Method for constructing non-structural network by simulating and extracting building contour in industrial park

Abstract

The invention relates to the technical field of three-dimensional building contour extraction, and discloses a method for constructing a non-structural network by simulating and extracting a building contour in an industrial park, so that efficient extraction of the building contour and output of a high-quality grid are realized, and reliable data support is provided for subsequent numerical simulation. The problems that in the prior art, building contour extraction precision is insufficient, three-dimensional model construction is complex, and grid generation quality is not high are effectively solved. Meanwhile, by integrating multiple advanced technologies, an integrated process from image processing to unstructured grid generation is constructed, and compared with a traditional fixed grid, the urban terrain and building details can be accurately reflected, higher flexibility is achieved, waste of computing resources can be effectively reduced, and the method is suitable for popularization and application. The precision and efficiency of numerical simulation are remarkably improved, and solid technical support is provided for fine analysis of pollution conditions in a complex environment of an industrial park.

Description

Technical Field

[0001] The present invention relates to the technical field of extracting three-dimensional building outlines. More specifically, it relates to a method for simulating and extracting building outlines for industrial parks and constructing an unstructured network. Background Art

[0002] With the rapid development of industrialization and urbanization, the number and scale of industrial parks are increasing day by day. While generating economic benefits, it also brings a series of environmental problems such as waste gas odor pollution and excessive particulate matter emissions. How to accurately simulate and evaluate the pollution situation in industrial parks has become a hot issue in current air pollution supervision. However, the complex building layout, irregular terrain features, diverse emission source types, and variable meteorological conditions in industrial parks make it difficult for traditional numerical simulation methods to meet the high-precision prediction requirements.

[0003] In numerical simulation, accurately extracting building outlines and constructing an unstructured grid are the keys to improving simulation accuracy. When the wind field flows through the boundaries of buildings and complex terrains, its flow pattern will show significant changes, such as flow around, wake flow, and vortex phenomena. Capturing these microscopic details requires high-resolution grid support. Traditional fixed grids are difficult to effectively adapt to complex geometric boundaries, resulting in uneven grid density distribution during the simulation process, a decrease in computational efficiency, and even problems such as simulation result deviations. In contrast, unstructured grids, due to their good adaptability and flexibility, can accurately capture the detailed characteristics of complex boundaries and provide strong support for the refined simulation of industrial parks.

[0004] As the basis for constructing an unstructured grid, the accurate extraction of building outlines directly determines the quality of the grid and the accuracy of numerical simulation. Currently, the methods for extracting building outlines are mainly divided into two categories: manual and automated. Manual methods rely on personnel to visually analyze image features and manually plot outlines. Although the accuracy is relatively high, the efficiency is extremely low and it depends on the experience of interpreters, making it difficult to meet the requirements for rapid processing of large-scale and high-resolution images in industrial parks. Automated methods use algorithms to analyze images to extract building outlines. Common techniques include edge detection and machine learning. Edge detection methods are simple to calculate and have high efficiency, but are easily affected by image noise, resulting in insufficient extraction accuracy and regularity; although machine learning techniques have strong feature expression capabilities, their dependence on labeled data and high computing power makes them costly. In addition, the diversity of building forms and the complexity of image backgrounds (such as interference factors like vegetation, shadows, and roads) further increase the technical difficulty of automated extraction.

[0005] In summary, the existing methods have two main problems: First, the accuracy and efficiency of building contour extraction are insufficient. It mainly relies on manual interpretation or automated algorithms. The former has low efficiency and is difficult to meet the processing requirements of large-scale high-resolution images. The latter is easily affected by image noise and background complexity, resulting in insufficient accuracy and high costs. Second, the adaptability and optimization ability of traditional grids are limited. They fail to adjust the density according to the complex environmental characteristics of industrial parks, leading to an excessive number of grids, waste of computing resources, and difficulty in accurately capturing boundary details, thus limiting the simulation accuracy. Summary of the Invention

[0006] In order to overcome the above-mentioned defects in the prior art, the present invention provides a method for simulating and extracting building contours in industrial parks and constructing an unstructured network, which realizes high-precision building contour extraction. It integrates technologies such as high-resolution image processing, sample separability analysis, supervised classification, and post-processing optimization, and proposes effective solutions to the problems of insufficient building contour extraction accuracy, low data processing efficiency, high costs, and complex operations in the prior art. It realizes the efficient extraction of building contours and the comprehensive integration of three-dimensional space information, providing accurate and reliable data support for applications in related fields.

[0007] The above technical objectives of the present invention are achieved through the following technical solutions: A method for simulating and extracting building contours in industrial parks and constructing an unstructured network includes the following steps:

[0008] S1. Data preparation: Export high-resolution tif images of the required areas in the industrial park based on the basemap Tianditu - image data block of ArcGIS Pro software, and perform data preprocessing; the data preprocessing includes radiometric calibration and atmospheric correction to eliminate radiation errors and interference caused by sensors, lighting conditions, and atmospheric components.

[0009] S2. Supervised classification:

[0010] (1) Sample selection and evaluation: Select samples according to different categories of ground objects, such as buildings, roads, vegetation, and shadows, etc., set regions of interest respectively, and calculate the separability of training samples. The values of any two parameters are between 0 and 2.0. Among them, a value greater than 1.9 indicates good separability between samples and belongs to qualified samples; a value less than 1.8 requires editing samples or reselecting samples; if the value is less than 1, the two types of samples can be combined into one type of sample.

[0011] (2) Support vector machine classification: Use the support vector machine classification method in ENVI 5.3 software for supervised classification to divide the pixels in the satellite image into different categories based on the training samples.

[0012] (3) Data quality assessment: After classification, use the confusion matrix to evaluate the classification accuracy and reliability. By calculating indicators such as overall accuracy and Kappa coefficient, ensure that the classification results meet the requirements, and convert the final classification results into vector data format;

[0013] S3. Classification result processing:

[0014] (1) Data analysis: Conduct maximum / minimum value analysis on the results, and eliminate or reclassify some patches generated by supervised classification;

[0015] (2) Data simplification: Compare the original image and the results after classification processing, and use the editing function of ArcMap 10.8 software to simplify the classification results, deleting misclassified objects and redundant attributes;

[0016] (3) Graphic normalization: Use ArcMap 10.8 to normalize the geometric shapes of ground features, and correct irregular parts of the boundaries or misclassified areas;

[0017] S4. Data integration and improvement: Merge the vector data shpfile after block processing to form the overall building contour layer of the study area; According to the building parameters of the study area and the information of high-resolution satellite images, add a field representing the building height to the merged layer attributes, and input the building height data according to the corresponding longitude and latitude coordinates to ensure the integrity of the building spatial information and provide input data for subsequent geometric modeling;

[0018] S5. Geometric modeling: Call a pre-written Python program for geometric modeling, and use the fiona, collections, and math libraries for data processing; Among them, fiona is used to read geographic information, and collections and math are used for mathematical calculations; This program is responsible for reading the shpfile to extract building contour and height information, creating building objects, generating complete geometric graphic files, and setting the initial grid of the building to 10m for subsequent construction of unstructured grids;

[0019] S6. Construct unstructured grid: Use GMSH software to read the geometric graphic file, and construct the corresponding unstructured grid according to the position information, element distribution, and element resolution information in the file.

[0020] Furthermore, in step S1, radiometric calibration adopts the absolute calibration method based on the image metadata model to convert the image pixel values into ground object reflectance.

[0021] Furthermore, in step S1, atmospheric correction is achieved through the FLAASH module in ENVI 5.3 to correct the absorption and scattering effects between bands and generate physically consistent and classification-friendly normalized data.

[0022] Further, in step S2, the support vector machine classification classifies the pixels in the satellite image by selecting representative training samples and setting appropriate classification parameters.

[0023] Further, in step S3, data analysis uses a method similar to convolution filtering to classify false pixels in a larger category into that category. Define a kernel window, and replace the category of the central pixel with the category of the dominant pixel in the kernel window, or use the category of the subordinate pixel to replace the category of the central pixel.

[0024] In summary, the present invention has the following beneficial effects: The present invention provides a method for extracting building outlines and constructing unstructured grids for fine-scale simulation of industrial parks, realizing the efficient extraction of building outlines and the output of high-quality grids, and providing reliable data support for subsequent numerical simulations. By integrating technologies such as high-resolution image processing, supervised classification, geometric modeling, and unstructured grid generation, it effectively solves the problems of insufficient accuracy in building outline extraction, complex three-dimensional model construction, and low quality of grid generation in the prior art. Compared with the automated method relying on machine learning, the present invention avoids the dependence on a large amount of labeled data and complex model training, and can directly achieve high-resolution building outline extraction. At the same time, by integrating a variety of advanced technologies, an integrated process from image processing to unstructured grid generation is constructed. Compared with traditional fixed grids, the unstructured grids constructed by the present invention can not only accurately reflect the urban terrain and building details, have stronger flexibility, but also effectively reduce the waste of computing resources, significantly improve the accuracy and efficiency of numerical simulations, and provide solid technical support for the fine-scale analysis of pollution conditions in complex environments of industrial parks. Brief Description of the Drawings

[0025] Figure 1 is a schematic flowchart of a method for extracting building outlines and constructing unstructured grids for fine-scale simulation of industrial parks in an embodiment of the present invention;

[0026] Figure 2 is a schematic flowchart of the process for extracting building outlines in an embodiment of the present invention. In the figure: (a) Satellite image of an industrial park; (b) Satellite image of a partial area; (c) Supervised classification result map of a partial area; (d) Spatial distribution map of buildings in a partial area;

[0027] Figure 3 is a schematic diagram of constructing an unstructured grid in an embodiment of the present invention. In the figure: (a) Satellite image of an industrial park; (b) Distribution map of unstructured grids on the building surface. Detailed Embodiment

[0028] The following is a further detailed description of the present invention in conjunction with the attached Figures 1-3 drawings.

[0029] Example 1: A method for simulating the extraction of building outlines in an industrial park to construct an unstructured network, as Figure 1 described, includes the following steps:

[0030] 1. Obtain high-resolution satellite images: Based on the basemap of ArcGIS Pro software, export the high-resolution tif images of the required area in the industrial park from the image data block of Tianditu.

[0031] 2. Data preprocessing: Data preprocessing includes radiometric calibration and atmospheric correction to eliminate the radiometric errors and interferences caused by sensors, lighting conditions, and atmospheric components. Among them, radiometric calibration adopts the absolute calibration method based on the image metadata model to convert the image pixel values into the surface reflectance; atmospheric correction is achieved through the FLAASH module in ENVI 5.3 to correct the absorption and scattering effects between bands and generate physically consistent and easily classified normalized data.

[0032] 3. Supervised classification:

[0033] 3.1 Set regions of interest according to the categories of ground objects, such as buildings, roads, vegetation, and shadows, etc., and calculate the separability of training samples. The values of any two parameters are both between 0 and 2.0. Among them, if the value is greater than 1.9, it indicates good separability between samples and belongs to qualified samples; if it is less than 1.8, the samples need to be edited or reselected; if the value is less than 1, the two types of samples can be combined into one type of sample;

[0034] 3.2 Use the support vector machine classification method in ENVI 5.3 software for supervised classification. By selecting representative training samples and setting appropriate classification parameters, classify the pixels in the satellite image;

[0035] 3.3 After classification, use the confusion matrix to evaluate the classification accuracy and reliability. By calculating indicators such as overall accuracy and Kappa coefficient, ensure that the classification results meet the requirements;

[0036] 3.4 Convert the final classification result into a vector data format.

[0037] 4. Normalize the graphics:

[0038] 4.1 Conduct maximum / minimum value analysis on the results, and eliminate or reclassify some patches generated by supervised classification. Specifically, adopt a method similar to convolutional filtering to classify the false pixels in the larger category into this category, and then define a kernel window. Replace the category of the central pixel with the category of the dominant pixel in the kernel window, or replace the category of the central pixel with the category of the secondary pixel;

[0039] 4.2 Compare the original images of the industrial park with the results of classification processing, and use ArcMap 10.8 software to refine the classification results, deleting misclassified objects and redundant attributes;

[0040] 4.3 Further optimize the building graphics, and use ArcMap 10.8 software to standardize the geometric shapes of ground features, correcting irregular parts of the boundaries or misclassified areas.

[0041] 5. Data integration and improvement:

[0042] 5.1 Merge the vector data shpfile after block processing to form a building contour layer within the industrial park;

[0043] 5.2 According to the building parameters and high-resolution satellite image information, add a field height representing the building height in the shpfile layer attributes, and input the building height data according to the corresponding longitude and latitude coordinates.

[0044] 6. Geometric modeling:

[0045] 6.1 Call a pre-written Python program and use the fiona, collections, and math libraries for data processing. Among them, fiona is used to read geographic information, and collections and math are used for mathematical calculations;

[0046] 6.2 Extract the building contour and height information from the industrial park shpfile, create building objects, and generate a complete geometric graphics file; and set the initial grid of the building to 10m for use in constructing an unstructured grid in the next step.

[0047] 7. Construct an unstructured grid: Use GMSH software to read the industrial park geometric graphics file generated in the previous step, and construct the corresponding unstructured grid according to the location information, element distribution, and element resolution information in the file. During the grid generation process, by combining the element characteristics and resolution requirements, the grid size is set in layers to improve the simulation efficiency and ensure the accuracy of the simulation results.

[0048] As Figure 2 、 Figure 3 shown, taking an industrial park as an example, Figure 2 (a) is the satellite image of the industrial park, Figure 2 (b) is the satellite image of a part of the industrial park, Figure 2 (c) is the supervised classification result map obtained by the method provided in this application for this area; Figure 2 (d) is the building space distribution map obtained by the method provided in this application for this area. Among them, Figure 2 (b) toFigure 2 (c) is the supervised classification of satellite images of some areas of the industrial park; Figure 2 (c) to Figure 2 (d) is the building space distribution map of some areas of the industrial park extracted according to the results of the supervised classification.

[0049] Figure 3 (a) is the satellite image of a certain area of the industrial park; Figure 3 (b) is the unstructured grid distribution map of the building surface of a certain area of the industrial park obtained by the method provided in this application.

[0050] This specific embodiment is only an explanation of the present invention, and it is not a limitation of the present invention. After reading this specification, those skilled in the art can make modifications that do not contribute creatively to this embodiment as needed, but as long as they are within the scope of the claims of the present invention, they are protected by the patent law.

Claims

1. A method for simulating and extracting building outlines to construct an unstructured network for an industrial park, characterized in that: The following steps are involved: S1. Data preparation: Based on the base map of ArcGIS Pro software, the sky map-image data is divided into blocks to export high-resolution tif images of the required area of ​​the industrial park, and data preprocessing is performed; data preprocessing includes radiation calibration and atmospheric correction to eliminate radiation errors and interference caused by sensors, lighting conditions and atmospheric composition; S2. Supervised classification: (1) Sample selection and evaluation: Samples are selected according to different categories of objects, such as buildings, roads, vegetation, and shadows. Regions of interest are set for each sample and the separability of the training samples is calculated. The values ​​of any two parameters are between 0 and 2.

0. If the value is greater than 1.9, it means that the samples are well separable and are qualified samples. If the value is less than 1.8, the samples need to be edited or reselected. If the value is less than 1, the two types of samples can be combined into one type of sample. (2) Support vector machine classification: The support vector machine classification method in ENVI 5.3 software is used for supervised classification to classify the pixels in the satellite image into different categories based on the training samples. (3) Data quality assessment: After the classification is completed, the confusion matrix is ​​used to evaluate the classification accuracy and reliability. By calculating indicators such as overall accuracy and Kappa coefficient, it is ensured that the classification results meet the requirements, and the final classification results are converted into vector data format; S3. Classification result processing: (1) Data analysis: Perform maximum / minimum analysis on the results and remove or reclassify some spots generated by supervised classification; (2) Data simplification: Compare the original image with the classification results, simplify the classification results using the editing function of ArcMap10.8 software, and delete the incorrectly classified objects and redundant attributes; (3) Graphics normalization: ArcMap 10.8 was used to normalize the geometric shapes of features and correct irregular parts of boundaries or misclassified areas; S4, data integration and improvement: merge the vector data shpfile after block processing to form the overall building outline layer of the study area; according to the building parameters and high-resolution satellite image information of the study area, add a field representing the building height in the merged layer attributes, and input the building height data according to the corresponding latitude and longitude coordinates to ensure the integrity of the building spatial information and provide input data for subsequent geometric modeling; S5, Geometric modeling: Call the pre-written Python program for geometric modeling, and use the fiona, collections, and math libraries for data processing; fiona is used to read geographic information, and collections and math are used for mathematical calculations; the program is responsible for reading the shpfile to extract the building outline and height information, creating building objects, generating complete geometric graphics files, and setting the initial building grid to 10m for subsequent construction of unstructured grids; S6. Construct unstructured grid: Use GMSH software to read the geometry file and construct the corresponding unstructured grid according to the position information, feature distribution and feature resolution information in the file.

2. The method for simulating and extracting building outlines to construct a non-structured network for an industrial park according to claim 1 is characterized in that: In step S1, the radiation calibration uses an absolute calibration method based on the image metadata model to convert the image pixel value into the reflectivity of the ground object.

3. The method for simulating and extracting building outlines to construct a non-structured network for an industrial park according to claim 1 is characterized in that: Atmospheric correction is implemented through the FLAASH module in ENVI 5.3 to correct the absorption and scattering effects between bands and generate normalized data that is physically consistent and easy to classify.

4. The method for simulating and extracting building outlines to construct a non-structured network for an industrial park according to claim 1, characterized in that: In step S2, the support vector machine classification classifies the pixels in the satellite image by selecting representative training samples and setting appropriate classification parameters.

5. The method for simulating and extracting building outlines to construct a non-structured network for an industrial park according to claim 1, characterized in that: In step S3, the data analysis uses a convolution filtering method to classify the false pixels in the larger category into the category, define a kernel window, and replace the category of the central pixel with the category of the pixel that occupies a dominant position in the kernel window, or replace the category of the central pixel with the category of the pixel that occupies a secondary position.

Images