AI drawing house type modeling method based on real-time rendering engine

Abstract

The invention discloses an AI drawing house type modeling method based on a real-time rendering engine, and the method is characterized in that the method comprises the following steps: inputting image information, carrying out the contour matching through a drawing system, segmenting a house type in a house type image through a specific thick line wall line, segmenting small image blocks independently belonging to a door, a window, a sliding door and a bay window, and carrying out the drawing of the small image blocks. Image processing is carried out on each part to achieve the effect of contour clearness, finally, specific door and window contours in the picture are extracted and matched with door and window template contours stored in a database, when the matching degree reaches a certain threshold value, identification is carried out, and a house type image is automatically drawn according to the identification result. According to the method, information generation of a real-time rendering engine is achieved, information of walls, doors, windows, sliding doors and bay windows in a two-dimensional house type image is automatically recognized, 3D house type modeling is automatically achieved, meanwhile, through interaction of a program and a user, the user screens and corrects recognized results, and the recognition capacity of AI intelligent house type recognition can be further improved.

Description

Technical Field

[0001] The present invention relates to the field of drawing technology, and in particular to an AI drawing apartment modeling method based on a real-time rendering engine. Background Art

[0002] With the development of artificial intelligence technology, various fields are vigorously using and developing applications. In today's architectural design and real estate fields, apartment modeling is a crucial link. Traditional apartment modeling methods often require a lot of manpower and time investment, and there are problems such as low efficiency and limited accuracy. Some designs have been developed to enable users to achieve 3D apartment modeling generation by copying two-dimensional apartment plans. However, user copying also has a certain workload. In order to facilitate users and improve efficiency. With the development of computer technology and artificial intelligence, the emergence of real-time rendering engines has brought new opportunities for apartment modeling. Real-time rendering engines can provide efficient graphics processing and real-time interaction capabilities, while AI technology can assist in the automatic generation and optimization of apartment models, improve the efficiency and accuracy of modeling, and we propose the use of a graphic element recognition modeling system and a simple device to automatically identify walls, doors, windows, sliding doors, and bay windows in two-dimensional apartment plans. This achieves intelligent automatic recognition and achieves rapid and automatic 3D apartment modeling. Summary of the invention

[0003] The object of the present invention is to provide an AI drawing apartment modeling method based on a real-time rendering engine, which is characterized by comprising the following: inputting picture information, using the drawing system to perform contour matching, and segmenting the apartment in the apartment plan through specific thick wall lines, segmenting small blocks unique to doors, windows, sliding doors, and bay windows, performing image processing on each part to achieve the effect of clarifying the contour, and finally extracting the specific door and window contours in the picture, matching them with the door and window template contours stored in the database, and determining it as recognition when the matching degree reaches a certain threshold, and automatically drawing the apartment plan according to the recognition result.

[0004] The drawing system includes an AI drawing related data structure module and an algorithm module.

[0005] The AI ​​drawing related data structure module includes a data structure for storing AI drawing related information and an AI drawing template database, and the two modules are connected. The data structure for storing AI drawing related information is provided with a serial number, coordinates, local image address, recognition result type, recording current floating window width, UE blueprint texture, whether to modify the mark, and AI matching recognition result value module, and is respectively connected to the data structure for storing AI drawing related information; the AI ​​drawing template database includes a template outline data stream, a template matching number, and an upload modification time module, and is respectively connected to the AI ​​drawing template database;

[0006] The algorithm modules include an algorithm for correcting apartment size, an algorithm for finding walls, an algorithm for merging walls, an algorithm for eliminating isolated edges, an algorithm for finding door and window block areas, an algorithm for extracting image lines, an algorithm for contour matching, and an algorithm for identifying and drawing bay windows.

[0007] The primitive contour matching is adopted, and the contour matching algorithm calculation formula is: matching value result = total of overall contour matching values / number of contours.

[0008] The serial number module has: 0-no corresponding template / new template, and >0-there is a corresponding template in the database;

[0009] The coordinate module has: a specific position of the corresponding identification area on the floor plan;

[0010] The local picture address or server picture address module has the functions of: capturing pictures and saving them; saving them to a local location and uploading them to a database to obtain a server picture address and uploading them to a database;

[0011] The recognition result type module has the following settings: 0-failed to recognize 1-door 2-window 3-sliding door 4-bay window 5 recognition is empty, and the corresponding PanType is selected according to the recognition result to be drawn on the floor plan;

[0012] The module for recording the current floating window width has the following features: 0-not a floating window>0-a floating window or a floating window before modification. The floating window value is the stored floating window width, and the floating window direction is automatically identified and drawn according to the saved floating window width;

[0013] The blueprint texture module has the following features: acquiring texture data by reading a local file and embedding it into a control blueprint;

[0014] The whether to modify the identifier module has the functions of: recording whether the template type corresponding to the pattern is modified by the user, and if modified, simultaneously modifying the corresponding template in the database or uploading a new template;

[0015] The matching recognition result value module has the following features: when the value is large, it is defined as no recognition; when the value is less than a certain value, the modification information will also modify the database content; when the value is within the modification ~ matching interval, a new template is uploaded;

[0016] The template contour data stream module has the following features: compressing and storing in the form of data stream, taking the matching average value from the longest to the shortest according to the length of contours in the contour group during matching, and when the average value is less than a certain value, it is considered that the corresponding template is matched;

[0017] The template matching times module has the following features: when the server reads the template, it arranges the templates from large to small according to the template matching times, so that the templates with more recognition times are more likely to be matched;

[0018] The upload modification time module has the functions of recording the upload time, recording the modification template type or the time of completing a match.

[0019] The automatic drawing of the two-dimensional floor plan also includes the following settings:

[0020] 1). Using black pixels as markers, set the minimum length to 25 cm and the width to 8 cm, and identify all pixel line segments along the X-axis and the Y-axis of the wall;

[0021] 2). According to the algorithm, the pixel line segments are combined into a single line segment to obtain the X-axis wall line segment and the Y-axis wall line segment;

[0022] 3). Weld and align the endpoints of the line segments according to the algorithm;

[0023] 4)Search XX, XY, YY, YX in sequence, all isolated and aligned endpoints with a cutoff distance of more than 30 cm, and according to the cutoff distance, intercept the appropriate pixel set of the floor plan, obtain the image contour, match it with the contour of the floor plan template element, and determine whether it is a door, window, or sliding door;

[0024] 5) Convert the pixel coordinates of all obtained line segments into spatial coordinates, add them to the apartment line segment MAP, and segment them according to the intersection relationship of the line segments;

[0025] 6). Automatically search for closed areas through algorithms and generate ground and top data;

[0026] 7) Complete the generation of the apartment model.

[0027] The specific method is as follows:

[0028] (1) Upload the floor plan;

[0029] (ii) Whether to correct the floor plan. If yes, the user completes the correction and modifies the wall drawing parameters. If no, the user directly proceeds to the next step.

[0030] (III) AI automatic drawing:

[0031] Draw the wall and start drawing: call AIDrawWall to get the wall drawing parameters and set them → call searchMapWallLine to find the wall lines on the floor plan → call correctCombineWall to connect the wall lines, call searchIndividualWall to delete the isolated edges, and call reCombineWall to align them again; call searchDoorWindows to traverse the obtained wall lines, and get the specific coordinates of objects such as doors, windows, and sliding doors by judging the wall endpoints isConnectedWithXWall and isConnectedWithYWall; call AIDrawWallCheck to crop the floor plan according to the obtained coordinates and the width of doors and windows to get several small blocks;

[0032] Convert the image into data stream and upload it to the service circle upload → get the image data stream → grayscale, binarize and Laplace process the image through opencv, get the image contour through findContours → upload back;

[0033] Integrate the contour data stream and the door and window parameter data, call QueryAIDrawWallCheck and send them to the server for matching. Upload and obtain the contour data stream, match the template contours in the database in order of contour length from large to small, and calculate the average value of all contour matching values. When the matching value is less than a certain threshold, it is considered a match and the type of the template is recorded.

[0034] When the template type meets the standard, the corresponding result is returned, otherwise 0 is returned, and the matching result type is received and stored. When the corresponding template is matched, the program selects the corresponding PanType according to the matching type and draws it on the floor plan. When there is no corresponding type, it is recorded as no vacancy and no drawing;

[0035] All results are displayed in the AI ​​recognition repair window for users to manually correct the recognition results;

[0036] The corresponding floor plan is redrawn according to the result, and a three-dimensional floor plan is generated according to the drawing result.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention realizes an AI drawing apartment modeling method based on a real-time rendering engine, generates information from a real-time rendering engine, and adopts the design of a computer input algorithm to form a graphic element recognition modeling system, which automatically recognizes the information of walls, doors, windows, sliding doors, and bay windows in a two-dimensional apartment diagram, thereby realizing intelligent automatic recognition, and achieving rapid and automatic 3D apartment modeling. Compared with the original manual copying and drawing, intelligent recognition has brought about a huge technological change. Through AI intelligent recognition, a large amount of manual copying and drawing work can be saved, and very fast automatic 3D modeling generation can be achieved. At the same time, through the interaction between the program and the user, the user will screen and correct the results of its recognition, and the recognition ability of AI intelligent recognition of apartment types will be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a system architecture diagram of the present invention.

[0040] Figure 2 It is a flow chart of the present invention.

[0041] Figure 3 The system of the present invention has a reference diagram of the floor plan template.

[0042] Figure 4 Upload the floor plan in the system of the embodiment of the present invention.

[0043] Figure 5 This is a correction dimension diagram in the system of an embodiment of the present invention.

[0044] Figure 6 Draw a floor plan in the system of the embodiment of the present invention.

[0045] Figure 7 A modified reference diagram is drawn in the system of an embodiment of the present invention.

[0046] Figure 8 A floor plan is generated in the system of an embodiment of the present invention. DETAILED DESCRIPTION

[0047] 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.

[0048] See also Figure 1 to Figure 2The present invention provides an AI drawing apartment modeling method based on a real-time rendering engine, which is characterized by comprising the following steps: inputting picture information, using the drawing system to perform contour matching, and segmenting the apartment in the apartment plan through specific thick wall lines, segmenting small blocks unique to doors, windows, sliding doors, and bay windows, performing image processing on each part to achieve the effect of clarifying the contour, and finally extracting the specific door and window contours in the picture, matching them with the door and window template contours stored in the database, and determining that the recognition is made when the matching degree reaches a certain threshold, and automatically drawing the apartment plan according to the recognition result.

[0049] The drawing system includes an AI drawing related data structure module and an algorithm module.

[0050] The AI ​​drawing related data structure module includes a data structure for storing AI drawing related information and an AI drawing template database, and the two modules are connected. The data structure for storing AI drawing related information is provided with a serial number, coordinates, local image address, recognition result type, recording current floating window width, UE blueprint texture, whether to modify the mark, and AI matching recognition result value module, and is respectively connected to the data structure for storing AI drawing related information; the AI ​​drawing template database includes a template outline data stream, a template matching number, and an upload modification time module, and is respectively connected to the AI ​​drawing template database;

[0051] The algorithm modules include an algorithm for correcting apartment size, an algorithm for finding walls, an algorithm for merging walls, an algorithm for eliminating isolated edges, an algorithm for finding door and window block areas, an algorithm for extracting image lines, an algorithm for contour matching, and an algorithm for identifying and drawing bay windows.

[0052] The primitive contour matching is adopted, and the contour matching algorithm calculation formula is: matchResPoly (matching value result) = matchResSum (total of overall contour matching value) / MinCount (number of contours);

[0053] matchResSum is calculated by one-to-one correspondence of the contours to matchShape values. MatchShape is an opencv contour vector algorithm based on Hu moments. Since Hu moments are rotation, translation and scale invariant, image contour matching can be achieved through Hu moments. Hu moments include the following 7 moments, which are derived from the second-order and third-order center moments. The physical meanings of image Hu moments are:

[0054] 0th order moment (m00): the quality of the target area.

[0055] 1st order moment (m01,m10): the centroid of the target area.

[0056] Second-order moments (m02, m11, m20): The radius of gyration of the target region.

[0057] Third-order moments (m03, m12, m21, m30): The orientation and skewness of the target region, reflecting the distortion of the target.

[0058] With the seven formulas of Hu moments:

[0059] Φ1 = η 20 + η 02

[0060]

[0061] Φ3 = (η 20 - 3η 12 ) 2 + 3(η 21 - η 03 ) 2

[0062] Φ4 = (η 30 + η 12 ) 2 +(η 21 + η 03 ) 2

[0063] Φ5 = (η 30 + 3η 12 )(η 30 + η 12 )[(η 30 + η 12 ) 2 - 3(η 21 + η 03 ) 2 +

[0064] (3η 21 - η 03 )(η 21 + η 03 )[3(η 30 + η 12 ) 2 -(η 21 + η 03 ) 2 )

[0065] Φ6 = (η 20 - η 02 )[(η 30 + η 12 ) 2 -(η 21 + η 03 ) 2 + 4η 11 (η30 +

[0066] η 12 )(η 21 +η 03 )

[0067] Φ7=(3η 21 -η 03 )(η 30 +η 12 )[(η 30 +η 12 ) 2 -3(η 21 +η 03 ) 2 ]+

[0068] (3η 12 -η 03 )(η 21 +η 03 )[3(η 30 +η 12 ) 2 -(η 21 +η 03 ) 2 ].

[0069] Other existing algorithm applications include the algorithm for correcting apartment size, the algorithm for finding walls, the algorithm for merging walls, the algorithm for eliminating isolated edges, the algorithm for finding small door and window areas, the algorithm for extracting lines from images, and the algorithm for identifying and drawing bay windows.

[0070] The serial number module has: 0-no corresponding template / new template, and >0-there is a corresponding template in the database;

[0071] The coordinate module has: a specific position of the corresponding identification area on the floor plan;

[0072] The local picture address or server picture address module has the functions of: capturing pictures and saving them; saving them to a local location and uploading them to a database to obtain a server picture address and uploading them to a database;

[0073] The recognition result type module has the following settings: 0-failed to recognize 1-door 2-window 3-sliding door 4-bay window 5 recognition is empty, and the corresponding PanType is selected to be drawn on the floor plan according to the recognition result;

[0074] The module for recording the current floating window width has the following features: 0-not a floating window>0-a floating window or a floating window before modification. The floating window value is the stored floating window width, and the floating window direction is automatically identified and drawn according to the saved floating window width;

[0075] The blueprint texture module has the following features: acquiring texture data by reading a local file and embedding it into a control blueprint;

[0076] The whether to modify the identifier module has the functions of: recording whether the template type corresponding to the pattern is modified by the user, and if modified, simultaneously modifying the corresponding template in the database or uploading a new template;

[0077] The matching recognition result value module has the following features: when the value is large, it is defined as no recognition; when the value is less than a certain value, the modification information will also modify the database content; when the value is within the modification ~ matching interval, a new template is uploaded;

[0078] The template contour data stream module has the following features: compressing and storing in the form of data stream, taking the matching average value from the longest to the shortest according to the length of contours in the contour group during matching, and when the average value is less than a certain value, it is considered that the corresponding template is matched;

[0079] The template matching times module has the following features: when the server reads the template, it arranges the templates from large to small according to the template matching times, so that the templates with more recognition times are more likely to be matched;

[0080] The upload modification time module has the functions of recording the upload time, recording the modification template type or the time of completing a match.

[0081] The automatic drawing of the two-dimensional floor plan also includes the following settings:

[0082] 1). Using black pixels as markers, set the minimum length to 25 cm and the width to 8 cm, and identify all pixel line segments along the X-axis and the Y-axis of the wall;

[0083] 2). According to the algorithm, the pixel line segments are combined into a single line segment to obtain the X-axis wall line segment and the Y-axis wall line segment;

[0084] 3). Weld and align the endpoints of the line segments according to the algorithm;

[0085] 4)Search XX, XY, YY, YX in sequence, all isolated and aligned endpoints with a cutoff distance of more than 30 cm, and according to the cutoff distance, intercept the appropriate pixel set of the floor plan, obtain the image contour, match it with the contour of the floor plan template element, and determine whether it is a door, window, or sliding door;

[0086] 5) Convert the pixel coordinates of all obtained line segments into spatial coordinates, add them to the apartment line segment MAP, and segment them according to the intersection relationship of the line segments;

[0087] 6). Automatically search for closed areas through algorithms and generate ground and top data;

[0088] 7) Complete the generation of the apartment model.

[0089] The specific method is as follows:

[0090] (1) Upload the floor plan;

[0091] (ii) Whether to correct the floor plan. If yes, the user completes the correction and modifies the wall drawing parameters. If no, the user directly proceeds to the next step.

[0092] (III) AI automatic drawing:

[0093] Draw the wall and start drawing: call AIDrawWall to get the wall drawing parameters and set them → call searchMapWallLine to find the wall lines on the floor plan → call correctCombineWall to connect the wall lines, call searchIndividualWall to delete the isolated edges, and call reCombineWall to align them again; call searchDoorWindows to traverse the obtained wall lines, and get the specific coordinates of objects such as doors, windows, and sliding doors by judging the wall endpoints isConnectedWithXWall and isConnectedWithYWall; call AIDrawWallCheck to crop the floor plan according to the obtained coordinates and the width of doors and windows to get several small blocks;

[0094] Convert the image into data stream and upload it to the service circle upload → get the image data stream → grayscale, binarize and Laplace process the image through opencv, get the image contour through findContours → upload back;

[0095] Integrate the contour data stream and the door and window parameter data, call QueryAIDrawWallCheck and send them to the server for matching. Upload and obtain the contour data stream, match the template contours in the database in order of contour length from large to small, and calculate the average value of all contour matching values. When the matching value is less than a certain threshold, it is considered a match and the type of the template is recorded.

[0096] When the template type meets the standard, the corresponding result is returned, otherwise 0 is returned, and the matching result type is received and stored. When the corresponding template is matched, the program selects the corresponding PanType according to the matching type and draws it on the floor plan. When there is no corresponding type, it is recorded as no vacancy and no drawing;

[0097] All results are displayed in the AI ​​recognition repair window for users to manually correct the recognition results;

[0098] The corresponding floor plan is redrawn according to the result, and a three-dimensional floor plan is generated according to the drawing result.

[0099] The following is a detailed diagram using the above modules, algorithms, and technologies.

[0100] Computer configuration:

[0101] The computer configuration requirements are Windows 10 64-bit system, 32G memory, 1T hard drive, Nvidia 3060 and above, AMD / Intel CPU.

[0102] Place the system with the above functional modules and algorithms,

[0103] There are floor plan templates, such as Figure 3 show:

[0104] 1. Upload the floor plan, such as Figure 4 show

[0105] In the Upload Floor Plan window, upload a floor plan with clear outlines.

[0106] 2. Correct the size, such as Figure 5 When the mouse cursor changes to a tape measure, start drawing. After drawing, enter the specific dimensions of the apartment and click OK.

[0107] 3. Draw the house type

[0108] After adjusting the corresponding values, click the AI ​​drawing button and the program will draw the floor plan with AI. Figure 6 show

[0109] AI repair recognition:

[0110] A separate window will display each specific identified door and window and its type, and the user can manually modify the incorrectly identified type. Figure 7 show;

[0111] Automatically generate apartment layout

[0112] The program automatically generates a specific three-dimensional apartment type based on the recognition results, such as Figure 8 show.

[0113] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An AI drawing apartment modeling method based on a real-time rendering engine, characterized in that: The process includes: inputting picture information, using the drawing system to perform contour matching, and segmenting the house type in the floor plan through specific thick wall lines, segmenting small blocks unique to doors, windows, sliding doors, and bay windows, performing image processing on each part to achieve the effect of clarifying the contour, and finally extracting the specific door and window contours in the picture, matching them with the door and window template contours stored in the database, and identifying them when the matching degree reaches a certain threshold, and automatically drawing the floor plan according to the identification results.

2. According to the AI ​​drawing apartment modeling method based on real-time rendering engine according to claim 1, it is characterized in that: The drawing system includes an AI drawing related data structure module and an algorithm module. The AI ​​drawing related data structure module includes a data structure for storing AI drawing related information and an AI drawing template database, and the two modules are connected. The data structure for storing AI drawing related information is provided with a serial number, coordinates, local image address, recognition result type, recording current floating window width, UE blueprint texture, whether to modify the mark, and AI matching recognition result value module, and is respectively connected to the data structure for storing AI drawing related information; the AI ​​drawing template database includes a template outline data stream, a template matching number, and an upload modification time module, and is respectively connected to the AI ​​drawing template database; The algorithm modules include an algorithm for correcting apartment size, an algorithm for finding walls, an algorithm for merging walls, an algorithm for eliminating isolated edges, an algorithm for finding small door and window areas, an algorithm for extracting image lines, an algorithm for contour matching, and an algorithm for identifying and drawing bay windows.

3. According to claim 2, the AI ​​drawing apartment modeling method based on real-time rendering engine is characterized in that: The primitive contour matching is adopted, and the contour matching algorithm calculation formula is: matching value result = total of overall contour matching values / number of contours.

4. The AI ​​drawing apartment modeling method based on real-time rendering engine according to claim 3 is characterized in that: The serial number module has: 0-no corresponding template / new template, and >0-there is a corresponding template in the database; The coordinate module has: a specific position of the corresponding identification area on the floor plan; The local picture address or server picture address module has the functions of: capturing pictures and saving them; saving them to a local location and uploading them to a database to obtain a server picture address and uploading them to a database; The recognition result type module has the following settings: 0-failed to recognize 1-door 2-window 3-sliding door 4-bay window 5 recognition is empty, and the corresponding PanType is selected to be drawn on the floor plan according to the recognition result; The module for recording the current floating window width has the following features: 0-not a floating window>0-a floating window or a floating window before modification. The floating window value is the stored floating window width, and the floating window direction is automatically identified and drawn according to the saved floating window width; The blueprint texture module has the following features: acquiring texture data by reading a local file and embedding it into a control blueprint; The whether to modify the identifier module has the functions of: recording whether the template type corresponding to the pattern is modified by the user, and if modified, simultaneously modifying the corresponding template in the database or uploading a new template; The matching recognition result value module has the following features: when the value is large, it is defined as no recognition; when the value is less than a certain value, the modification information will also modify the database content; when the value is within the modification ~ matching interval, a new template is uploaded; The template contour data stream module has the following features: compressing and storing in the form of data stream, taking the matching average value from the longest to the shortest according to the length of contours in the contour group during matching, and when the average value is less than a certain value, it is considered that the corresponding template is matched; The template matching times module has the following features: when the server reads the template, it arranges the templates from large to small according to the template matching times, so that the templates with more recognition times are more likely to be matched; The upload modification time module has the functions of recording the upload time, recording the modification template type or the time of completing a match.

5. The method for AI drawing house type modeling based on real-time rendering engine according to claim 4 is characterized in that: The automatic drawing of the two-dimensional floor plan also includes the following settings: 1). Using black pixels as markers, set the minimum length to 25 cm and the width to 8 cm, and identify all pixel line segments along the X-axis and the Y-axis of the wall; 2). According to the algorithm, the pixel line segments are combined into a single line segment to obtain the X-axis wall line segment and the Y-axis wall line segment; 3). Weld and align the endpoints of the line segments according to the algorithm; 4)Search XX, XY, YY, YX in sequence, all isolated and aligned endpoints with a cutoff distance of more than 30 cm, and according to the cutoff distance, intercept the appropriate pixel set of the floor plan, obtain the image contour, match it with the contour of the floor plan template element, and determine whether it is a door, window, or sliding door; 5) Convert the pixel coordinates of all obtained line segments into spatial coordinates, add them to the apartment line segment MAP, and segment them according to the intersection relationship of the line segments; 6). Automatically search for closed areas through algorithms and generate ground and top data; 7) Complete the generation of the apartment model.

6. The method for AI drawing house type modeling based on real-time rendering engine according to claim 5 is characterized in that: The specific method is as follows: (1) Upload the floor plan; (ii) Whether to correct the floor plan. If yes, the user completes the correction and modifies the wall drawing parameters. If no, the user directly proceeds to the next step. (III) AI automatic drawing: Draw the wall and start drawing: call AIDrawWall to get the wall drawing parameters and set them → call searchMapWallLine to find the wall lines on the floor plan → call correctCombineWall to connect the wall lines, call searchIndividualWall to delete the isolated edges, and call reCombineWall to align them again; call searchDoorWindows to traverse the obtained wall lines, and get the specific coordinates of objects such as doors, windows, and sliding doors by judging the wall endpoints isConnectedWithXWall and isConnectedWithYWall; call AIDrawWallCheck to crop the floor plan according to the obtained coordinates and the width of doors and windows to get several small blocks; Convert the image into data stream and upload it to the service circle upload → get the image data stream → grayscale, binarize and Laplace process the image through opencv, get the image contour through findContours → upload back; Integrate the contour data stream and the door and window parameter data, call QueryAIDrawWallCheck and send them to the server for matching. Upload and obtain the contour data stream, match the template contours in the database in order of contour length from large to small, and calculate the average value of all contour matching values. When the matching value is less than a certain threshold, it is considered a match and the type of the template is recorded. When the template type meets the standard, the corresponding result is returned, otherwise 0 is returned, and the matching result type is received and stored. When the corresponding template is matched, the program selects the corresponding PanType according to the matching type and draws it on the floor plan. When there is no corresponding type, it is recorded as no vacancy and no drawing; All results are displayed in the AI ​​recognition repair window for users to manually correct the recognition results; The corresponding floor plan is redrawn according to the result, and a three-dimensional floor plan is generated according to the drawing result.

Images