Urban design scheme image generation method and device

Abstract

The invention provides a city design scheme image generation method and device, and the method is characterized in that the method comprises the steps: S1, constructing a training data set according to an existing city planning image; s2, constructing a planning parameter prediction model and a design generation model, and performing training according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model; s3, inputting a specified area image into the trained planning parameter prediction model to obtain a corresponding prediction planning parameter; and S4, inputting the prediction planning parameters and the specified area image into the trained design generation model to obtain a design scheme image. In a word, the method can generate an effective design scheme image with specified features and styles according to the specified image.

Description

Technical Field

[0001] The invention belongs to the field of urban planning, and in particular relates to a method and device for generating an image of an urban design scheme. Background Art

[0002] Traditional urban design mainly relies on manual spatial form design. With the development of computer-aided design technology, a variety of computer-aided form generation tools have emerged since 1980, such as parametric urban design and urban form simulation tools, which have improved the efficiency and scientificity of design generation in different links. However, these classic urban space form generation tools mainly realize the automatic generation and modeling of three-dimensional space based on manually set form rules, and these rules are difficult to fully reflect the logic of urban space form. Therefore, they are weak in generating urban space form, especially generating block-scale space form, which is a more precise and complex task.

[0003] The development of new data environment and artificial intelligence technology has brought new perspectives to the generation of block-scale morphology. Compared with macro-scale research, block-scale morphological research is characterized by small scale, diversity, and discretization. It not only focuses on the subtle relationship of urban spatial morphology, but also reflects the characteristics and style of design, which brings challenges to the generative model based on preset rules. Artificial intelligence methods based on artificial neural networks, represented by machine learning and deep learning, can effectively identify invisible features in structured data or image data, thus making up for this shortcoming. These models usually use pre-training to extract data features from a large number of samples and store them for prediction on new data sets.

[0004] Such methods can provide fast and personalized proposals, but in practice they still suffer from problems such as complex training process, weak controllability, and difficulty in interpretation. Therefore, there is an urgent need for a method that can effectively generate urban design proposal images. Summary of the invention

[0005] The present invention is made to solve the above-mentioned problem, and its purpose is to provide a method and device for generating an urban design scheme image.

[0006] The present invention provides a method for generating an urban design scheme image, which is used to generate a corresponding design scheme image according to an image of a designated area, and has the following characteristics, comprising the following steps: step S1, constructing a training data set according to existing urban planning images; step S2, constructing a planning parameter prediction model and a design generation model, and training them respectively according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model; step S3, inputting the designated area image into the trained planning parameter prediction model to obtain corresponding predicted planning parameters; step S4, inputting the predicted planning parameters and the designated area image into the trained design generation model to obtain a design scheme image.

[0007] The method for generating an urban design scheme image provided by the present invention may also have the following feature: wherein the predicted planning parameters are parameters related to the urban design space.

[0008] The urban design scheme image generation method provided by the present invention may also have the following features: wherein the trained planning parameter prediction model is provided with the range and distribution characteristics of the planning parameters, and parameter extraction is performed on all training data of the same image type as the specified area in the training data set to obtain the range and distribution characteristics of the planning parameters.

[0009] The urban design scheme image generation method provided by the present invention may also have the following features: wherein the trained design generation model generates an optimized image based on the image of the specified area, and adjusts the parameters and layout of the buildings in the optimized image according to the predicted planning parameters to obtain a design scheme image.

[0010] The urban design scheme image generation method provided by the present invention may also have the following features: wherein, the existing urban planning images include satellite images and aerial images, and step S1 includes the following sub-steps: step S1-1, classifying all existing urban planning images according to the urban design space element type to obtain multiple sub-classes; step S1-2, for each sub-class, slicing each existing urban planning image in the sub-class according to a preset interception frame to obtain multiple first samples; step S1-3, for each first sample, translating the corresponding preset interception frame up, down, left, and right by a preset pixel distance, and intercepting the corresponding four second samples; step S1-4, preprocessing each first sample and second sample to obtain a training data set.

[0011] The present invention also provides an urban design scheme image generation device, which is used to generate a corresponding design scheme image according to a designated area image, and has the following characteristics, including: a training data construction module, which is used to construct a training data set according to an existing urban planning image; a model generation module, which is used to construct a planning parameter prediction model and a design generation model, and train them respectively according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model; a parameter prediction module, which is used to input the designated area image into the trained planning parameter prediction model to obtain corresponding predicted planning parameters; a scheme generation module, which is used to input the predicted planning parameters and the designated area image into the trained design generation model to obtain a design scheme image.

[0012] Functions and Effects of the Invention

[0013] According to the method and device for generating an urban design scheme image involved in the present invention, on the one hand, the planning parameter prediction model is trained through the training data set, so that the planning parameter prediction model can learn the range and distribution of the predicted planning parameters of various buildings in the existing urban planning image, and then the trained planning parameter prediction model can generate the predicted planning parameters corresponding to the designated area image that meet the characteristics of the training data; on the other hand, the design generation model learns the existing urban planning images in the training data set to perform pixel learning, obtain the layout characteristics and other characteristics, optimize the layout planning in the designated area image, and then generate the parameters of each building in the optimized image according to the predicted planning parameters, so as to obtain the design scheme image. Therefore, the method and device for generating an urban design scheme image of the present invention can generate an effective design scheme image with specified characteristics and style according to the specified image. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a block diagram of a device for generating an urban design scheme image in an embodiment of the present invention;

[0015] Figure 2 is a schematic diagram of a process for constructing a training data set in an embodiment of the present invention;

[0016] Figure 3 It is a schematic diagram of the flow of the method for generating an urban design scheme image in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the following embodiments and the accompanying drawings specifically illustrate the method and device for generating an urban design scheme image of the present invention.

[0018] This embodiment provides a device for generating an urban design scheme image, which is used to generate a corresponding design scheme image according to an image of a designated area.

[0019] Figure 1 4 is a block diagram of a device for generating an urban design scheme image in an embodiment of the present invention.

[0020] like Figure 1 As shown, the urban design scheme image generation device 100 includes a training data construction module 10, a model generation module 20, a parameter prediction module 30, a scheme generation module 40 and a control module 50 for controlling the operation of the above modules.

[0021] The training data construction module 10 is used to construct a training data set according to existing urban planning images.

[0022] Figure 2 It is a schematic diagram of the process of constructing a training data set in an embodiment of the present invention.

[0023] like Figure 2 As shown, the training data construction module 10 constructs a training data set according to the existing urban planning images, including the following steps:

[0024] Step S1-1, classify all existing urban planning images according to the types of urban design space elements to obtain multiple subclasses.

[0025] Step S1-2: for each sub-category, slice each existing urban planning image in the sub-category according to a preset interception frame to obtain a plurality of first samples. In this embodiment, the size of the preset interception frame is 800m*800m.

[0026] Step S1-3: for each first sample, the corresponding preset capture frame is translated up, down, left, and right by preset pixel distances to capture four corresponding second samples.

[0027] Step S1-4, preprocessing each first sample and second sample to obtain a training data set. In this embodiment, the non-building covered area in the sample is set to white through preprocessing, and the area covered by the corresponding building is set to the corresponding grayscale according to the height of the building, so as to obtain a grayscale image as training data.

[0028] The model generation module 20 is used to construct a planning parameter prediction model and a design generation model, and train them respectively according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model. In this embodiment, the planning parameter prediction model and the design generation model are both relatively lightweight models, which can be obtained by adjusting existing lightweight models such as linear models, decision trees, support vector machines, and lightweight neural networks. In the training process of this embodiment, the planning parameter prediction model and the design generation model are optimized by parameter pruning, quantization, model distillation and other methods, so as to reduce the model size and computing requirements while maintaining the performance of the model as much as possible.

[0029] The parameter prediction module 30 is used to input the designated area image into the trained planning parameter prediction model to obtain the corresponding predicted planning parameters.

[0030] The trained planning parameter prediction model is provided with the range and distribution characteristics of the planning parameters. The range and distribution characteristics of the planning parameters are obtained by extracting parameters from all training data of the same image type as the specified area in the training data set. The predicted planning parameters are parameters related to the urban design space. In this embodiment, the predicted planning parameters include waterfront landscape openness, open space ratio, development intensity, average building height, etc.

[0031] The solution generation module 40 is used to input the predicted planning parameters and the designated area image into the trained design generation model to obtain the design solution image.

[0032] Among them, the trained design generation model generates an optimized image based on the image of the specified area, and adjusts the parameters and layout of the buildings in the optimized image according to the predicted planning parameters to obtain the design plan image.

[0033] The control module 50 stores a control program for controlling the operation of each module.

[0034] The process of performing a method for generating an urban design scheme image by using the urban design scheme image generating device 100 will be described below in conjunction with the accompanying drawings.

[0035] Figure 3 It is a schematic diagram of the flow of the method for generating an urban design scheme image in an embodiment of the present invention.

[0036] like Figure 3 As shown, the method for generating an urban design scheme image includes the following steps:

[0037] Step S1, using the training data construction module 10 to construct a training data set according to existing urban planning images.

[0038] Step S2, using the model generation module 20 to construct a planning parameter prediction model and a design generation model, and training them respectively according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model.

[0039] Step S3, using the parameter prediction module 30 to input the designated area image into the trained planning parameter prediction model to obtain the corresponding predicted planning parameters.

[0040] Step S4, using the solution generation module 40 to input the predicted planning parameters and the designated area image into the trained design generation model to obtain a design solution image.

[0041] In this embodiment, an image of a waterfront area of ​​a city is used as a designated area image, and a city image with excellent landscape design is selected as an existing urban planning image, and a corresponding design scheme image is generated by the urban design scheme image generation device 100. Compared with the original image, the design scheme image adds multiple waterfront open spaces and corridors extending from the coastline to the hinterland. It can be seen that the design scheme image can optimize the design of the existing area and improve the landscape value. In other embodiments, by selecting appropriate existing urban planning images to construct a training data set, a design scheme image with corresponding features can be generated.

[0042] Functions and Effects of the Embodiments

[0043] According to the urban design scheme image generation method and device involved in this embodiment, on the one hand, the planning parameter prediction model is trained through the training data set, so that the planning parameter prediction model can learn the range and distribution of the predicted planning parameters of various buildings in the existing urban planning image, and then the predicted planning parameters that meet the characteristics of the training data corresponding to the designated area image can be generated through the trained planning parameter prediction model; on the other hand, the existing urban planning images in the training data set are learned by the design generation model for pixel learning to obtain layout features and other characteristics, and the layout planning in the designated area image is optimized, and then the parameters of each building in the optimized image are generated according to the predicted planning parameters, so as to obtain the design scheme image. In short, this method can generate an effective design scheme image with specified characteristics and style according to the specified image.

[0044] Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for generating an urban design scheme image, for generating a corresponding design scheme image according to an image of a designated area, characterized in that: The following steps are involved: Step S1, constructing a training data set based on existing urban planning images; Step S2, constructing a planning parameter prediction model and a design generation model, and training them respectively according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model; Step S3, inputting the designated area image into the trained planning parameter prediction model to obtain corresponding predicted planning parameters; Step S4, inputting the predicted planning parameters and the designated area image into the trained design generation model to obtain the design solution image.

2. The method for generating an urban design scheme image according to claim 1, characterized in that: in, The predicted planning parameters are parameters related to the urban design space.

3. The method for generating an urban design scheme image according to claim 1, characterized in that: in, The trained planning parameter prediction model is provided with the range and distribution characteristics of the planning parameters, Parameters are extracted from all training data of the same type as the designated area image in the training data set to obtain the range and distribution characteristics of the planning parameters.

4. The method for generating an urban design scheme image according to claim 1, characterized in that: in, The trained design generation model generates an optimized image according to the designated area image, and adjusts the parameters and layout of the buildings in the optimized image according to the predicted planning parameters to obtain the design solution image.

5. The method for generating an urban design scheme image according to claim 1, characterized in that: in, The existing urban planning images include satellite images and aerial images. The step S1 comprises the following sub-steps: Step S1-1, classifying all the existing urban planning images according to the types of urban design space elements to obtain multiple subcategories; Step S1-2, for each of the subclasses, slicing each of the existing urban planning images in the subclass according to a preset clipping frame to obtain a plurality of first samples; Step S1-3, for each of the first samples, the corresponding preset capture frame is translated up, down, left, and right by preset pixel distances to capture four corresponding second samples; Step S1-4: preprocess each of the first samples and the second samples to obtain the training data set.

6. A device for generating an urban design scheme image, for generating a corresponding design scheme image according to an image of a designated area, characterized in that: include: A training data construction module is used to construct a training data set based on existing urban planning images; A model generation module, used to construct a planning parameter prediction model and a design generation model, and to perform training respectively according to the training data set to obtain a trained planning parameter prediction model and a trained design generation model; A parameter prediction module, used for inputting the designated area image into the trained planning parameter prediction model to obtain corresponding predicted planning parameters; The scheme generation module is used to input the predicted planning parameters and the designated area image into the trained design generation model to obtain the design scheme image.

Images