A roof landscape evaluation method based on city model view content traversal recognition

Abstract

The application discloses a kind of based on city model view content traversal identification's roof landscape evaluation method, belong to city planning technical field.The present application is in view of prior art deficiency, by in the city three-dimensional model batch export the view image of the city space of the view to be evaluated, and in considering export view, fully consider that city model can be traversed and positioned roof space viewpoint and look direction;Then by digging the label color data information in view, and then objectively measure the city each view landscape index under the view of this direction to reflect the visual features of roof view.The present application can quantitatively evaluate the look condition of each landscape element in roof view, provide scientific basis for city planning design.

Description

Technical Field

[0001] This invention relates to the field of urban planning technology, and in particular to a rooftop landscape evaluation method based on urban model visual content traversal recognition. Background Technology

[0002] In densely populated urban areas, people's perception of urban space is increasingly becoming a focus of attention. Rooftops, as a unique spatial resource and a novel perspective for alleviating visual congestion in the city, are receiving growing attention. Through urban rooftops, citizens can enjoy a broader view of the urban landscape, thereby gaining a deeper understanding of the city's character and enhancing its recognizability and texture, ultimately creating a more beautiful and captivating urban scene. Therefore, deeply exploring the functions of urban rooftops as a new perspective is of crucial value in promoting the understanding of urban landscape and elevating the visual realm of the city.

[0003] Past urban landscape research has primarily focused on streetscapes, often treating rooftops as part of the urban landscape, lacking in-depth analysis from this perspective. In recent years, research on urban rooftop views has gained increasing attention. With expanding urban scale leading to visual congestion and limited urban space, the demand for maximizing the use of rooftop space and extracting unique visual experiences is growing, inevitably giving rise to new research trends in urban rooftop studies. In this new direction, urban rooftops may be viewed as activated visual platforms, becoming an innovative way to rethink and reshape the urban landscape. Researchers are exploring how to transform urban rooftops into viewing areas for citizens, allowing for a fresh perspective on the city, distinct from traditional streetscapes. This new research direction not only challenges the understanding of urban landscape but also injects more creative elements into urban design and planning.

[0004] In general, most current research on urban rooftops views them as a resource for urban landscape design, emphasizing their decorative and beautifying role. This perspective often treats rooftops as a single visual element, limiting its focus to the landscape effect they present. However, research analyzing urban landscape from the perspective of rooftops is relatively lacking. Past urban landscape studies have primarily focused on the visual aspects of streetscapes, concentrating on ground-level visual elements such as streets and building facades, with limited understanding of the visual potential of urban rooftops. Especially in high-density urban areas, due to limited ground space and the pressure of population growth, it is increasingly difficult to find sufficient space for innovative streetscape visuals. As urban landscape research deepens, there is an urgent need to explore and present the unique urban landscape from new perspectives. Against this backdrop, the potential of urban rooftops is gradually being recognized; they represent a higher vantage point, elevating the field of vision and providing a new urban landscape different from traditional streetscapes.

[0005] Existing urban visual assessment methods and their main problems include:

[0006] (1) Analytic Hierarchy Process (AHP) Evaluation Method. This method primarily involves relevant experts and technical personnel using their understanding of urban landscapes, combined with their professional experience, to comprehensively consider factors such as naturalness, safety, aesthetics, and accessibility to conduct a multi-level systematic evaluation of urban landscapes. The key to this method lies in the evaluation system itself; however, the design of each level and indicator within the system often relies heavily on subjective cognition and experience, making it difficult to develop a universally accepted and reasonable evaluation system. Furthermore, because this AHP system often includes both subjective and objective indicators, each designed AHP system is difficult to apply elsewhere due to the different objectives of each project.

[0007] (2) Scenery Evaluation Method. This method involves collecting environmental photographs in a specific spatial environment and having test subjects from different backgrounds evaluate the spatial environment depicted in the photographs. The photographs are then divided into several pixel grids, and the number of pixel grids for each label element is counted and used to build a model with the test subject scores to analyze their correlation. This method involves extracting the proportion of content reflected in the scene and combines subjective evaluation factors with objective content proportion indicators. However, this method requires manual collection of images of the specific spatial scene to be evaluated, making it difficult to evaluate a large number of urban spatial scenes of a certain type. Furthermore, the evaluation space is essentially a subjective evaluation by the test subjects, and the background and aesthetic perception of the test subjects cannot be objectively defined and used to select them. Therefore, its reliability is often limited by the selection of test subjects.

[0008] (3) Map Street View Evaluation Method. This method is mainly based on street view map crawling and image semantic segmentation technology. With the popularization of online digital sharing platforms and the rapid development of artificial intelligence technology, image analysis technology represented by convolutional neural network semantic segmentation has brought opportunities for big data analysis and evaluation of urban scene images. Using the labeled elements after semantic segmentation as factors to evaluate urban road street views is objective, but this method can only be used to evaluate urban road street views. Road street view images are essentially road driving views, so it is difficult to use this method to reflect the urban space that citizens can feel and experience when not driving.

[0009] In summary, existing methods for evaluating urban spatial visual appearance have significant shortcomings, and there is an urgent need to provide a more scientific, faster, and more accurate evaluation index method for urban spatial visual appearance. Summary of the Invention

[0010] The technical problem this invention aims to solve is to overcome the shortcomings of existing technologies and provide a rooftop landscape evaluation method based on urban model visual content traversal and recognition. Using urban 3D models and GIS data as a foundation, it batch-exports visual images of urban rooftops, enabling rapid and accurate acquisition of relevant landscape data. By calculating the proportion of various landscape element labels in the visual images, it provides a scientific basis for urban planning and design.

[0011] The present invention specifically adopts the following technical solutions to solve the above-mentioned technical problems:

[0012] A rooftop landscape evaluation method based on urban model visual content traversal recognition includes the following steps:

[0013] Step 1: Identify the urban rooftop space to be evaluated and locate it as urban rooftop viewpoint A in the urban 3D model. i ;

[0014] Step 2: Assign landscape visual labels to the city area and surrounding areas in the city 3D model described in Step 1 using the scope and label information of GIS (Geographic Information System).

[0015] Step 3: Determine the number of directions n in a 360° panoramic view, and obtain the urban rooftop viewpoint A by taking a direction vector every 360° / n. i Each view direction vector For from A i Given the view direction vector of point j in the j-th direction, j≤n, derive the viewpoint A of all city rooftops. i Each view direction vector The city view image seen;

[0016] Step 4: Read and calculate the vectors of each direction. The proportion data of various landscape labels in the following visual image;

[0017] Step 5: Based on the proportion data of each landscape view label from Step 4, analyze the urban rooftop viewpoint A. i To conduct evaluation and screening.

[0018] Furthermore, step 3 exports all city rooftop viewpoints (A). i Each view direction vector The city view image seen includes the following sub-steps:

[0019] Step 3-1: Set the view frame;

[0020] Step 3-2: Set the camera focal length for the view. The focal length is determined using the following formula:

[0021]

[0022] Where θ is the lateral field of view, which is determined by 2π / n; L is the length of the view frame; f is the focal length; and B is the width of the view frame.

[0023] Step 3-3: Viewpoint A from the city rooftop i Camera placement point A in the city's 3D model i Then, taking that as the starting point, vectors are directed towards the corresponding viewpoint directions. Point A after moving any distance i "The target point in the view is located at camera placement point A." i 'and visual target point A i Export the rooftop view image.

[0024] Furthermore, in step 3-3, the viewpoint A is taken from the city rooftop. i Point A is translated along each directional vector to the edge of the rooftop and then moved vertically upwards by 1.6m. i "' represents the camera placement point in the 3D model, and the camera placement point A is used for navigation." i "' and visual target point A i Export the rooftop view image.

[0025] Furthermore, step 4 specifically includes:

[0026] Step 4-1: Calculate the view direction vector The percentage of each landscape view label in the lower view image relative to the total image size, including:

[0027] View direction vector Given a B×L pixel frame view image v ij={p ab |1≤a≤B,1≤b≤L}, and define a finite set of landscape and visual label F; the set of landscape and visual label F is defined according to specific needs, for example, F={“sky”, “Pearl River”, “historical district”, “arcade street facade”, “modern high-rise”, “historical landmark”, “Baiyun Mountain”, “water and green landscape”}.

[0028] The rooftop numbered i has a view direction vector The lower f-scenery type factor RVF ijf The ratio is:

[0029]

[0030] Where λ(p) = f is the landscape view label of pixel p, and || is the radix operator representing the total number of pixels, and RVF ijf ∈[0,1],f∈F,1≤j≤n,1≤i≤K,K is the number of rooftops;

[0031] Step 4-2: View point A from the city rooftop i Sum the landscape view labels from n directions:

[0032]

[0033] Step 4-3: Remove the data with a value of 100% to obtain the RVI of each label. if :

[0034] Z+ represents a positive integer, where Std() represents the value of the exponent after standard normalization.

[0035] Furthermore, in step 5, based on the aforementioned landscape visual labels (RVI)... if Viewpoint A from the city rooftop i The evaluation and screening process involves using the Delphi method to assign weights to each data point, or evaluating viewpoint A using only one data point. i .

[0036] Furthermore, using the indicator RVI if The normalized value is used as a view factor to measure and evaluate the rooftop view. The normalization formula is as follows:

[0037]

[0038] Where M is the normalized value, and X is the index to be normalized. min X is the minimum value among this set of indicators. max This represents the maximum value of this set of indicators.

[0039] The above technical solution can obtain an objective evaluation of the appearance index of the urban space by determining the urban rooftop space and exporting the visual images of the urban rooftop space from various directions in batches.

[0040] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0041] This invention utilizes a three-dimensional urban model and GIS data to export visual images and obtain landscape index data through large-scale batch processing. It then evaluates the visual views from urban rooftops based on visual line segments. This approach better aligns with the objectivity of landscape evaluation and reflects the rationality of an objective, human-centered perspective.

[0042] When considering data mining of visual images in a 3D model, this invention not only involves determining the viewpoints of urban rooftops in the 3D model and exporting visual images in batches, but also involves determining the image frame, focal length, and direction of the visual image, resulting in visual images that are more adaptable to changes in requirements. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the visual content recognition and evaluation method of the present invention in a specific implementation.

[0044] Figure 2 The above describes a visual model of the historical urban area of ​​Guangzhou obtained using the method of this invention, along with some exported visual images.

[0045] Figure 3 This document presents schematic diagrams of the view images from different directions, using a building as an example, and the exported view images from eight rooftops.

[0046] Figure 4 The method of this invention is used to obtain the urban rooftop view RVI of Guangzhou's historical urban area. sky Result: 3D plot. Detailed Implementation

[0047] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings:

[0048] This invention addresses the shortcomings of existing technologies by batch exporting visual images of urban rooftop spaces to be evaluated from a 3D urban model. During image export, it not only fully considers the ability of the 3D urban model to freely locate the viewpoint and viewing direction of the space to be evaluated, but also addresses the determination of the image frame, focal length, and direction of the visual image. Then, by extracting the exported visual images and identifying and calculating various indices, the visual characteristics of the urban rooftops under that direction are measured to evaluate and select the desired urban rooftops.

[0049] Step 1: Define and extract the urban rooftop viewpoint A from the 3D model to obtain the urban rooftop viewpoint A. iThe urban rooftop space is extracted from the attributes of the 3D model itself, such as area and height, as well as the urban attribute data imported into the 3D model by GIS.

[0050] Step 2: Determine the export view direction Define the number of directions n in a 360° panoramic view, and obtain the corresponding landing point A for each rooftop by taking a direction vector every 360° / n. i A set of n view direction vectors ( For from A i The view direction vector of point j in the j-th direction, j≤n);

[0051] Step 3: Batch export rooftop viewpoints A for each city i Each view direction vector The specific visual image is as follows: Step 3-1, set the view frame, that is, the view height and width in pixels of a single view direction;

[0052] Step 3-2: Set the camera focal length for the view. When the frame is fixed, the focal length setting is related to the horizontal field of view. The focal length can be calculated using the formula for focal length and field of view.

[0053]

[0054] Where θ is the horizontal field of view, and in this setting, θ is determined by 2π / n, L is the length of the view frame, B is the width of the view frame, and f is the focal length;

[0055] Step 3-3: Viewpoint A from the city rooftop i Point A after moving vertically upwards by 1.6m i 'This is the camera placement point in the 3D model, and from that point, vectors are generated towards the corresponding view directions.' Point A after moving any distance i "As the target point of the view, camera point A is placed..." i 'and target point A i "This will allow you to export the rooftop view image;

[0056] Step 4: Calculate the Relative Vision Index (RVI) for each landscape feature, as detailed below:

[0057]

[0058] Where Std() represents the value of the exponent after standard normalization; RVI if For RVF if_sum The visual index obtained by removing data with a value of 100% (a label of 100% indicates complete occlusion and is meaningless) is the RVF (Vision Value). if_sum The specific calculation is as follows:

[0059]

[0060] Among them, RVF ijf For a rooftop numbered i, in the direction The factor for the landscape view type is calculated as follows:

[0061]

[0062] Where λ(p) = f is the landscape view label of pixel p, such as "sky" or "historical landmark", and || is the radix operator representing the total number of pixels. And VF ijf ∈[0,1],f∈F,1≤j≤n,1≤i≤K. v ij ={p ab |1≤a≤B,1≤b≤L}.

[0063] Step 5: If a comprehensive score is required or the rooftop is selected based on actual needs, the above-mentioned view indicators need to be weighted according to practical requirements to obtain the final weighted score. If only objective indicators are required for evaluation, the above-mentioned indicators or their normalized values ​​can be used directly as hierarchical factors to measure and evaluate the rooftop view. The normalization formula is as follows:

[0064]

[0065] Where M is the normalized value, and X is the index to be normalized. min X is the minimum value among this set of indicators. max This represents the maximum value of this set of indicators.

[0066] The technical solution of the present invention will be described in detail below with reference to a specific embodiment.

[0067] The method of the present invention includes the following steps:

[0068] Step 1: Generate a 3D urban massing model of Guangzhou's historic city district from GIS data, including building plans, building foundation heights, building heights, and terrain elevations. Then, assign landscape visual labels to the model according to the ten categories of labels F = {"Sky", "Pearl River", "Historic Blocks", "Arcade Street Facades", "Modern High-Rise Buildings", "Historic Landmarks", "Baiyun Mountain", and "Water and Green Landscape"}. Figure 2 (As shown); This visual model serves as the basis for subsequent urban rooftop visual images. Unlike traditional street view images that reflect non-building three-dimensional morphological information such as street trees in the city, the visual images derived from it can reflect more the visual landscape phenomena caused by the three-dimensional building shapes themselves in the city. In comparison, it can more accurately evaluate and analyze the research and design of urban three-dimensional morphology.

[0069] Step 2: Select 26,571 rooftop spaces from all buildings in the historic city area of ​​Guangzhou to be evaluated, and locate them as urban rooftop viewpoints A in the city's 3D model. i Among them, the 26,571 urban rooftops selected were all located within the historical city area with an area greater than 20 square meters. 2 The rooftop of the building has a height greater than 8m; the viewpoint of the rooftop space is the geometric center point of the building's rooftop, but the subsequent placement of the view camera is not based on this geometric center point, but will be determined by the subsequent direction vector to point A. i The camera is moved along each directional vector to the edge of the rooftop and raised to the viewpoint of the viewer as the camera placement point.

[0070] Step 3: Determine the exported view direction vector We define 8 directions for a 360° panoramic view, and obtain a corresponding spatial landing point A by taking a direction vector every 45°. i A set of 8 view direction vectors ( For from A i The view direction vector of the j-th direction of the point (j≤8);

[0071] Step 4: Batch export rooftop viewpoints A for each city i directional vectors The visual image is as follows:

[0072] Step 4-1: Set the view frame to 1083*1083 (1:1 aspect ratio for a single direction). The frame size is determined to better reproduce the human field of vision. Together with the focal length and other factors, it determines the viewing range. When observing from a rooftop, the vertical viewing range of the human eye also needs to be considered. Therefore, with a fixed focal length, a smaller frame width will result in a larger vertical field of vision. The vertical field of vision of the human eye is usually 50° upward and 70° downward. However, when observing urban scenes from a rooftop, the human observation behavior is often dynamic, with the head tilting up and down. Therefore, this factor also needs to be considered when determining the viewing range.

[0073] Step 4-2: Set the camera focal length for the view. When the frame is fixed, the focal length setting is related to the horizontal field of view. The focal length can be calculated using the formula for focal length and field of view.

[0074]

[0075] In the formula, θ is the horizontal field of view. In this setting, θ is determined by 2π / n, L is the diagonal length of the view frame, and f is the focal length. According to the formula, the focal length of the required horizontal 45° field of view is approximately 50mm.

[0076] Step 4-3: Viewpoint A from the city rooftop i Vectors in each viewing direction Point A is moved horizontally to the edge of the rooftop and then vertically upwards by 1.6m. ij Let this be the camera placement point in the 3D model, and then, using this as the starting point, vectors are drawn towards the corresponding view directions. Point A after moving any distance ij 'For the target point in the view, place the camera points A by traversing the view.' ij and target point A ij The visual images can be exported in batches; a total of 26,571 rooftops and 212,568 visual images were obtained; the point of view is moved to the edge of the rooftop and raised to a height of 1.6m to avoid the view being blocked by the rooftop surface at the center point of the rooftop, and it is also more in line with the mode of people observing the urban space from the rooftop.

[0077] Step 5: Calculate the direction vector The proportion of each landscape view label in the following view image is calculated, and the RVI (Rooftop View Index) of each landscape label is obtained as follows:

[0078]

[0079] Where Std() represents the value of the exponent after standard normalization; RVI if For RVF if_sum The visual index obtained by removing data with a value of 100% (a label of 100% indicates complete occlusion and is meaningless) is the RVF (Vision Value). if_sum The specific calculation is as follows:

[0080]

[0081] Among them, RVF ijf For a rooftop numbered i, in the direction The factor for the landscape view type is calculated as follows:

[0082]

[0083] Where λ(p) = f is the landscape view label of pixel p, such as "sky" or "historical landmark", and || is the radix operator representing the total number of pixels. And RVF ijf ∈[0,1],f∈F,1≤j≤n,1≤i≤K. v ij ={p ab |1≤a≤B,1≤b≤L};

[0084] Step 6: This example uses the above ten RVI indices and three hierarchical objective indicators to evaluate the rooftop views in Guangzhou's historical urban area. The above indicators or their normalized values ​​are used as factors to measure and evaluate the viewing level and hierarchy of various aspects of the rooftop views.

[0085] The entire algorithm described above is as follows: Figure 1 As shown.

[0086] To verify the effectiveness of the technical solution of this invention, a verification experiment was conducted. The experiment first assigned visual style tags (such as...) to the 3D model of the historical city of Guangzhou and its surrounding area. Figure 2 The study comprehensively examined the rooftops of buildings within a 25-square-kilometer area of ​​Guangzhou's historic city. The initial screening phase focused on all buildings within the historic city with an area greater than 20 square meters. 2 Furthermore, the rooftops of buildings with a height greater than 8 meters were used to conduct a comprehensive assessment of these rooftops of evaluation value. The pattern of citizens viewing the city from urban rooftops was reconstructed by defining the view frame, focal length, camera placement point, and target point (e.g., Figure 3 Then, by batch exporting over 210,000 rooftop viewpoints from various directions across the city, rich and detailed rooftop view data for Guangzhou's historical urban area was obtained. Further data mining and processing of these images yielded an assessment result of the city's spatial visual landscape index. Verification and comparative analysis were conducted using on-site survey data and online survey images of rooftops in Guangzhou's historical urban area, ultimately resulting in a relatively reliable and objective assessment of the city's rooftops. Based on this, the aforementioned index can be used to filter and extract assessment results for specific needs.

[0087] The final experimental results are as follows Figure 4 As shown in the figure, this method provides objective factors for the visual assessment of urban space, including appearance index and hierarchy, and realizes the connection between three-dimensional urban model information and visual landscape evaluation, rather than a subjective questionnaire-based evaluation with predetermined survey subjects.

Claims

1. A rooftop landscape evaluation method based on urban model visual content traversal recognition, characterized in that, Includes the following steps: Step 1: Identify the urban rooftop space to be evaluated and locate it as urban rooftop viewpoint A in the urban 3D model. i ; Step 2: Assign landscape visual labels to the city area and surrounding areas in the city 3D model described in Step 1 using GIS scope and label information; Step 3: Determine the number of directions n in a 360° panoramic view, and obtain the urban rooftop viewpoint A by taking a direction vector every 360° / n. i Each view direction vector , For from A i Given the view direction vector of point j in the j-th direction, j≤n, derive all city rooftop viewpoints A. i Each view direction vector The city view image seen; Step 4: Read and calculate the vectors of each view direction. The proportion data of various landscape labels in the following visual image; Step 5: Based on the proportion data of each landscape view label from Step 4, analyze the urban rooftop viewpoint A. i Conduct evaluation and screening; Furthermore, step 4 specifically includes: Step 4-1: Calculate the view direction vector The percentage of each landscape view label in the lower view image relative to the total image size, including: View direction vector Given a B×L pixel frame view image And define a finite set of landscape view labels F; The rooftop numbered i has a view direction vector Down Factors of landscape and visual type The ratio is: ； Where λ(p)=f is the landscape view label of pixel p, and This is the radix operator representing the total number of pixels, and , The number of rooftops; Step 4-2: View point A from the city rooftop i Sum the landscape view labels from n directions: ； Step 4-3: Remove the data with a value of 100% to obtain the values ​​of each label. : ； in, This represents the value after standard normalization.

2. The rooftop landscape evaluation method based on urban model visual content traversal recognition according to claim 1, characterized in that, In step 3, export all city rooftop viewpoints (A). i Each view direction vector The city view image seen includes the following sub-steps: Step 3-1: Set the view frame; Step 3-2: Set the camera focal length for the view. The focal length is determined using the following formula: ； Where θ is the lateral field of view, which is determined by 2π / n; L is the length of the view frame; f is the focal length; and B is the width of the view frame. Step 3-3: Viewpoint A from the city rooftop i Camera placement point A in the city's 3D model i ’ Then, taking that as the starting point, vectors are directed towards the corresponding viewpoint directions. Point A after moving any distance i ’’ As the target point in the view, through camera placement point A i ’ and visual target point A i ’’ Export the rooftop view image.

3. The rooftop landscape evaluation method based on urban model visual content traversal recognition according to claim 2, characterized in that, In step 3-3, viewpoint A is taken from the city rooftop. i Point A is translated along each directional vector to the edge of the rooftop and then moved vertically upwards by 1.6m. i ’’’ Let A be the camera placement point in the 3D model. i ’’’ and visual target point A i ’’ Export the rooftop view image.

4. The rooftop landscape evaluation method based on urban model visual content traversal recognition according to claim 1, characterized in that, In step 5, based on the above-mentioned landscape visual labels... Viewpoint A from the city rooftop i Evaluation and screening were conducted, with the Delphi method used to assign weights to each data point, or a single data point was used to evaluate the urban rooftop viewpoint A. i .

5. The rooftop landscape evaluation method based on urban model visual content traversal recognition according to claim 4, characterized in that, Indicators The normalized value is used as a view factor to measure and evaluate the rooftop view. The normalization formula is as follows: ； Where M is the normalized value, and X is the index to be normalized. The minimum value in the index. This represents the maximum value in the indicator.

Images