameter is calculated. If the variance is greater than or equal to a predetermined variance threshold, the three-dimensional space is determined to be a visually correlated fuzzy category. If the variance is less than a predetermined variance threshold, the three-dimensional space is determined to be a visually associated non-fuzzy category.

[0014] Furthermore, when pushing model objects based on visual association fuzzy categories, If the three-dimensional space is a visually associated fuzzy category, the corresponding type of the model body is determined according to the predetermined instructions, the field of view is adjusted to extract scene visual features, and the visual features of each dimension are reconstructed to correct the selected association database. The model body to be pushed is determined based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the association database. If the three-dimensional space is a visually associated non-fuzzy category, then the model body to be pushed is determined directly based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the associated database.

[0015] Furthermore, the process of adjusting the framing range to extract scene visual features and reconstructing visual features across various dimensions to correct the selected associated database includes... The model body is determined according to the predetermined instructions, the field of view of the viewfinder is reduced, the scene visual features are extracted, and the visual features of each dimension are determined. Based on the visual features of each dimension, match the corresponding visual constraint range to combine the portrait; The associated database for mapping is determined based on the combined portraits within the aforementioned visual constraint range; Specifically, the two-dimensional image collected during scene visual feature extraction must contain the location corresponding to the model body.

[0016] Furthermore, the process of determining the model body to be pushed based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding model bodies in the associated database includes: The type corresponding to the retrieval model body is determined according to the predetermined instructions; Determine the average value of the association parameters between the same type of model body and the remaining model bodies in the three-dimensional space in the association database; Sort the model bodies of the same type in the associated database in descending order based on the mean of the associated parameters; A predetermined number of model bodies are selected from the beginning of the sorted sequence and pushed out.

[0017] Compared with existing technologies, this invention extracts visual features from the 3D space within a design sample library, constructs a visual constraint range combination profile corresponding to the 3D space by selecting feature constraint ranges in each dimension, and builds a dedicated association database for different visual constraint range combination profiles. The association database stores the association parameters between different model bodies. Subsequently, during the creation of 3D space on the user's end, after the user executes a predetermined command, the association database is selected based on the current 3D space, and the visual association fuzziness category of the 3D space is verified. Then, model bodies are adaptively pushed. This invention constructs a visual feature understanding mechanism during the creation of 3D space on the user's end, and adaptively pushes model bodies that conform to the visual features of the current 3D space, reducing the browsing and filtering burden when the user calls model bodies and improving design efficiency.

[0018] In particular, this invention extracts scene visual features based on the three-dimensional space in the design sample library, determines visual features in different dimensions, and forms a feature description system for the visual style of three-dimensional space by extracting visual features in four dimensions: texture distribution, lines, geometry, and color. In practice, the introduction of texture dimension features can effectively capture the texture of the surface of objects in space; the line dimension features use the kurtosis of the gradient direction histogram as a quantitative indicator to characterize the directional concentration of the edge lines of objects in space; the geometry dimension features reflect the regularity of the geometric shape of objects in space through the proportion of straight lines to the total length of the object's outline; and the color dimension features can intuitively reflect the color tendency of space. Through the multi-dimensional feature description system of three-dimensional spatial visual style, data support is provided for the subsequent creation of visual constraint range combination portraits.

[0019] In particular, this invention constructs a visual constraint range combination portrait in three-dimensional space. In actual design scenarios, different design styles exhibit significant differences in visual feature distribution patterns in three-dimensional space. For example, minimalist style typically corresponds to a combination of visual features such as high geometric linear ratio, low texture density, and low color style offset, while classical style often exhibits low geometric linear ratio, high texture density, and high color offset. The visual constraint range combination portrait achieves a multi-dimensional quantitative definition of design styles. Furthermore, the matching habits and co-occurrence patterns between model bodies in the three-dimensional space of different design styles are fundamentally different. Based on this, the visual constraint range combination portrait is used to summarize the corresponding three-dimensional... The association parameters between model bodies in 3D space comprehensively reflect the co-occurrence probability and visual compatibility of model bodies under specific visual constraint range combinations in historical design samples. A dedicated association database for visual constraint range combinations is constructed to reflect the matching of model bodies in different 3D spaces. Furthermore, since the 3D space is progressively formed during the user's design process, the use of visual constraint range combinations enables real-time perception of the current 3D spatial scene. This facilitates the user's subsequent use of visual feature understanding mechanisms to adaptively push model bodies that conform to the current 3D spatial visual features, reducing the amount of browsing required when the user calls model bodies and improving design efficiency.

[0020] In particular, this invention determines the visual association ambiguity category by considering the gradual nature of the user's construction of the 3D space. During the process, there may be discrete relationships between model objects, which are identified by the system as visual association ambiguity categories. In this case, it indicates that the current spatial style may not have converged clearly. The invention adopts the method of adjusting the framing range to use a smaller framing range, reducing interference features, and re-extracting scene visual features, focusing on the visual features of the space near the model object that the user wants to create or change. Based on this, the selected association database is corrected, which facilitates more accurate push of model objects in the future. As a result, when the user searches for models, models that are more in line with the visual features of the 3D space are pushed, reducing the burden of searching and filtering a large model library on the user's end and improving design efficiency.

[0021] In particular, by using the remaining model bodies that already exist in the three-dimensional space as contextual constraints, and by searching the associated database for candidate model bodies of the same type that have high correlation parameters with the remaining model bodies, since the correlation parameters comprehensively reflect the co-occurrence probability and visual compatibility between model bodies in historical design samples, after selecting the associated database, the present invention can prioritize presenting the model bodies that are most reasonably matched with the existing model bodies in the current space and have the most harmonious style, thereby reducing the burden of searching and filtering in a large model library on the user end and improving design efficiency. Attached Figure Description

[0022] Figure 1A schematic diagram illustrating the steps of a data processing method based on a building structure model design, as described in an embodiment of the invention. Figure 2 This is a logic block diagram for scene feature extraction based on the current three-dimensional space, as shown in the embodiment of the invention. Figure 3 A logic block diagram for verifying the visual association fuzzy category in three-dimensional space in an embodiment of the invention; Figure 4 This is a logic block diagram of a visual association-based fuzzy category push model body, which is an embodiment of the invention. Detailed Implementation

[0023] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0024] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0025] Please see Figure 1 as well as Figure 2 As shown, Figure 1 This is a schematic diagram illustrating the steps of a data processing method based on a building structure model design, as described in an embodiment of the invention. Figure 2 The present invention provides a logical flowchart for scene feature extraction based on the current three-dimensional space. The data processing method based on building structure model design in this invention includes: Step S1: Obtain several three-dimensional spaces from the design sample library, and extract scene visual features for the three-dimensional spaces, including obtaining two-dimensional images from different perspectives based on the viewfinder, and obtaining several dimensions of visual features based on each of the two-dimensional images. Step S2: Select feature constraint ranges in each dimension to construct a visual constraint range combination portrait corresponding to the three-dimensional space. Cluster the three-dimensional space corresponding to the visual constraint range combination portrait, identify the model bodies in the three-dimensional space, determine the correlation parameters between the model bodies, and construct the correlation database of the visual constraint range combination portrait mapping. Step S3: In response to detecting that the user terminal executes a predetermined command, scene features are extracted based on the current three-dimensional space to match the corresponding visual constraint range combined image and select the associated database; Step S4: Verify the visual association fuzzy category in 3D space based on the association database, and push the model volume based on the visual association fuzzy category, including: The model body type is determined according to the predetermined instructions, the field of view is adjusted to extract scene visual features, and the visual features of each dimension are reconstructed to correct the selected association database. The model body to be pushed is determined based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the association database. Alternatively, the model body to be pushed can be determined directly based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding model bodies in the associated database; The predetermined instruction includes retrieving the model body.

[0026] Specifically, there are no restrictions on how to create a 3D space. For example, users can create a blank 3D space scene using existing 3D design software, or they can open a saved historical design project file in the aforementioned software and edit and modify the existing 3D space.

[0027] The viewfinder is a constructed virtual image acquisition device used to capture scene images in three-dimensional space at a set viewpoint position and viewing direction, and generate corresponding two-dimensional images.

[0028] There are no restrictions on how the design sample library is constructed. The design sample library is used to store several completed 3D spaces as a priori data basis. The 3D spaces in the design sample library can come from various channels, such as historical project files accumulated by authorized design companies or individuals in their long-term design practice, exemplary 3D space cases with marked style types obtained from public design resource platforms, and of course, other legal sources can also be used, which will not be elaborated here.

[0029] Specifically, in the actual operation of 3D design software, one of the most common ways for users to add model objects to space is to search the model library by keywords and then select a model object to create a model object in 3D space or replace a model object in 3D space.

[0030] Specifically, when monitoring user-side instructions, user-side authorization is required. There are no restrictions on the method of pushing model objects. For example, the model objects to be pushed can be displayed in the user interface in the form of a list. Users can select the required model objects from the list and load them into the specified position in the current three-dimensional space. Of course, those skilled in the art can choose appropriate front-end display and interaction methods according to the actual application scenario, which will not be elaborated here.

[0031] In the design field, model objects can be categorized based on the scene. Taking interior design as an example, possible categories for model objects include chairs, tables, sofas, background paintings, etc., which will not be elaborated further. Specifically, the process of acquiring visual features in several dimensions includes, Create a viewfinder at the center of a three-dimensional space to obtain two-dimensional images from different perspectives; Identify several object contours in each of the two-dimensional images, extract the texture density of local images within the object contours, and set the mean texture density as a texture dimension feature. Extract the gradient direction histogram of the local image within the contour of each object to determine the kurtosis of the gradient direction histogram, and determine the mean kurtosis as the line dimension feature; Line detection is performed on the object's outline, and the ratio of the straight line length to the total length of the object's outline is determined as the geometric dimension feature. Extract the joint distribution histogram of saturation and brightness of the two-dimensional image, calculate the color style shift based on the distance between the joint distribution peak point and the preset neutral reference point, and set the mean of the color style shift as the color dimension feature. The visual features include texture distribution dimension features, line dimension features, geometric dimension features, and color dimension features.

[0032] In practice, the viewfinder can be set at the center of the three-dimensional space and rotated in the horizontal plane at preset angular intervals to acquire a sequence of two-dimensional images from multiple perspectives covering the entire three-dimensional space. For example, the rotation angle can be 90°.

[0033] There are no restrictions on the method for recognizing object contours. For example, image segmentation algorithms can be used to recognize the edges of object contours. Of course, other methods can also be used, which will not be elaborated here.

[0034] When determining texture density, first determine the local image of the object's outline, and use the contrast of the pixel gray-level co-occurrence matrix of the local image as the texture density. The larger the contrast value, the more significant the pixel gray-level difference in the image, the deeper the texture grooves, and the rougher the visual appearance; the smaller the contrast value, the smaller the pixel gray-level difference, the shallower the texture, and the smoother the visual appearance.

[0035] Kurtosis is a descriptive statistical measure. The kurtosis of a gradient direction histogram is used to describe the steepness of the gradient direction in a local area. The higher the kurtosis, the more concentrated the gradient direction is, and there are a large number of parallel and perpendicular lines in the space. The lower the kurtosis, the more uniform the distribution of the gradient direction is, and there are more curved lines in the space.

[0036] There are no restrictions on the method for line detection. For example, the Hough transform algorithm can be used to detect line segments on the object contour, extract the set of pixels that satisfy the geometric constraints of the line in the contour edge, count the sum of the pixel lengths of all detected line segments, and use the ratio of this to the total pixel length of the object contour to represent the ratio of the line length to the total length of the object contour, thus obtaining the geometric dimension features.

[0037] The process of constructing a joint histogram of saturation and brightness distribution includes, The two-dimensional image is converted from the RGB color space to the HSV color space. For each pixel in the image, its saturation channel value S and lightness channel value V are extracted. The saturation and brightness values ​​of all pixels in the image are statistically analyzed, and a joint distribution histogram of saturation and brightness is constructed. The horizontal axis of the histogram represents the saturation interval, and the vertical axis represents the brightness interval. The value of each cell represents the number of pixels falling into that interval. The number of saturation intervals and brightness intervals is 20 each, which can form a 20×20 joint distribution histogram. When determining the peak point of the joint distribution, the cell with the most occurrences in the histogram of the joint distribution of saturation and brightness is identified, and the median value of the saturation interval and the median value of the brightness interval corresponding to the cell are used as the coordinates of the peak point of the joint distribution, reflecting the dominant color attribute in three-dimensional space; The preset neutral reference point is set as the coordinate point corresponding to a saturation of 0 and a brightness of 0.5. A saturation of 0 represents no color, and a brightness of 0.5 represents medium brightness, in order to reflect a neutral visual state without color bias.

[0038] The Euclidean distance between the joint distribution peak point and the neutral reference point is calculated and used as the color style offset. Different color style offsets can reflect different color styles when they are in different intervals.

[0039] Specifically, feature constraint ranges are selected in each dimension to construct a combined image of visual constraint ranges in three-dimensional space. Determine the distribution range of visual features for each dimension, and divide the distribution range into several sub-intervals to obtain several feature constraint ranges corresponding to each dimension; Select individual feature constraint ranges in each dimension to form a combined image of several visual constraint ranges; The combination of feature constraint ranges contained in each of the aforementioned visual constraint range combination images is different.

[0040] In implementation, the visual features of each dimension are divided into three distribution intervals, which is sufficient to cover the visual feature combination patterns of common design styles. Furthermore, it avoids the problem of the number of visual constraint range combination images surging due to overly fine interval division. For example, if divided into 4 or 5 distribution intervals, the number of visual constraint range combination images surges to 256 or 625, while if only two intervals are divided, the number of visual constraint range combination images is only 16, which is difficult to cover the visual feature combination patterns of common design styles.

[0041] For the boundaries of the distribution intervals of visual features in each dimension, a pre-defined statistical method is used to statistically analyze the visual features in each dimension as sample values. For a single-dimensional visual feature, the 33rd percentile and 67th percentile of the corresponding sample value distribution are used as two dividing points to divide the distribution into three intervals.

[0042] Specifically, the process of identifying model objects in three-dimensional space, determining the correlation parameters between the model objects, and constructing an associated database of the visual constraint range combined image mapping includes: Determine the probability of simultaneous occurrence of model volumes in three-dimensional space; Extract the visual features of each dimension corresponding to the model body, and determine the visual feature compatibility parameters between model bodies; The correlation parameter is obtained by normalizing and weighting the probability and visual feature compatibility parameters; Record the corresponding correlation parameters between each model body, construct a correlation database, and establish a mapping relationship between the correlation database and the combined image of visual constraint range.

[0043] Specifically, during normalization, the probability and visual feature compatibility parameters are normalized to the interval [0, 1]. During weighted summation, in order to comprehensively consider the influence of the visual feature compatibility parameters, the weights are set to 0.5 respectively.

[0044] This invention constructs a visual constraint range combination portrait in three-dimensional space. In actual design scenarios, different design styles exhibit significant differences in the distribution patterns of visual features in three-dimensional space. For example, minimalist style typically corresponds to a combination of visual features such as high geometric linear ratio, low texture density, and low color style offset, while classical style often exhibits low geometric linear ratio, high texture density, and high color offset. The visual constraint range combination portrait achieves a multi-dimensional quantitative definition of design styles. Furthermore, the matching habits and co-occurrence patterns between model bodies in the three-dimensional space of different design styles are fundamentally different. Based on this, the visual constraint range combination portrait is used to summarize the corresponding three-dimensional... The association parameters between model bodies in the space comprehensively reflect the co-occurrence probability and visual compatibility of model bodies under specific visual constraint range combinations in historical design samples. A dedicated association database for visual constraint range combinations is constructed to reflect the matching situation between model bodies in different 3D spaces. Furthermore, since the 3D space is progressively formed during the user's design process, the use of visual constraint range combinations enables real-time perception of the current 3D spatial scene's visuals. This facilitates the user's subsequent use of visual feature understanding mechanisms to adaptively push model bodies that conform to the current 3D spatial visual features, reducing the amount of browsing required when the user accesses model bodies and improving design efficiency.

[0045] Specifically, the process of determining the visual feature compatibility parameters between model bodies includes, The mean ratio of visual feature differences among several model objects in each three-dimensional space within the statistical design sample library is used as the calibration quantile. For two model bodies, determine the offset ratio of the visual feature difference ratio between the two model bodies in each three-dimensional space relative to the calibration value, and solve for the mean of the offset ratio as the visual feature compatibility parameter between the two model bodies.

[0046] The offset ratio is the ratio of the absolute difference between two values ​​to the mean of the two values.

[0047] It is understandable that there are many model objects, and the corresponding visual feature compatibility parameters need to be determined for each model object.

[0048] The visual feature difference ratio between model objects is the mean of the visual feature difference ratios of each dimension. The difference ratios of visual features for each dimension of the two model objects are calculated separately, and the mean of each difference ratio is taken as the difference ratio of visual features between the model objects.

[0049] The visual features of the model objects in each dimension are obtained based on the two-dimensional images containing the model objects. The process includes: Identify the outline of the object corresponding to the model body, and extract the texture density of the local image within the outline of the object body as the texture dimension feature of the model body. Extract the gradient direction histogram of the local image within the contour of each object to determine the kurtosis of the gradient direction histogram, and use the kurtosis as the line dimension feature of the model body. Line detection is performed on the object's outline, and the ratio of the straight line length to the total length of the object's outline is determined as the geometric dimension feature of the model body. Extract the saturation and brightness joint distribution histogram of the local image within the body outline, calculate the color style shift based on the distance between the joint distribution peak point and the preset neutral reference point, and determine the color style shift as a color dimension feature. Specifically, the process of extracting scene features based on the current 3D space and matching them with corresponding visual constraints to create a composite image includes, By extracting scene visual features, the feature constraint range of each dimension of visual features is determined, and the visual constraint range combination image corresponding to the current 3D scene is determined. Determine the associated database for the visual constraint range combination image mapping, and select the associated database.

[0050] Specifically, please refer to Figure 3 The diagram shown is a logical block diagram of a method for verifying the visual association fuzzy category of a three-dimensional space according to an embodiment of the invention. The process of verifying the visual association fuzzy category of a three-dimensional space based on an association database includes: Determine the correlation parameters between model bodies in 3D space based on the associated database; The variance of the correlation parameter is calculated. If the variance is greater than or equal to a predetermined variance threshold, the three-dimensional space is determined to be a visually correlated fuzzy category. If the variance is less than a predetermined variance threshold, the three-dimensional space is determined to be a visually associated non-fuzzy category.

[0051] Specifically, the variance threshold is a pre-set value. It can be obtained by taking the normal distribution of the variance based on the correlation parameter variance between each model body in each three-dimensional space in the design sample library, determining the 95% confidence interval, and setting the upper limit of the 95% confidence interval as the variance threshold.

[0052] Specifically, please refer to Figure 4 As shown, it is a logic block diagram of a visual association fuzzy category push model body according to an embodiment of the invention. When pushing model bodies based on visual association fuzzy categories, If the three-dimensional space is a visually associated fuzzy category, the corresponding type of the model body is determined according to the predetermined instructions, the field of view is adjusted to extract scene visual features, and the visual features of each dimension are reconstructed to correct the selected association database. The model body to be pushed is determined based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the association database. If the three-dimensional space is a visually associated non-fuzzy category, then the model body to be pushed is determined directly based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the associated database.

[0053] This invention identifies visual association ambiguity categories. Considering the gradual nature of users constructing 3D space, there may be discrete relationships between model objects during the process, which the system identifies as visual association ambiguity categories. In this case, it indicates that the current spatial style may not have converged clearly. The system adjusts the framing range to a smaller range, reducing interfering features, and re-extracts scene visual features, focusing on the visual features of the space near the model object that the user wants to create or change. Based on this, the selected association database is corrected, which facilitates more accurate model object recommendations in the future. As a result, when users search for models, models that better match the visual features of 3D space are recommended, reducing the burden of searching and filtering a large model library on the user's end and improving design efficiency.

[0054] By using the remaining model bodies that already exist in the three-dimensional space as contextual constraints, and by searching the associated database for candidate model bodies of the same type that have high correlation parameters with the remaining model bodies, since the correlation parameters comprehensively reflect the co-occurrence probability and visual compatibility between model bodies in historical design samples, this invention can prioritize presenting the model bodies that are most reasonably matched with the existing model bodies in the current space and have the most harmonious style after selecting the associated database. This reduces the burden of searching and filtering a large number of model libraries on the user end and improves design efficiency.

[0055] Specifically, the process of adjusting the framing range to extract scene visual features and reconstructing visual features in various dimensions to correct the selected associated database includes, The model body is determined according to the predetermined instructions, the field of view of the viewfinder is reduced, the scene visual features are extracted, and the visual features of each dimension are determined. Based on the visual features of each dimension, match the corresponding visual constraint range to combine the portrait; The associated database for mapping is determined based on the combined portraits within the aforementioned visual constraint range; Specifically, the two-dimensional image collected during scene visual feature extraction must contain the location corresponding to the model body.

[0056] In practice, optionally, the initial framing range is set based on the ability to completely cover at least one wall and the area it encloses in the current 3D space, in order to acquire the visual features of the global scene. When local feature reconstruction is required, the framing range is reduced to approximately 50% of the initial framing range, centered on the model body targeted by the predetermined command. The reduced framing range focuses on the model body and its adjacent contextual space region, effectively filtering out visual interference from distant areas and irrelevant walls while ensuring that the acquired image includes the model body.

[0057] In practice, after the viewfinder narrows its range, it can use the geometric center of the model as a reference point so that the viewfinder's line of sight points towards the model, and the framing range corresponds to the narrowed framing range, so as to obtain two-dimensional images from multiple perspectives for scene visual feature extraction.

[0058] Specifically, the process of determining the model body to be pushed based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding model bodies in the associated database includes: The type corresponding to the retrieval model body is determined according to the predetermined instructions; Determine the average value of the association parameters between the same type of model body and the remaining model bodies in the three-dimensional space in the association database; Sort the model bodies of the same type in the associated database in descending order based on the mean of the associated parameters; A predetermined number of model bodies are selected from the beginning of the sorted sequence and pushed out.

[0059] It is understandable that when users search for model objects, the search terms they enter usually match the keywords corresponding to each category during classification, which can directly determine the corresponding type. For example, if the categories include chairs and tables, users can enter chairs as the search term.

[0060] In practice, the preset quantity can be determined based on the preset display position of the current user interface, or it can be a fixed value set by those skilled in the art, which will not be elaborated here.

[0061] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A data processing method based on building structure model design, characterized in that, include: Several three-dimensional spaces are obtained from the design sample library, and scene visual features are extracted for the three-dimensional spaces, including obtaining two-dimensional images from different perspectives based on the viewfinder, and obtaining several dimensions of visual features based on each of the two-dimensional images. Feature constraint ranges are selected in each dimension to construct a visual constraint range combination portrait in three-dimensional space. The three-dimensional space corresponding to the visual constraint range combination portrait is clustered to identify the model bodies in the three-dimensional space, so as to determine the correlation parameters between the model bodies and construct the correlation database of the visual constraint range combination portrait mapping. In response to the detection of a user executing a predetermined command, scene features are extracted based on the current three-dimensional space to match the corresponding visual constraint range combined image and select the associated database; Verify the visual association fuzzy category in 3D space based on the association database, and push model volumes based on the visual association fuzzy category, including, The model body type is determined according to the predetermined instructions, the field of view is adjusted to extract scene visual features, and the visual features of each dimension are reconstructed to correct the selected association database. The model body to be pushed is determined based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the association database. Alternatively, the model body to be pushed can be determined directly based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding model bodies in the associated database; The predetermined instruction includes retrieving the model body.

2. The data processing method based on building structure model design according to claim 1, characterized in that, The process of acquiring visual features in several dimensions includes, Create a viewfinder at the center of a three-dimensional space to obtain two-dimensional images from different perspectives; Identify several object contours in each of the two-dimensional images, extract the texture density of local images within the object contours, and set the mean texture density as a texture dimension feature. Extract the gradient direction histogram of the local image within the contour of each object to determine the kurtosis of the gradient direction histogram, and determine the mean kurtosis as the line dimension feature; Line detection is performed on the object's outline, and the ratio of the straight line length to the total length of the object's outline is determined as the geometric dimension feature. Extract the joint distribution histogram of saturation and brightness of the two-dimensional image, calculate the color style shift based on the distance between the joint distribution peak point and the preset neutral reference point, and set the mean of the color style shift as the color dimension feature. The visual features include texture distribution dimension features, line dimension features, geometric dimension features, and color dimension features.

3. The data processing method based on building structure model design according to claim 2, characterized in that, By selecting feature constraint ranges in each dimension, a combined image of visual constraint ranges corresponding to the three-dimensional space is constructed. Determine the distribution range of visual features for each dimension, and divide the distribution range into several sub-intervals to obtain several feature constraint ranges corresponding to each dimension; Select individual feature constraint ranges in each dimension to form a combined image of several visual constraint ranges; The combination of feature constraint ranges contained in each of the aforementioned visual constraint range combination images is different.

4. The data processing method based on building structure model design according to claim 3, characterized in that, The process of identifying model objects in three-dimensional space, determining the correlation parameters between the model objects, and constructing a correlation database of the combined image mapping of the visual constraint range includes: Determine the probability of simultaneous occurrence of model volumes in three-dimensional space; Extract the visual features of each dimension corresponding to the model body, and determine the visual feature compatibility parameters between model bodies; The correlation parameter is obtained by normalizing and weighting the probability and visual feature compatibility parameters; Record the corresponding correlation parameters between each model body, construct a correlation database, and establish a mapping relationship between the correlation database and the combined image of visual constraint range.

5. The data processing method based on building structure model design according to claim 4, characterized in that, The process of determining the visual feature compatibility parameters between model objects includes, The mean ratio of visual feature differences among several model objects in each three-dimensional space within the statistical design sample library is used as the calibration quantile. For two model bodies, determine the offset ratio of the visual feature difference ratio between the two model bodies in each three-dimensional space relative to the calibration value, and solve for the mean of the offset ratio as the visual feature compatibility parameter between the two model bodies.

6. The data processing method based on building structure model design according to claim 1, characterized in that, The process of extracting scene features based on the current 3D space and matching them with corresponding visual constraints to create a composite image includes: By extracting scene visual features, the feature constraint range of each dimension of visual features is determined, and the visual constraint range combination image corresponding to the current 3D scene is determined. Determine the associated database for the visual constraint range combination image mapping, and select the associated database.

7. The data processing method based on building structure model design according to claim 1, characterized in that, The process of verifying visual association fuzzy categories in 3D space based on an association database includes: Determine the correlation parameters between model bodies in 3D space based on the associated database; The variance of the correlation parameter is calculated. If the variance is greater than or equal to a predetermined variance threshold, the three-dimensional space is determined to be a visually correlated fuzzy category. If the variance is less than a predetermined variance threshold, the three-dimensional space is determined to be a visually associated non-fuzzy category.

8. The data processing method based on building structure model design according to claim 7, characterized in that, When pushing model objects based on visual association fuzzy categories If the three-dimensional space is a visually associated fuzzy category, the corresponding type of the model body is determined according to the predetermined instructions, the field of view is adjusted to extract scene visual features, and the visual features of each dimension are reconstructed to correct the selected association database. The model body to be pushed is determined based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the association database. If the three-dimensional space is a visually associated non-fuzzy category, then the model body to be pushed is determined directly based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding type model bodies in the associated database.

9. The data processing method based on building structure model design according to claim 1, characterized in that, The process of adjusting the framing range to extract scene visual features and reconstructing visual features in various dimensions to correct the selected associated database includes: Reduce the field of view of the viewfinder, extract the visual features of the scene, and determine the visual features in each dimension; Based on the visual features of each dimension, match the corresponding visual constraint range to combine the portrait; The associated database for mapping is determined based on the combined portraits within the aforementioned visual constraint range; Specifically, the two-dimensional image collected during scene visual feature extraction must contain the location corresponding to the model body.

10. The data processing method based on building structure model design according to claim 9, characterized in that, The process of determining the model body to be pushed based on the association parameters between the remaining model bodies in the three-dimensional space and the corresponding model bodies in the associated database includes: The type corresponding to the retrieval model body is determined according to the predetermined instructions; Determine the average value of the association parameters between the same type of model body and the remaining model bodies in the three-dimensional space in the association database; Sort the model bodies of the same type in the associated database in descending order based on the mean of the associated parameters; A predetermined number of model bodies are selected from the beginning of the sorted sequence and pushed out.

Images