Large-scale road pavement type identification method, medium and equipment

Abstract

The invention discloses a large-scale road pavement type identification method, a medium and equipment, and relates to the field of earth science. The large-scale road pavement type identification method mainly comprises the following steps: sampling an open street map to obtain sampling points; the method comprises the steps of obtaining input features and pavement type labels of all sampling points in combination with multi-source geographic space data to obtain a training set, training a deep learning model, obtaining a road pavement type of each sampling point by using the trained deep learning model, and further obtaining a pavement type of each road section. By implementing the large-scale road surface type identification method, the medium and the equipment provided by the invention, the road surface type information of each road in a large-scale range can be automatically obtained.

Description

Technical Field

[0001] The present invention relates to the field of earth science, and more specifically, to a large-scale road pavement type identification method, medium and equipment. Background Art

[0002] Roads are important transportation infrastructure and a key component of national infrastructure. Identifying road pavement types (such as hardened and unhardened roads) is of great significance for pavement maintenance, traffic management, and road network accessibility analysis. At present, there are three main methods for identifying road pavement types: methods based on vehicle-mounted sensor equipment, methods based on street view images, and methods based on high-resolution remote sensing images (Sabapathy & Biswas 2023; Lee et al. 2024; Zhou et al. 2024). For example, Sabapathy & Biswas (2023) used a vehicle-mounted speed monitoring system to collect data and used support vector machines and convolutional neural networks to identify road pavement types, but this method only supports small-scale road collection and is difficult to apply to large-scale research areas; Lee et al. (2024) used street view data and artificial neural networks to identify road pavement types, but street view data is only available in some countries and regions; Zhou et al. (2024) used high-resolution remote sensing images and deep learning models to identify road pavement types, but in areas with high vegetation coverage, this method has low accuracy in identifying road pavement types.

[0003] In recent years, open geospatial data, such as road data OpenStreetMap (OSM), population data LandScan, and land cover data ESRI Land Cover, have become increasingly accessible. They are suitable for large-scale studies such as national and regional studies due to their advantages of open access, global coverage, and dynamic updates. On the other hand, road pavement types may be related to factors such as road network morphology, socioeconomics, and geographical environment. Therefore, it is feasible to use open geospatial data to identify road pavement types. How to use multi-source geospatial data to identify the pavement type of each road in a large-scale road network such as a country or region is a technical problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to provide a large-scale road pavement type identification method, medium and equipment, which can automatically obtain the pavement type information of each road in a large-scale range.

[0005] The present invention provides a large-scale road pavement type recognition method, comprising the following steps: S1: sampling an open street map at preset intervals to obtain sampling points; S2: obtaining input features and pavement type labels of each sampling point based on multi-source geospatial data and the sampling points; S3: obtaining a training set based on the sampling points and the pavement type labels; S4: training a deep learning model using the training set to obtain a trained deep learning model; S5: obtaining the road pavement type of each sampling point based on the sampling points and the trained deep learning model; S6: obtaining the pavement type of each road section based on the road pavement type of each sampling point.

[0006] Furthermore, step S1 specifically includes: when the length of the road segment in the open street map is not greater than the preset interval, the center point of the road segment is used as the sampling point; when the length of the road segment is greater than the preset interval, the sampling points are uniformly generated along the road direction according to the preset interval.

[0007] Furthermore, the multi-source geospatial data include: OSM road data, GDP grid data, population grid data, building height grid data and land cover data.

[0008] Further, step S2 specifically includes: S21: assigning the road grade in the OSM road data as an attribute value to each sampling point to obtain the road grade feature of each sampling point; S22: constructing adjacent road segments in the road segments in the OSM road data whose angle is less than a preset angle threshold into road sections; according to the road sections, obtaining the length feature and topological feature of each road section, assigning the length feature and topological feature to the sampling point of the road, and obtaining the road section feature of each sampling point; S23: assigning the grid value of the sampling point falling into the GDP grid data to each sampling point to obtain the GDP feature of each sampling point; S24: assigning the grid value of the sampling point falling into the population grid data to each sampling point to obtain the GDP feature of each sampling point. The population quantity characteristics of the sampling point; S25: assigning the grid value of the sampling point falling into the building height grid data to each sampling point to obtain the building height characteristics of each sampling point; S26: assigning the grid value of the sampling point falling into the land cover data to each sampling point to obtain the land cover characteristics of each sampling point; S27: standardizing the road grade characteristics, the road marking characteristics, the GDP characteristics, the population quantity characteristics, the building height characteristics and the land cover characteristics to obtain the input characteristics of each sampling point; S28: If the OSM road data where the sampling point is located contains a road pavement type value, then assigning the road pavement type value as a label to the sampling point to obtain the pavement type label of each sampling point.

[0009] Furthermore, topological features include degree centrality, closeness centrality, and betweenness centrality.

[0010] Furthermore, the deep learning model is a tabular deep learning model.

[0011] Furthermore, the tabular deep learning model is TabNet.

[0012] Furthermore, step S6 specifically includes: S61: counting the number of sampling points of each road pavement type of each road segment, taking the pavement type with the largest number as the pavement type of the road segment, and obtaining the pavement type of each road segment; S62: if there are two or more pavement types with the largest number on the road segment, randomly adding a sampling point on the road segment, using the trained deep learning model to obtain the road pavement type of each sampling point, and returning to step S61.

[0013] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned large-scale road pavement type identification method are implemented.

[0014] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned large-scale road surface type identification method when executing the program.

[0015] The large-scale road surface type identification method, medium and device provided by the present invention have the following beneficial effects: The present invention generates sampling points at certain intervals based on OpenStreetMap road data, calculates input features of each sampling point based on multi-source geospatial data, randomly and equally extracts sampling points of various types from road segments containing road surface type labels in the OpenStreetMap road data as training sets, applies the input features of the training set and the corresponding road surface type labels to training a deep learning model, applies the trained model to all sampling points and identifies their road surface types, and finally determines the road surface type of each road segment according to the road surface type of the sampling points, thereby realizing the identification of the road surface type of each road in large-scale road networks such as national and regional ones based on open geospatial data, and the source data used has the advantages of global coverage and open access, which helps users automatically obtain the road surface type information of each road at national and regional scales. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 is a flow chart of a large-scale road pavement type identification method provided by the present invention; Figure 2 is a schematic diagram of generating sampling points along a road provided by the present invention; Figure 3 It is a schematic diagram of calculating the characteristic values ​​of sampling points based on multi-source geographic spatial data provided by the present invention; Figure 4 is a schematic diagram of the first step of determining the pavement type of a road segment provided by the present invention; Figure 5 is a schematic diagram of the second step of determining the road surface type of a road segment provided by the present invention; Figure 6 It is a schematic diagram of the spatial distribution of road pavement types at the national scale in Cameroon provided by the present invention; Figure 7 It is a structural block diagram of the computer device provided by the present invention. DETAILED DESCRIPTION

[0017] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0018] Figure 1 A schematic diagram of a large-scale road pavement type recognition method of this embodiment is shown. In this embodiment, a large-scale road pavement type intelligent recognition method based on multi-source geographic spatial data includes the following steps: S1: Sampling the open street map at preset intervals to obtain sampling points; Specifically, step S1 includes: when the length of the road segment in the open street map is not greater than the preset interval, the center point of the road segment is used as a sampling point; when the length of the road segment is greater than the preset interval, the sampling points are uniformly generated along the road direction according to the preset interval; In an exemplary embodiment, the preset interval is 100 meters; In one embodiment, in step S1, sampling points are generated at a certain interval M along the road direction of the OpenStreetMap (OSM) road data; (1) if the length of the road segment is not greater than M, the center point of the road segment is used as the sampling point; (2) if the length of the road segment is greater than M, sampling points are uniformly generated along the road direction at the set interval; It should be noted that the sampling interval M is recommended to be set to 50~1000 meters; if the sampling interval is too large, it is difficult to identify different road surface types on the same road; if the sampling interval is too small, the number of sampling points is too large, resulting in an increase in the calculation time cost of the method; S2: obtaining input features and road surface type labels of each sampling point according to the multi-source geospatial data and the sampling points; Specifically, the multi-source geospatial data include: OSM road data, GDP grid data, population grid data, building height grid data and land cover data; Specifically, step S2 includes: S21: assigning the road grade in the OSM road data as an attribute value to each sampling point to obtain the road grade feature of each sampling point; S22: constructing adjacent road segments in the road segments in the OSM road data whose angles are less than a preset angle threshold into road markings; obtaining a length feature and a topological feature of each road marking according to the road markings, assigning the length feature and the topological feature to sampling points of the road, and obtaining road marking features of each sampling point; Specifically, the above topological features include degree centrality, closeness centrality, and betweenness centrality; In an exemplary embodiment, the preset angle threshold is 70°; S23: assigning the grid value of the sampling point falling into the GDP grid data to each sampling point to obtain the GDP feature of each sampling point; S24: assigning the grid value of the sampling point falling into the population grid data to each sampling point to obtain the population quantity characteristics of each sampling point; S25: assigning the grid value of the building height grid data where the sampling point falls to each sampling point to obtain the building height feature of each sampling point; S26: assigning the grid value of the land cover data where the sampling point falls to each sampling point to obtain the land cover characteristics of each sampling point; S27: Standardizing the road grade feature, the road marking feature, the GDP feature, the population feature, the building height feature, and the land cover feature to obtain input features of each sampling point; S28: If the OSM road data where the sampling point is located contains a road pavement type value, the road pavement type value is assigned to the sampling point as a label to obtain a pavement type label of each sampling point; Specifically, the standardization process is one of Min-Max Scaling, Z-score Standardization, Decimal Scaling, Normalization, Log Transformation, and Skewness Correction; In an exemplary embodiment, step S2 calculates input features of each sampling point based on multi-source geospatial data; specifically, the features of each sampling point are calculated based on multi-source geospatial data such as OSM road data, gross domestic product (GDP) grid data, population grid data, building height grid data, land cover data, etc.; if the OSM road data where the sampling point is located contains a road surface type value, then the road surface type value is assigned to the sampling point as a label; step S2 specifically includes the following steps: (1) obtaining multi-source geospatial data such as OSM road data, GDP grid data, population grid data, building height grid data, land cover data, etc., and superimposing them with all sampling points; (2) The road grade in the OSM road data is assigned as an attribute value to the sampling point to obtain the road grade feature of the sampling point; (3) The adjacent road segments in the OSM road segment with an angle less than a certain threshold A are constructed into "road sections", the length feature of each road section is calculated, and the topological features such as degree centrality, closeness centrality and betweenness centrality of the road section are calculated using the topological relationship of the road network; and the feature value is assigned to the sampling point on the road; (4) The grid value of the sampling point falling into the GDP grid data is assigned to the sampling point to obtain the GDP feature of the sampling point; (5) The raster value of the sampling point falling into the population grid data is assigned to the sampling point to obtain the population quantity feature of the sampling point; (6) The raster value of the sampling point falling into the building height grid data is assigned to the sampling point to obtain the building height feature of the sampling point; (7) The raster value of the sampling point falling into the land cover data is assigned to the sampling point to obtain the land cover feature of the sampling point; (8) The multiple features calculated above are standardized; S3: obtaining a training set according to the sampling points and the road surface type labels; Specifically, in step S3, sampling points of each road surface type are randomly and equally selected from the sampling points containing road surface type labels in the OSM road data on the road segment as a training set; S4: Using the training set to train the deep learning model to obtain a trained deep learning model; Specifically, the above deep learning model is a table deep learning model; In an exemplary embodiment, the above-mentioned table deep learning model is TabNet; Specifically, in step S4, the input features of each sampling point in the training set, i.e., the input variables, and the corresponding road surface type labels, i.e., the output variables, are applied to the training deep learning model to obtain the trained deep learning model; S5: According to the sampling points, using the trained deep learning model, obtaining the road surface type of each sampling point; Specifically, in step S5, the model obtained after training is applied to all sampling points to identify the road surface type of each sampling point; S6: Obtaining the road surface type of each road section according to the road surface type of each sampling point; Specifically, step S6 includes: S61: counting the number of sampling points of each road surface type of each road segment, taking the road surface type with the largest number as the road surface type of the road segment, and obtaining the road surface type of each road segment; S62: if there are two or more road surface types with the largest number on the road segment, randomly adding a sampling point on the road segment, using the trained deep learning model to obtain the road surface type of each sampling point, and returning to step S61; In one embodiment, in step S6, the pavement type of each road segment is determined according to the road pavement type of the sampling point, and the specific steps are: 1) for each road segment, the pavement types of its different sampling points are counted, and the pavement type with the largest number is used as the pavement type of the road segment; 2) if there are two or more most common pavement types, a sampling point is randomly added to the road segment, and its pavement type is determined using the trained model, and step 1) is repeated.

[0019] In another embodiment, it is assumed that the road surface type is related to the importance of the road in the road network, the level of social and economic development of the region, and the geographical environment; based on the above assumption, the large-scale road surface type identification method of this embodiment includes the following steps: Step 1: Generate sampling points at certain intervals from OSM road data; Step 2: Calculate the input features of each sampling point based on multi-source geospatial data; Step 3: Randomly select sampling points of each type from the road segments containing road surface type labels in the OSM road data as the training set; Step 4: Apply the input features of the training set and the corresponding road type labels to train the deep learning model; Step 5: Apply the trained model to all sampling points and identify their road surface types; Step 6: Determine the road surface type of each road section based on the road surface type of the sampling point; Specifically, step 1 includes: generating sampling points along the road direction of the OSM data at a certain interval M. If the length of the road segment is not greater than M, the center point of the road segment is used as the sampling point; if the length of the road segment is greater than M, the sampling points are uniformly generated along the road direction at the set interval; It should be noted that the sampling interval M is set to 50-1000 meters; if the sampling interval is too large, it is difficult to identify different road surface types on the same road; if the sampling interval is too small, the number of sampling points is too large, resulting in an increase in the computational time cost of the method; Specifically, step 2 includes: calculating the characteristics of each sampling point based on multi-source geospatial data such as OSM road data, GDP grid data, population grid data, building height grid data, and land cover data; if the OSM road data where the sampling point is located contains the road surface type, assign it as a label to the sampling point; the specific steps for calculating the characteristics of the sampling point are as follows: (a) Obtain multi-source geospatial data such as OSM road data, GDP grid data, population grid data, building height data, land cover data, and overlay them with all sampling points; (b) Assign the road grade in the OSM road data as an attribute value to the sampling point to obtain the road grade characteristics of the sampling point; (c) Adjacent road segments with an angle less than a certain threshold A in OSM road segments are constructed into “road blocks” and the length characteristics of each road block are calculated. The topological characteristics of the road block, such as degree centrality, closeness centrality and betweenness centrality, are calculated using the topological relationship of the road network, and the characteristic values ​​are assigned to the sampling points on the road. (d) Assign the grid value of the sampling point that falls into the GDP grid data to the sampling point to obtain the GDP characteristics of the sampling point; (e) Assign the grid value of the sampling point that falls into the population grid data to the sampling point to obtain the population quantity characteristics of the sampling point; (f) assigning the grid value of the sampling point falling into the building height grid data to the sampling point to obtain the building height feature of the sampling point; (g) assigning the grid value of the land cover data where the sampling point falls to the sampling point to obtain the land cover characteristics of the sampling point; (h) normalizing the multiple features calculated above; Specifically, step 3 includes: randomly and equally proportionally selecting sampling points of different road surface types from sampling points containing road surface type labels as a training set; Specifically, step 4 includes: applying the features of the training set (i.e., input variables) and the corresponding road surface types (i.e., output variables) to the training deep learning model; Specifically, the accuracy of road surface type recognition is higher when using tabular deep learning models such as TabNet; Specifically, step 5 includes: applying the model obtained after training in step 4 to all sampling points to identify the road surface type of each sampling point; Specifically, step 6 includes: determining the road surface type of the road segment according to the road surface type of the sampling point, and the specific steps are as follows: a) For each road section, the pavement types at different sampling points are counted, and the pavement type with the largest number is taken as the pavement type of the road section; b) If there are two or more most common road surface types, a new sampling point is randomly added to the road segment, its road surface type is determined using the trained model, and step a) is repeated.

[0020] In an exemplary embodiment, the above-mentioned large-scale road surface type recognition method can also be implemented in the following manner.

[0021] In this embodiment, the African country Cameroon is taken as the research area. The open data used in this embodiment include: OSM road data (http: / / download.geofabrik.de / index.html), GDP grid data (Chen et al. 2022, https: / / figshare.com / articles / dataset / 17004523 / 1), population grid data (https: / / landscan.ornl.gov / ), building height grid data (https: / / ghsl.jrc.ec.europa.eu / ghs_buH2022.php) and land cover data (https: / / livingatlas.ar cgis.com / landcover); In this embodiment, the large-scale road pavement type recognition method includes the following steps: Step 01: When the length of the road segment is not greater than 100 meters, the center point of the road segment is selected as the sampling point; Figure 2 Generate a schematic diagram of sampling points at 100-meter intervals along the OSM road data; Step 02: Calculate the characteristics of each sampling point based on OSM road data, GDP grid data, population grid data, building height grid data and land cover data; Figure 3 The figure shows that, taking the population grid data as an example, the value of each sampling point falling into the population grid unit is used as the characteristic value of the sampling point; if the OSM road data where the sampling point is located contains the road surface type, it is assigned to the sampling point as a label; specifically, this embodiment divides the road surface type into two categories: "paved" and "unpaved", that is, the road marked as "paved" in the OSM road data is regarded as a hardened road; the road marked as "unpaved" is regarded as an unpaved road; Step 03: Randomly and equally proportionally extract sampling points of different road surface types from the sampling points containing the road surface types as a training set; in this embodiment, 5,000 hardened road sampling points and 5,000 unhardened road sampling points are randomly selected as a training set, and 2,000 sampling points are additionally randomly selected as a validation set; the road surface type of each sampling point in the validation set is used as a reference based on the visual interpretation of high-resolution remote sensing images; Step 04: Apply the features of the training set (i.e. input variables) and the corresponding road surface types (i.e. output variables) to train the deep learning model; Step 05: Apply the model trained in step 04 to the 2000 sampling points in the validation set and identify the road surface type of each sampling point; This example compares the effectiveness of five different machine learning and deep learning models (including CatBoost, XGBoost, LightGBM, FT-Transformer, and TabNet) in road surface type recognition. The accuracy of different models is evaluated using accuracy, precision, recall, and F1 score. The results are shown in Table 1. Table 1: Classification accuracy of five different models

[0022] As can be seen from Table 1, the accuracy of various models is close to or exceeds 85%, indicating that the present invention has a good effect in identifying road surface types; most indicators of the TabNet model are optimal, so it is recommended to use the TabNet table deep learning model; Step 06: Determine the road surface type of each road section according to the road surface type of the sampling point. The specific steps include: A) For each road segment, count the pavement types at different sampling points, and take the pavement type with the largest number as the pavement type of the road segment; Figure 4 As shown, when the number of sampling points identified as hardened roads (p1, p2, and p4) on a road segment is greater than the number of sampling points identified as unhardened roads (p3), the road segment is classified as a hardened road; B) If there are two or more road surface types, a random sampling point is added to the road segment, the road surface type of the newly added sampling point is determined using the trained model, and step A is repeated); Figure 5 As shown in the figure, the number of sampling points classified as hardened roads (p1 and p4) and the number of sampling points classified as non-hardened roads (p2 and p3) are equal, so a new sampling point (p5) is randomly generated on the road segment, and the trained model is used to identify the newly added sampling point as a hardened road; therefore, the road segment is classified as a hardened road; The results of using the method of this embodiment to identify the road surface types of Cameroon national scale are as follows: Figure 6 As shown; it can be seen that the present invention not only has high accuracy in identifying road pavement types, but is also applicable to large-scale research areas such as countries and regions, thereby verifying the effectiveness of the method of the present invention.

[0023] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned large-scale road surface type identification method are implemented. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium may also include a combination of the above-mentioned types of memory.

[0024] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned large-scale road pavement type identification method are implemented.

[0025] like Figure 7 As shown, the computer device may include: at least one processor 121, such as a central processing unit (CPU), at least one communication interface 123, a memory 124, and at least one communication bus 122. Among them, the communication bus 122 is used to realize the connection and communication between these components. Among them, the communication interface 123 may include a display screen (Display), a keyboard (Keyboard), and the optional communication interface 123 may also include a standard wired interface and a wireless interface. The memory 124 may be a high-speed random access memory (Random Access Memory, RAM), or a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 124 may also be at least one storage device located away from the aforementioned processor 121. Among them, the memory 124 stores an application program, and the processor 121 calls the program code stored in the memory 124 to perform any of the above method steps. Among them, the communication bus 122 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 122 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one line is used to represent it, but it does not mean that there is only one bus or one type of bus. Among them, the memory 124 may include a volatile memory (volatile memory), such as a random-access memory (random-access memory, RAM); the memory may also include a non-volatile memory (non-volatile memory), such as a flash memory (flash memory), a hard disk drive (hard disk drive, HDD) or a solid-state drive (solid-state drive, SSD); the memory 124 may also include a combination of the above-mentioned types of memory. Among them, the processor 121 may be a central processing unit (central processing unit, CPU), a network processor (network processor, NP) or a combination of CPU and NP. Among them, the processor 121 may further include a hardware chip. The above-mentioned hardware chip may be an application-specific integrated circuit (application-specific integrated circuit, ASIC), a programmable logic device (programmable logic device, PLD) or a combination thereof. The above-mentioned PLD may be a complex programmable logic device (complex programmable logic device, CPLD), a field-programmable gate array (field-programmable gate array, FPGA), a generic array logic (generic array logic, GAL) or any combination thereof. Optionally, the memory 124 is also used to store program instructions. The processor 121 can call the program instructions to implement the large-scale road surface type recognition method of this embodiment.

[0026] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A large-scale road surface type recognition method, characterized in that: The following steps are involved: S1: Sampling the open street map at preset intervals to obtain sampling points; S2: obtaining input features and road surface type labels of each sampling point according to the multi-source geospatial data and the sampling points; S3: obtaining a training set according to the sampling points and the road surface type labels; S4: Using the training set to train the deep learning model to obtain a trained deep learning model; S5: According to the sampling points, using the trained deep learning model, obtaining the road surface type of each sampling point; S6: Obtain the road surface type of each road section according to the road surface types of each sampling point.

2. The large-scale road pavement type recognition method according to claim 1 is characterized in that: Step S1 specifically includes: when the length of a road segment in the open street map is not greater than a preset interval, the center point of the road segment is used as a sampling point; when the length of the road segment is greater than the preset interval, the sampling points are uniformly generated along the road direction according to the preset interval.

3. The large-scale road pavement type recognition method according to claim 1, characterized in that: The multi-source geospatial data include: OSM road data, GDP grid data, population grid data, building height grid data and land cover data.

4. The large-scale road surface type recognition method according to claim 3 is characterized in that: Step S2 specifically includes: S21: assigning the road grade in the OSM road data as an attribute value to each sampling point to obtain the road grade feature of each sampling point; S22: constructing adjacent road segments in the road segments in the OSM road data whose angles are less than a preset angle threshold into road markings; obtaining a length feature and a topological feature of each road marking according to the road markings, assigning the length feature and the topological feature to sampling points of the road, and obtaining road marking features of each sampling point; S23: assigning the grid value of the sampling point falling into the GDP grid data to each sampling point to obtain the GDP feature of each sampling point; S24: assigning the grid value of the sampling point falling into the population grid data to each sampling point to obtain the population quantity characteristics of each sampling point; S25: assigning the grid value of the building height grid data where the sampling point falls to each sampling point to obtain the building height feature of each sampling point; S26: assigning the grid value of the land cover data where the sampling point falls to each sampling point to obtain the land cover characteristics of each sampling point; S27: Standardizing the road grade feature, the road marking feature, the GDP feature, the population feature, the building height feature, and the land cover feature to obtain input features of each sampling point; S28: If the OSM road data at the sampling point contains a road pavement type value, the road pavement type value is assigned to the sampling point as a label to obtain a pavement type label for each sampling point.

5. The large-scale road surface type recognition method according to claim 4 is characterized in that: The topological features include degree centrality, closeness centrality and betweenness centrality.

6. The large-scale road surface type recognition method according to claim 1 is characterized in that: The deep learning model is a tabular deep learning model.

7. The large-scale road surface type recognition method according to claim 6 is characterized in that: The tabular deep learning model is TabNet.

8. The large-scale road surface type recognition method according to claim 1 is characterized in that: Step S6 specifically includes: S61: Counting the number of sampling points of each road pavement type of each road segment, taking the pavement type with the largest number as the pavement type of the road segment, and obtaining the pavement type of each road segment; S62: If there are two or more road surface types with the largest number on the road segment, a new sampling point is randomly added to the road segment, and the road surface type of each sampling point is obtained using the trained deep learning model, and the process returns to step S61.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the large-scale road pavement type recognition method as described in any one of claims 1 to 8 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the large-scale road pavement type recognition method as described in any one of claims 1-8 are implemented.

Images