Urban functional area fine classification method based on multi-modal data collaboration

Abstract

The invention relates to the technical field of urban space information processing, in particular to an urban functional area fine classification method based on multi-modal data collaboration. According to the method, POI data, AOI data, land utilization data, remote sensing images and other multi-source heterogeneous data are fused, a multi-dimensional feature system is constructed, and an urban functional area classification model covering a single functional area, a composite functional area, a mixed functional area and a non-functional area is provided. According to the method, firstly, block-level basic units are constructed based on an open street map (OSM), then coordinate transformation and function classification are carried out on POI and AOI data, features such as POI density, AOI area proportion and land utilization structure are extracted respectively, and finally, function type identification is carried out on urban block units according to set classification rules. Compared with the prior art, the method has the advantages that through multi-source information fusion and discrimination logic optimization, the accuracy and applicability of urban functional area recognition are remarkably improved, and the method is suitable for scenes such as urban planning, land utilization evaluation and urban management.

Description

Technical Field

[0001] This invention belongs to the field of geospatial data application technology, specifically a method for fine-grained classification of urban functional zones based on multimodal data collaboration. By integrating point-of-interest (POI) data, area-of-interest (AOI) data, land use data, and high-resolution remote sensing imagery, a classification system for single-function zones, composite-function zones, and non-function zones is constructed, enabling precise analysis of urban spatial structure and in-depth exploration of functional semantics. This method can provide scientific decision-making support for urban planning, thermal environment assessment, and carbon sequestration management. Background Art

[0002] As the fundamental unit of urban planning, urban functional zones comprehensively reflect the multiple characteristics of urban spatial form and human activities, centrally reflecting attributes such as socioeconomic activity, the spatial concentration of resources, and the organization of transportation functions. Accurately distinguishing and identifying different types of functional zones is crucial for a deeper understanding of their internal characteristics and can effectively support research in areas such as urban thermal environment and carbon sequestration.

[0003] Traditional functional area identification methods, which mainly rely on a common single data source approach, have certain limitations. POI data abstracts entities into point data, which can reflect specific location information. However, since it cannot reflect the scale and spatial structure of the entity, it can easily lead to misclassification of functional areas. In contrast, AOI data covers the scale and land use nature of the entity and can accurately divide functional areas based on area proportion, but its distribution range is limited, especially in urban fringe areas, where recognition blind spots are obvious. In addition, POI and AOI data are often missing in undeveloped areas or farmland in cities. It is difficult to obtain accurate recognition results using any one of these data alone. Remote sensing images need to be combined for proofreading and supplementation to further improve the accuracy and comprehensiveness of recognition. Therefore, combining multimodal data such as POI data, AOI data, and remote sensing images for comprehensive identification and verification is a necessary means to improve the accuracy of urban functional area division.

[0004] Currently, the classification of urban functional zones is mostly based on urban land use classification standards, dividing them into single-function zones and mixed-function zones. Single-function zones typically include commercial, residential, and transportation zones, while mixed-function zones are not clearly defined. This classification system is relatively crude and fails to accurately capture the multiple characteristics of complex functional zones. There is an urgent need to develop a more refined functional zone classification system that encompasses multiple levels of functional zones, including single-function zones, composite zones, and mixed zones. This would provide a theoretical basis and technical support for the precise classification and comprehensive analysis of urban functional zones. Summary of the Invention

[0005] The present invention provides a method for fine classification of urban functional areas based on multimodal data collaboration, and the fused data includes point of interest (POI) data, area of ​​interest (AOI) data, land use data and remote sensing images. Based on this multi-source data fusion method, the present invention constructs a set of urban functional area classification systems covering single functional areas, compound functional areas, mixed functional areas and non-functional areas. The method first constructs the basic units of urban blocks based on Open Street Map (OSM) data, and then fuses multi-source heterogeneous data to extract and comprehensively analyze the spatial characteristics of each functional area unit, thereby realizing fine identification of urban functional area types. This method effectively improves the recognition accuracy and processing efficiency of urban functional area division, and overcomes the problems of strong data dependence and low recognition accuracy existing in traditional functional area recognition methods.

[0006] The present invention solves its technical problems by adopting the following technical solutions:

[0007] The method for fine-grained identification of urban functional zones by integrating multi-source data provided by the present invention is as follows: first, the OSM road network is preprocessed to generate basic units of urban functional zones; then, the proportion of land use types in each functional zone unit is identified, and cultivated land and water areas are identified; then, multi-dimensional features are extracted from each functional zone unit, including the frequency density of each POI type and the area proportion of each AOI type; finally, the extracted feature variables are used to identify the type of urban functional zones.

[0008] The present invention can use the following method to generate the basic unit of urban functional area:

[0009] (1) Projection transformation to unify the spatial reference system;

[0010] (2) Road data cleaning to remove redundant line segments and hanging nodes;

[0011] (3) Road merging to form complete block boundaries;

[0012] (4) Convert elements to surfaces to generate closed blocks;

[0013] (5) Compare with high-resolution remote sensing images to ensure spatial accuracy.

[0014] After the processing is completed, the basic unit for functional identification, namely the urban functional area unit, is formed.

[0015] The present invention can use the following methods to extract features of various types of data:

[0016] Land use data feature extraction:

[0017] (1) Based on the land use map data, calculate the ratio of cultivated land to water area in each functional area unit:

[0018] (2) If a certain category accounts for more than 70%, it will be directly identified as a cultivated land or water functional area; otherwise, it will enter the functional area type identification process.

[0019] AOI data feature extraction:

[0020] (1) According to the attribute labels of AOI, it is divided into six categories: residential, public, commercial, industrial, road, and green space.

[0021] (2) Calculate the AOI area ratio of each research unit by the area ratio method, that is, the proportion of AOI data of each functional area unit. The specific calculation formula is as follows:

[0022]

[0023] In the formula, i is the AOI type, j is the urban functional area, is the area of ​​the i-th type AOI, is the area of ​​the jth functional area, and T is the type of functional area.

[0024] (3) Record the AOI types corresponding to the maximum and second largest area ratios.

[0025] POI data feature extraction:

[0026] (1) According to the attribute labels of POIs, they are divided into six categories: residential, public, commercial, industrial, road, and green space.

[0027] (2) The POI density of each research unit is calculated by the quadrat density method, that is, the number of POIs per unit area. The specific calculation formula is as follows:

[0028]

[0029] Where i is the POI type, t is the total number of types, and n i is the number of POIs of type i in the unit, S i is the unit area, F i is the density of POI of category i (per square kilometer).

[0030] (3) The POI types corresponding to the maximum and second largest values ​​of record density.

[0031] The present invention can use the following classification rules to identify the type of urban functional area, first identifying the functional area type based on the area ratio of AOI data, and then identifying the functional area type based on the density ratio of POI. The process includes:

[0032] (1) If the maximum value of each AOI area ratio is 0, the judgment is made based on the POI density ratio:

[0033] If the maximum POI density is 0, it is determined to be a non-functional area;

[0034] If the maximum POI density is >50%, it is judged as a single-function area;

[0035] If the maximum value is between 30% and 50%, and the second largest value is also between 30% and 50%, it is judged as a composite functional area;

[0036] If the second largest value is ≤30%, it is judged as a mixed functional area;

[0037] If the maximum value is <30%, it is directly judged as a mixed functional area.

[0038] (2) If the maximum area of ​​AOI is greater than 50%, it is judged as a single functional area according to its corresponding functional type;

[0039] (3) If the maximum AOI value is between 30% and 50%, and the second largest value is also between 30% and 50%, it is judged as a composite functional area;

[0040] (4) If the second largest AOI value is ≤30%, it is judged as a mixed functional area.

[0041] When the AOI and POI recognition results are inconsistent, the functional area type is determined based on the AOI result first.

[0042] To verify the effectiveness of the proposed method, a confusion matrix model was used to evaluate the recognition accuracy of urban functional zones (UFZs). The specific method is as follows:

[0043] (1) 20 sample points were selected from each type of single functional area as the validation set;

[0044] (2) Calculate producer accuracy, user accuracy, overall accuracy, and Kappa coefficient;

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] This paper proposes a method for identifying urban functional areas that integrates satellite remote sensing data, POI data, and AOI data. AOI data is prioritized during identification, mitigating the misclassification problem of POI data as point-based data. Furthermore, compared to the traditional functional area classification system, this paper adds the category of composite functional areas in addition to single and mixed functional areas, further refining the classification of urban functional areas and improving classification accuracy and precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a flow chart of the present invention for identifying urban functional areas by combining POI data, AOI data, and land use data;

[0048] Figure 2 This is a POI distribution map used in the implementation case of the present invention;

[0049] Figure 3 This is the AOI distribution map used in the implementation case of the present invention;

[0050] Figure 4 A distribution map of urban functional area types identified by the method of the present invention; DETAILED DESCRIPTION

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0052] This example uses the Nangang, Xiangfang, Daoli, Daowai, Songbei, and Pingfang districts of Harbin, Heilongjiang Province, as the research area. The proposed method for identifying urban functional areas based on multi-source data fusion is used to divide the area into functional zones. The research area encompasses a variety of typical urban functional types. The following is a detailed implementation process.

[0053] (1) Construction of urban block units

[0054] The basic identification units for urban functional zones were derived from OpenStreetMap (OSM) road network data. Vector road data for the Harbin area (.osm.pbf format) was downloaded from the OSM official website (https: / / download.geofabrik.de) and preprocessed using QGIS software (version 3.28).

[0055] First, we imported the road network data into QGIS and used the "Linear Feature Extraction" tool to select primary highways, secondary highways, and tertiary highways to form the road network skeleton. Using the "Projection Conversion" function, we converted the original WGS84 coordinate system to the CGCS2000 national geodetic coordinate system to ensure consistency with subsequent land use data.

[0056] Next, the QGIS plug-in "Polygonize" was used to automatically convert closed road loops into polygons, generating block patches. These patches were then subjected to "geometric repair" and area screening, removing outliers smaller than 2,000 square meters. Ultimately, 364 block patches were retained as the basic analysis units for urban functional zones. Manual spot checks and remote sensing image comparisons confirmed that the patches largely matched actual block boundaries.

[0057] (2) POI data processing

[0058] POI data comes from the AutoNavi Map open platform. Using its open API (http: / / lbs.amap.com / api / webservice / summary / ), we wrote a Python script to extract batch data. Using keyword searches and administrative boundary restrictions, we extracted all POI information for Nangang, Daoli, and Xiangfang Districts. Extracted fields included name, longitude and latitude, category and subcategory, and address.

[0059] The captured POI data is initially in the GCJ-02 coordinate system and requires coordinate conversion. Using the Python pyproj library, convert the coordinates from GCJ-02 to WGS-84 to CGCS2000 in the following order to ensure consistency with the block units.

[0060] POIs were originally categorized into over 20 major groups. These groups have been reclassified into six categories, based on the "Guidelines for the Classification of Land and Sea Use for National Land Survey, Planning, and Use Control (Trial Implementation)": Residential, Public Service, Commercial, Industrial, Road Traffic, and Green Space. Categories such as "Gate," "Entrance and Exit," and "Administrative Signage" that lack clear functional descriptions have been removed.

[0061] The "Kernel Density" tool in ArcGIS Pro (version 3.1) was used to estimate kernel density for each of the six POI categories. The grid size was set to 100 meters and the bandwidth to 300 meters. The bandwidth value was determined through repeated testing to ensure smoothness while preserving spatial variation. The kernel density raster was then averaged across each block using the Zonal Statistics tool to obtain density characteristics for the six POI categories.

[0062] (3) AOI data processing

[0063] AOI data is extracted from the vector layer of Amap. An automated crawler script (Selenium+BeautifulSoup) is used to download the AOI surface data of the corresponding area, including residential areas, business districts, industrial parks, transportation hubs, parks and green spaces.

[0064] In QGIS, we categorized and coded the AOI data into six main groups, consistent with the POIs. We used the "Intersect" tool to overlay the AOIs with the block cells and calculated the area of ​​each AOI type within each cell. We used the "Field Calculator" to calculate the area percentages of the six AOI types and recorded the functional types corresponding to the maximum and second-largest values ​​for subsequent identification.

[0065] (4) Land use data processing

[0066] Land use data is sourced from the Harbin Municipal Natural Resources Bureau. This data is highly accurate and categorized. In ArcGIS, we filtered out cultivated land and water areas based on the patch attribute fields and overlaid them with the block units using the "SpatialJoin" tool.

[0067] Calculate the proportion of cultivated land or water in each block unit. If the proportion of either type exceeds 70%, the unit is directly classified as cultivated land or water, skipping the subsequent analysis process; the remaining areas continue to perform POI / AOI fusion judgment.

[0068] (5) Implementation of urban functional area identification rules

[0069] Functional area recognition is implemented in Python by building a logical judgment model. The judgment logic is as follows:

[0070] First determine the area ratio of AOI data:

[0071] If the maximum proportion is greater than 50%, it is identified as a single functional area;

[0072] If the maximum and second-largest values ​​are both between 30% and 50%, it is identified as a composite functional area;

[0073] If the maximum value is less than 30%, or the second largest value is less than 30%, it is identified as a mixed functional area;

[0074] If the AOI data is missing or all are 0, the POI density ratio is used for judgment:

[0075] If the maximum density is greater than 50%, it is identified as a single functional area;

[0076] If the maximum and second-largest densities are both 30% to 50%, it is identified as a composite functional area;

[0077] If the maximum density is less than 30%, it is identified as a mixed-function area;

[0078] If the density of all six types of POIs is close to 0, it is identified as a non-functional area.

[0079] When the AOI and POI results are inconsistent, the AOI data result is used first. This judgment logic is implemented in Python using an if-elif-else structure and is dynamically updated with the attribute table.

[0080] (6) Classification accuracy verification

[0081] To verify the accuracy of the recognition results, 20 blocks were randomly selected from each type of single-function area. Combining remote sensing image interpretation and field investigation, the real functional types were manually annotated and a confusion matrix model was constructed.

[0082] Using the confusion_matrix and classification_report modules in scikit-learn, we calculated producer accuracy, user accuracy, overall accuracy, and the Kappa coefficient. Verification results showed an overall accuracy of 89.2% and a Kappa coefficient of 0.853, demonstrating the applicability and stability of this method in real-world urban data environments.

Claims

1. A fine classification method of urban functional areas based on multimodal data collaboration, characterized by: The steps include: S1: Obtain the Open Street Map (OSM) road network data of the urban area, perform projection conversion, road cleaning, road merging, and construct urban block units through line-to-surface operations; S2: Obtain POI (points of interest), AOI (areas of interest), land use data and remote sensing image data of the study area; S3: Convert the coordinate system of POI and AOI data, and classify the POI and AOI into six functional categories: residential, public service, commercial, industrial, road traffic, and green space. S4: Use the kernel density estimation method to process the six types of POI data respectively to obtain the POI kernel density raster layer; S5: Spatially overlay the POI points, AOI layer, and POI kernel density map with the block unit to count the number of POIs, POI density, and AOI area ratio of each type in each block; S6: Determine the area ratio of cultivated land or water bodies in the block. If one type exceeds 70%, the block is identified as a cultivated land functional area or a water body functional area. S7: The remaining blocks are functionally identified based on the proportion of AOI area and POI density, and a discrimination rule based on the proportion of dominant functions is constructed to divide the block units into single-function areas, composite-function areas, mixed-function areas, or non-function areas.

2. The method according to claim 1, wherein: The construction of the block unit is based on the main roads, secondary roads and connecting roads in the OSM data, and the "line to surface" operation in the geographic information system (GIS) platform is used to generate closed block patches.

3. The method according to claim 1, wherein: The POI kernel density estimation uses a Gaussian kernel function, a bandwidth of 300 meters, and a grid size of 100 meters.

4. The method according to claim 1, wherein: The AOI data is calculated by functional type and its area ratio in each block is calculated as follows: In the formula, i is the AOI type, j is the urban functional area, is the area of ​​the i-th type AOI, is the area of ​​the jth functional area, and T is the type of functional area.

5. The method according to claim 1, wherein: The POI density is calculated using the quadrat density method. The density calculation formula for the i-th type of POI per unit area is: Where i is the POI type, t is the total number of types, and n i is the number of POIs of type i in the unit, S i is the unit area, F i is the density of POI of category i (per square kilometer).

6. The method according to claim 1, wherein: If the largest AOI area in a block accounts for more than 50%, the block is identified as a single-functional area corresponding to this type; if the largest and second largest AOI areas both account for between 30% and 50%, it is identified as a composite functional area; if both accounts for less than 30%, it is identified as a mixed functional area.

7. The method according to claim 1, wherein: When AOI data is missing or the area ratio is zero, function judgment is performed based on the POI density ratio. The classification rules are as follows: -When the maximum POI density is greater than 50%, it is identified as a single functional area; - The largest and second largest proportions are both between 30% and 50% and are identified as composite functional areas; -When the maximum value is less than 30% or the density of the six categories is close to 0, it is identified as a mixed functional area or a non-functional area.

8. The method according to claim 1, wherein: The accuracy of the urban functional area classification results was verified by constructing a confusion matrix model, and typical sample points were selected to calculate the producer accuracy, user accuracy, overall accuracy and Kappa coefficient.

9. The method according to claim 1, wherein: The block identification results are finally divided into the following categories: cultivated areas, water areas, single-function areas, composite-function areas (such as industrial-commercial composite-function areas), mixed-function areas, and non-function areas.

Images