Origin-destination matrix generation method, device and equipment and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for generating origin-endpoint OD matrices, relating to the field of urban data processing and analysis technology. The method includes: responding to user-inputted information about a target city, dividing the target city into multiple sub-regions; for each sub-region, acquiring population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data; extracting structured population features from the population structure data, high-dimensional semantic features from the satellite remote sensing images, and spatial distribution features of POIs from the POI distribution data; and generating a target OD matrix for the target city based on the distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs for each sub-region. The technical solution provided in this application effectively improves the generation efficiency of the OD matrix.

Description

Technical Field

[0001] This application relates to the field of urban data processing and analysis technology, and in particular to a method, apparatus, device and storage medium for generating an origin-endpoint OD matrix. Background Technology

[0002] Urban mobility data generation technology aims to estimate the flow of people between different areas of a city, typically presented in the form of an origin-destination (OD) matrix.

[0003] To generate an OD matrix, it is usually necessary to manually obtain functional attribute features that characterize urban space from different data sources, such as population spatial distribution, structured demographic data, high-resolution satellite remote sensing images, and spatial distribution of points of interest (POIs). The functional attribute features from different data sources are then manually preprocessed, and the OD matrix is ​​generated based on these preprocessed functional attribute features.

[0004] However, the acquisition and preprocessing of the functional attribute characteristics of the above-mentioned different data sources are highly dependent on manual intervention, and the process is complex, resulting in low efficiency in generating the OD matrix. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for generating origin and destination OD matrices, which solves the problem that the low generation efficiency of OD matrices in the prior art is caused by the high dependence of manual intervention on the acquisition and preprocessing of functional attribute characteristics of different data sources, thereby effectively improving the generation efficiency of OD matrices.

[0006] This application provides a method for generating a start-end point OD matrix, including: In response to the target city information input by the user, the target city is divided into regions to obtain multiple sub-regions; For each sub-region, acquire population structure data within the sub-region, satellite remote sensing images corresponding to the sub-region, and point-of-interest (POI) distribution data within the sub-region; and extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of POIs corresponding to the POI distribution data. Based on the distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of points of interest corresponding to each of the multiple sub-regions, a target OD matrix corresponding to the target city is generated; wherein, the target OD matrix is ​​used to characterize the population flow between different regions in the target city.

[0007] According to the origin-destination OD matrix generation method provided in this application, the step of generating the target OD matrix corresponding to the target city based on the distance between the plurality of sub-regions, the structured artificial features corresponding to each of the plurality of sub-regions, the high-dimensional semantic features, and the spatial distribution features of interest points includes: The distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each of the multiple sub-regions are input multiple times into a graph-based denoising diffusion model. The graph-based denoising diffusion model is used to randomly sample the flow of people in different areas of the target city multiple times to obtain multiple OD matrices. The target OD matrix is ​​generated based on the multiple OD matrices.

[0008] According to the method for generating a start-endpoint OD matrix provided in this application, the step of generating the target OD matrix based on the plurality of OD matrices includes: The average OD matrix is ​​obtained by averaging the multiple OD matrices. The average OD matrix is ​​determined as the target OD matrix.

[0009] According to the origin-endpoint OD matrix generation method provided in this application, high-dimensional semantic features corresponding to the satellite remote sensing image are extracted, including: The satellite remote sensing image is preprocessed to obtain a preprocessed target satellite remote sensing image; The target satellite remote sensing image is input into a multimodal visual model, and the high-dimensional semantic features are output through the multimodal visual model.

[0010] According to the method for generating an origin-endpoint OD matrix provided in this application, obtaining population structure data within the sub-region includes: Spatially overlay the target city with the raster pixels in the population raster dataset to determine the raster data covering the target city; Based on the boundary of the sub-region, the population structure data within the raster data falling into the sub-region from the raster data covering the target city are weighted and aggregated to obtain the population structure data within the sub-region.

[0011] According to the method for generating an origin-endpoint OD matrix provided in this application, the satellite remote sensing image corresponding to the sub-region is obtained, including: Obtain satellite remote sensing image tiles corresponding to the target city; The satellite remote sensing image tiles corresponding to the target city are stitched together and cropped to obtain the satellite remote sensing image corresponding to the target city. Based on the boundary of the sub-region, the satellite remote sensing image corresponding to the sub-region is obtained from the satellite remote sensing image corresponding to the target city.

[0012] This application also provides a start-endpoint OD matrix generation device, comprising: A partitioning unit is used to divide the target city into multiple sub-regions in response to user-inputted target city information. The processing unit is configured to acquire population structure data, satellite remote sensing images corresponding to the sub-region, and point-of-interest distribution data within the sub-region for each sub-region; and extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of points of interest corresponding to the point-of-interest distribution data. The generation unit is used to generate a target OD matrix corresponding to the target city based on the distance between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each of the multiple sub-regions; wherein, the target OD matrix is ​​used to characterize the population flow between different regions in the target city.

[0013] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the start-endpoint OD matrix generation method as described in any of the preceding claims.

[0014] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the start-endpoint OD matrix generation method as described in any of the preceding claims.

[0015] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the start-endpoint OD matrix generation method as described in any of the preceding claims.

[0016] The origin-endpoint OD matrix generation method, apparatus, device, and storage medium provided in this application, in response to user-inputted target city information, divide the target city into multiple sub-regions; for each sub-region, acquire population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data; extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of POIs corresponding to the POI distribution data; and generate the target OD matrix corresponding to the target city based on the distances between multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs corresponding to each sub-region. This method, which divides the target city into sub-regions, automatically acquires population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data within each sub-region. It then extracts structured population features from the population structure data, high-dimensional semantic features from the satellite remote sensing images, and spatial distribution features from the POI distribution data. This allows for the combined generation of the target OD matrix for the target city by combining the structured artificial features, high-dimensional semantic features, and spatial distribution features of multiple sub-regions. This automated OD matrix generation eliminates the need for manual intervention in acquiring and preprocessing functional attribute features from different data sources, simplifying the process and overcoming the low efficiency of OD matrix generation in existing technologies due to the heavy reliance on manual intervention in acquiring and preprocessing functional attribute features from different data sources. This significantly improves the efficiency of OD matrix generation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of a framework for generating a start-endpoint OD matrix, provided in an embodiment of this application.

[0019] Figure 2 This is a flowchart illustrating a method for generating a start-endpoint OD matrix according to an embodiment of this application.

[0020] Figure 3 This is a schematic diagram illustrating a gridded area division within a city, as provided in an embodiment of this application.

[0021] Figure 4 This is a schematic diagram illustrating the extraction of high-dimensional semantic features as provided in an embodiment of this application.

[0022] Figure 5 This is a schematic diagram illustrating how a target OD matrix for a target city is generated by combining key urban area features, as provided in an embodiment of this application.

[0023] Figure 6 This is a schematic diagram illustrating how to obtain population structure data within a sub-region, as provided in an embodiment of this application.

[0024] Figure 7 This is a schematic diagram illustrating an embodiment of the present application for obtaining satellite remote sensing images corresponding to a sub-region.

[0025] Figure 8 This is a schematic diagram of a process for generating a target OD matrix corresponding to a target city, provided in an embodiment of this application.

[0026] Figure 9 This is a schematic diagram showing the comparison between an OD matrix generated based on a graph denoising diffusion model and the actual OD matrix, provided in an embodiment of this application.

[0027] Figure 10 This is a schematic diagram of a starting and ending point OD matrix generation device provided in an embodiment of this application.

[0028] Figure 11 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] In the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone, where A and B can be singular or plural. In the textual description of this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0031] The technical solution provided in this application can be applied to scenarios such as urban planning, traffic optimization, emergency management, and digital twin cities. The urban population mobility data generation technology aims to estimate the flow of people between different areas of a city, typically presented in the form of an OD matrix. Using the generated population mobility data, city managers can systematically optimize the layout of transportation facilities, urban spatial structure, and allocation of public service resources from a sustainable development perspective, thereby improving urban operational efficiency and reducing resource waste.

[0032] Currently, in order to generate an OD matrix, it is usually necessary to first manually obtain functional attribute features that characterize urban space from different data sources, such as population spatial distribution, structured demographic data, high-resolution satellite remote sensing images, and spatial distribution of points of interest (POIs). Then, the functional attribute features from different data sources are manually preprocessed, and the OD matrix is ​​generated based on these preprocessed functional attribute features.

[0033] However, the acquisition and preprocessing of the functional attribute characteristics of the above-mentioned different data sources are highly dependent on manual intervention, and the process is complex, resulting in low efficiency in generating the OD matrix.

[0034] To address the shortcomings of existing technologies where the acquisition and preprocessing of functional attributes from different data sources heavily rely on manual intervention, resulting in low efficiency in OD matrix generation, and to effectively improve the efficiency of OD matrix generation, the generation process in this application embodiment may include three core modules: regional division of the selected target city, global public data crawling acquisition, and OD matrix generation based on a graph denoising diffusion model, thereby constituting an end-to-end fully automated OD matrix generation process.

[0035] For example, see Figure 1 As shown, Figure 1 This document provides a schematic diagram of a framework for generating an origin-endpoint OD matrix, as illustrated in an embodiment of this application. For a user-selected target city, the city is first divided into multiple sub-regions. For each sub-region, population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data can be obtained using global public data crawling. Based on these data, the urban regional characteristics of the sub-region are determined, including structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of POIs corresponding to the POI distribution data. A target OD matrix for the target city is generated based on the distances between multiple sub-regions, their respective structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs. The target OD matrix characterizes the population flow between different areas within the target city.

[0036] This method, which divides the target city into sub-regions, automatically acquires population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data within each sub-region. It then extracts structured population features from the population structure data, high-dimensional semantic features from the satellite remote sensing images, and spatial distribution features from the POI distribution data. This allows for the combined generation of the target OD matrix for the target city by combining the structured artificial features, high-dimensional semantic features, and spatial distribution features of multiple sub-regions. This automated OD matrix generation eliminates the need for manual intervention in acquiring and preprocessing functional attribute features from different data sources, simplifying the process and overcoming the low efficiency of OD matrix generation in existing technologies due to the heavy reliance on manual intervention in acquiring and preprocessing functional attribute features from different data sources. This significantly improves the efficiency of OD matrix generation.

[0037] It is understood that the execution subject of the start-end point OD matrix generation method provided in this application can be a computer, a server, or a specially set start-end point OD matrix generation device or other electronic equipment, or it can be a start-end point OD matrix generation device set in such electronic equipment. The start-end point OD matrix generation device can be implemented by software, hardware or a combination of both, and can be set according to actual needs.

[0038] The starting and ending point OD matrix generation method provided in this application will be described in detail below through several specific embodiments. It is understood that these specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0039] Figure 2 This application provides a flowchart illustrating a method for generating a start-endpoint OD matrix. For example, please refer to [link to flowchart illustration]. Figure 2 As shown, the method for generating the start and end point OD matrix can include: S201. In response to the target city information input by the user, the target city is divided into regions to obtain multiple sub-regions.

[0040] The target city can be a city that the user is interested in, and the specific city can be set according to actual needs.

[0041] For example, the information input by the user regarding the target city can be the city's identifier or its boundary information, etc., which can be set according to actual needs. For example, the boundary information can be polygon coordinates, etc.

[0042] For example, when dividing a target city into regions, an adaptive grid partitioning algorithm can be used to divide the entire target city into multiple sub-regions of controllable size. See also... Figure 3 As shown, Figure 3 This is a schematic diagram of a gridded area division within a city provided in an embodiment of this application. Specifically, when dividing the area, the spatial resolution and the efficiency of subsequent OD matrix generation can be considered. For cities with smaller areas, smaller grids can be used for area division to preserve the spatial details of the small city; for cities with larger areas, larger grids can be used for area division to improve the efficiency of OD matrix generation.

[0043] For example, the adaptive meshing algorithm can be an error- or density-based adaptive meshing algorithm, or a clustering or graph-based adaptive meshing algorithm, or an adaptive meshing algorithm based on spatiotemporal data and artificial intelligence (AI), etc., which can be set according to actual needs.

[0044] The core idea of ​​the adaptive mesh partitioning algorithm based on error or density is to dynamically adjust the mesh density according to the local error estimate, data density or gradient magnitude, so that key areas, such as areas with dense data or drastic changes, can be partitioned more finely.

[0045] The core idea of ​​adaptive grid algorithms based on clustering or graph partitioning is to abstract urban space into a graph or point cloud and use algorithms such as clustering and graph partitioning to divide the region, which can be combined with features such as spatial autocorrelation and POI distribution.

[0046] The core idea of ​​the adaptive grid partitioning algorithm based on spatiotemporal data and AI is to dynamically adjust the grid by using multi-source spatiotemporal data, such as mobile phone signaling, traffic flow, and social media check-ins, combined with technologies such as machine learning and deep learning.

[0047] For example, the grid size can be automatically calculated based on the size of the city's bounding box, with a default setting of 5% of the boundary length. It also has an upper limit, such as 5 kilometers, and a lower limit, such as 500 meters, to ensure that the division results not only have the ability to express spatial structure and population flow, but also facilitate the subsequent generation of the OD matrix.

[0048] In summary, this method of dividing areas based on city size avoids noise interference caused by excessive fragmentation, while also reasonably capturing the spatial structure and population flow within the city. The entire area division process is automated and requires no manual intervention, effectively improving the efficiency and consistency of urban area division, thereby providing high-quality spatial base units for subsequent OD matrix generation.

[0049] S202. For each sub-region, acquire population structure data, satellite remote sensing images of the sub-region, and point-of-interest (POI) distribution data within the sub-region; and extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of POIs corresponding to the POI distribution data.

[0050] Among them, the high-dimensional semantic features corresponding to satellite remote sensing images are used to characterize the spatial functions and semantic attributes of sub-regions.

[0051] For example, population structure data can include total population, age, gender, occupational structure, and other structured data, which can be set according to actual needs.

[0052] For example, the distribution data of points of interest can include data such as the category and size of the points of interest, which can be set according to actual needs.

[0053] For example, in the embodiments of this application, when extracting the structured population features corresponding to the population structure data, automated crawling and application programming interface (API) calling algorithms can be used to extract the structured population features corresponding to the population structure data; data cleaning and spatialization processing can also be used to extract the structured population features corresponding to the population structure data, etc. The specific settings can be made according to actual needs.

[0054] For example, in the embodiments of this application, when extracting high-dimensional semantic features corresponding to satellite remote sensing images, the satellite remote sensing images can be preprocessed first to obtain preprocessed target satellite remote sensing images; the target satellite remote sensing images are then input into a multimodal visual model, and the multimodal visual model outputs high-dimensional semantic features used to characterize the spatial functions and semantic attributes of sub-regions. For example, see [link to relevant documentation]. Figure 4 As shown, Figure 4 This is a schematic diagram of high-dimensional semantic feature extraction provided in an embodiment of this application. Target satellite remote sensing images including multiple sub-regions can be input into a multimodal visual model, and the multimodal visual model can output high-dimensional semantic features to characterize the spatial functions and semantic attributes of the sub-regions.

[0055] For example, the multimodal vision model can be the RemoteCLIP model, or the CLIP (Contrastive Language-Image Pre-training) model, or FLAVA (Foundational Language and VisionAlignment), etc., which can be set according to actual needs.

[0056] For example, in the embodiments of this application, when extracting the spatial distribution features of interest points corresponding to the interest point distribution data, spatial statistical analysis methods can be used to extract the spatial distribution features of interest points corresponding to the interest point distribution data, or machine learning methods, such as K-Means clustering or DBSCAN clustering, can be used to extract the spatial distribution features of interest points corresponding to the interest point distribution data. Alternatively, deep learning methods can be used to extract the spatial distribution features of interest points corresponding to the interest point distribution data. The specific settings can be made according to actual needs.

[0057] This automated acquisition and extraction process automatically extracts structured artificial features, high-dimensional semantic features, and spatial distribution features of points of interest (POIs) from multiple sub-regions. Based on the distances between these sub-regions, it integrates these three key urban region features—structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs—and combines them with the key urban region features to generate the target OD matrix for the target city. For an example, see [link to example]. Figure 5 As shown, Figure 5 This is a schematic diagram illustrating how a target OD matrix for a target city is generated by combining key urban area features, thereby reducing manual workload and improving the efficiency of OD matrix generation.

[0058] Among them, the distance between multiple sub-regions, the structured artificial features corresponding to each sub-region, the high-dimensional semantic features, and the spatial distribution features of interest points can comprehensively characterize the spatial functional attributes of the target region, providing a basis for the subsequent generation of the OD matrix.

[0059] S203. Based on the distance between multiple sub-regions, the structured artificial features corresponding to each sub-region, the high-dimensional semantic features, and the spatial distribution features of interest points, generate the target OD matrix corresponding to the target city.

[0060] The target OD matrix is ​​used to characterize the flow of people between different areas in the target city.

[0061] As can be seen, in this embodiment, in response to the user-inputted target city information, the target city is divided into multiple sub-regions. For each sub-region, population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data within the sub-region are automatically acquired. The structured population features corresponding to the population structure data, the high-dimensional semantic features corresponding to the satellite remote sensing images, and the spatial distribution features of POIs corresponding to the POI distribution data are extracted. This allows the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs corresponding to each of the multiple sub-regions to jointly generate the target OD matrix corresponding to the target city. This achieves automated OD matrix generation, eliminating the need for manual intervention in the acquisition and preprocessing of functional attribute features from different data sources. The process is simple and solves the problem of low OD matrix generation efficiency caused by the high reliance on manual intervention in the acquisition and preprocessing of functional attribute features from different data sources in the prior art. This effectively improves the generation efficiency of the OD matrix.

[0062] Based on the above Figure 2 In the illustrated embodiment, for example, when obtaining population structure data within a sub-region in S202 above, see, for example, [the following is an example]. Figure 6 As shown, Figure 6 This illustration shows a method for obtaining population structure data within a sub-region, as provided in this application embodiment. First, the target city and the raster pixels in the population raster dataset are spatially overlaid to determine the raster data covering the target city. Then, based on the boundary of the sub-region, the population structure data falling within the raster data covering the target city and falling within the sub-region is weighted and aggregated to obtain the population structure data within the sub-region. The entire acquisition process requires no manual intervention, ensuring the consistency and accuracy of the population feature data and providing reliable basic population information for subsequent OD matrix generation.

[0063] For example, the population raster dataset can be the WorldPop dataset, and the specific settings can be configured according to actual needs.

[0064] It is understood that, in the embodiments of this application, in addition to using the above-mentioned methods to obtain population structure data within a sub-region, spatial interpolation and statistical downscaling algorithms or Geographic Information System (GIS) and spatial analysis algorithms can also be used to obtain population structure data within a sub-region. The specific settings can be made according to actual needs.

[0065] For example, in S202 above, when acquiring the satellite remote sensing image corresponding to the sub-region, see, for example, [the following text is missing]. Figure 7 As shown, Figure 7This is a schematic diagram illustrating an embodiment of the present application for obtaining a satellite remote sensing image corresponding to a sub-region. First, satellite remote sensing image tiles corresponding to the target city are obtained; then, these tiles are stitched and cropped to obtain a satellite remote sensing image corresponding to the target city; finally, based on the boundaries of the sub-region, a satellite remote sensing image corresponding to the sub-region is obtained from the satellite remote sensing image corresponding to the target city.

[0066] For example, when acquiring satellite remote sensing image tiles corresponding to a target city, the range of satellite remote sensing image tiles corresponding to the target city can be automatically identified based on the boundary information of the target city. Based on the high-resolution satellite image interface provided by Esri World Imagery, all satellite remote sensing image tiles containing the target city can be automatically downloaded according to the latitude and longitude covered by the regional boundary, thereby acquiring the satellite remote sensing image tiles corresponding to the target city.

[0067] For example, when stitching and cropping satellite remote sensing image tiles corresponding to a target city to obtain a satellite remote sensing image of the target city, the stitching stage can seamlessly combine adjacent satellite remote sensing image tiles into a large image covering the entire target city; in the cropping stage, based on the boundary information of the target city, the large image covering the entire target city can be precisely cropped. For the parts outside the boundary of the target city, a zero-fill operation is automatically performed to ensure that the final image is completely aligned and contains only information of the region of interest, thereby obtaining a satellite remote sensing image that conforms to the true boundary shape of the target city. The entire process is based on the boundary of the target city and adopts a highly integrated automated workflow. The entire acquisition process does not require manual intervention and provides reliable satellite remote sensing images for subsequent OD matrix generation.

[0068] It is understood that, in the embodiments of this application, in addition to using the above-mentioned methods to obtain satellite remote sensing images corresponding to sub-regions, methods such as direct cropping based on vector boundaries or automated and deep learning cropping can also be used to obtain satellite remote sensing images corresponding to sub-regions. The specific settings can be made according to actual needs.

[0069] For example, in S202 above, when obtaining the distribution data of points of interest within a sub-region, the OpenStreetMap (OSM) open application programming interface (API) can be used to query and download the data in real time based on the sub-region's boundary information. The API automatically counts the number of each type of point of interest, such as residences, schools, hospitals, shops, and office buildings, and categorizes and summarizes them by function type, thus obtaining the distribution data of points of interest within the sub-region. The entire process requires no manual intervention and can automatically adjust the query scope and classification logic according to the actual spatial structure of different regions, thereby ensuring the accuracy and universality of the distribution data of points of interest within the sub-region and providing reliable distribution data for subsequent OD matrix generation.

[0070] It is understood that, in the embodiments of this application, in addition to using the above-mentioned methods to obtain the distribution data of points of interest within a sub-region, commercial map API call algorithms or GIS software and spatial analysis algorithms can also be used to obtain the distribution data of points of interest within a sub-region. The specific settings can be made according to actual needs.

[0071] Based on the above description, for each sub-region after dividing the target city into regions, population structure data, corresponding satellite remote sensing images, and point-of-interest (POI) distribution data within the sub-region can be obtained by crawling publicly available global data. The system automatically extracts structured population features from the population structure data, high-dimensional semantic features from the satellite remote sensing images, and spatial distribution features of POIs from the POI distribution data. This allows for a comprehensive characterization of the target region's spatial functional attributes, based on the distances between multiple sub-regions and combining the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs corresponding to each sub-region. This serves as a guiding condition for generating the OD matrix, providing a basis for subsequent OD matrix generation.

[0072] Based on any of the above embodiments, for example, in S203 above, the specific implementation of generating the target OD matrix corresponding to the target city based on the distance between multiple sub-regions, the structured artificial features corresponding to each of the multiple sub-regions, the high-dimensional semantic features, and the spatial distribution features of interest points can be found below. Figure 8 The example shown.

[0073] Figure 8 This application provides a flowchart illustrating the process of generating a target OD matrix corresponding to a target city. For example, please refer to [link to relevant documentation]. Figure 8 As shown, the method may include: S801. The distance between multiple sub-regions, the structured artificial features, high-dimensional semantic features and spatial distribution features of interest points corresponding to each sub-region are input multiple times into the graph denoising diffusion model. The graph denoising diffusion model is used to randomly sample the flow of people in different areas of the target city multiple times to obtain multiple OD matrices.

[0074] The graph denoising diffusion model is a probabilistic model that does not rely on additional training. It can obtain published graph denoising diffusion model weights from the HuggingFace platform and automatically deploy them to the local environment, providing plug-and-play inference support and ensuring system usability and deployment efficiency. During deployment, an initial noise matrix is ​​first randomly generated using a standard normal distribution, representing an unknown, purely noisy OD (Original Distance Flow) state. Then, using a pre-trained graph denoising diffusion model, combined with the spatial semantic features of each sub-region, the model is guided to progressively denoise the initial noise, thereby recovering an OD matrix that conforms to the actual population flow patterns in the region. No model training or manual intervention is required; all inference steps are automated.

[0075] For example, when the distances between multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each sub-region are input into the graph-based denoising diffusion model, the distances between multiple sub-regions can be expressed as a distance matrix. The graph-based denoising diffusion model then randomly samples the population flow in different areas of the target city to obtain the randomly sampled OD matrix.

[0076] Since the graph denoising diffusion model is a probabilistic model, by inputting the distance between multiple sub-regions, the structured artificial features corresponding to each sub-region, the high-dimensional semantic features, and the spatial distribution features of interest points into the graph denoising diffusion model multiple times, the flow of people in different areas of the target city can be randomly sampled multiple times through the graph denoising diffusion model; among them, the OD matrix obtained after different random samplings is different.

[0077] After performing multiple random samplings of population flow patterns in different areas of the target city using a graph-based denoising diffusion model to obtain multiple OD matrices, the following S802 can be executed: S802. Generate the target OD matrix based on multiple OD matrices.

[0078] For example, in an embodiment of this application, when generating a target OD matrix based on multiple OD matrices, the average of the multiple OD matrices can be calculated to obtain an average OD matrix; and the average OD matrix can be determined as the target OD matrix.

[0079] As can be seen, in this embodiment of the application, when generating the target OD matrix corresponding to the target city, the distance between multiple sub-regions, the structured artificial features, high-dimensional semantic features and spatial distribution features of interest points corresponding to each sub-region are input into the graph denoising diffusion model multiple times. The graph denoising diffusion model performs multiple random samplings of the population flow in different areas of the target city to obtain multiple OD matrices. Based on the multiple OD matrices, the target OD matrix is ​​determined together. This can effectively enhance the stability and reliability of the output results, effectively reduce the uncertainty in the OD matrix generation process, and improve the robustness of the results.

[0080] Furthermore, after generating the target OD matrix for the target city, this target OD matrix can be further visualized. For example, see [link to example]. Figure 9 As shown, Figure 9 This is a schematic diagram showing the comparison between an OD matrix generated based on a graph denoising diffusion model and a real OD matrix, provided in an embodiment of this application. It can be seen that the OD matrix generated based on the graph denoising diffusion model provided in this embodiment is basically the same as the real OD matrix, and can well represent the flow of people between different areas in the city.

[0081] It should be noted that the OD matrix generation module provided in this application embodiment can be designed as a standardized service interface, which can be flexibly embedded into other urban computing systems or deployed as an independent component. It supports localized operation, parallel processing, and batch inference of large-scale urban areas, thereby greatly improving the efficiency and usability of OD matrix generation in practical applications.

[0082] To verify the advancement and effectiveness of the OD matrix generation method provided in this application, it can be compared with existing gravity models, radiation models, deep gravity models, and graph representation learning machine models. Specifically, the root mean square error (RMSE) and commuting common component (CPC) of each model can be calculated to compare and verify their advancement and effectiveness, thereby validating the advancement and effectiveness of this invention. RMSE measures the error between predicted and true values; a smaller value indicates a more accurate prediction. CPC measures the correlation between predicted and true values; a value closer to 1 indicates a better model fit.

[0083] in, Represents the first in multiple sub-regions Sub-regions Represents the first in multiple sub-regions Sub-regions Indicates the target city. Indicates the first generation generated by the model Sub-regions and the first OD matrix elements corresponding to each sub-region Indicates the first Sub-regions and the first The labels of the OD matrix elements corresponding to each sub-region.

[0084] Based on the above two formulas, the RMSE and CPC corresponding to the gravity model, radiation model, depth gravity model, graph representation learning machine model, and OD matrix generation method provided in this application can be calculated respectively, as shown in Table 1 below: Table 1 Model RMSE CPC Gravity model 101.0 0.240 Radiation model 211.4 0.323 Deep Gravity Model 157.0 0.359 Graph representation learning machine model 149.1 0.362 The technical solution provided in this application 72.6 0.485 As can be seen from Table 1 above, the OD matrix generation method provided in this application has a significant advantage in prediction accuracy. Its prediction results have the strongest correlation with the true values, making it the optimal OD matrix generation method among the aforementioned models.

[0085] The starting and ending point OD matrix generation apparatus provided in this application is described below. The starting and ending point OD matrix generation apparatus described below can be referred to in correspondence with the starting and ending point OD matrix generation method described above.

[0086] Figure 10 This application provides a schematic diagram of the structure of a start-endpoint OD matrix generation device. For example, please refer to [link to relevant documentation]. Figure 10 As shown, the start-end point OD matrix generation device 100 may include: The partitioning unit 1001 is used to partition the target city into multiple sub-regions in response to the target city information input by the user. The processing unit 1002 is configured to acquire, for each sub-region, population structure data within the sub-region, satellite remote sensing images corresponding to the sub-region, and point-of-interest distribution data within the sub-region; and extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of point-of-interest corresponding to the point-of-interest distribution data. The generation unit 1003 is used to generate a target OD matrix corresponding to the target city based on the distance between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each of the multiple sub-regions; wherein, the target OD matrix is ​​used to characterize the flow of people between different regions in the target city.

[0087] For example, in this embodiment of the application, the generation unit 1003 is used to generate a target OD matrix corresponding to the target city based on the distance between the plurality of sub-regions, the structured artificial features corresponding to each of the plurality of sub-regions, the high-dimensional semantic features, and the spatial distribution features of interest points, including: The distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each of the multiple sub-regions are input multiple times into a graph-based denoising diffusion model. The graph-based denoising diffusion model is used to randomly sample the flow of people in different areas of the target city multiple times to obtain multiple OD matrices. The target OD matrix is ​​generated based on the multiple OD matrices.

[0088] For example, in an embodiment of this application, the generation unit 1003 is used to generate the target OD matrix based on the plurality of OD matrices, including: The average OD matrix is ​​obtained by averaging the multiple OD matrices. The average OD matrix is ​​determined as the target OD matrix.

[0089] For example, in this embodiment of the application, the processing unit 1002 is used to extract high-dimensional semantic features corresponding to the satellite remote sensing image, including: The satellite remote sensing image is preprocessed to obtain a preprocessed target satellite remote sensing image; The target satellite remote sensing image is input into a multimodal visual model, and the high-dimensional semantic features are output through the multimodal visual model.

[0090] For example, in an embodiment of this application, the processing unit 1002 is used to acquire population structure data within the sub-region, including: Spatially overlay the target city with the raster pixels in the population raster dataset to determine the raster data covering the target city; Based on the boundary of the sub-region, the population structure data within the raster data falling into the sub-region from the raster data covering the target city are weighted and aggregated to obtain the population structure data within the sub-region.

[0091] For example, in an embodiment of this application, the processing unit 1002 is used to acquire satellite remote sensing images corresponding to the sub-region, including: Obtain satellite remote sensing image tiles corresponding to the target city; The satellite remote sensing image tiles corresponding to the target city are stitched together and cropped to obtain the satellite remote sensing image corresponding to the target city. Based on the boundary of the sub-region, the satellite remote sensing image corresponding to the sub-region is obtained from the satellite remote sensing image corresponding to the target city.

[0092] The start-end point OD matrix generation device provided in this application embodiment can execute the technical solution of the start-end point OD matrix generation method in any of the above embodiments. Its implementation principle and beneficial effects are similar to those of the start-end point OD matrix generation method. Please refer to the implementation principle and beneficial effects of the start-end point OD matrix generation method. It will not be repeated here.

[0093] Figure 11 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application, such as... Figure 11 As shown, the electronic device may include: a processor 1110, a communications interface 1120, a memory 1130, and a communications bus 1140, wherein the processor 1110, the communications interface 1120, and the memory 1130 communicate with each other through the communications bus 1140. The processor 1110 can call logical instructions in the memory 1130 to execute a method for generating an origin-endpoint OD matrix. This method includes: responding to user-inputted information about a target city, dividing the target city into multiple sub-regions; for each sub-region, acquiring population structure data, a corresponding satellite remote sensing image, and point-of-interest (POI) distribution data within the sub-region; extracting structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing image, and spatial distribution features of POIs corresponding to the POI distribution data; and generating a target OD matrix corresponding to the target city based on the distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs corresponding to each of the multiple sub-regions; wherein the target OD matrix is ​​used to characterize the population flow between different regions within the target city.

[0094] Furthermore, the logical instructions in the aforementioned memory 1130 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0095] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the origin-endpoint OD matrix generation method provided by the above methods. The method includes: responding to user-inputted information about a target city, dividing the target city into multiple sub-regions; for each sub-region, acquiring population structure data, satellite remote sensing images corresponding to the sub-region, and point-of-interest (POI) distribution data within the sub-region; and extracting structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of POIs corresponding to the POI distribution data; and generating a target OD matrix corresponding to the target city based on the distance between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs corresponding to each of the multiple sub-regions; wherein the target OD matrix is ​​used to characterize the population flow between different regions in the target city.

[0096] In another aspect, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for generating origin-endpoint OD matrices provided by the methods described above. This method includes: in response to user-inputted information about a target city, dividing the target city into multiple sub-regions; for each sub-region, acquiring population structure data, a satellite remote sensing image corresponding to the sub-region, and point-of-interest (POI) distribution data within the sub-region; and extracting structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing image, and spatial distribution features of POIs corresponding to the POI distribution data; and generating a target OD matrix corresponding to the target city based on the distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of POIs corresponding to each of the multiple sub-regions; wherein the target OD matrix is ​​used to characterize the population flow between different regions in the target city.

[0097] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0098] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for generating a start-endpoint OD matrix, characterized in that, include: In response to the target city information input by the user, the target city is divided into regions to obtain multiple sub-regions; For each sub-region, acquire population structure data within the sub-region, satellite remote sensing images corresponding to the sub-region, and point-of-interest (POI) distribution data within the sub-region; and extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of POIs corresponding to the POI distribution data. Based on the distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of points of interest corresponding to each of the multiple sub-regions, a target OD matrix corresponding to the target city is generated; wherein, the target OD matrix is ​​used to characterize the population flow between different regions in the target city.

2. The method according to claim 1, characterized in that, The step of generating the target OD matrix corresponding to the target city based on the distances between the multiple sub-regions, the structured artificial features corresponding to each of the multiple sub-regions, the high-dimensional semantic features, and the spatial distribution features of interest points includes: The distances between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each of the multiple sub-regions are input multiple times into a graph-based denoising diffusion model. The graph-based denoising diffusion model is used to randomly sample the flow of people in different areas of the target city multiple times to obtain multiple OD matrices. The target OD matrix is ​​generated based on the multiple OD matrices.

3. The method according to claim 2, characterized in that, The step of generating the target OD matrix based on the plurality of OD matrices includes: The average OD matrix is ​​obtained by averaging the multiple OD matrices. The average OD matrix is ​​determined as the target OD matrix.

4. The method according to any one of claims 1-3, characterized in that, Extracting high-dimensional semantic features corresponding to the satellite remote sensing image, including: The satellite remote sensing image is preprocessed to obtain a preprocessed target satellite remote sensing image; The target satellite remote sensing image is input into a multimodal visual model, and the high-dimensional semantic features are output through the multimodal visual model.

5. The method according to any one of claims 1-3, characterized in that, The acquisition of population structure data within the sub-region includes: Spatially overlay the target city with the raster pixels in the population raster dataset to determine the raster data covering the target city; Based on the boundary of the sub-region, the population structure data within the raster data falling into the sub-region from the raster data covering the target city are weighted and aggregated to obtain the population structure data within the sub-region.

6. The method according to any one of claims 1-3, characterized in that, Obtaining the satellite remote sensing image corresponding to the sub-region includes: Obtain satellite remote sensing image tiles corresponding to the target city; The satellite remote sensing image tiles corresponding to the target city are stitched together and cropped to obtain the satellite remote sensing image corresponding to the target city. Based on the boundary of the sub-region, the satellite remote sensing image corresponding to the sub-region is obtained from the satellite remote sensing image corresponding to the target city.

7. A device for generating a start-endpoint OD matrix, characterized in that, include: A partitioning unit is used to divide the target city into multiple sub-regions in response to user-inputted target city information. The processing unit is configured to acquire population structure data, satellite remote sensing images corresponding to the sub-region, and point-of-interest distribution data within the sub-region for each sub-region; and extract structured population features corresponding to the population structure data, high-dimensional semantic features corresponding to the satellite remote sensing images, and spatial distribution features of points of interest corresponding to the point-of-interest distribution data. The generation unit is used to generate a target OD matrix corresponding to the target city based on the distance between the multiple sub-regions, the structured artificial features, high-dimensional semantic features, and spatial distribution features of interest points corresponding to each of the multiple sub-regions; wherein, the target OD matrix is ​​used to characterize the population flow between different regions in the target city.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the start-endpoint OD matrix generation method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the start-endpoint OD matrix generation method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the start-endpoint OD matrix generation method as described in any one of claims 1 to 6.

Images