Functional area classification method, system, medium and equipment based on multi-source data fusion

Abstract

The present invention belongs to the field of urban functional area classification, and in particular relates to a functional area classification method, system, medium and equipment based on multi-source data fusion. Combining the two- and three-dimensional morphological features of the city and POI data, supplemented by road network data, a functional area classification study based on multi-source data fusion is carried out. Based on the differences in spatial landscape structure and socio-economic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities; the two- and three-dimensional morphological features of the city and POI features are fused, different functional area classification schemes are designed, and the impact of different features on the accuracy of functional area classification is compared and analyzed, which significantly improves the urban functional area classification results.

Description

Technical Field

[0001] The present invention belongs to the field of urban functional area classification, and in particular relates to a functional area classification method, system, medium and equipment based on multi-source data fusion. Background Art

[0002] Urban functional zones divide cities into spatially distinct areas with distinct attributes based on relevant standards or functional types. These areas are both independent and interconnected, representing complex and diverse land use entities with similar spatial landscape structures and socioeconomic activities. Different functional zones often have different spatial landscape structures and socioeconomic activities. The classification of urban functional zones yields varying results based on different classification criteria. In current research, modern cities are primarily divided into residential, industrial, commercial, public, and cultural and tourism areas. Traditional functional zone classification methods rely on expert surveys or expert judgment, which are highly subjective, time-consuming, prone to errors, and require significant effort. With the rapid development of information science and technology, big data-based urban points of interest (POIs) describe the spatial and attribute information of geographic entities, greatly enhancing the ability to capture entity locations and thus better reflecting human activity in cities. POI data has attracted widespread attention from researchers, such as using POI data to conduct research on the quantitative identification and visualization of urban functional areas, identifying and classifying commercial centers based on POI data, integrating road network data and POI data to identify and apply urban functional areas, and constructing urban functional area analysis models based on POI and road network data using kernel density analysis methods to improve the classification accuracy of single-functional areas and mixed-functional areas. Although the above functional area identification based on POI data has shown initial success, it is highly dependent on POI data. When the study area is too large, areas that cannot be covered or are insufficiently covered by POI data will result in low accuracy in functional area division. In addition, urban functional area identification based on mobile phone data and trajectory data also has the problem of strong data dependence. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a functional area classification method, system, medium and equipment based on multi-source data fusion.

[0004] The present invention solves the above technical problems with the following technical solutions: A functional area classification method based on multi-source data fusion, comprising:

[0005] Step 1: Obtain object-level land use classification results, road network data, and POI data;

[0006] Step 2: Determine the block range based on the road network data, determine the POI kernel density characteristics within each block range, and determine the two-dimensional morphological characteristics and three-dimensional morphological characteristics corresponding to each category in the object-level land use classification results;

[0007] Step 3: Determine functional area classification schemes based on different feature combinations according to the kernel density features, two-dimensional morphological features, and three-dimensional morphological features of all POIs, and use the functional area classification schemes based on different feature combinations as input features of the classification model to construct the classification model;

[0008] Step 4: classify the study area based on the classification model.

[0009] The beneficial effects of this invention include extracting two- and three-dimensional urban morphological features, combining these features with POI data, and supplementing them with road network data to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0010] On the basis of the above technical solution, the present invention can also be improved as follows.

[0011] Furthermore, the process of obtaining the object-level land use classification results specifically includes:

[0012] Obtain LiDAR point cloud data and aerial image data at the same time through the airborne LiDAR measurement platform;

[0013] Processing the LiDAR point cloud data and the aerial image data using a multi-source data fusion land use classification method to obtain an object-level land use classification result;

[0014] The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

[0015] Furthermore, the process of determining the block range based on the road network data includes:

[0016] Topological errors are repaired so that each block is surrounded by streets to form a complete area.

[0017] Furthermore, the process of determining the two-dimensional morphological features and three-dimensional morphological features corresponding to each category in the object-level land use classification results includes:

[0018] The two-dimensional morphological features and three-dimensional morphological features are determined by combining the basic information corresponding to each category in the object-level land use classification results with the mapping table.

[0019] Furthermore, the construction of the classification model specifically includes:

[0020] Sample data were selected, 70% of the sample data were used as training sample data sets, 30% of the sample data were used as testing sample data sets, and a classification model was constructed by combining KNN, RF, XGBoost and input features.

[0021] Another technical solution of the present invention to solve the above technical problems is as follows: a functional area classification system based on multi-source data fusion, comprising:

[0022] The acquisition module is used to obtain object-level land use classification results, road network data, and POI data;

[0023] The determination module is used to: determine the block range according to the road network data, determine the POI kernel density characteristics within each block range, and determine the two-dimensional morphological characteristics and three-dimensional morphological characteristics corresponding to each category in the object-level land use classification result;

[0024] The construction module is used to: determine different feature combination functional area classification schemes based on all POI kernel density features, two-dimensional morphological features, and three-dimensional morphological features, and use the different feature combination functional area classification schemes as input features of the classification model to construct the classification model;

[0025] The classification module is used to classify the study area based on the classification model.

[0026] The beneficial effects of this invention include extracting two- and three-dimensional urban morphological features, combining these features with POI data, and supplementing them with road network data to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0027] Furthermore, the process of obtaining the object-level land use classification results specifically includes:

[0028] Obtain LiDAR point cloud data and aerial image data at the same time through the airborne LiDAR measurement platform;

[0029] Processing the LiDAR point cloud data and the aerial image data using a multi-source data fusion land use classification method to obtain an object-level land use classification result;

[0030] The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

[0031] Furthermore, the process of determining the block range based on the road network data includes:

[0032] Topological errors are repaired so that each block is surrounded by streets to form a complete area.

[0033] Furthermore, the process of determining the two-dimensional morphological features and three-dimensional morphological features corresponding to each category in the object-level land use classification results includes:

[0034] The two-dimensional morphological features and three-dimensional morphological features are determined by combining the basic information corresponding to each category in the object-level land use classification results with the mapping table.

[0035] Furthermore, the construction of the classification model specifically includes:

[0036] Sample data were selected, 70% of the sample data were used as training sample data sets, 30% of the sample data were used as testing sample data sets, and a classification model was constructed by combining KNN, RF, XGBoost and input features.

[0037] Another technical solution of the present invention to solve the above technical problem is as follows: a storage medium, wherein instructions are stored in the storage medium, and when a computer reads the instructions, the computer executes any of the methods described above.

[0038] The beneficial effects of this invention include extracting two- and three-dimensional urban morphological features, combining these features with POI data, and supplementing them with road network data to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0039] Another technical solution of the present invention to solve the above technical problem is as follows: an electronic device includes the above storage medium and a processor that executes instructions in the above storage medium.

[0040] The beneficial effects of this invention include extracting two- and three-dimensional urban morphological features, combining these features with POI data, and supplementing them with road network data to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of an embodiment of a functional area classification method based on multi-source data fusion according to the present invention is provided;

[0042] Figure 2 A structural framework diagram of an embodiment of a functional area classification system based on multi-source data fusion according to the present invention;

[0043] Figure 3 A schematic diagram of a functional area classification research workflow provided by an embodiment of a functional area classification method based on multi-source data fusion according to the present invention;

[0044] Figure 4 A flowchart of feature extraction of POI data in a research area provided by an embodiment of a functional area classification method based on multi-source data fusion of the present invention;

[0045] Figure 5 A schematic diagram of a classification model construction provided in an embodiment of a functional area classification method based on multi-source data fusion of the present invention. DETAILED DESCRIPTION

[0046] The principles and features of the present invention are described below. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0047] like Figure 1 As shown in FIG, a functional area classification method based on multi-source data fusion includes:

[0048] Step 1: Obtain object-level land use classification results, road network data, and POI data;

[0049] Step 2: Determine the block range based on the road network data, determine the POI kernel density characteristics within each block range, and determine the two-dimensional morphological characteristics and three-dimensional morphological characteristics corresponding to each category in the object-level land use classification results;

[0050] Step 3: Determine functional area classification schemes based on different feature combinations according to the kernel density features, two-dimensional morphological features, and three-dimensional morphological features of all POIs, and use the functional area classification schemes based on different feature combinations as input features of the classification model to construct the classification model;

[0051] Step 4: classify the study area based on the classification model.

[0052] In some possible implementations, two- and three-dimensional urban morphological features are extracted and then combined with POI data, supplemented by road network data, to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0053] like Figure 3 As shown in S1, the process of obtaining object-level land use classification results, road network data and POI data is as follows:

[0054] S11, acquiring airborne LiDAR point cloud data and airborne aerial image data;

[0055] Airborne LiDAR point cloud data and airborne aerial image data are collected through the same flight equipment based on the principles of time synchronization and space synchronization.

[0056] S12, performing multi-source data fusion land use classification processing on the airborne LiDAR point cloud data and the airborne aerial image data to obtain object-level land use classification results. The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

[0057] S2, determining the block range based on the road network data, determining the POI kernel density feature within each block range, and determining the two-dimensional morphological features and three-dimensional morphological features corresponding to each category in the object-level land use classification result are as follows:

[0058] S21, determining the block range based on the road network data, and completing the street information by patching the topological errors, so that the block range that is not a complete closed area due to missing street information becomes a complete closed area;

[0059] The reasonable division of blocks is the basic premise for functional area classification. OSM road network data is used to divide blocks, correct some topological errors, and combine the airborne aerial image data obtained in this study to supplement some remote roads to achieve reasonable division of blocks.

[0060] To fix some topology errors:

[0061] OSM road network data is vector data, consisting of points, lines, and surfaces. Since OSM road network data is downloaded from the internet, it contains some topological errors. These errors primarily include overlapping vector surfaces, gaps between surfaces, intersecting line segments, and dangling line segments. Topological errors are primarily corrected using ArcGIS editing tools combined with manual inspection. Only when OSM vector data is free of topological errors can the road network data be correctly partitioned and further research conducted.

[0062] For the additional part of the road:

[0063] The true color images displayed by the red, green and blue bands contained in the airborne images are used to assist in the block division.

[0064] Remote roads refer to areas not covered by OSM road network data based on the distribution of features shown in airborne aerial images.

[0065] S22, determining the POI kernel density characteristics within each block, determining the two-dimensional morphological characteristics of building land, bare land, cultivated land, grassland, road, woodland or water body within each block, and determining the three-dimensional morphological characteristics of building land, bare land, cultivated land, grassland, road, woodland or water body within each block.

[0066] The process of determining the POI kernel density characteristics within each block is as follows:

[0067] S221, classifying functional areas according to POI data types, where the functional areas include residential areas, commercial areas, industrial areas, public services, and open spaces;

[0068] The POI data that can express human activities, social, economic and other phenomena are spatially transformed, and the kernel density analysis method is used to convert discrete point data into raster data that can be spatially analyzed and expressed.

[0069] Figure 4 The following figure shows a flowchart for extracting POI data features from the study area. First, the POI data needs to be reclassified. The POI data types in the study area mainly include 20 categories: automotive services, car sales, car repair, motorcycle services, catering services, shopping services, lifestyle services, sports and leisure services, healthcare services, accommodation services, scenic spots, commercial residences, government agencies and social groups, science, education and culture services, transportation facilities services, financial and insurance services, companies and enterprises, road ancillary facilities, place names and addresses, and public facilities. It can be seen that this data is mainly distributed in built-up areas. During the reclassification process, these 20 categories of data are reclassified according to functional areas into residential areas, commercial areas, industrial areas, public services, and open spaces. POI data is obtained from the AutoNavi open platform.

[0070] S222, performing kernel density analysis on the POI data of each functional area to obtain POI kernel density features;

[0071] A kernel density analysis was performed on the POI data containing functional area category attributes. After multiple experiments, the search radius was set to 500m, and finally a 1m resolution raster image was generated, which was consistent with the resolution of the land use classification results data. It is also called a POI distribution heat map. According to the range of the block, the mean (mn), standard deviation (std) and sum (sum) of the kernel density of residential areas (Res_**), commercial areas (Comm_**), industrial areas (Ind_**), public services (Ins_**) and open spaces (Open_**) in each block were extracted. Finally, 15 POI kernel density features were obtained.

[0072] Kernel density analysis can effectively express the spatial distribution of POI data and is a commonly used method for expressing POI data. This solution is based on the quartic function of Silverman (1986). The predicted density of the new (x, y) position is determined by the following formula:

[0073]

[0074] Where i = 1, 2, ..., n are input points, and only points in the sum are included if they are within the radius of the (x, y) position; D(x, y) is the density prediction value of the new (x, y) point; r is the search radius; pop i is the weight value of the point (negligible); d i The distance between point i and (x,y). The population (pop) field assigns greater weight to certain features than others and allows a single point to represent multiple observations. Note that this field is generally not used in calculations and is set to 1.

[0075] Using blocks as units, we extract two- and three-dimensional morphological features and extract point of interest (POI) features based on kernel density analysis. Within each block, we extract the two-dimensional morphological characteristics (landscape pattern index) of buildings, bare land, cultivated land, grassland, roads, woodlands, and water bodies, as well as the three-dimensional morphological characteristics of buildings and trees. Finally, these extracted features are used as attribute information for each block to participate in functional area classification research.

[0076] The process of determining the two-dimensional morphological characteristics of built-up land, bare land, cultivated land, grassland, road, forest land or water body within each block is as follows:

[0077] Extract the object-level land use classification results, i.e., the distribution and area of ​​building land, bare land, cultivated land, grassland, roads, woodlands, and water bodies within the street area. This basic information can be found in the parameter descriptions in Table 1. This information serves as the target for 2D morphological feature extraction, i.e., the types of patches in the landscape index. Patches refer to the distribution of each feature category in the object-level land use classification results. At the object level, features are distributed as objects, not pixels. These types include:

[0078] Area indicators: type coverage (Percentage of Landscape, PLAND), edge density (EdgeDensity, ED);

[0079] Shape indices: Area-weighted Mean Shape Index (SHAPE_AM), Area-weighted Mean Fractal Dimension Index (FRAC_AM);

[0080] Aggregation / dispersion indicators: Patch Density (PD), Landscape Shape Index (LSI), Mean Proximity Index (PROX_MN), Euclidean Nearest-Neighbor Mean Distance (ENN_MN), Patch Cohesion Index (COHESION) and Shannon's Diversity Index (SHDI).

[0081] The calculation methods and descriptions of each two-dimensional morphological characteristic index are shown in the following table.

[0082] Table 1 Calculation method and description of two-dimensional morphological features

[0083]

[0084]

[0085]

[0086] The process of determining the three-dimensional morphological characteristics of building land, bare land, cultivated land, grassland, road, forest land or water body within each block is as follows:

[0087] The 3D morphological characteristics of a city primarily refer to the spatial distribution, variations, and relationships of 3D information about buildings and trees. Based on the 2D distribution of 3D features and the characteristics of the elevation model, we obtain 16 3D building morphological characteristics and 12 3D tree morphological characteristics, including 3D elevation, area, shape, and spatial distribution metrics. The specific calculation methods are described below.

[0088] Based on the two-dimensional distribution of buildings in the study area and the elevation information in the building elevation model, 16 different three-dimensional building morphological features are extracted, including:

[0089] Building Mean Height (BMH), Building Max Height (BMaH), Building Height Variance (BHV), Normalized Building height variance (NBHV), Building Height Range (BHR), Building Surface Area (BSA), Building Volume (BV), Ratio of Street Height and Building Width (BHW), Ratio of Street Height and Length (BHL), Percentage of Building Surface Area (PBSA), Percentage of Building Volume (PBV), Building Edge Density in 3D Space (BED), Building Shape coefficient (BSC), Building Landscape Shape Index in 3D Space (BLSI), Frontal Area Index (FAI) The calculation method and description of the Building Sky View Factor (BSVF) are shown in Table 2:

[0090] Table 2 Calculation method and description of three-dimensional building morphological characteristics

[0091]

[0092]

[0093] The extraction of 3D tree morphological parameters is similar to that of building morphological parameters. Based on the 2D distribution of trees and the elevation information provided by the crown elevation model, 12 types of 3D tree morphological characteristics are extracted, including:

[0094] Table 3 lists the definition formulas and parameter explanations of 12 three-dimensional tree morphological characteristics, including tree height variation, three-dimensional surface area and volume, morphological indices, and spatial distribution.

[0095] Table 3 Calculation method and description of three-dimensional building morphological characteristics

[0096]

[0097]

[0098] S3, determining different feature combination functional area classification schemes based on all POI kernel density features, two-dimensional morphological features, and three-dimensional morphological features, and using the different feature combination functional area classification schemes as input features of the classification model to construct the classification model in the following process:

[0099] S31, fusing all POI kernel density features, two-dimensional morphological features, and three-dimensional morphological features to obtain functional area classification schemes with different feature combinations;

[0100] In the design of the urban functional area data fusion scheme, each block is taken as the most basic classification unit, and the corresponding two- and three-dimensional urban morphological features and POI kernel density features within each block are extracted respectively. Based on the above three categories of features, a functional area classification study is conducted on multi-source data fusion. The impact of input features from different data sources (one is the two- and three-dimensional morphological features extracted based on the object-level classification results of airborne LiDAR land use classification, and the other is the POI kernel density features extracted from POI data) on the functional area classification is analyzed and compared. This study designed six different experimental schemes (abbreviated as Exp.#) based on both single-source and multi-source data fusion approaches. First, from a single-source perspective, three comparative experiments were designed to compare and analyze the impact of POI kernel density features, two-dimensional morphological features, and three-dimensional morphological features on functional area classification accuracy. Second, from a multi-source data fusion perspective, three comparative experiments were designed to add two-dimensional morphological features, three-dimensional building morphological features, and three-dimensional tree morphological features to the POI kernel density features. The contributions of these features to improving classification accuracy were compared and analyzed. The results were also compared with the classification results from a single data source, analyzing their respective advantages and disadvantages. Considering the potential for data redundancy in high-dimensional input bands, feature optimization was performed on all input features, and optimized feature combination schemes were extracted. Table 4 shows the functional area classification schemes for different input feature combinations and the corresponding number of input bands (features) for each scheme.

[0101] Table 4 Experimental scheme for functional area classification with different feature combinations

[0102]

[0103] S32, selecting sample data, and constructing a classification model based on the sample data, the classifier, and the input features.

[0104] The process of selecting sample data is:

[0105] The study area primarily consists of residential buildings, factories, hotels, shopping centers, urban villages, and rural settlements. Combining airborne imagery and high-precision Google Maps, the study area was divided into built-up and non-built-up areas based on the social and economic functional characteristics of each block. Built-up areas include residential areas, commercial areas, industrial areas, public services, and open spaces; non-built-up areas include agricultural areas, green spaces, water bodies, and unused areas. Table 5 provides a detailed description of each functional area category in the study area. Furthermore, using existing imagery and map data as reference, we selected sample datasets for each functional area and divided the sample data into 70% training and 30% testing datasets (with no overlap between the training and testing datasets).

[0106] Table 5 Classification and description of functional areas in the study area

[0107]

[0108]

[0109] The classifier selection process is:

[0110] This study selected three machine learning classifiers: KNN, RF, and XGBoost, to conduct a functional area classification study. By analyzing the classification accuracy of the three classifiers, we tested the consistency of the impact of feature addition on classification accuracy. Furthermore, by comparing the classification results of the three classifiers, we selected the classification scheme with the highest classification accuracy and output the classification results for the land use classification of the study area. The land use classification based on the three classifiers was implemented in Python 3.7.

[0111] In practical applications, the confusion matrix (CM) is often used to evaluate the accuracy of remote sensing image classification results. It is a standard format for accuracy assessment, primarily based on comparing the degree of confusion between the classification results and the measured values. It has been widely used in remote sensing image classification research. Using the CM, we can determine the total number of samples for each feature category, as well as the number of misclassified and missed samples, making the classification results more intuitive.

[0112] In this study, accuracy evaluation runs through the process of classifier model optimization and final classification result analysis and evaluation. Figure 5 The flow chart for the classification accuracy evaluation and analysis scheme is presented. First, the sample dataset is randomly divided into a 70% training sample dataset and a 30% validation sample dataset. Second, three classifier algorithms, KNN, RF, and XGBoost, are used to learn and train the training sample dataset and build a training model. Next, the validation sample dataset is used to calculate the accuracy of the model, and the model parameters are optimized by analyzing the classification results. This process is repeated until the optimal training model and the best classification accuracy are obtained. Finally, the final classification results of all classification schemes are evaluated and analyzed. By comparing and analyzing evaluation indicators such as Overall Accuracy (OA), User's Accuracy (UA), Producer's Accuracy (PA), and kappa coefficient, the impact of each input feature and classification scheme on the functional area classification results is evaluated and analyzed.

[0113] Through the accuracy evaluation method of this study, the effects of six different data fusion and optimization classification schemes on the accuracy of functional area classification are comprehensively analyzed to achieve the output of functional area classification results.

[0114] Step 4: classify the study area based on the classification model.

[0115] The input of the classification model is a combination of different features, such as those shown in Table 4, and the output is the classification result of the functional area corresponding to the input parameters, that is, which category it belongs to.

[0116] Human activity is a major driver of changes in the structure and properties of Earth's surface morphology. Through continuous production, transformation, and practice, it influences the two- and three-dimensional spatial form of cities. While the definition of different functional zones reflects, to a certain extent, human activities such as living, consumption, and leisure, there are also spatial structural differences between different functional zones, reflected in the two- and three-dimensional spatial morphology presented by different land cover types. Therefore, relying solely on a single data feature cannot fully describe the differences between functional zones. Compared with existing functional zone classification methods, this method comprehensively considers the spatial two- and three-dimensional morphological differences between different functional zones and information on socioeconomic activities. Taking into account the importance of temporal and spatial synchronization, this method extracts detailed two- and three-dimensional urban morphological features based on high-resolution airborne LiDAR point clouds and aerial imagery data. By integrating these two- and three-dimensional urban morphological features with point-of-interest (POI) features, urban functional zone classification can significantly improve the accuracy of functional zone classification. Understanding the distribution of urban functional zones is of great significance for urban development and optimization of urban patterns.

[0117] Preferably, in any of the above embodiments, the process of obtaining the object-level land use classification result specifically includes:

[0118] Obtain LiDAR point cloud data and aerial image data at the same time through the airborne LiDAR measurement platform;

[0119] Processing the LiDAR point cloud data and the aerial image data using a multi-source data fusion land use classification method to obtain an object-level land use classification result;

[0120] The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

[0121] The airborne LiDAR measurement platform is: The airborne LiDAR system is also called the airborne laser scanning system. It is a high-precision measurement system integrated with multiple sensors installed on manned or unmanned aircraft. The system is mainly composed of hardware integration such as laser scanner, global positioning system (GPS), inertial measurement unit (IMU), control system, high-resolution digital camera, etc.

[0122] Preferably, in any of the above embodiments, the process of determining the block range based on the road network data includes:

[0123] Topological errors are repaired so that each block is surrounded by streets to form a complete area.

[0124] Preferably, in any of the above embodiments, the process of determining the two-dimensional morphological features and the three-dimensional morphological features corresponding to each category in the object-level land use classification result includes:

[0125] The two-dimensional morphological features and three-dimensional morphological features are determined by combining the basic information corresponding to each category in the object-level land use classification results with the mapping table.

[0126] Preferably, in any of the above embodiments, the construction of the classification model specifically includes:

[0127] Sample data were selected, 70% of the sample data were used as training sample data sets, 30% of the sample data were used as testing sample data sets, and a classification model was constructed by combining KNN, RF, XGBoost and input features.

[0128] like Figure 2 As shown, a functional area classification system based on multi-source data fusion includes:

[0129] The acquisition module 100 is used to obtain object-level land use classification results, road network data and POI data;

[0130] The determination module 200 is used to: determine the block range according to the road network data, determine the POI kernel density characteristics within each block range, and determine the two-dimensional morphological characteristics and three-dimensional morphological characteristics corresponding to each category in the object-level land use classification result;

[0131] The construction module 300 is used to determine different feature combination functional area classification schemes based on all POI kernel density features, two-dimensional morphological features, and three-dimensional morphological features, and use the different feature combination functional area classification schemes as input features of the classification model to construct the classification model;

[0132] The classification module 400 is used to classify the research area based on the classification model.

[0133] In some possible implementations, two- and three-dimensional urban morphological features are extracted and then combined with POI data, supplemented by road network data, to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0134] Preferably, in any of the above embodiments, the process of obtaining the object-level land use classification result specifically includes:

[0135] Obtain LiDAR point cloud data and aerial image data at the same time through the airborne LiDAR measurement platform;

[0136] Processing the LiDAR point cloud data and the aerial image data using a multi-source data fusion land use classification method to obtain an object-level land use classification result;

[0137] The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

[0138] Preferably, in any of the above embodiments, the process of determining the block range based on the road network data includes:

[0139] Topological errors are repaired so that each block is surrounded by streets to form a complete area.

[0140] Preferably, in any of the above embodiments, the process of determining the two-dimensional morphological features and the three-dimensional morphological features corresponding to each category in the object-level land use classification result includes:

[0141] The two-dimensional morphological features and three-dimensional morphological features are determined by combining the basic information corresponding to each category in the object-level land use classification results with the mapping table.

[0142] Preferably, in any of the above embodiments, the construction of the classification model specifically includes:

[0143] Sample data were selected, 70% of the sample data were used as training sample data sets, 30% of the sample data were used as testing sample data sets, and a classification model was constructed by combining KNN, RF, XGBoost and input features.

[0144] Another technical solution of the present invention to solve the above technical problem is as follows: a storage medium, wherein instructions are stored in the storage medium, and when a computer reads the instructions, the computer executes any of the methods described above.

[0145] In some possible implementations, two- and three-dimensional urban morphological features are extracted and then combined with POI data, supplemented by road network data, to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0146] Another technical solution of the present invention to solve the above technical problem is as follows: an electronic device includes the above storage medium and a processor that executes instructions in the above storage medium.

[0147] In some possible implementations, two- and three-dimensional urban morphological features are extracted and then combined with POI data, supplemented by road network data, to conduct a multi-source data fusion study on functional area classification. Based on the differences in spatial landscape structure and socioeconomic activities reflected by different functional areas, the two- and three-dimensional morphological features of each functional area are extracted to describe its spatial landscape structure, and POI data is used to describe basic human activities. By fusing the two- and three-dimensional urban morphological features with POI features, different functional area classification schemes are designed, and the impact of different features on functional area classification accuracy is compared and analyzed, significantly improving the results of urban functional area classification.

[0148] The reader should understand that in the description of this specification, reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the method embodiments described above are merely illustrative. For example, the division of steps is merely a logical function division. In actual implementation, other division methods may be used. For example, multiple steps may be combined or integrated into another step, or some features may be ignored or not performed.

[0150] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0151] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A functional area classification method based on multi-source data fusion, characterized in that: include: Step 1: Obtain object-level land use classification results, road network data, and POI data; Step 2: Determine the block range based on the road network data, determine the POI kernel density characteristics within each block range, and determine the two-dimensional morphological characteristics and three-dimensional morphological characteristics corresponding to each category in the object-level land use classification results; Step 3: Determine functional area classification schemes based on different feature combinations according to the kernel density features, two-dimensional morphological features, and three-dimensional morphological features of all POIs, and use the functional area classification schemes based on different feature combinations as input features of the classification model to construct the classification model; Step 4, classifying the study area based on the classification model; S2, determining the block range based on the road network data, determining the POI kernel density feature within each block range, and determining the two-dimensional morphological features and three-dimensional morphological features corresponding to each category in the object-level land use classification result are as follows: S21, determining the block range based on the road network data, and completing the street information by patching the topological errors, so that the block range that is not a complete closed area due to missing street information becomes a complete closed area; The rational division of blocks is the basic premise for functional area classification. OSM road network data is used to divide blocks, correct some topological errors, and combine aerial image data to supplement some remote roads to achieve a reasonable division of blocks. To fix some topology errors: OSM road network data is vector data, which includes points, lines, and surfaces. Since OSM road network data is vector data downloaded from the Internet, it may contain some topological errors, including overlapping vector surfaces, gaps between surfaces, intersections between line segments, and dangling line segments. Topological errors can be repaired using the editing tools in ArcGIS software combined with manual inspection. Only when OSM vector data is free of topological errors can the road network data be correctly divided. For the additional part of the road: The true color images displayed by the red, green and blue bands contained in the airborne images are used to assist in the block division; Remote roads refer to areas where the OSM road network data does not cover the distribution of features shown in airborne aerial images. S22, determining the POI kernel density characteristics within each block, determining the two-dimensional morphological characteristics of building land, bare land, cultivated land, grassland, road, woodland, or water body within each block, and determining the three-dimensional morphological characteristics of building land, bare land, cultivated land, grassland, road, woodland, or water body within each block; The process of determining the POI kernel density characteristics within each block is as follows: S221, classifying functional areas according to POI data types, where the functional areas include residential areas, commercial areas, industrial areas, public services, and open spaces; The POI data that can express human activities, social and economic phenomena are spatially transformed, and the kernel density analysis method is used to convert the discrete point data into raster data that can be spatially analyzed and expressed; S222, performing kernel density analysis on the POI data of each functional area to obtain POI kernel density features; Kernel density analysis was performed on POI data containing functional area category attributes. After multiple experiments, the search radius was set to 500m, and a 1m resolution raster image, also known as a POI distribution heat map, was generated to match the resolution of the land use classification results. Based on the block range, the mean mn, standard deviation std, and sum of the kernel density of residential areas, commercial areas, industrial areas, public services, and open spaces within each block were extracted, resulting in 15 POI kernel density features. Kernel density analysis can effectively express the spatial distribution of POI data and is a commonly used POI data expression method. Based on Silverman's quartic function, the predicted density of a new (x, y) location is determined by the following formula: Where i = 1, 2, ..., n are input points, and only points in the sum are included if they are within the radius of the (x, y) position; D(x, y) is the density prediction value of the new (x, y) point; r is the search radius; pop i is the weight value of the point; d i is the distance between point i and (x, y); pop i The field gives some features more weight than others and also allows multiple observations to be represented by a single point.

2. The functional area classification method based on multi-source data fusion according to claim 1 is characterized in that: The process of obtaining the object-level land use classification results specifically includes: Obtain LiDAR point cloud data and aerial image data at the same time through the airborne LiDAR measurement platform; Processing the LiDAR point cloud data and the aerial image data using a multi-source data fusion land use classification method to obtain an object-level land use classification result; The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

3. The functional area classification method based on multi-source data fusion according to claim 1 is characterized in that: The process of determining the block range based on the road network data includes: Topological errors are repaired so that each block is surrounded by streets to form a complete area.

4. The functional area classification method based on multi-source data fusion according to claim 1 is characterized in that: The process of determining the two-dimensional and three-dimensional morphological features corresponding to each category in the object-level land use classification results includes: The two-dimensional morphological features and three-dimensional morphological features are determined by combining the basic information corresponding to each category in the object-level land use classification results with the mapping table.

5. The functional area classification method based on multi-source data fusion according to claim 1 is characterized in that: The construction of the classification model specifically includes: Sample data were selected, 70% of the sample data were used as training sample data sets, 30% of the sample data were used as testing sample data sets, and a classification model was constructed by combining KNN, RF, XGBoost and input features.

6. A functional area classification system based on multi-source data fusion, using the functional area classification method based on multi-source data fusion according to claim 1, characterized in that: The system includes: The acquisition module is used to obtain object-level land use classification results, road network data, and POI data; The determination module is used to: determine the block range according to the road network data, determine the POI kernel density characteristics within each block range, and determine the two-dimensional morphological characteristics and three-dimensional morphological characteristics corresponding to each category in the object-level land use classification result; The construction module is used to: determine different feature combination functional area classification schemes based on all POI kernel density features, two-dimensional morphological features, and three-dimensional morphological features, and use the different feature combination functional area classification schemes as input features of the classification model to construct the classification model; The classification module is used to classify the study area based on the classification model.

7. The functional area classification system based on multi-source data fusion according to claim 6 is characterized in that: The process of obtaining the object-level land use classification results specifically includes: Obtain LiDAR point cloud data and aerial image data at the same time through the airborne LiDAR measurement platform; Processing the LiDAR point cloud data and the aerial image data using a multi-source data fusion land use classification method to obtain an object-level land use classification result; The object-level land use classification results include: building land, bare land, cultivated land, grassland, road, woodland and water body.

8. The functional area classification system based on multi-source data fusion according to claim 6 is characterized in that: The process of determining the block range based on the road network data includes: Topological errors are repaired so that each block is surrounded by streets to form a complete area.

9. A storage medium, characterized in that: The medium stores instructions, and when a computer reads the instructions, the computer executes the method according to any one of claims 1 to 5.

10. An electronic device, characterized in that: The invention comprises the storage medium according to claim 9 and a processor for executing instructions in the storage medium.

Images