A fast vectorization method for real estate surveying and mapping based on deep learning

Abstract

The present invention provides a method for rapid vectorization of real estate survey maps based on deep learning, belonging to the field of image processing technology. The method specifically comprises the following steps: Step 1: dividing the real estate survey map data to obtain a table area and a main figure area; Step 2: extracting the coordinate data of the main figure from the table area using high-precision text recognition technology; Step 3: automatically vectorizing the main figure area using a trained UNET++ deep learning model to obtain a vectorized result; Step 4: accurately matching the vectorized result with a building line map by establishing a corresponding relationship between the coordinate data, and applying affine transformation technology to standardize the vectorized result based on the actual spatial dimension information in the building line map to obtain the final vector data with actual geographic coordinates. The present invention improves processing efficiency and accuracy.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of image processing technology, and in particular to a method for fast vectorization of real estate surveying maps based on deep learning. Background Art

[0002] Real estate survey maps are crucial documents in real estate management, recording and conveying key information such as a property's floor plan, dimensions, and ownership. These drawings serve not only as the foundation for real estate transactions and property registration, but also as crucial reference material for urban planning and real estate management. Vectorization of real estate survey maps, which converts traditional paper or scanned drawings into a computer-readable and processable vector format, is crucial for establishing modern real estate information management systems and improving property management efficiency. Vectorization enables digital storage, rapid retrieval, and intelligent analysis of real estate data, providing technical support for the standardized development of the real estate market.

[0003] Currently, the vectorization of real estate survey maps faces numerous technical challenges. Firstly, the quality of survey maps varies greatly, with significant variations in clarity between drawings from different eras and sources. The longevity of these drawings can also lead to blurred or broken lines. Secondly, survey maps are often overlaid with a large amount of textual information such as dimensions, room numbers, and various official seals. This additional information not only obscures the original graphic elements but also varies significantly in complexity and completeness, posing significant challenges to automated processing.

[0004] It can be seen that there is an urgent need for a fast vectorization method for real estate surveying and mapping based on deep learning with high processing efficiency and accuracy. Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides a deep learning-based rapid vectorization method for real estate surveying and mapping, which at least partially solves the problems of poor processing efficiency and accuracy in the prior art.

[0006] The embodiment of the present invention provides a method for fast vectorization of real estate surveying maps based on deep learning, including:

[0007] Step 1: Divide the real estate surveying map data into a table area and a main graphic area;

[0008] Step 2: Using high-precision text recognition technology to extract the coordinate data of the main shape from the table area;

[0009] The step 2 specifically includes:

[0010] Step 2.1: By traversing all the recognized texts in the table area and applying the preset filtering rules, the system prioritizes identifying text content that does not contain Chinese characters and selects the longest character string as the candidate measurement number;

[0011] Step 2.2, use the preset regular expression to verify the format of the candidate measurement number;

[0012] Step 2.3, determine the number of layers by identifying pure numeric text and screening the maximum value not greater than 99, and identify integers containing negative signs to determine the hierarchical information;

[0013] In step 2.4, when some information cannot be recognized, the system will automatically mark the corresponding field as N and synthesize the coordinate number data;

[0014] Step 3: Use the trained UNET++ deep learning model to automatically vectorize the main graphic area to obtain the vectorized result;

[0015] Step 4: Accurately match the vectorized result with the building line map by establishing a corresponding relationship between the coordinate number data. Based on the actual spatial size information in the building line map, the vectorized result is normalized using affine transformation technology to obtain the final vector data with actual geographic coordinates.

[0016] The step 4 specifically includes:

[0017] Step 4.1, matching the measurement number of the attribute information of the table area with the measurement number in the attribute table of the building line diagram to generate a mapping relationship table;

[0018] Step 4.2: Select at least 4 evenly distributed corner points from the mapping table as control points. , construct the transformation matrix to solve the equation:

[0019] ;

[0020] in, are the pixel coordinates in the source image, are the coordinates in the target geographic coordinate system, is the affine transformation coefficient to be solved;

[0021] In step 4.3, the least squares method is used to solve the transformation matrix solution equation to obtain the optimal affine transformation matrix T. The affine transformation matrix T is applied to all vertex coordinates of the vectorized results generated by the UNET++ model. The transformed vertex coordinates are converted from the pixel coordinate system to the actual geographic coordinate system to obtain the final vector data with actual geographic coordinates.

[0022] According to a specific implementation of the embodiment of the present invention, step 1 specifically includes:

[0023] Step 1.1, calculate the row and column ratio of the image in the real estate surveying map data ,when When it is greater than 1, the image is rotated 90 degrees clockwise;

[0024] Step 1.2: Apply Gaussian filter to reduce noise on the image, and then use Canny edge detection operator to extract the image contour.

[0025] ;

[0026] ;

[0027] ;

[0028] in, is the Gaussian kernel formula, is the pixel coordinate, is the standard deviation of the Gaussian kernel, controlling the degree of smoothing, is the gradient strength, is the horizontal gradient, is the vertical gradient, For the gradient direction, it retains the local maximum along the gradient direction, suppresses non-maximum points, and refines the edges;

[0029] Step 1.3, for each detected contour, calculate its area and perimeter, set an area threshold and retain contours with an area greater than the area threshold;

[0030] Step 1.4, calculate the vertical position ratio of the contour, and select the optimal segmentation threshold based on the analysis of the real estate surveying and mapping data , and according to the optimal segmentation threshold Each contour with an area greater than the area threshold is divided into a table area and a main graphic area.

[0031] According to a specific implementation of the embodiment of the present invention, after step 1.2, the method further includes:

[0032] Use morphological operations to remove small noise in the extracted contours:

[0033] ;

[0034] in, is a structural element, represents the dilation operation, represents the erosion operation, is the processed binary image, is the extracted contour.

[0035] According to a specific implementation of an embodiment of the present invention, step 1.4 specifically includes:

[0036] When the height of the contour When , the outline is divided into table areas, containing attribute information;

[0037] When the height of the contour When , the outline is divided into the main graphic area, which contains the geometric shape information of the house. is the total height of the image.

[0038] According to a specific implementation of the embodiment of the present invention, step 4.1 specifically includes:

[0039] Step 4.1.1: construct the measurement number information extracted from the table area into a first association table;

[0040] Step 4.1.2, extracting the measurement number from the building line drawing attribute table and constructing it into a second association table;

[0041] In step 4.1.3, a spatial join operation is performed based on the measurement ID as the join key to generate a mapping relationship table.

[0042] The deep learning-based rapid vectorization solution for real estate surveying and mapping in an embodiment of the present invention includes: step 1, dividing the real estate surveying and mapping data to obtain a table area and a main body graphic area; step 2, extracting the coordinate number data of the main body shape from the table area using high-precision text recognition technology; step 3, using a trained UNET++ deep learning model to automatically vectorize the main body graphic area to obtain a vectorized result; step 4, by establishing a corresponding relationship between the coordinate number data, accurately matching the vectorized result with the house building line map, and applying affine transformation technology to standardize the vectorized result based on the actual spatial size information in the house building line map to obtain final vector data with actual geographic coordinates.

[0043] The beneficial effects of the embodiments of the present invention are as follows: through the solution of the present invention, through the innovative integration of deep learning and text recognition technology, fully automatic vectorization processing of real estate surveying maps is realized, which greatly improves the processing efficiency and accuracy compared with the semi-automatic method relying on manual intervention in the existing technology; the affine transformation method based on coordinate number matching is applied to solve the problem of accurate conversion from pixel coordinates to actual geographic coordinates, overcome the defects of relying on manual alignment in the existing technology, and realize high-precision automatic alignment of vectorization results with the actual outer contour of the house. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 A flowchart of a method for rapid vectorization of real estate surveying maps based on deep learning provided by an embodiment of the present invention;

[0046] Figure 2 A schematic diagram of a specific implementation process of a method for rapid vectorization of real estate surveying and mapping based on deep learning provided by an embodiment of the present invention;

[0047] Figure 3 This is a vectorization result of a real estate surveying map provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0049] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0050] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present invention, those skilled in the art will appreciate that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this apparatus and / or practice this method.

[0051] It should also be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. The illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0052] Additionally, in the following description, specific details are provided to provide a thorough understanding of the examples. However, one skilled in the art will appreciate that the aspects described can be practiced without these specific details.

[0053] Traditional methods for vectorizing real estate survey maps rely primarily on manual drawing or semi-automated processing. Manual drawing requires operators to manually draw the outlines and internal structures of buildings using a digitizer or computer software. This method is not only time-consuming but also susceptible to subjective judgment, making it difficult to ensure measurement accuracy and consistency. Furthermore, manual processing is inefficient and cannot meet the needs of large-scale real estate data processing. With the advancement of image processing technology, various automatic vectorization methods have emerged. Common methods include edge detection and image segmentation. Edge detection methods primarily use operators such as Canny and Sobel to detect edge information in images and extract building outlines through morphological processing and line tracing techniques. However, these methods place high demands on image quality and are prone to incomplete edge detection or false edge detection when processing images with high noise levels or blurred lines. Image segmentation methods attempt to divide images into distinct semantic regions using techniques such as region growing and watershed algorithms. However, these methods often struggle to accurately identify and segment detailed features of buildings when processing complex real estate survey maps. Segmentation results are particularly suboptimal when processing areas with overlapping stamps or densely labeled areas.

[0054] In recent years, deep learning technology has achieved breakthroughs in image processing. In particular, methods based on convolutional neural networks have demonstrated strong performance in tasks such as image classification, object detection, and semantic segmentation. This has provided new technical solutions for the vectorization of real estate survey maps. However, research on the application of deep learning technology to real estate survey map vectorization is relatively limited, and a fast deep learning-based vectorization method for real estate survey maps is still lacking.

[0055] It can be seen that (1) the traditional manual vectorization method is time-consuming and labor-intensive, and inefficient; the existing automatic vectorization method has high requirements for image quality and poor anti-interference ability. (2) It lacks the ability to recognize and process the attribute table area unique to real estate surveying maps; the vectorization results have poor correlation with the actual spatial coordinate system, making it difficult to perform actual size calibration.

[0056] An embodiment of the present invention provides a method for rapid vectorization of real estate surveying maps based on deep learning, which can be applied to the surveying map management process in real estate management scenarios.

[0057] See also Figure 1 , which is a flow chart of a method for fast vectorization of real estate surveying and mapping based on deep learning provided by an embodiment of the present invention. Figure 1 and Figure 2 As shown, the method mainly includes the following steps:

[0058] Step 1: Divide the real estate surveying map data into a table area and a main graphic area;

[0059] In specific implementation, the specific process of dividing the real estate surveying map can be as follows:

[0060] 1.1 Image preprocessing and contour extraction

[0061] First, by analyzing the characteristics of the real estate surveying map data, we found that its width is always greater than its height, so we calculated the row and column ratio of the image. ,when When it is greater than 1, that is, when the image height is greater than the width, the image is rotated 90 degrees clockwise to ensure that the image always remains horizontal:

[0062] ;

[0063] in, 、 are the number of rows and columns of the image, respectively.

[0064] Furthermore, a Gaussian filter is applied to the image for noise reduction, and then the Canny edge detection operator is used to extract the image edge. The calculation formula is as follows:

[0065] ;

[0066] ;

[0067] ;

[0068] in, is the Gaussian kernel formula, is the pixel coordinate, is the standard deviation of the Gaussian kernel, which controls the degree of smoothing; is the gradient strength, is the horizontal gradient, is the vertical gradient; It is the gradient direction, retains the local maximum along the gradient direction, suppresses non-maximum points, and refines the edges.

[0069] In order to further optimize the detection results, morphological operations are used to eliminate small noises:

[0070] ;

[0071] in, is a structural element; represents the dilation operation; represents the corrosion operation; is the processed binary image.

[0072] For each detected contour, calculate its area and perimeter. Set area threshold , only retaining contours whose area is larger than a threshold:

[0073] ;

[0074] in, is the area of the contour.

[0075] 1.2 Regional Division

[0076] Adaptive region division is performed based on the geometric features of the rectangle. First, the vertical position ratio of the rectangle is calculated. , by analyzing the real estate survey map, select the optimal segmentation threshold , this threshold can effectively distinguish the table area and the main graphic area. Specifically, when the vertical coordinate When the vertical coordinate is , the area is divided into the main graphic area, which contains the geometric shape information of the house.

[0077] ;

[0078] in, is the ordinate of the center point of the rectangle, is the total height of the image.

[0079] Step 2: Using high-precision text recognition technology to extract the coordinate data of the main shape from the table area;

[0080] In specific implementations, a high-precision text recognition system based on PaddleOCR can be used. By loading the PP-OCRv4 pre-trained model, combined with a text orientation classification model and a Chinese language recognition model, a complete recognition framework is constructed. The system automatically corrects text orientation during processing and performs intelligent preprocessing on the input image to ensure that the image always maintains the optimal recognition posture. In particular, when it detects that the image height is greater than the width, the system automatically performs a 90-degree clockwise rotation to ensure accurate text recognition.

[0081] During the recognition process, the system employs a multi-layered information extraction strategy. First, by traversing all recognized text and applying specific filtering rules, the system prioritizes text content that does not contain Chinese characters and selects the longest character string as a candidate measurement number. Subsequently, the candidate measurement numbers are formatted using strict regular expressions to ensure compliance with standard specifications. The system also intelligently extracts other key information, such as determining the number of layers by identifying pure numeric text and filtering for a maximum value no greater than 99, and identifying integers containing negative signs to determine hierarchical information. When certain information cannot be recognized, the system automatically marks the corresponding field with an "N" to ensure data integrity.

[0082] Step 3: Use the trained UNET++ deep learning model to automatically vectorize the main graphic area to obtain the vectorized result;

[0083] In specific implementation, the embodiment of the present invention adopts the UNet++ network architecture to realize automatic vectorization of the main graphics. , its feature map calculation expression is:

[0084] ;

[0085] in, Represents the dense convolution block operation of the i-th node in the l-th layer.

[0086] Feature map after convolution operation It can be expressed as:

[0087] ;

[0088] in, The convolution kernel is at position The weight of is the input feature map, is the bias term, is the half-width of the convolution kernel.

[0089] During network training, the cross entropy loss function is used:

[0090] ;

[0091] in, is the true label, is the probability predicted by the model, is the sample size.

[0092] After training the UNET++ deep learning model, for the input image , its feature extraction process is achieved through multi-scale convolution operation:

[0093] ;

[0094] in, For the The feature map of the layer, For the The convolution kernel weights of the layer, For the The bias term of the layer, is a non-linear activation function.

[0095] Step 4: By establishing a corresponding relationship between the coordinate number data, the vectorized results are accurately matched with the building line map. Based on the actual spatial size information in the building line map, the vectorized results are standardized using affine transformation technology to obtain the final vector data with actual geographic coordinates.

[0096] In specific implementation, the process of measurement number matching and affine transformation can be as follows:

[0097] 4.1 Measurement number matching:

[0098] The measurement number data of the main shape extracted from the table area through the above steps is matched with the measurement number information in the attribute table of the building line drawing. The building line drawing is vector data, and the measurement number information in the attribute table of this data is the outer outline of the real house in the real world. Its size and orientation are consistent with the real world. The main steps include: ① Constructing the measurement number information extracted from the table area into an association table {measurement ID, pixel coordinates (x, y)}; ② Extract the measurement ID information from the building line map attribute table and construct it into an association table {survey ID, geographic coordinates (lon, lat)}; ③ Based on the survey ID as the connection key, perform a spatial join operation to generate a mapping relationship table M {survey ID, pixel coordinates (x, y), geographic coordinates (lon, lat)};

[0099] 4.2 Affine transformation calculation based on control points

[0100] In order to achieve accurate transformation from pixel space to actual geographic space, this method uses affine transformation technology. Select at least 4 evenly distributed corner points from the mapping relationship table in step 4.1 as control points; for the selected control point set , construct the transformation matrix to solve the equation:

[0101] ;

[0102] in, are the pixel coordinates in the source image, are the coordinates in the target geographic coordinate system, is the affine transformation coefficient to be solved.

[0103] The least squares method is used to solve the equations to obtain the optimal affine transformation matrix T. The affine transformation matrix T is applied to all vertex coordinates of the preliminary vectorization results generated by the UNET++ model. The transformed vertex coordinates are converted from the pixel coordinate system to the actual geographic coordinate system.

[0104] The deep learning-based rapid vectorization method for real estate surveying and mapping provided in this embodiment realizes fully automatic vectorization processing of real estate surveying and mapping through the innovative integration of deep learning and text recognition technology. Compared with the semi-automatic method in the existing technology that relies on manual intervention, it greatly improves processing efficiency and accuracy. It applies an affine transformation method based on coordinate number matching to solve the problem of accurate conversion from pixel coordinates to actual geographic coordinates, overcomes the defect of relying on manual alignment in the existing technology, and realizes high-precision automatic alignment of vectorization results with the actual outer contour of the house.

[0105] The method of the present invention will be further described below in conjunction with a specific embodiment, using the real estate surveying map data and building line data of A to illustrate the specific implementation process of the present invention:

[0106] (1) In this embodiment, a real estate survey map of a certain area in area A is selected as input data. First, the input image is preprocessed, including Gaussian filtering and Canny edge detection, image edge extraction, and morphological operations to eliminate noise. Direction correction is performed by calculating the image row-column ratio to ensure that the image always remains horizontal. Then, the image is divided into regions based on rectangular geometric features and adaptive thresholds, separating the real estate survey map into the main graphic area and the table area.

[0107] (2) The table area is processed using a high-precision text recognition system based on PaddleOCR. The system loads the PP-OCRv4 pre-trained model and combines the text direction classification model and the Chinese recognition model to build a complete recognition framework. Through a multi-level information extraction strategy, the system identifies key information such as the measurement number and the number of layers. For the measurement number, the system prioritizes the recognition of text content that does not contain Chinese characters and uses regular expressions for format verification to ensure compliance with standard specifications.

[0108] (3) Automatic vectorization of the main graphic area was performed using the trained UNet++ deep learning model. The model extracts image features through multi-scale convolution operations and is optimized using a cross-entropy loss function. During implementation, 180 real estate survey maps of Site A were used as a dataset, 150 of which were used for model training and 30 for validation. The model was applied to the pre-processed main graphic area to generate preliminary vectorization results, which contain the geometric shape information of the houses but lack the actual geographic coordinates.

[0109] (4) Match the measurement number information extracted in step (2) with the measurement number data in the attribute table of the building line map. Perform a spatial join operation using the measurement number ID as the connection key to establish a correspondence between the preliminary vectorization result and the actual geographic coordinates. Based on the correspondence, select four evenly distributed corner points as control points, construct and solve the affine transformation matrix, and convert the vectorization result from the pixel coordinate system to the actual geographic coordinate system to obtain the final vectorization result as shown in the figure. Figure 3 shown.

[0110] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware or a combination thereof.

[0111] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for fast vectorization of real estate surveying maps based on deep learning, characterized in that: include: Step 1: Divide the real estate surveying map data into a table area and a main graphic area; Step 2: Using high-precision text recognition technology to extract the coordinate data of the main shape from the table area; The step 2 specifically includes: Step 2.1: By traversing all the recognized texts in the table area and applying the preset filtering rules, the system prioritizes identifying text content that does not contain Chinese characters and selects the longest character string as the candidate measurement number; Step 2.2, use the preset regular expression to verify the format of the candidate measurement number; Step 2.3, determine the number of layers by identifying pure numeric text and screening the maximum value not greater than 99, and identify integers containing negative signs to determine the hierarchical information; In step 2.4, when some information cannot be recognized, the system will automatically mark the corresponding field as N and synthesize the coordinate number data; Step 3: Use the trained UNET++ deep learning model to automatically vectorize the main graphic area to obtain the vectorized result; Step 4: Accurately match the vectorized result with the building line map by establishing a corresponding relationship between the coordinate number data. Based on the actual spatial size information in the building line map, the vectorized result is normalized using affine transformation technology to obtain the final vector data with actual geographic coordinates. The step 4 specifically includes: Step 4.1, matching the measurement number of the attribute information of the table area with the measurement number in the attribute table of the building line map, and generating a mapping relationship table M {measurement number ID, pixel coordinates (x, y), geographic coordinates (lon, lat)}; Step 4.2, select at least 4 evenly distributed corner points from the mapping relationship table as control points, and for the selected control point set P = {(x1, y1, lon1, lat1), (x2, y2, lon2, lat2), ..., (x n ,y n ,lon n ,lat n )}, construct the transformation matrix to solve the equation: Where (x, y) is the pixel coordinate in the source image, (x', y') is the coordinate in the target geographic coordinate system, and a ij is the affine transformation coefficient to be solved; In step 4.3, the least squares method is used to solve the transformation matrix solution equation to obtain the optimal affine transformation matrix T. The affine transformation matrix T is applied to all vertex coordinates of the vectorized results generated by the UNET++ model. The transformed vertex coordinates are converted from the pixel coordinate system to the actual geographic coordinate system to obtain the final vector data with actual geographic coordinates.

2. The method according to claim 1, characterized in that The step 1 specifically includes: Step 1.1, calculate the row-column ratio α of the image in the real estate surveying map data. When α is greater than 1, rotate the image 90 degrees clockwise; Step 1.2: Apply Gaussian filter to reduce noise on the image, and then use Canny edge detection operator to extract the image contour. Among them, G(x,y) is the Gaussian kernel formula, (x,y) is the pixel coordinate, σ is the standard deviation of the Gaussian kernel, which controls the degree of smoothness, and G is the gradient strength. is the horizontal gradient, is the vertical gradient, θ is the gradient direction, and the local maximum value is retained along the gradient direction, non-maximum points are suppressed, and edges are refined; Step 1.3, for each detected contour, calculate its area and perimeter, set an area threshold and retain contours with an area greater than the area threshold; Step 1.4, calculate the longitudinal position ratio of the contour, and select the optimal segmentation threshold k based on the analysis of the real estate surveying and mapping data. threshold , and according to the optimal segmentation threshold k threshold Each contour with an area greater than the area threshold is divided into a table area and a main graphic area.

3. The method according to claim 2, characterized in that After step 1.2, the method further comprises: Use morphological operations to remove small noise in the extracted contours: Among them, S is the structural element, represents the dilation operation, represents the erosion operation, B' is the processed binary image, and B is the extracted contour.

4. The method according to claim 3, characterized in that The step 1.4 specifically includes: When the height y of the contour <k threshold ×H, the outline is divided into table areas, which contain attribute information; When the height of the contour y ≥ k threshold ×H, the outline is divided into the main graphic area, which contains the geometric shape information of the house, and H is the total height of the image.

5. The method according to claim 4, characterized in that The step 4.1 specifically includes: Step 4.1.1: construct the measurement number information extracted from the table area into a first association table; Step 4.1.2, extracting the measurement number from the building line drawing attribute table and constructing it into a second association table; In step 4.1.3, a spatial join operation is performed based on the measurement ID as the join key to generate a mapping relationship table.

Images