Building transformation method, system and equipment based on style migration and medium

Abstract

The invention discloses a building reconstruction method, system and equipment based on style migration, and a medium. The method comprises the steps of obtaining a first building facade image corresponding to a to-be-reconstructed building and a style building facade image used for style migration reference; extracting building features corresponding to the first building facade image and style building features corresponding to the style building facade image; fusing the building features and the style building features to obtain target building image features; and obtaining a target facade image of the target building according to the target building image features so as to improve the building reconstruction effect.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, system, device and medium for architectural reconstruction based on style migration. Background Art

[0002] With the development of social economy, the unified planning of architectural style has become a difficult problem in the contest between urban expansion and the protection of traditional buildings in villages and towns.

[0003] Different from traditional architectural renovation, style transfer is a technology that transfers high-level abstract features of a building's style, such as color, line style, texture, etc., to another building, making the visual style of the other building similar to that of the building.

[0004] In the existing building renovation, the existing buildings are often directly renovated based on the appearance of the reference building, without considering the characteristics of the building itself, such as the functional zoning of the building itself. As a result, the renovated buildings cannot take into account the living needs of the residents well, which in turn leads to very poor renovation results. Summary of the invention

[0005] In order to solve the above technical problems, the present invention discloses a building renovation method, system, equipment and medium based on style migration to solve the above technical problems and improve the renovation effect.

[0006] In order to achieve the above objectives, in a first aspect, the present invention discloses a method for architectural reconstruction based on style migration, comprising:

[0007] Obtaining a first building facade image corresponding to the building to be transformed and a style building facade image used as a reference for style migration;

[0008] Extracting architectural features corresponding to the first building facade image and architectural style features corresponding to the architectural style facade image;

[0009] The architectural features and the style architectural features are integrated to obtain target architectural image features;

[0010] According to the target building image features, a target facade image of the target building is acquired.

[0011] The present invention discloses a building renovation method based on style transfer, firstly, a facade image of a building to be renovated and a style building facade image used as a reference for style transfer are obtained, so as to perform style transfer on the building to be renovated according to the style building facade image. Secondly, the features of the building facade image and the features of the style building facade image are extracted respectively, and the features are fused, and the target building image of the building corresponding to the building to be renovated after the style transfer is performed is constructed using the fused image features, so that the target building image not only fits the transferred style, but also adapts to the features of the original building, thereby improving the effect of building renovation.

[0012] As a preferred example, the step of obtaining a first building facade image corresponding to the building to be transformed and a style building facade image used as a reference for style transfer includes:

[0013] Acquire a first initial high-resolution facade image corresponding to the building to be renovated;

[0014] Obtaining a second initial high-resolution facade image of a style building for style transfer reference;

[0015] Perform image preprocessing, denoising and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image respectively to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

[0016] The present invention obtains high-frequency images to provide accurate data basis for feature extraction, and then performs preprocessing, denoising and image enhancement operations on the images to further enhance the features of the images themselves, thereby ensuring the accuracy of the features extracted subsequently.

[0017] As a preferred example, before extracting the architectural features corresponding to the first building facade image and the style architectural features corresponding to the style building facade image, the method includes:

[0018] Generate an initial style transfer model based on a generative adversarial network;

[0019] Obtain a high-resolution building facade image dataset, and draw the building outline, functional zoning map, and building facade map corresponding to each high-resolution facade image;

[0020] Processing the high-resolution facade image according to the initial style transfer model, and outputting a training functional zoning map and a training building facade map corresponding to each of the high-resolution facade images;

[0021] Obtaining a first error between the training functional zoning diagram and the functional zoning diagram and a second error between the training building elevation diagram and the building elevation diagram;

[0022] The model parameters of the initial style transfer model are continuously adjusted according to the first error and the second error until the first error and the second error meet a preset error condition, thereby obtaining a style transfer model.

[0023] The present invention constructs the initial model based on the adversarial network to ensure the accuracy of the image output by the model. Further, the model is trained using the constructed data set to improve the accuracy of the model in extracting image features.

[0024] As a preferred example, the extracting of architectural features corresponding to the first building facade image and style architectural features corresponding to the style building facade image includes:

[0025] The first building facade image and the style building facade image are respectively convolved by the style transfer model to obtain building features corresponding to the first building facade image and style building features corresponding to the style building facade image.

[0026] The present invention performs feature extraction of images based on a pre-trained model, which not only ensures the accuracy of feature extraction, but also improves the efficiency of image output through the model.

[0027] As a preferred example, the fusing of the architectural features and the style architectural features to obtain the target architectural image features includes:

[0028] The style building features are fused into the building features by a style transfer algorithm preset in the style transfer model to obtain the target building image features.

[0029] The present invention utilizes a style migration algorithm to fuse the style architectural features into the architectural features, so that the fused features contain both the features of the building itself and the features of the migrated style, so that the renovated building completes the style migration without destroying the original building itself, thereby improving the effect of the renovation.

[0030] As a preferred example, acquiring a target facade image of a target building according to the target building image feature includes:

[0031] According to the target building image features, a corresponding functional zoning map is generated by a generator preset in the style transfer model;

[0032] Based on the functional partition diagram, the generator is used to make detailed adjustments to obtain a rendering corresponding to each functional partition;

[0033] A target facade image is obtained according to the effect diagram and the functional zoning diagram.

[0034] The present invention first generates a functional zoning map according to the image features of the target building, and then generates a rendering based on the functional zoning map, showing the details and overall design style of different functional areas, thereby improving the effect of the renovation.

[0035] In a second aspect, the present invention discloses a building transformation system based on style migration, including an image acquisition module, a feature extraction module, a feature fusion module and a building transformation module;

[0036] The image acquisition module is used to acquire a first building facade image corresponding to the building to be transformed and a style building facade image used as a reference for style migration;

[0037] The feature extraction module is used to extract the building features corresponding to the first building facade image and the style building features corresponding to the style building facade image;

[0038] The feature fusion module is used to fuse the architectural features and the style architectural features to obtain target architectural image features;

[0039] The building reconstruction module is used to obtain a target facade image of a target building according to the image features of the target building.

[0040] The present invention discloses a building renovation system based on style transfer, which first obtains a facade image of a building to be renovated and a style building facade image used as a reference for style transfer, so as to perform style transfer on the building to be renovated according to the style building facade image. Secondly, the features of the building facade image and the features of the style building facade image are extracted respectively, and the features are fused, and the target building image of the building corresponding to the building to be renovated after the style transfer is constructed using the fused image features, so that the target building image not only fits the transferred style, but also adapts to the features of the original building, thereby improving the effect of building renovation.

[0041] As a preferred example, the image acquisition module includes an acquisition unit and a processing unit;

[0042] The acquisition unit is used to acquire a first initial high-resolution facade image corresponding to the building to be renovated; and acquire a second initial high-resolution facade image of the style building used as a reference for style migration;

[0043] The processing unit is used to perform image preprocessing operations, denoising operations and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image, respectively, to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

[0044] The present invention obtains high-frequency images to provide accurate data basis for feature extraction, and then performs preprocessing, denoising and image enhancement operations on the images to further enhance the features of the images themselves, thereby ensuring the accuracy of the features extracted subsequently.

[0045] In a third aspect, the present invention discloses a computing device, comprising: one or more processors; a memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute a style transfer-based architectural transformation method as described in the first aspect.

[0046] In a fourth aspect, the present invention discloses a computer-readable storage medium, in which a program code is stored. The program code can be called by a processor to execute a building transformation method based on style transfer as described in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 : A schematic flow chart of a building renovation method based on style transfer disclosed in an embodiment of the present invention.

[0048] Figure 2 : A structural schematic diagram of a building renovation system based on style migration disclosed in an embodiment of the present invention.

[0049] Figure 3 : A schematic diagram of a flow chart of a building renovation method based on style transfer disclosed in another embodiment of the present invention;

[0050] Figure 4 : is a structural schematic diagram of a model discriminator disclosed in another embodiment of the present invention;

[0051] Figure 5 : A flowchart of an adaptive learning mechanism disclosed in another embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Embodiment 1

[0054] This embodiment discloses a building reconstruction method based on style migration. For the specific implementation process of the reconstruction method, please refer to Figure 1, mainly including steps 101 to 104, the steps include:

[0055] Step 101: Obtain a first building facade image corresponding to a building to be renovated and a style building facade image used as a reference for style transfer.

[0056] In this embodiment, the step mainly includes: obtaining a first initial high-resolution facade image corresponding to the building to be renovated; obtaining a second initial high-resolution facade image of the style building used as a reference for style migration; performing image preprocessing operations, denoising operations and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image, respectively, to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

[0057] In this embodiment, this step provides accurate data basis for feature extraction by acquiring high-frequency images, and then performs preprocessing, denoising and image enhancement operations on the image to further enhance the features of the image itself, thereby ensuring the accuracy of the features extracted subsequently.

[0058] Step 102: extracting architectural features corresponding to the first building facade image and architectural style features corresponding to the architectural style facade image.

[0059] In this embodiment, this step mainly includes: constructing a style transfer model, wherein the training process of the style transfer model is: generating an initial style transfer model based on a generative adversarial network; obtaining a high-resolution facade image data set of a building, and drawing a building outline map, a functional zoning map, and a building rendering map corresponding to each high-resolution facade image; performing feature extraction on the high-resolution facade image according to the initial style transfer model, so as to output a training functional zoning map and a training building rendering map corresponding to each high-resolution facade image according to the extracted image features; obtaining a first error between the training functional zoning map and the functional zoning map and a second error between the training building rendering map and the building rendering map; and continuously adjusting the model parameters of the initial style transfer model according to the first error and the second error until the first error and the second error meet a preset error condition, thereby obtaining a style transfer model.

[0060] Next, convolution processing is performed on the first building facade image and the style building facade image respectively through the style transfer model to obtain building features corresponding to the first building facade image and style building features corresponding to the style building facade image.

[0061] In this embodiment, this step constructs the initial model based on the adversarial network to ensure the accuracy of the image output by the model. Further, the model is trained using the constructed data set to improve the accuracy of the model in extracting image features. Further, image feature extraction is performed based on the pre-trained model, which not only ensures the accuracy of feature extraction, but also improves the efficiency of image output through the model.

[0062] Step 103: The architectural features and the style architectural features are integrated to obtain target architectural image features.

[0063] In this embodiment, this step mainly includes: fusing the style building features into the building features through a style transfer algorithm preset in the style transfer model to obtain the target building image features.

[0064] In this embodiment, this step utilizes a style transfer algorithm to fuse the style architectural features into the building features, so that the fused features contain both the features of the building itself and the features of the migrated style, so that the renovated building completes the style transfer without destroying the original building itself, thereby improving the effect of the renovation.

[0065] Step 104: Acquire a target facade image of the target building according to the target building image features.

[0066] In this embodiment, this step mainly includes: according to the target building image features, generating a corresponding functional zoning map through a generator preset in the style transfer model; based on the functional zoning map, making detail adjustments through the generator to obtain a rendering corresponding to each functional zoning; according to the rendering and the functional zoning map, obtaining a target facade image.

[0067] In this embodiment, this step first generates a functional zoning map according to the target building image features, and then generates a rendering based on the functional zoning map, showing the details and overall design style of different functional areas to improve the effect of the renovation.

[0068] On the other hand, this embodiment also discloses a building transformation system based on style migration. For the specific structure of the transformation system, please refer to Figure 2 , including an image acquisition module 201, a feature extraction module 202, a feature fusion module 203 and a building transformation module 204.

[0069] The image acquisition module 201 is used to acquire a first building facade image corresponding to the building to be renovated and a style building facade image used as a reference for style transfer.

[0070] The feature extraction module 202 is used to extract the building features corresponding to the first building facade image and the style building features corresponding to the style building facade image.

[0071] The feature fusion module 203 is used to fuse the architectural features and the style architectural features to obtain target architectural image features.

[0072] The building reconstruction module 204 is used to obtain a target facade image of a target building according to the target building image features.

[0073] In this embodiment, the image acquisition module 201 includes an acquisition unit and a processing unit.

[0074] The acquisition unit is used to acquire a first initial high-resolution facade image corresponding to the building to be renovated; and acquire a second initial high-resolution facade image of the style building used as a reference for style migration.

[0075] The processing unit is used to perform image preprocessing operations, denoising operations and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image, respectively, to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

[0076] In addition to the above method and system, this embodiment also discloses a computing device and a computer-readable storage medium. The computing device includes: one or more processors; a memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute a building renovation method based on style transfer as described in this embodiment; and the computer-readable storage medium stores program code, which can be called by the processor to execute a building renovation method based on style transfer as described in this embodiment.

[0077] Embodiment 2

[0078] Specifically, refer to Figure 3 , is a flow chart of a building renovation method based on style migration provided in this embodiment, which mainly includes steps 301 to 304, and the steps are mainly:

[0079] Step 301: Construct a style transfer model based on a generative adversarial network.

[0080] In this embodiment, this step mainly includes: generating an initial style transfer model based on a generative adversarial network; obtaining a high-resolution facade image dataset of a building to train the initial style transfer model until the error between the result output by the model and the corresponding image in the high-resolution facade image dataset reaches a preset error condition, and then outputting the corresponding style transfer model.

[0081] Specifically, in some implementations of this embodiment, an initial style transfer model is first constructed based on a generative adversarial network. The English name of the style transfer model may be Architectur l Facade Generat i ve Adversarial Nets. The style transfer model includes a generator and a discriminator. The structure of the discriminator can refer to Figure 4 , mainly including convolutional layer, activation layer, pooling layer, batch normalization layer, dense layer and output layer.

[0082] like Figure 4 As shown, in the discriminator of the style transfer model, the input layer receives the image generated by the generator or the real image as the starting point of the discrimination; convolution layers 1 and 2 extract the features of the image by using convolution kernels. Each convolution layer extracts features at different levels. The first layer focuses on basic graphics and edges, using larger filters (5x5), while the second layer recognizes more complex image features and uses smaller filters (3x3); the activation layer ReLU introduces nonlinearity to the model to help the model learn complex image features; the activation function f(x) = max(0, γx) is used to increase the nonlinearity of the network, where x is the input of this layer. Among them, γ is automatically adjusted according to the architectural style; pooling layers (such as maximum pooling) are used to reduce the number of parameters and the complexity of calculation, while making feature detection more stable to small changes in the input image. Function f(s) = max(s), which converts non-overlapping subregions of the input s into a single value, that is, the maximum value of each subregion. A 2x2 pooling kernel with a stride of 2 is used to reduce the dimension of the feature map; the batch normalization layer performs a batch normalization operation, which helps the stability of the network during training. It standardizes the input of each layer to make the activation distribution more stable. For the layer input x, batch normalization standardizes the features by the following formula:

[0083]

[0084] Where μB is the mean of batch B, is the variance, and ∈ is a small constant to avoid division by zero.

[0085] The dense layer (multilayer perceptron) is a fully connected layer, which further processes the output data of the flattening layer, extracts and combines features to make the final judgment; function f(x) = W x+b, where W is the weight matrix, b is the bias vector, and x is the input vector. The number of neurons can be adjusted to different numbers, such as 64, 128, 256, etc., depending on the complexity of the model and the amount of training data; the output layer outputs the discrimination result, that is, whether the image is real or generated by the generator. The final output layer uses the sigmoid activation function

[0086]

[0087] Convert the discriminator's score to a probability value between 0 and 1.

[0088] For inclusion Figure 4 The discriminator style transfer model shown in FIG. is trained, and during the training process of the model, an adaptive learning mechanism is used to continuously adjust the generation strategy according to the feedback during the training process to improve the quality and relevance of the generated image. The training process of the model is as follows: Figure 5 As shown, including:

[0089] ① Initial training: The training phase at the beginning of the model uses the initial data set and strategy for training. When constructing the initial data set, high-resolution facade images of the corresponding buildings can be obtained based on the determined renovation style. For example, when the renovation style is Wuyi style, multiple high-resolution facade images of buildings in Wuyi style are obtained. Then, in order to reduce the time of model training and improve the accuracy of the model, the collected high-resolution facade images are redrawn into building outline maps (BEP), functional zoning maps (FSL), and building facade maps (BE), that is, building effects. Figure 3 The building outline map can be a grayscale image with a black color (R0, G0, B0). The functional zoning map is used to mark various elements in the building facade with different colors, and the functional zoning map is presented in the form of a color picture. For example, the door is red, the window is green, the wall is yellow, the guardrail is blue, and the roof is brown. In summary, a data set for training is constructed based on the building outline map (BEP), the functional zoning map (FSL), and the building facade map (BE).

[0090] ② Generate images: Generate images of building facades using the trained model. The model is trained using paired images, i.e., the three types of images after data processing are trained from building outlines to functional zoning maps and from functional zoning maps to building facades. Based on the above training, the model corresponds to two weight models, mode l 1 and mode l2. The above two models cover all data for generating building facades from building outlines, thereby generating building images.

[0091] ③Evaluation images: Evaluate the quality and relevance of the generated images. This includes the loss function and recall rate of the automated evaluation index model. By looking at the loss value (Loss) and the pictures of the training process, you can have a basic judgment on the accuracy of the model. The CGAN loss value line chart shows the stability of the generator and discriminator during the training process. For better training results, the loss value will usually continue to decrease and stabilize within a small range of fluctuations. The pictures of the training process can be used to check whether the generator has generated image noise (image noise). This image noise is likely to deceive the true and false judgment of the discriminator, resulting in a better result in the loss value, but the actual generated effect is poor.

[0092] ④Feedback analysis: Analyze the evaluation results and identify the advantages and disadvantages of the model.

[0093] ⑤ Adjustment strategy: According to the feedback results, adjust the model's learning strategy, including modifying the model's learning rate and the batch size of the training samples.

[0094] ⑥Update model: Apply the adjusted strategy and update the model to improve its performance.

[0095] Step 302: Obtain a first facade image corresponding to the building to be renovated and a second facade image used as a reference for style transfer.

[0096] In this embodiment, this step mainly includes: obtaining a first initial high-resolution facade image corresponding to the building to be renovated and a second initial high-resolution facade image of the style building used as a reference for style migration; performing image preprocessing operations, denoising operations and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

[0097] Specifically, collect original building facade images, including but not limited to various building types such as historical buildings and modern buildings. Use high-resolution cameras to ensure image quality and take into account different lighting conditions and viewing angles. Further, perform preprocessing operations on the images, such as resizing and color adjustment: use OpenCV or Pillow library to resize and standardize the colors. Adjust all collected images to a uniform resolution to ensure consistency in model processing. Standardize the colors of the images to ensure that the images input to the model have a consistent color baseline. Perform denoising and enhancement operations on the images, such as applying filters to denoise the images, and use image enhancement techniques to highlight key features and enhance key details in the images, such as building lines and textures.

[0098] Optionally, OpenCV feature detection can be used to automatically annotate important architectural elements in the image, such as windows, doors, balconies, etc.

[0099] Step 303: extracting first architectural features corresponding to the first facade image and second architectural features corresponding to the second facade image based on the style transfer model.

[0100] In this embodiment, this step mainly includes: performing convolution processing on the first facade image and the second facade image respectively through the style transfer model to obtain a first building feature corresponding to the first facade image and a second building feature corresponding to the second facade image.

[0101] Step 304: Based on the style transfer model, the first building feature and the second building feature are integrated to generate a zoning function map of the building to be renovated after style transfer, and a building rendering is output based on the zoning function map.

[0102] In this embodiment, this step mainly includes: fusing the style architectural features into the architectural features through the style transfer algorithm preset in the style transfer model to obtain the target architectural image features. According to the target architectural image features, a corresponding functional zoning map is generated through a generator preset in the style transfer model; based on the functional zoning map, the generator is used to make detail adjustments to obtain the architectural rendering corresponding to the functional zoning map; the target facade image includes the functional zoning map and the architectural rendering.

[0103] Specifically, the style transfer algorithm is used to fuse different styles into architectural features, and a specific style is integrated into the architectural features. Then, the generator in AFGAN is used to generate a partition function map of the building facade based on the extracted features. The model generates a partition function map based on the extracted features and identifies different functional areas of the building facade, such as the entrance area, window area, etc. Based on the partition function map, the model further generates a rendering that shows the details and overall design style of different functional areas.

[0104] Optionally, in the process of the generator generating the effect diagram and the partition function diagram, as Figure 4 The discriminator shown in the figure discriminates the fake images input by the generator, marking the real images as 1 and the fake images as 0. If the generated image is judged to be fake, the discriminator returns the deviation value between the fake image and the real image to the generator, so that the generator can be upgraded and generate images closer to the real image.

[0105] The architectural reconstruction method based on style transfer disclosed in this embodiment has significant efficiency advantages for the mass generation of hundreds of village and town buildings, and the generated effect is relatively ideal. In addition, considering the saving of machine load and learning time cost, the prediction accuracy of the model after a small number of iterations of learning and training can meet the requirements of target prediction.

[0106] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A building transformation method based on style transfer, characterized in that: include: Obtaining a first building facade image corresponding to the building to be transformed and a style building facade image used as a reference for style migration; Extracting architectural features corresponding to the first building facade image and architectural style features corresponding to the architectural style facade image; The architectural features and the style architectural features are integrated to obtain target architectural image features; According to the target building image features, a target facade image of the target building is acquired.

2. The architectural transformation method based on style transfer according to claim 1, characterized in that: The step of obtaining a first building facade image corresponding to the building to be transformed and a style building facade image used as a reference for style migration includes: Acquire a first initial high-resolution facade image corresponding to the building to be renovated; Obtaining a second initial high-resolution facade image of a style building for style transfer reference; Perform image preprocessing, denoising and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image respectively to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

3. The architectural transformation method based on style transfer according to claim 1, characterized in that: Before extracting the architectural features corresponding to the first building facade image and the style architectural features corresponding to the style building facade image, the method includes: Generate an initial style transfer model based on a generative adversarial network; Obtain a high-resolution building facade image dataset, and draw the building outline, functional zoning map, and building rendering corresponding to each high-resolution facade image; Performing feature extraction on the high-resolution facade image according to the initial style transfer model, so as to output a training functional zoning map and a training building rendering corresponding to each of the high-resolution facade images according to the extracted image features; Obtaining a first error between the training functional zoning diagram and the functional zoning diagram and a second error between the training building rendering and the building rendering; The model parameters of the initial style transfer model are continuously adjusted according to the first error and the second error until the first error and the second error meet a preset error condition, thereby obtaining a style transfer model.

4. The architectural reconstruction method based on style transfer according to claim 3, characterized in that: The extracting the architectural features corresponding to the first building facade image and the style architectural features corresponding to the style building facade image includes: The first building facade image and the style building facade image are respectively convolved by the style transfer model to obtain building features corresponding to the first building facade image and style building features corresponding to the style building facade image.

5. The architectural reconstruction method based on style transfer according to claim 3, characterized in that: The step of fusing the architectural features and the style architectural features to obtain target architectural image features includes: The style building features are fused into the building features by a style transfer algorithm preset in the style transfer model to obtain the target building image features.

6. The architectural reconstruction method based on style transfer according to claim 3, characterized in that: The step of acquiring a target facade image of a target building according to the target building image features includes: According to the target building image features, a corresponding functional zoning map is generated by a generator preset in the style transfer model; Based on the functional zoning diagram, the generator performs detail adjustments to obtain a building rendering corresponding to the functional zoning diagram; the target facade image includes the functional zoning diagram and the building rendering.

7. A building transformation system based on style migration, characterized in that: It includes image acquisition module, feature extraction module, feature fusion module and building transformation module; The image acquisition module is used to acquire a first building facade image corresponding to the building to be transformed and a style building facade image used as a reference for style migration; The feature extraction module is used to extract the building features corresponding to the first building facade image and the style building features corresponding to the style building facade image; The feature fusion module is used to fuse the architectural features and the style architectural features to obtain target architectural image features; The building reconstruction module is used to obtain a target facade image of a target building according to the image features of the target building.

8. The architectural transformation system based on style transfer according to claim 7, characterized in that: The image acquisition module includes an acquisition unit and a processing unit; The acquisition unit is used to acquire a first initial high-resolution facade image corresponding to the building to be renovated; and acquire a second initial high-resolution facade image of the style building used as a reference for style migration; The processing unit is used to perform image preprocessing operations, denoising operations and image enhancement operations on the first initial high-resolution facade image and the second initial high-resolution facade image, respectively, to obtain a first building facade image corresponding to the building to be renovated and a style building facade image corresponding to the style building.

9. A computing device, characterized in that include: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute a style transfer-based architectural transformation method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, and the program code can be called by a processor to execute the architectural transformation method based on style transfer as described in any one of claims 1 to 6.

Images