Urban form dynamic evolution automatic modeling and updating method based on parameter iteration

Abstract

The invention provides an urban form dynamic evolution automatic modeling and updating method based on parameter iteration. The method comprises six steps of data acquisition and three-dimensional model library construction, building height information calibration, building change area identification, change area high-precision modeling, building function information identification and abnormal area identification and detection updating. According to the invention, the dynamic monitoring and accurate modeling of the city form are realized through the collection and processing of the high-resolution remote sensing image, the city three-dimensional model and the land function data in combination with the neural network model and the image processing algorithm, and the effective technical support is provided for the city planning and management.

Description

Technical Field

[0001] The present invention relates to the fields of urban planning, geographic information system (GIS), remote sensing technology, computer vision, machine learning, 3D modeling, and urban planning. Specifically, it combines remote sensing image processing, neural network model application, image recognition and segmentation technology, and lidar scanning technology to achieve dynamic monitoring, automatic modeling, and updating of urban morphology. Background Art

[0002] With the acceleration of the urbanization process, the dynamic evolution monitoring and modeling of urban morphology have become important requirements for urban planning and management. Traditional methods for urban morphology monitoring and modeling mainly rely on manual field surveys and two-dimensional map analysis. These methods are not only time-consuming and laborious but also difficult to capture the subtle changes in the urban three-dimensional morphology. In recent years, the development of remote sensing technology and geographic information system (GIS) has provided new technical means for the dynamic monitoring of urban morphology. High-resolution remote sensing images can provide large-scale urban surface information, while 3D modeling technology can visually display the urban spatial structure. However, how to efficiently and automatically extract urban morphology change information from remote sensing data and update the 3D model in real time remains a technical challenge. In existing technologies, although there are some methods that can achieve the construction and update of urban 3D models, these methods often have the following deficiencies: (1) Low data processing efficiency: Traditional data processing methods are difficult to meet the high-efficiency processing requirements of a large amount of remote sensing data. (2) Low accuracy: Existing automatic modeling technologies are difficult to meet the high-precision requirements for urban morphology modeling. (3) Insufficient recognition of functional information: Existing technologies are difficult to accurately identify and distinguish the functional information of buildings. (3) Weak dynamic update ability: Existing systems often lack effective dynamic monitoring and automatic update mechanisms. Summary of the Invention

[0003] Object of the Invention: Aiming at the deficiencies in the above background art, the present invention proposes an automatic modeling and updating system for the dynamic evolution of urban morphology based on parameter iteration, aiming to improve data processing efficiency, enhance modeling accuracy, strengthen the ability to recognize functional information, and achieve dynamic monitoring and automatic updating of urban morphology, which can be widely applied to fields such as urban planning, urban change analysis, and urban electronic map production.

[0004] Technical Solution Adopted by the Present Invention: An automatic modeling and updating method for the dynamic evolution of urban morphology based on parameter iteration, comprising the following steps:

[0005] S1: Implement data collection and construction of the 3D urban form model database; for a specified research period and area, collect dual-temporal high-resolution remote sensing images, 3D urban models, and urban land use function data; the image database integrates multi-city historical dual-temporal high-resolution remote sensing images; the 3D urban form model database stores the urban 3D model shapefile that matches the remote sensing images.

[0006] S2: Calibrate building height information; use the UNet neural network model to extract remote sensing information of buildings and shadows from the image database, estimate the building height by using the shadows extracted from the images and combining with the imaging time of the remote sensing images.

[0007] S3: Identify building change areas; use the region growing algorithm to compare the changes in satellite images, and mark the areas where the building outline changes exceed the threshold as change areas.

[0008] S4: High-precision modeling of change areas; perform edge detection on satellite images through the Sobel algorithm to accurately capture the edges, corners, and texture features of buildings; use the watershed algorithm to perform region segmentation based on the image topology structure to separate individual buildings; use the Marching Squares algorithm to track the contours of the segmented areas to form the accurate contour lines of buildings; optimize the contours to remove noise and small-area irrelevant details; finally, further refine the shape and continuity of the contours through morphological dilation and erosion operations.

[0009] S5: Identify building function information; perform high-precision modeling of the change areas extracted in step S4 according to land use functions to obtain building function features, and use different color blocks to distinguish building functions; construct an index system for describing building geometric features, use the entropy weight method to assign weights to the indicators, calculate the high-precision modeling of the change areas extracted in step S4 to obtain building geometric features, and then segment the measured values and assign corresponding colors to each segment of the values.

[0010] Further, in step S2, a UNet neural network model is used to extract remote sensing information of buildings and shadows from the image database, including: using the original UNet convolutional neural network model to extract urban building information. First, the remote sensing images in the image database are data-annotated to generate label data for training and validation. The annotated content includes the boundaries of buildings and shadows, which are represented in the form of pixel-level binary masks. The remote sensing image data is cropped to 256 pixels * 256 pixels, and then the image data is normalized to improve the stability and convergence speed of network training. The dataset is augmented through operations such as rotation, flipping, cropping, and scaling to obtain enhanced sample data. The sample data is divided into a training set and a validation set in a ratio of 4:1. When the accuracy and loss rate of the validation set tend to be stable and the model converges during model training, the optimal model is saved for extracting building and shadow information.

[0011] Further, step S2 estimates the building height by using the shadows extracted from the image and combining the imaging time of the remote sensing image, including: estimating the building height by means of the directional relationship between the building and the shadow when the image is taken; generating several parallel lines along the solar azimuth direction and calculating the average value of the lengths of the lines inside the building shadow to estimate the length of the building shadow; there are two geometric relationships between the solar altitude angle and the building shadow, namely, the cases where the sun and the satellite are on the same side and on different sides of the building; when the sun and the satellite are on the same side of the building, the estimation formula for the building height H is where ED is the length of the shadow on the image, EC is the actual length of the building shadow, CD is the length of the shadow that cannot be shown on the image due to building occlusion, α is the solar altitude angle, and β is the satellite altitude angle; when the sun and the satellite are on different sides of the building, the estimation formula for the building height is H = EC × tanα = ED × tanα.

[0012] Further, step S3 uses the region growing algorithm to compare the changes in satellite images, including:

[0013] Select a seed point located at the changes in the building's functional characteristics and geometric characteristics; then set the growth criteria, including the similarity criterion and the change threshold; then start from the seed point and judge the adjacent pixels or regions according to the growth criteria. If the similarity criterion is met, it is included in the changed area and region growing is performed; during the growth process, the adjacent regions that meet the conditions are merged to form a larger changed area; when the outer contour of the changed area changes by more than the preset threshold, it is marked as a changed area; finally, when no new pixels or regions meet the growth criteria, the region growing process terminates, thereby realizing the comparison of the changes in the building's functional characteristics and geometric characteristics and effectively marking the areas with significant changes in the outer contour.

[0014] Further, the Sobel algorithm in step S4 performs edge detection on the satellite image, including:

[0015] Step S4-1-1 Satellite image preprocessing: Perform Gaussian filtering preprocessing on the input satellite image of the changed area, use a 5×5 convolution kernel for noise suppression, and set the standard deviation σ = 1.5 to balance the denoising effect and edge retention;

[0016] Step S4-1-2 Gradient calculation: Construct the horizontal Sobel operator G x and the vertical Sobel operator G y , and perform two-dimensional convolution operations with the preprocessed image respectively to obtain the horizontal gradient matrix I x and the vertical gradient matrix I y , where the G x and G y are:

[0017]

[0018] I x and I y are:

[0019]

[0020] where A is the original satellite image; Calculate the pixel-level gradient magnitude G and gradient direction θ based on the gradient matrix, and establish a gradient feature field. The G and θ are:

[0021]

[0022] Step S4-1-3 Edge determination: Use the non-maximum suppression algorithm to refine the edges along the gradient direction, retain the pixel points with the local maximum gradient value, and eliminate the edge blurring effect; Set the dynamic double thresholds T low and T high , and divide the pixel points into three categories: strong edges, weak edges, and non-edges. The T low and T high are:

[0023] T low = 0.05×G max , T high = 0.15×G max

[0024] where G max is the maximum value of the gradient magnitude G;

[0025] Pixel classification rules

[0026] Category Condition Processing method Physical meaning Strong edge <![CDATA[G≥T high > Directly retained as the final edge Significant structural edges (such as building outlines, road boundaries) Weak edge <![CDATA[T low ≤G<T high > Connectivity needs to be verified Possible edge fragments or noise (such as texture details, local shadows) Non-edge <![CDATA[G<T low > Directly removed Flat areas or noise (such as sky, wall surface) ;

[0027] Step S4-1-4 Edge connection optimization: Perform eight-neighborhood connectivity detection on weak edge pixels. When it is adjacent to strong edges, it is retained as a valid edge; otherwise, it is removed. Output a binary edge feature map, and at the same time record the gradient direction matrix for subsequent building corner detection, where the corner determination criterion is that the gradient direction has a direction mutation of more than 45° within a 3×3 window.

[0028] Further, the region segmentation based on the image topology using the watershed algorithm in step S4 includes:

[0029] Step S4-2-1 Gradient magnitude normalization and image enhancement: Normalize the gradient magnitude matrix G output by the Sobel edge detection to the range [0, 255] to generate the input gradient image for the watershed algorithm. The specific method is:

[0030]

[0031] where G norm (i,j) is the value of the normalized gradient magnitude matrix at position (i,j), G(i,j) is the value of the original gradient magnitude matrix at position (i,j), G min and G max are the global minimum and maximum values of the gradient magnitude respectively; perform histogram equalization on the normalized image to enhance the edge features in low-contrast regions;

[0032] Step S4-2-2 Distance field map generation and seed marker extraction: Binarize the gradient image with a threshold T binary =0.1×G max , generate a building base mask, and obtain the Euclidean distance transform map D(x,y), which represents the geometric distance from each pixel to the nearest background; search for local maximum points in the distance field map as initial seeds, satisfying:

[0033] D(x,y)>D(x±1,y±1) and D(x,y)≥5 pixels. The marker point spacing needs to be greater than the minimum building interval to avoid over-dense seeds; where D(x±1,y±1) represents the distance values of the pixels in the 8-neighborhood of the pixel point;

[0034] Step S4-2-3 Seed marker optimization: Perform morphological opening operation on the seed marker image using a 3×3 circular structuring element to eliminate isolated noise markers with an area less than 10 pixels, and retain candidate building cores with a connected region area S≥20 pixels 2 ;

[0035] Step S4-2-4 Watershed segmentation: Superimpose the optimized seed marker image and the normalized gradient image to generate the input image I with marker constraints marked, where the seed area is forced to be set to the minimum gray value of 0; initialize the water level to 0, traverse the pixels in ascending order of gray value, take the seed area as the initial catchment basin, when the water level rises to the watershed line (adjacent basins meet), record the barrier position, and the flooding termination condition is that all basin boundaries are extended until all basins complete boundary division;

[0036] Step S4-2-5 Region merging optimization; Implement merging for over-segmented regions, which need to meet the following conditions simultaneously:

[0037] (a) The average gradient magnitude of the watershed line between adjacent regions Satisfies:

[0038]

[0039] (b) Area ratio:

[0040]

[0041] For region pairs that meet the conditions, delete the watershed barrier and update the region labels, output the segmentation result map with topological labels, where each independent labeled region corresponds to a single building entity, and record the region adjacency relationship matrix for subsequent contour tracing; among them, min(S1, S2) and max(S1, S2) respectively represent the smaller value and the larger value of the areas of the two regions.

[0042] Furthermore, in the construction of the building geometric feature description index system in step S5, using the entropy weight method to assign weights to the indicators includes:

[0043] Geometric feature description index system

[0044]

[0045]

[0046] Advantages of the present invention:

[0047] 1. Constructed an automatic modeling and updating method process for the dynamic evolution of urban form based on parameter iteration. Compared with traditional methods, the recognition efficiency has been increased by at least 80%, reducing manual intervention and realizing rapid and automated monitoring of urban form changes.

[0048] 2. Reduced the misrecognition rate: By optimizing the threshold setting and growth criteria of the region growing algorithm, the present invention significantly reduces the misrecognition rate in building change detection. Compared with traditional methods, the misrecognition rate has decreased by at least 30%, thereby improving the accuracy and reliability of the marked change regions and reducing the workload of subsequent manual review. Description of the Drawings

[0049] Figure 1This is the flowchart of the automatic modeling and updating method for the dynamic evolution of urban form based on parameter iteration of the present invention. Detailed implementation manners

[0050] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation manners:

[0051] As Figure 1 shown, an automatic modeling and updating method for the dynamic evolution of urban form based on parameter iteration includes the following steps:

[0052] S1: Implement data collection and construction of the urban form three-dimensional model database; for a specified research period and area, collect dual-temporal high-resolution remote sensing images, urban three-dimensional models, and urban land use function data; the image database integrates multi-city historical dual-temporal high-resolution remote sensing images; the urban form three-dimensional model database stores the urban three-dimensional model shapefile files that match the remote sensing images.

[0053] S2: Calibrate building height information; use the UNet neural network model to extract remote sensing information of buildings and shadows from the image database, and estimate the building height by using the shadows extracted from the images and combining with the imaging time of the remote sensing images.

[0054] The step S2 uses the UNet neural network model to extract remote sensing information of buildings and shadows from the image database, including: using the original UNet convolutional neural network model to extract urban building information. First, label the remote sensing images in the image database to generate label data for training and verification. The labeled content includes the boundaries of buildings and shadows, represented in the form of pixel-level binary masks; crop the remote sensing image data to 256 pixels * 256 pixels, and then normalize the image data to improve the stability and convergence speed of network training; expand the data set through operations such as rotation, flipping, cropping, and scaling to enhance and obtain sample data; divide the sample data into a training set and a verification set according to a ratio of 4:1. When the accuracy and loss rate of the verification set tend to be stable and the model converges during model training, save the optimal model for extracting building and shadow information.

[0055] The step S2 estimates the building height by using the shadows extracted from the images and combining with the imaging time of the remote sensing images, including: estimating the building height by means of the directional relationship between the building and the shadow when the image is taken; generating several parallel lines along the solar azimuth angle direction, and calculating the average value of the lengths of the lines inside the building shadow to estimate the building shadow length; there are two geometric relationships between the solar altitude angle and the building shadow, namely the cases where the sun and the satellite are on the same side and on different sides of the building; when the sun and the satellite are on the same side of the building, the estimation formula for the building height H is Where ED is the length of the shadow on the image, EC is the actual length of the building's shadow, CD is the length of the shadow that cannot be displayed on the image due to building occlusion, α is the solar altitude angle, and β is the satellite altitude angle; when the sun and the satellite are on opposite sides of the building, the building height estimation formula is H = EC × tanα = ED × tanα;

[0056] S3: Identification of building change areas; Using the region growing algorithm to compare the changes in satellite images, and marking the areas where the outer contour of the building changes beyond the threshold as change areas;

[0057] The comparison of the changes in satellite images by the region growing algorithm in step S3 includes:

[0058] Select a seed point located at the changes in the building's functional characteristics and geometric characteristics; then set the growth criteria, including the similarity criterion and the change threshold; then start from the seed point, and judge the adjacent pixels or regions according to the growth criteria. If the similarity criterion is met, include them in the change area and perform region growth; during the growth process, merge the adjacent regions that meet the conditions to form a larger change area; when the change in the outer contour of the change area exceeds the preset threshold, mark it as a change area; finally, when no new pixels or regions meet the growth criteria, the region growth process terminates, so as to realize the comparison of the changes in the building's functional characteristics and geometric characteristics, and effectively mark the areas with significant changes in the outer contour;

[0059] S4: High-precision modeling of change areas; Perform edge detection on satellite images through the Sobel algorithm to accurately capture the edges, corners and texture features of buildings; use the watershed algorithm to perform region segmentation based on the image topology structure to separate individual buildings; use the Marching Squares algorithm to perform contour tracking on the segmented regions to form an accurate contour line of the building; optimize the contour to remove noise and small-area irrelevant details; finally, through morphological dilation and erosion operations, further refine the shape and continuity of the contour;

[0060] The edge detection of satellite images by the Sobel algorithm in step S4 includes:

[0061] Step S4-1-1 Satellite image preprocessing; Perform Gaussian filtering preprocessing on the input satellite image of the change area, use a 5×5 convolution kernel for noise suppression, and set the standard deviation σ = 1.5 to balance the denoising effect and edge preservation;

[0062] Step S4-1-2 Gradient calculation; Construct the horizontal Sobel operator G x and the vertical Sobel operator G y , and perform two-dimensional convolution operations with the preprocessed image respectively to obtain the horizontal gradient matrix I x and the longitudinal gradient matrix Iy , where the G x and G y are:

[0063]

[0064] I x and I y are:

[0065]

[0066] Among them, A is the original satellite image; the pixel-level gradient magnitude G and gradient direction θ are calculated based on the gradient matrix to establish a gradient feature field, and the G and θ are:

[0067]

[0068] Step S4-1-3 Edge determination; the non-maximum suppression algorithm is used to refine the edge along the gradient direction, retaining the pixel points with the local gradient maximum value and eliminating the edge blurring effect; set the dynamic double thresholds T low and T high , and the pixel points are divided into three categories: strong edges, weak edges, and non-edges. The T low and T high are:

[0069] T low = 0.05×G max , T high = 0.15×G max

[0070] Among them, G max is the maximum value of the gradient magnitude G;

[0071] Pixel classification rule

[0072] Category Condition Processing method Physical meaning Strong edge <![CDATA[G≥T high > Directly retained as the final edge Significant structural edges (such as building outlines, road boundaries) Weak edge <![CDATA[T low ≤G<T high > Connectivity needs to be verified Possible edge fragments or noise (such as texture details, local shadows) Non-edge <![CDATA[G<T low > Directly removed Flat areas or noise (such as sky, wall surface) ;

[0073] Step S4-1-4 Edge connection optimization; perform an eight-neighborhood connectivity detection on the weak edge pixel points. When they are adjacent to the strong edge pixels, they are retained as valid edges, otherwise they are eliminated; output a binary edge feature map, and at the same time record the gradient direction matrix for subsequent building corner detection, where the corner determination criterion is that the gradient direction has a direction mutation of more than 45° within a 3×3 window;

[0074] The implementation of region segmentation based on the image topology using the watershed algorithm in step S4 includes:

[0075] Step S4-2-1 Gradient Magnitude Normalization and Image Enhancement; Normalize the gradient magnitude matrix G output by Sobel edge detection to the interval [0, 255] to generate the input gradient image for the watershed algorithm. The specific method is as follows:

[0076]

[0077] where G norm (i,j) is the value of the normalized gradient magnitude matrix at position (i,j), G(i,j) is the value of the original gradient magnitude matrix at position (i,j), G min and G max are the global minimum and maximum values of the gradient magnitude respectively; Perform histogram equalization on the normalized image to enhance the edge features in low-contrast regions;

[0078] Step S4-2-2 Distance Field Map Generation and Seed Marker Extraction; Binarize the gradient image with a threshold T binary = 0.1×G max to generate a building base mask, and obtain the Euclidean distance transform map D(x,y), which represents the geometric distance from each pixel to the nearest background; Search for local maximum points in the distance field map as initial seeds, satisfying:

[0079] D(x,y)>D(x±1,y±1) and D(x,y)≥5 pixels. The marker point spacing needs to be greater than the minimum building interval to avoid over-dense seeds; where D(x±1,y±1) represents the distance values of the pixels in the 8-neighborhood of the pixel point;

[0080] Step S4-2-3 Seed Marker Optimization; Perform morphological opening on the seed marker image using a 3×3 circular structuring element to eliminate isolated noise markers with an area less than 10 pixels, and retain candidate building cores with a connected area S≥20 pixels 2 ;

[0081] Step S4-2-4 Watershed Segmentation; Superimpose the optimized seed marker image and the normalized gradient image to generate the input image I marked constrained by markers, where the seed region is forced to be set to the minimum gray value 0; Initialize the water level to 0, traverse the pixels in ascending gray value, use the seed region as the initial catchment basin, and when the water level rises to the watershed line (adjacent basins meet), record the barrier position. The flooding termination condition is that all basin boundaries expand until all basin boundaries are divided;

[0082] Step S4-2-5 Region Merging Optimization; Merge the over-segmented regions, which need to meet the following conditions simultaneously:

[0083] (a) The average gradient magnitude of the watershed line between adjacent regions Satisfies:

[0084]

[0085] (b) Area ratio:

[0086]

[0087] For the region pairs that meet the conditions, delete the watershed barriers and update the region labels, and output the segmentation result map with topological labels, where each independent labeled region corresponds to a single building entity, and record the region adjacency relationship matrix for subsequent contour tracing; among them, min(S1, S2) and max(S1, S2) respectively represent the smaller value and the larger value of the areas of the two regions.

[0088] S5: Identification of building function information; perform high-precision modeling on the changed regions extracted in step S4 according to the land use function, obtain the building function characteristics, and use different color blocks to distinguish the building functions; construct an index system for describing building geometric characteristics, use the entropy weight method to assign weights to the indexes, calculate the high-precision modeling of the changed regions extracted in step S4 to obtain the building geometric characteristics, and then segment the measured values and assign corresponding colors to each segment of the values;

[0089] The construction of the index system for describing building geometric characteristics and the use of the entropy weight method to assign weights to the indexes in step S5 include:

[0090] Index system for describing geometric characteristics

[0091]

[0092]

[0093] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. An automatic modeling and updating method for the dynamic evolution of urban form based on parameter iteration, characterized in that The steps include: S1: Implement data collection and construction of the urban form 3D model database; for a specified research period and area, collect dual-temporal high-resolution remote sensing images, urban 3D models, and urban land use function data; the image database integrates multi-city historical dual-temporal high-resolution remote sensing images; the urban form 3D model database stores the urban 3D model shapefile that matches the remote sensing images. S2: Calibrate building height information; use the UNet neural network model to extract remote sensing information of buildings and shadows from the image database, and estimate the building height by using the extracted shadows from the images and combining with the imaging time of the remote sensing images. S3: Identify building change areas; use the region growing algorithm to compare the changes in satellite images, and mark the areas where the building outline changes exceed the threshold as change areas. S4: High-precision modeling of change areas; perform edge detection on satellite images through the Sobel algorithm to accurately capture the edges, corners, and texture features of buildings; use the watershed algorithm to perform region segmentation based on the image topology structure to separate individual buildings; use the Marching Squares algorithm to perform contour tracking on the segmented areas to form the accurate contour lines of buildings; optimize the contours to remove noise and small-area irrelevant details; finally, through morphological dilation and erosion operations, further refine the shape and continuity of the contours. S5: Identify building function information; perform high-precision modeling of the change areas extracted in step S4 according to the land use function, obtain the building function characteristics, and use different color blocks to distinguish the building functions. Construct an index system for describing building geometric features, use the entropy weight method to assign weights to the indexes, calculate the high-precision modeling of the change areas extracted in step S4 to obtain the building geometric features, and then segment the measured values and assign corresponding colors to each segment of the values.

2. The automatic modeling and updating method for dynamic evolution of urban form based on parameter iteration according to claim 1, characterized in that In step S2, the UNet neural network model is used to extract remote sensing information of buildings and shadows from the image database, including: using the original UNet convolutional neural network model to extract urban building information. First, perform data annotation on the remote sensing images in the image database to generate label data for training and verification. The annotation content includes the boundaries of buildings and shadows, represented in the form of pixel-level binary masks; crop the remote sensing image data to 256 pixels * 256 pixels, and then perform normalization processing on the image data to improve the stability and convergence speed of network training; expand the data set through operations such as rotation, flipping, cropping, and scaling to enhance the obtained sample data; divide the sample data into a training set and a verification set according to a ratio of 4:

1. When the accuracy and loss rate of the verification set tend to be stable and the model converges during model training, save the optimal model for extracting building and shadow information.

3. The automatic modeling and updating method for dynamic evolution of urban form based on parameter iteration according to claim 2, characterized in that, The step S2 estimates the building height by using the shadow extracted from the image and combining the imaging time of the remote sensing image, including: estimating the building height by means of the directional relationship between the building and the shadow during image imaging; generating a number of parallel lines along the solar azimuth direction and calculating the average value of the lengths of the lines inside the building shadow to estimate the length of the building shadow; there are two geometric relationships between the solar altitude angle and the building shadow, namely, the cases where the sun and the satellite are on the same side and on different sides of the building; when the sun and the satellite are on the same side of the building, the estimation formula for the building height H is where ED is the length of the shadow on the image, EC is the actual length of the building shadow, CD is the length of the shadow that cannot be shown on the image due to building occlusion, α is the solar altitude angle, and β is the satellite altitude angle; when the sun and the satellite are on different sides of the building, the estimation formula for the building height is H = EC × tanα = ED × tanα.

4. The automatic modeling and updating method for the dynamic evolution of urban form based on parameter iteration according to claim 3, characterized in that The region growing algorithm in step S3 for comparing the changes in satellite images includes: Select a seed point located at the changes in building functional features and geometric features; then set the growth criteria, including the similarity criterion and the change threshold; then start from the seed point and judge the adjacent pixels or regions according to the growth criteria. If the similarity criterion is met, include them in the changed area and perform region growing; during the growing process, merge the adjacent regions that meet the conditions to form a larger changed area; when the outer contour of the changed area changes more than the preset threshold, mark it as the changed area; finally, when no new pixels or regions meet the growth criteria, the region growing process terminates, so as to realize the comparison of the changes in building functional features and geometric features and effectively mark the areas with significant changes in the outer contour.

5. The automatic modeling and updating method for dynamic evolution of urban form based on parameter iteration according to claim 4, characterized in that The edge detection of the satellite image by the Sobel algorithm in step S4 includes: Step S4-1-1 Satellite image preprocessing; perform Gaussian filtering preprocessing on the input satellite image of the changed area, use a 5×5 convolution kernel for noise suppression, and set the standard deviation σ = 1.5 to balance the denoising effect and edge preservation; Step S4-1-2 Gradient calculation; construct the horizontal Sobel operator G x and the vertical Sobel operator G y , and perform two-dimensional convolution operations with the preprocessed image respectively to obtain the horizontal gradient matrix I x and the vertical gradient matrix I y , where the G x and G y are as follows: I x and I y is as follows: Among them, A is the original satellite image; calculate the pixel-level gradient magnitude G and gradient direction θ based on the gradient matrix, and establish a gradient feature field. The G and θ are: Step S4-1-3 Edge determination: Use the non-maximum suppression algorithm to refine the edges along the gradient direction, retain the pixel points with the maximum local gradient, and eliminate the edge blurring effect; set the dynamic double thresholds T low and T high , divide the pixel points into three categories: strong edges, weak edges, and non-edges, where the T low and T high are as follows: T low = 0.05 × G max ,T high = 0.15 × G max Among them, G max is the maximum value of the gradient magnitude G; Pixel classification rule ； Step S4-1-4 Edge connection optimization; perform eight-neighborhood connectivity detection on weak edge pixel points. When it is adjacent to strong edge pixels, retain it as a valid edge, otherwise remove it; output a binary edge feature map, and record the gradient direction matrix for subsequent building corner detection at the same time. The corner determination criterion is that the gradient direction has a direction mutation of more than 45° within a 3×3 window.

6. The automatic modeling and updating method for dynamic evolution of urban form based on parameter iteration according to claim 5, characterized in that The region segmentation based on the image topology by using the watershed algorithm in step S4 includes: Step S4-2-1 Gradient magnitude normalization and image enhancement; normalize the gradient magnitude matrix G output by the Sobel edge detection to the [0, 255] interval to generate the input gradient image for the watershed algorithm. The specific method is: Among them, G norm (i,j) is the value of the normalized gradient magnitude matrix at position (i,j), G(i,j) is the value of the original gradient magnitude matrix at position (i,j), G min and G max are the global minimum and maximum values of the gradient magnitude respectively; perform histogram equalization on the normalized image to enhance the edge features of the low-contrast regions; Step S4-2-2 Distance field map generation and seed marker extraction; Binarize the gradient image with a threshold T binary = 0.1×G max , generate a building base mask, obtain the Euclidean distance transform map D(x, y), which represents the geometric distance from each pixel to the nearest background; Search for local maximum points in the distance field map as initial seeds, satisfying: D(x,y)>D(x±1,y±1) and D(x,y)≥5 pixels The distance between the marked points needs to be greater than the minimum building interval to avoid over-dense seeds; where D(x±1,y±1) represents the distance value of the pixel points in the 8-neighborhood of the pixel point. Step S4-2-3 Seed Marker Optimization; Perform morphological opening operation on the seed marker image using a 3×3 circular structural element to eliminate isolated noise markers with an area less than 10 pixels, and retain candidate building cores with a connected area S≥20 pixels 2 of the candidate building core; Step S4-2-4 Watershed segmentation; overlay the optimized seed-labeled image with the normalized gradient image to generate the input image I with label constraints marked , where the seed region is forced to be set to the minimum gray value of 0; initialize the water level to 0, traverse the pixels in ascending order of gray value, take the seed region as the initial catchment basin, when the water level rises to the watershed line, record the barrier position, and the flooding termination condition is that all basin boundaries are extended until all basins complete boundary division; Step S4-2-5 Region merging optimization; merge the over-segmented regions, and the following conditions need to be met simultaneously: (a) Mean of gradient magnitudes of watershed lines between adjacent regions Satisfy: (b) Area ratio: For the region pairs that meet the conditions, delete the watershed barrier and update the region labels, and output the segmentation result map with topological labels. Each independent label region corresponds to a single building entity, and record the region adjacency relationship matrix for subsequent contour tracking; among them, min(S1,S2) and max(S1,S2) respectively represent the smaller value and the larger value of the areas of the two regions.

7. The automatic modeling and updating method for dynamic evolution of urban form based on parameter iteration according to claim 6, characterized in that The construction of the building geometric feature description index system in step S5 and the assignment of weights to the indexes by using the entropy weight method include: Geometric feature description index system

Images