Building scheme generation system and method based on artificial intelligence

Abstract

The invention discloses a building scheme generation system and method based on artificial intelligence, and relates to the technical field of artificial intelligence, the system comprises a preprocessing module, a picture classification module, an input analysis module and a scheme generation module; the preprocessing module is used for collecting a building scheme and a building picture, and determining building scheme elements and elements and element values of a manual annotation picture; the picture classification module is used for performing preliminary classification and re-classification on the pictures according to the elements and clustering the pictures of the same classification to determine representative pictures; the input analysis module is used for analyzing user input; and the scheme generation module is used for generating a building scheme according to user input and feedback and providing a selection and judgment mechanism to optimize scheme generation. The invention also provides a method for realizing the system. According to the method, the conditions that text analysis and picture processing are difficult to effectively fuse and picture and data classification standards are relatively single in the prior art can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based building plan generation system and method. Background Art

[0002] Artificial intelligence is widely used in the generation of architectural solutions. It can quickly analyze factors such as site conditions and functional requirements based on a large amount of data. It generates a variety of architectural layouts and form solutions through algorithms, providing designers with rich creative inspiration. For example, it can accurately plan space in projects with complex terrain. At the same time, artificial intelligence can simulate building performance, such as lighting, ventilation and energy consumption, to assist in optimizing the design. It can also combine virtual reality technology to allow designers to immerse themselves in the effects of the solution. In terms of collaborative design, it promotes communication between different professionals, improves design efficiency and quality, and promotes the transformation of architectural solutions from traditional human conception to intelligent assisted generation.

[0003] In the existing architectural plan generation technology, when classifying relevant pictures or data, the classification standards or methods used are relatively simple, resulting in a one-sided fusion of pictures and text elements, which cannot meet the classification requirements of complex objects such as architectural plans in different dimensions; in addition, some existing technologies are based on initial settings or fixed modes, and lack the ability to dynamically adjust and generate plans based on user feedback. Summary of the invention

[0004] The purpose of the present invention is to provide a system and method for generating a building plan based on artificial intelligence to solve the problems raised in the prior art.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an artificial intelligence-based building plan generation system, the system comprising a preprocessing module, an image classification module, an input analysis module and a plan generation module; The pre-processing module is used to collect architectural plans and architectural pictures, determine architectural plan elements, and manually annotate the elements and element values ​​of the pictures; the picture classification module is used to preliminarily classify and reclassify the pictures according to the elements, store the classified picture information, and cluster the pictures of the same classification to determine the representative pictures; the input analysis module is used to analyze the user input; the solution generation module is used to generate the architectural plan according to the user input and feedback and provide a selection and judgment mechanism to optimize the solution generation; The output end of the preprocessing module is connected to the input end of the picture classification module; the output end of the picture classification module is connected to the input ends of the input analysis module and the solution generation module; the output end of the input analysis module is connected to the input end of the solution generation module.

[0006] The preprocessing module includes a solution collection unit, an element determination unit, a picture collection unit and a picture annotation unit; The scheme collection unit is used to collect building schemes; the element determination unit is used to determine building scheme elements according to the collected building schemes; the picture collection unit is used to collect building pictures; the picture annotation unit is used to manually annotate picture elements and element values; The output end of the scheme collection unit is connected to the input end of the factor determination unit; the output end of the factor determination unit is connected to the input end of the picture collection unit; the output end of the picture collection unit is connected to the input end of the picture annotation unit; the output end of the picture annotation unit is connected to the input end of the picture classification module.

[0007] The picture classification module includes a picture classification unit, a type determination unit, a range division unit, a database unit, a picture clustering unit and a representative graph determination unit; The picture classification unit is used to classify pictures according to elements; the type determination unit is used to determine the element value category according to the picture annotation situation; the range division unit is used to divide the numerical element value according to the manual annotation situation; the database unit is used to store the classification of pictures after classification according to the category type and the numerical element value; the picture clustering unit is used to cluster pictures of the same category; the representative image determination unit is used to select a random picture in the category with the largest number of pictures after the pictures are clustered as a representative image; The output end of the picture classification unit is connected to the input end of the type determination unit and the range division unit; the output end of the type determination unit is connected to the input end of the database unit; the output end of the range division unit is connected to the input end of the database unit; the output end of the database unit is connected to the input end of the picture clustering unit and the input analysis module; the output end of the picture clustering unit is connected to the input end of the representative graph determination unit; the output end of the representative graph determination unit is connected to the input end of the scheme generation module.

[0008] The input analysis module includes a user input unit and an input analysis unit; The user input unit is used for the user to input elements and corresponding element values; the input analysis unit is used to make analysis based on the matching between the user input and the elements and element values ​​in the database; The output end of the user input unit is connected to the input end of the input analysis unit; the output end of the input analysis unit is connected to the input end of the solution generation module.

[0009] The solution generation module includes an output selection unit, an output judgment unit, a user feedback unit and a solution generation unit; The output selection unit is used to output the corresponding picture when the user input matches the classification situation for the user to judge whether it meets the expectations; the output judgment unit is used to output the corresponding picture when the user input does not match the classification situation for the user to make a choice; the user feedback unit is used for the user to give feedback on the output; the solution generation unit is used to generate a building solution in combination with the user input, feedback and selection; The output end of the output selection unit is connected to the input end of the user feedback unit; the output end of the output judgment unit is connected to the input end of the user feedback unit; the output end of the user feedback unit is connected to the input end of the solution generation unit.

[0010] A method for generating a building plan based on artificial intelligence, the method comprising the following steps: Step 1: Collect a large number of architectural plans and determine the elements of the architectural plans; collect a large number of architectural pictures, manually determine the architectural plan elements in the pictures and mark the element values; Step 2: Classify and cluster the images by elements and element value types, determine representative images, and build a database to store relevant information; Step 3: Analyze the user input to obtain the matching between the user input and the elements and element values; Step 4: Provide a selection or judgment mechanism based on the matching between user input and elements and element values, and generate a building plan based on user feedback.

[0011] In step 1, a large number of architectural plans are collected to ensure sample diversity; Use the word segmentation tool to segment each architectural plan text; create a stop word list and remove stop words; For each sample, calculate the frequency of all the words in the sample after removing the stop words in all samples: frequency of occurrence = number of samples with the word / total number of samples; set a threshold A, and when the frequency of a word is greater than A, set the word as a building plan element; Calculate the frequency of all word segments after removing stop words in all samples, and compare them with the threshold A to obtain all building plan elements, expressed as: [E1, E2, …, E a ]; Where a is a positive integer, indicating the number of building plan elements, E a represents the a-th building plan element; Collect a large number of architectural pictures and annotate them manually: including identifying elements and annotating element values: [Pic α :{E i :V(E i ),E j :V(Ej ),…,E k :V(E k )}]; Among them, α is a positive integer, representing the picture sequence, Pic α represents the αth picture; i, j, k are positive integers, representing the sequence of building plan elements, E i ,E j ,…,E k ∈{E1,E2,…,E a}; V(E k ) indicates a picture Pic α The element value of .

[0012] In step 2, the images are preliminarily classified according to their elements to obtain images corresponding to each element: i , the corresponding picture is represented as [Pic α ,Pic β ,…,Pic γ ]; Among them, β and γ are positive integers, representing the image sequence; Reclassify the images according to the type of feature value: For elements whose element values ​​are categorical: i , summarize the feature value categories based on manual annotation [T1(E i ),T2(E i ),…,T n (E i )]; Where n is a positive integer, representing the factor E i The number of categories of values, T n (E i ) represents element E i The nth category of the value; Classify the images according to the category of the feature value; for the feature value category T n (E i ), the corresponding picture is represented as [Pic o ,Pic p ,…,Pic q ]; Among them, o, p, q are positive integers, representing the picture sequence; Elements whose element values ​​are numerical: For element E j , obtain element E according to the manual annotation jThe maximum value max and the minimum value min of the value; divide the feature value into N parts: {[min,min+(max-min) / N],(min+(max-min) / N,min+(max-min)2 / N],…,(min+(max-min)(N-1) / N,max]}; classify the images according to different intervals; Wherein, N is a positive integer; (min+(max-min) / N,min+(max-min)2 / N] represents the interval min+(max-min) / N to min+(max-min)2 / N, excluding min+(max-min) / N; Construct a database to store different elements and corresponding element values ​​of different categories and element values ​​of different intervals, including different element value categories and their corresponding pictures and pictures corresponding to different element value intervals; For images corresponding to a certain feature value category or feature value interval, clustering is performed on them: the images are normalized; the pre-trained convolutional neural network model is used to extract image features and form feature vectors; the number of clusters K is set, and the K-means clustering algorithm is used to divide the images into K classes; According to the classification situation, the number of images in each class is counted, and a picture is randomly selected from the class with the largest number of pictures as the representative picture of the feature value category or feature value range.

[0013] In Step 3 and Step 4, the user input is analyzed: the user input is segmented; word vectors are assigned to the user input word vectors, elements, and element type values ​​using Word2Vec; and the cosine similarity between each user input word vector and each element word vector is calculated; For element E whose element value is categorical i , when there is a user input word vector and element E i The word vector cosine similarity value is 1 and there is a user input word vector and element E i When the cosine similarity of the word vector of a certain element type value is 1, the representative graph corresponding to the element type value is selected and output to the user, and the user makes a judgment whether it meets expectations and gives feedback; when there is a word vector of the user input segmentation and the element E i The word vector cosine similarity value is 1, but there is no user input word vector and element E i When the cosine similarity of any element type value word vector is 1 or when all user input word vectors are similar to element E i When the cosine similarity values ​​of the word vectors are not 1, the representative graphs corresponding to all the element value categories of the element Ei are output to the user for selection and feedback; For element E whose element value is numeric j, when there is a user input word vector and element E j When the word vector cosine similarity value is 1 and the context analysis of the user input checks that there is numerical data before and after the corresponding word segmentation, select element E j The image corresponding to the element value with the smallest absolute value of the difference with the checked numerical data is output to the user, and the user makes a judgment whether it meets the expectations and gives feedback; when there is a word vector of the user input segmentation and the element E j When the cosine similarity value of the word vector is 1, but the context analysis of the user input shows that there is no numerical data before and after the corresponding word segmentation, or when all user input word segmentation word vectors are consistent with the element E j When the cosine similarity values ​​of the word vectors are not 1, the output element E j The representative graphs corresponding to all factor value intervals are given to the user for selection and feedback; Combine user input, feedback, and choices to generate a building plan: all elements and their corresponding values.

[0014] Compared with the prior art, the beneficial effects of the present invention are: the multi-level classification method of the present invention can classify images more carefully according to different features, and can provide more accurate classification results that are more in line with actual needs, so as to facilitate the subsequent accurate matching of images according to user needs; the present invention extracts feature vectors through a pre-trained convolutional neural network model and uses the K-means clustering algorithm, which can mine the intrinsic similarities between images. The selected representative images can represent the typical characteristics of the element value category or interval to a certain extent, and can better present representative image examples to users, assisting users in understanding and selecting. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a process of a building plan generation system based on artificial intelligence according to the present invention; Figure 2 The present invention is a schematic diagram of the steps of a method for generating a building plan based on artificial intelligence. DETAILED DESCRIPTION

[0016] 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.

[0017] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a building solution generation system based on artificial intelligence, the system includes a preprocessing module, a picture classification module, an input analysis module and a solution generation module; The pre-processing module is used to collect architectural plans and architectural pictures, determine architectural plan elements, and manually annotate the elements and element values ​​of the pictures; the picture classification module is used to preliminarily classify and reclassify the pictures according to the elements, store the classified picture information, and cluster the pictures of the same classification to determine the representative pictures; the input analysis module is used to analyze the user input; the solution generation module is used to generate the architectural plan according to the user input and feedback and provide a selection and judgment mechanism to optimize the solution generation; The output end of the preprocessing module is connected to the input end of the picture classification module; the output end of the picture classification module is connected to the input ends of the input analysis module and the solution generation module; the output end of the input analysis module is connected to the input end of the solution generation module.

[0018] The preprocessing module includes a solution collection unit, an element determination unit, a picture collection unit and a picture annotation unit; The scheme collection unit is used to collect building schemes; the element determination unit is used to determine building scheme elements according to the collected building schemes; the picture collection unit is used to collect building pictures; the picture annotation unit is used to manually annotate picture elements and element values; The output end of the scheme collection unit is connected to the input end of the factor determination unit; the output end of the factor determination unit is connected to the input end of the picture collection unit; the output end of the picture collection unit is connected to the input end of the picture annotation unit; the output end of the picture annotation unit is connected to the input end of the picture classification module.

[0019] The picture classification module includes a picture classification unit, a type determination unit, a range division unit, a database unit, a picture clustering unit and a representative graph determination unit; The picture classification unit is used to classify pictures according to elements; the type determination unit is used to determine the element value category according to the picture annotation situation; the range division unit is used to divide the numerical element value according to the manual annotation situation; the database unit is used to store the classification of pictures after classification according to the category type and the numerical element value; the picture clustering unit is used to cluster pictures of the same category; the representative image determination unit is used to select a random picture in the category with the largest number of pictures after the pictures are clustered as a representative image; The output end of the picture classification unit is connected to the input end of the type determination unit and the range division unit; the output end of the type determination unit is connected to the input end of the database unit; the output end of the range division unit is connected to the input end of the database unit; the output end of the database unit is connected to the input end of the picture clustering unit and the input analysis module; the output end of the picture clustering unit is connected to the input end of the representative graph determination unit; the output end of the representative graph determination unit is connected to the input end of the scheme generation module.

[0020] The input analysis module includes a user input unit and an input analysis unit; The user input unit is used for the user to input elements and corresponding element values; the input analysis unit is used to make analysis based on the matching between the user input and the elements and element values ​​in the database; The output end of the user input unit is connected to the input end of the input analysis unit; the output end of the input analysis unit is connected to the input end of the solution generation module.

[0021] The solution generation module includes an output selection unit, an output judgment unit, a user feedback unit and a solution generation unit; The output selection unit is used to output the corresponding picture when the user input matches the classification situation for the user to judge whether it meets the expectations; the output judgment unit is used to output the corresponding picture when the user input does not match the classification situation for the user to make a choice; the user feedback unit is used for the user to give feedback on the output; the solution generation unit is used to generate a building solution in combination with the user input, feedback and selection; The output end of the output selection unit is connected to the input end of the user feedback unit; the output end of the output judgment unit is connected to the input end of the user feedback unit; the output end of the user feedback unit is connected to the input end of the solution generation unit.

[0022] A method for generating a building plan based on artificial intelligence, the method comprising the following steps: Step 1: Collect a large number of architectural plans and determine the elements of the architectural plans; collect a large number of architectural pictures, manually determine the architectural plan elements in the pictures and mark the element values; Step 2: Classify and cluster the images by elements and element value types, determine representative images, and build a database to store relevant information; Step 3: Analyze the user input to obtain the matching between the user input and the elements and element values; Step 4: Provide a selection or judgment mechanism based on the matching between user input and elements and element values, and generate a building plan based on user feedback.

[0023] In step 1, a large number of architectural plans are collected to ensure sample diversity; Use the word segmentation tool to segment each architectural plan text; create a stop word list and remove stop words; For each sample, calculate the frequency of all the words in the sample after removing the stop words in all samples: frequency of occurrence = number of samples with the word / total number of samples; set a threshold A, and when the frequency of a word is greater than A, set the word as a building plan element; Calculate the frequency of all word segments after removing stop words in all samples, and compare them with the threshold A to obtain all building plan elements, expressed as: [E1, E2, …, E a ]; Where a is a positive integer, indicating the number of building plan elements, E a represents the a-th building plan element; Collect a large number of architectural pictures and annotate them manually: including identifying elements and annotating element values: [Pic α :{E i :V(E i ),E j :V(E j ),…,E k :V(E k )}]; Among them, α is a positive integer, representing the picture sequence, Pic α represents the αth picture; i, j, k are positive integers, representing the sequence of building plan elements, E i ,E j ,…,E k ∈{E1,E2,…,E a}; V(E k ) indicates a picture Pic α The element value of .

[0024] In step 2, the images are preliminarily classified according to their elements to obtain images corresponding to each element: i , the corresponding picture is represented as [Pic α ,Pic β ,…,Pic γ ]; Among them, β and γ are positive integers, representing the image sequence; Reclassify the images according to the type of feature value: For elements whose element values ​​are categorical: i , summarize the feature value categories based on manual annotation [T1(E i ),T2(E i ),…,Tn (E i )]; Where n is a positive integer, representing the factor E i The number of categories of values, T n (E i ) represents element E i The nth category of the value; Classify the images according to the category of the feature value; for the feature value category T n (E i ), the corresponding picture is represented as [Pic o ,Pic p ,…,Pic q ]; Among them, o, p, q are positive integers, representing the picture sequence; Elements whose element values ​​are numerical: For element E j , obtain element E according to the manual annotation j The maximum value max and the minimum value min of the value; divide the feature value into N parts: {[min,min+(max-min) / N],(min+(max-min) / N,min+(max-min)2 / N],…,(min+(max-min)(N-1) / N,max]}; classify the images according to different intervals; Wherein, N is a positive integer; (min+(max-min) / N,min+(max-min)2 / N] represents the interval min+(max-min) / N to min+(max-min)2 / N, excluding min+(max-min) / N; Construct a database to store different elements and corresponding element values ​​of different categories and element values ​​of different intervals, including different element value categories and their corresponding pictures and pictures corresponding to different element value intervals; For images corresponding to a certain feature value category or feature value interval, clustering is performed on them: the images are normalized; the pre-trained convolutional neural network model is used to extract image features and form feature vectors; the number of clusters K is set, and the K-means clustering algorithm is used to divide the images into K classes; According to the classification situation, the number of images in each class is counted, and a picture is randomly selected from the class with the largest number of pictures as the representative picture of the feature value category or feature value range.

[0025] In Step 3 and Step 4, the user input is analyzed: the user input is segmented; word vectors are assigned to the user input word vectors, elements, and element type values ​​using Word2Vec; and the cosine similarity between each user input word vector and each element word vector is calculated; For element E whose element value is categorical i , when there is a user input word vector and element E i The word vector cosine similarity value is 1 and there is a user input word vector and element E i When the cosine similarity of the word vector of a certain element type value is 1, the representative graph corresponding to the element type value is selected and output to the user, and the user makes a judgment whether it meets expectations and gives feedback; when there is a word vector of the user input segmentation and the element E i The word vector cosine similarity value is 1, but there is no user input word vector and element E i When the cosine similarity of any element type value word vector is 1 or when all user input word vectors are similar to element E i When the cosine similarity values ​​of the word vectors are not 1, the representative graphs corresponding to all the element value categories of the element Ei are output to the user for selection and feedback; For element E whose element value is numeric j , when there is a user input word vector and element E j When the word vector cosine similarity value is 1 and the context analysis of the user input checks that there is numerical data before and after the corresponding word segmentation, select element E j The image corresponding to the element value with the smallest absolute value of the difference with the checked numerical data is output to the user, and the user makes a judgment whether it meets the expectations and gives feedback; when there is a word vector of the user input segmentation and the element E j When the cosine similarity value of the word vector is 1, but the context analysis of the user input shows that there is no numerical data before and after the corresponding word segmentation, or when all user input word segmentation word vectors are consistent with the element E j When the cosine similarity values ​​of the word vectors are not 1, the output element E j The representative graphs corresponding to all factor value intervals are given to the user for selection and feedback; Combine user input, feedback, and choices to generate a building plan: all elements and their corresponding values.

[0026] In this embodiment: Step 1: Collect a large number of architectural plans to ensure sample diversity; Use the word segmentation tool to segment each architectural plan text; create a stop word list and remove stop words; For each sample, calculate the frequency of all the words in the sample after removing the stop words in all samples: frequency of occurrence = number of samples with the word / total number of samples; set a threshold, and when the frequency of a word is greater than the threshold, set the word as a building plan element; Calculate the frequency of all word segments after removing stop words in all samples, and compare with threshold A to obtain all architectural plan elements, expressed as: [style, area]; Collect a large number of architectural pictures and manually annotate them, including determining the elements of the pictures and annotating the element values: [pic1:{style: Chinese, area: 200}], [pic2:{style: European, area: 150}], [pic3:{style: Chinese, area: 120}], [pic4:{style: European, area: 100}]; Step 2: preliminarily classify the images according to their elements to obtain the images corresponding to each element. For element style, the corresponding images are represented as [pic1, pic2, pic3, pic4]; Reclassify the images according to the type of feature value: Elements with categorical feature values: For feature styles, the feature value categories [Chinese, European] are summarized based on manual annotations; Classify the images according to the category of the feature value; for the feature value category, the corresponding image is represented as [pic1, pic3]; For elements with numerical values: for element area, the maximum value 200 and the minimum value 100 of the element area value are obtained according to the manual annotation; the element value is divided into two parts: {[100,150], (150,200]}; the images are classified according to different intervals; Construct a database to store different elements and corresponding element values ​​of different categories and element values ​​of different intervals, including different element value categories and their corresponding pictures and pictures corresponding to different element value intervals; For images corresponding to a certain feature value category or feature value interval, clustering is performed on them: the images are normalized; the pre-trained convolutional neural network model is used to extract image features and form feature vectors; the number of clusters is set to 2, and the K-means clustering algorithm is used to divide the images into 2 classes; According to the classification situation, the number of images in each class is counted, and a picture is randomly selected from the class with the largest number of pictures as the representative picture of the feature value category or feature value range.

[0027] Step 3 and Step 4 analyze the user input: perform word segmentation on the user input; use Word2Vec to assign word vectors to the user input word vectors, elements, and element type values; calculate the cosine similarity between each user input word vector and each element word vector; For the element style whose element value is categorical, when there is a cosine similarity value of 1 between the word vector of the user input segmentation and the word vector of the element style, and there is a cosine similarity value of 1 between the word vector of the user input segmentation and the word vector of the element style, a representative graph corresponding to the element type value is selected and output to the user, and the user makes a judgment whether it meets expectations and gives feedback; when there is a cosine similarity value of 1 between the word vector of the user input segmentation and the word vector of the element style, but there is no cosine similarity value of 1 between the word vector of the user input segmentation and the word vector of any element type value of the element style, or when all the cosine similarities between the word vector of the user input segmentation and the word vector of the element style are not 1, the representative graphs corresponding to all element value categories of the element style are output to the user for the user to select and give feedback; For the factor area whose factor value is numerical, when there is a cosine similarity value of 1 between the word vector of the user input segmentation and the word vector of the factor area, and the context analysis of the user input checks that there is numerical data before and after the corresponding segmentation, the picture corresponding to the factor value with the smallest absolute value of the difference between the factor area and the checked numerical data is selected and output to the user, and the user makes a judgment whether it meets expectations and gives feedback; when there is a cosine similarity value of 1 between the word vector of the user input segmentation and the word vector of the factor area, but the context analysis of the user input checks that there is no numerical data before and after the corresponding segmentation, or when all the cosine similarity values ​​of the word vector of the user input segmentation and the word vector of the factor area are not 1, the representative pictures corresponding to all the factor value intervals of the factor area are output to the user for the user to select and give feedback; Combine user input, feedback, and choices to generate a building plan: all elements and their corresponding values.

[0028] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. An artificial intelligence-based building plan generation system, characterized by: The system includes a preprocessing module, an image classification module, an input analysis module, and a solution generation module; The pre-processing module is used to collect architectural plans and architectural pictures, determine architectural plan elements, and manually annotate elements and element values ​​of pictures; The picture classification module is used to preliminarily classify and reclassify pictures according to the elements, store the classified picture information, and cluster the pictures of the same classification to determine the representative pictures; The input analysis module is used to analyze user input; The scheme generation module is used to generate building schemes based on user input and feedback and provide a selection and judgment mechanism to optimize scheme generation; The output end of the preprocessing module is connected to the input end of the picture classification module; The output end of the picture classification module is connected to the input end of the input analysis module and the input end of the solution generation module; the output end of the input analysis module is connected to the input end of the solution generation module.

2. The artificial intelligence-based building plan generation system according to claim 1, characterized in that: The preprocessing module includes a solution collection unit, an element determination unit, a picture collection unit and a picture annotation unit; The scheme collection unit is used to collect building schemes; The element determination unit is used to determine the building plan elements according to the collected building plans; The picture collecting unit is used to collect building pictures; The picture annotation unit is used to manually annotate picture elements and element values; The output end of the scheme collection unit is connected to the input end of the factor determination unit; the output end of the factor determination unit is connected to the input end of the picture collection unit; the output end of the picture collection unit is connected to the input end of the picture annotation unit; the output end of the picture annotation unit is connected to the input end of the picture classification module.

3. The artificial intelligence-based building plan generation system according to claim 2, characterized in that: The picture classification module includes a picture classification unit, a type determination unit, a range division unit, a database unit, a picture clustering unit and a representative graph determination unit; The picture classification unit is used to classify pictures according to elements; the type determination unit is used to determine the element value category according to the picture annotation situation; the range division unit is used to divide the numerical element value according to the manual annotation situation; the database unit is used to store the classification of pictures after classification according to the category type and the numerical element value; the picture clustering unit is used to cluster pictures of the same category; the representative image determination unit is used to select a random picture in the category with the largest number of pictures after the pictures are clustered as a representative image; The output end of the picture classification unit is connected to the input ends of the type determination unit and the range division unit; The output end of the type determination unit is connected to the input end of the database unit; the output end of the range division unit is connected to the input end of the database unit; the output end of the database unit is connected to the input end of the picture clustering unit and the input analysis module; The output end of the picture clustering unit is connected to the input end of the representative graph determination unit; the output end of the representative graph determination unit is connected to the input end of the solution generation module.

4. The artificial intelligence-based building plan generation system according to claim 3, characterized in that: The input analysis module includes a user input unit and an input analysis unit; The user input unit is used for the user to input elements and corresponding element values; the input analysis unit is used to make analysis based on the matching between the user input and the elements and element values ​​in the database; The output end of the user input unit is connected to the input end of the input analysis unit; the output end of the input analysis unit is connected to the input end of the solution generation module.

5. The artificial intelligence-based building plan generation system according to claim 4, characterized in that: The solution generation module includes an output selection unit, an output judgment unit, a user feedback unit and a solution generation unit; The output selection unit is used to output the corresponding picture for the user to judge whether it meets the expectations when the user input matches the classification situation; the output judgment unit is used to output the corresponding picture for the user to make a choice when the user input does not match the classification situation; The user feedback unit is used for the user to give feedback on the output; The scheme generating unit is used to generate a building scheme in combination with user input, feedback and selection; The output end of the output selection unit is connected to the input end of the user feedback unit; the output end of the output judgment unit is connected to the input end of the user feedback unit; The output end of the user feedback unit is connected to the input end of the solution generation unit.

6. A method for generating a building plan based on artificial intelligence, characterized in that: The method comprises the following steps: Step 1: Collect a large number of architectural plans and determine the elements of the architectural plans; collect a large number of architectural pictures, manually determine the architectural plan elements in the pictures and mark the element values; Step 2: Classify and cluster the images by elements and element value types, determine representative images, and build a database to store relevant information; Step 3: Analyze the user input to obtain the matching between the user input and the elements and element values; Step 4: Provide a selection or judgment mechanism based on the matching of user input with elements and element values, and generate a building plan based on user feedback.

7. The method for generating a building plan based on artificial intelligence according to claim 6, characterized in that: In step 1, a large number of architectural plans are collected to ensure sample diversity; Use the word segmentation tool to segment each architectural plan text; create a stop word list and remove stop words; For each sample, calculate the frequency of all the words in the sample after removing the stop words in all samples: frequency of occurrence = number of samples with the word / total number of samples; Set a threshold A. When the frequency of a word is greater than A, set the word as a building plan element. Calculate the frequency of all word segments after removing stop words in all samples, and compare them with the threshold A to obtain all building plan elements, expressed as: [E1, E2, …, E a ]; Where a is a positive integer, indicating the number of building plan elements, E a represents the a-th building plan element; Collect a large number of architectural pictures and annotate them manually: including identifying elements and annotating element values: [Pic α :{E i :V(E i ),E j :V(E j ),…,E k :V(E k )}]; Among them, α is a positive integer, representing the picture sequence, Pic α represents the αth picture; i, j, k are positive integers, representing the sequence of building plan elements, E i ,E j ,…,E k ∈{E1,E2,…,E a }; V(E k ) indicates a picture Pic α The element value of .

8. The method for generating a building plan based on artificial intelligence according to claim 7, characterized in that: In step 2, the images are preliminarily classified according to their elements to obtain images corresponding to each element: i , the corresponding picture is represented as [Pic α ,Pic β ,…,Pic γ ]; Among them, β and γ are positive integers, representing the image sequence; Reclassify the images according to the type of feature value: For elements whose element values ​​are categorical: i , summarize the feature value categories based on manual annotation [T1(E i ),T2(E i ),…,T n (E i )]; Where n is a positive integer, representing the factor E i The number of categories of values, T n (E i ) represents element E i The nth category of the value; Classify the images according to the category of the feature value; for the feature value category T n (E i ), the corresponding picture is represented as [Pic o ,Pic p ,…,Pic q ]; Among them, o, p, q are positive integers, representing the picture sequence; Elements whose element values ​​are numerical: For element E j , obtain element E according to the manual annotation j The maximum value max and the minimum value min of the value; divide the feature value into N parts: {[min,min+(max-min) / N],(min+(max-min) / N,min+(max-min)2 / N],…,(min+(max-min)(N-1) / N,max]}; classify the images according to different intervals; Wherein, N is a positive integer; (min+(max-min) / N,min+(max-min)2 / N] represents the interval min+(max-min) / N to min+(max-min)2 / N, excluding min+(max-min) / N; Construct a database to store different elements and corresponding element values ​​of different categories and element values ​​of different intervals, including different element value categories and their corresponding pictures and pictures corresponding to different element value intervals; For images corresponding to a certain feature value category or feature value interval, clustering is performed on them: the images are normalized; the pre-trained convolutional neural network model is used to extract image features and form feature vectors; the number of clusters K is set, and the K-means clustering algorithm is used to divide the images into K classes; According to the classification situation, the number of images in each class is counted, and a picture is randomly selected from the class with the largest number of pictures as the representative picture of the feature value category or feature value range.

9. The method for generating a building plan based on artificial intelligence according to claim 8, characterized in that: In step 3 and step 4, the user input is analyzed: the user input is segmented; Use Word2Vec to assign word vectors to user-input word segments, features, and feature type values; calculate the cosine similarity between each user-input word vector and each feature vector; For element E whose element value is categorical i , when there is a user input word vector and element E i The word vector cosine similarity value is 1 and there is a user input word vector and element E i When the cosine similarity of the word vector of a certain element type value is 1, the representative graph corresponding to the element type value is selected and output to the user, and the user makes a judgment whether it meets expectations and gives feedback; when there is a word vector of the user input segmentation and the element E i The word vector cosine similarity value is 1, but there is no user input word vector and element E i When the cosine similarity of any element type value word vector is 1 or when all user input word vectors are similar to element E i When the cosine similarity values ​​of the word vectors are not 1, the representative graphs corresponding to all the element value categories of the element Ei are output to the user for selection and feedback; For element E whose element value is numeric j , when there is a user input word vector and element E j When the word vector cosine similarity value is 1 and the context analysis of the user input checks that there is numerical data before and after the corresponding word segmentation, select element E j The image corresponding to the element value with the smallest absolute value of the difference with the checked numerical data is output to the user, and the user makes a judgment whether it meets the expectations and gives feedback; when there is a word vector of the user input segmentation and the element E j When the word vector cosine similarity value is 1, but the context analysis of the user input shows that there is no numerical data before and after the corresponding word segmentation, or when all user input word segmentation word vectors are consistent with the element E j When the cosine similarity values ​​of the word vectors are not 1, the output element E j The representative graphs corresponding to all factor value intervals are given to the user for selection and feedback; Combine user input, feedback, and choices to generate a building plan: all elements and their corresponding values.

Images