Method and system for updating residential place information in multi-mode crowd-sourced space-time vector product
By preprocessing and multi-dimensional feature extraction of mass source space-time vector data, and using heterogeneous machine learning models for data matching and update, the automatic fusion and real-time update of mass source heterogeneous data is solved, and efficient and accurate residents are achieved Data updates reduce costs and resource consumption.
Patent Information
- Application Number
- CN202411425009.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-12
- Publication Date
- 2025-01-07
AI Technical Summary
The existing technology is difficult to realize the automatic fusion and real-time update of mass-source heterogeneous spatiotemporal data, resulting in time delays in data updates and cannot promptly reflect changes in the geographical environment. Especially in dense urban environments, the sensitivity is insufficient, affecting urban planning and decision-making.
By preprocessing the multi-source spatiotemporal vector data, extracting multi-dimensional spatiotemporal features, using heterogeneous machine learning models and machine ensemble learning methods to match and update data, and using multi-pattern matching algorithms and incremental update mechanisms to achieve efficient data integration and real-time response.
It significantly improves the timeliness and accuracy of residential data, reduces manual intervention, optimizes system resource consumption, improves data processing efficiency, and reduces costs.
Smart Images
Figure CN119271682A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of spatial data processing technology, and in particular to a method and system for updating settlement information in multi-source spatiotemporal vector products under multiple modes. Background Art
[0002] Crowd-source spatiotemporal datasets, as open resources of geospatial information, are collected and provided by the public, and have significant advantages such as large data volume, strong timeliness, rich information and low cost. These datasets have become a hot topic of research at home and abroad in the field of geographic information science, especially in the application and automatic update of product sets. Settlement vector data, as an important part of basic geographic information data, plays an indispensable supporting role in urban planning, social management, economic development and other fields. Although crowd-source datasets have significant advantages in the field of geographic information science, the differences in application requirements between different datasets lead to inconsistencies in data content and accuracy. The existing technology lacks a unified data framework to integrate and fuse heterogeneous spatiotemporal data from crowd-sources, so as to give full play to the advantages of different versions of data. In addition, the continuous update characteristics of crowd-source open spatial data platforms such as OSM make traditional spatial data update processes face challenges. These processes usually rely on periodic manual surveys or remote sensing data collection, resulting in time delays in data updates and failure to reflect actual changes in the geographical environment in a timely manner.
[0003] In order to achieve automatic fusion of multi-source data, automatic matching algorithms are the key. Existing algorithms mainly perform automatic matching on data from a single data source or the same architecture, but the effect is not ideal when processing multi-source heterogeneous data such as OSM. Vector data matching methods, including spatial matching of entities with the same name based on area similarity, vector feature matching based on spatial feature codes, vectorized buffer matching, and triple-based matching methods, all have certain limitations. For example, methods based on area similarity may lead to false detection or missed detection due to differences in data quality; and methods based on triples are not efficient in the face of large data sets and are difficult to meet real-time requirements. The irregular shape, small area, and slight changes of residential vector data make existing matching algorithms insensitive in dense urban environments, and may ignore geographical changes that have an important impact on urban planning and decision-making. Summary of the invention
[0004] To this end, the present invention provides a method and system for updating settlement information in multi-source spatiotemporal vector products under multi-mode, so as to solve the problem of automatic matching and updating of settlement information in multi-source vector data under long time series.
[0005] According to the design scheme provided by the present invention, on the one hand, a method for updating settlement information in multi-source spatiotemporal vector products in a multi-mode is provided, comprising:
[0006] Preprocessing the multi-source spatiotemporal vector data to obtain the data of each residential area in the multi-source spatiotemporal vector data, wherein the preprocessing includes data cleaning processing, data conversion processing and data standardization processing;
[0007] Extracting multi-dimensional spatiotemporal features of each residential plot data, wherein the multi-dimensional spatiotemporal features include time features, spatial features, and metadata features;
[0008] Calculate the similarity between each residential area plot based on multi-dimensional spatiotemporal characteristics, and use the similarity to match and align residential areas in different time series and / or different data sources;
[0009] The surface feature similarity of residential plot data and the matching alignment results of residential plots are used as training data, and the specified heterogeneous machine learning model is trained using the training data. An integrated learning model is constructed based on the heterogeneous machine learning model and using a machine integrated learning method, so as to obtain the final matching output through a voting mechanism or a stacking method in the machine integrated learning.
[0010] The integrated learning model is used to match the residential data to be detected. If the matching detection result exceeds the residential feature update threshold, the residential data to be detected will be updated.
[0011] As a method for updating settlement information in a crowd-source space-time vector product under multiple modes of the present invention, further, preprocessing the crowd-source space-time vector data comprises:
[0012] Perform data cleaning on crowd-source spatiotemporal vector data to remove invalid and erroneous data;
[0013] The vector data after data cleaning is standardized, and the standardization of each vector data is ensured by unifying the data format and coordinate system.
[0014] As a method for updating residential area information in a multi-source spatiotemporal vector product under multiple modes of the present invention, further, extracting multi-dimensional spatiotemporal features of each residential area plot data includes:
[0015] Extracting relevant time features of residential land plot data, wherein the relevant time features include land plot update timestamp, land plot feature planning time and land plot construction time;
[0016] Extracting relevant spatial features of residential plot data, the relevant spatial features include plot building area, building boundary shape, plot location, plot direction, topological relationship of features, feature corner information and order of points on the boundary of surface elements;
[0017] Extract relevant metadata features of residential land parcel data, wherein the relevant metadata features include land parcel data source, land parcel data source credibility and land parcel code.
[0018] As a method for updating residential area information in a multi-source spatiotemporal vector product under multiple modes of the present invention, further, the similarity between each residential area plot is calculated based on multi-dimensional spatiotemporal features, including:
[0019] Setting feature matching indicators for judging whether residential land plot elements need to be matched and updated, the feature matching indicators include direction matching, position matching, topology matching, overlapping area matching and shape matching;
[0020] The matching degree of each indicator between residential plots in different time series and / or different data sources is calculated in the order of direction, position, topology, overlapping area and shape, and the total matching degree of residential plots in different data sources is obtained by weighted summation.
[0021] As a method for updating settlement information in multi-source spatiotemporal vector products under multi-mode of the present invention, further, using similarity to match and align settlement plots under different time series and / or different data sources, it includes:
[0022] The residential plot with the highest total matching degree is selected, and the total matching degree of the residential plot is compared with the preset threshold. If the total matching degree is greater than the preset threshold, the residential plots of different time series and / or different data sources currently matched are matched and aligned, and the matching degree of each feature matching indicator is recorded.
[0023] As a method for updating settlement information in multi-source spatiotemporal vector products under multiple modes of the present invention, further, using training data to train a specified heterogeneous machine learning model, it includes:
[0024] Constructing a heterogeneous machine learning model, wherein the heterogeneous machine learning model includes: a decision tree, a SVM, and a BP neural network;
[0025] The matching degree of each feature matching index and the multi-dimensional spatiotemporal characteristics of residential plots are used as the input of each heterogeneous machine learning model, and the heterogeneous machine learning model is iteratively trained to learn the weight of each feature matching index, the threshold of a single feature matching index, and the total matching judgment matching threshold;
[0026] Until the model converges, obtain the converged weight matrix and store it.
[0027] As a method for updating settlement information in multi-source spatiotemporal vector products under multi-mode of the present invention, further, obtaining the final matching output through a voting mechanism or a stacking method in machine ensemble learning also includes:
[0028] In machine ensemble learning, the feature matching index cut-off thresholds of each heterogeneous machine learning model are recorded, and the total matching judgment matching threshold is obtained based on the feature matching index cut-off thresholds;
[0029] The residential ground feature matching results are graded according to the size of the total matching degree and the total matching judgment matching threshold.
[0030] On the other hand, the present invention also provides a system for updating settlement information in multi-source spatiotemporal vector products under multiple modes, comprising: a data processing module, a feature extraction module, a matching alignment module, an integrated learning module and a target matching module, wherein:
[0031] A data processing module is used to pre-process the multi-source spatiotemporal vector data to obtain the data of each residential area in the multi-source spatiotemporal vector data, wherein the pre-processing includes data cleaning processing, data conversion processing and data standardization processing;
[0032] A feature extraction module is used to extract multi-dimensional spatiotemporal features of each residential plot data, wherein the multi-dimensional spatiotemporal features include time features, spatial features and metadata features;
[0033] A matching and alignment module is used to calculate the similarity between each residential area plot based on multi-dimensional spatiotemporal features, and use the similarity to match and align residential areas plots in different time series and / or different data sources;
[0034] An integrated learning module is used to use the surface feature similarity of residential plot data and the matching alignment results of residential plots as training data, use the training data to train a specified heterogeneous machine learning model, and build an integrated learning model based on the heterogeneous machine learning model and using a machine integrated learning method to obtain the final matching output through a voting mechanism or a stacking method in the machine integrated learning;
[0035] The target matching module is used to use the integrated learning model to match the residential data to be detected. If the matching detection result exceeds the residential feature update threshold, the residential data to be detected will be updated.
[0036] Beneficial effects of the present invention:
[0037] The present invention realizes the efficient integration of multi-source data through a unified spatiotemporal data model and an automated adaptation process. At the same time, it introduces an incremental update mechanism to ensure that the data set can quickly respond to real-time changes in the geographical environment, significantly improving the timeliness and accuracy of residential data. The multi-mode matching algorithm significantly improves the automatic matching accuracy of residential information and reduces manual intervention. At the same time, by only updating the changed parts of the data set, the system resource consumption is optimized, the data processing efficiency is improved, and the cost is reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 The figure is a schematic diagram of the process of updating the settlement information in the multi-source space-time vector product in the embodiment. DETAILED DESCRIPTION
[0039] In order to make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention is further described in detail below in conjunction with the accompanying drawings and technical solutions.
[0040] Embodiments of the present invention, see Figure 1 As shown, a method for updating settlement information in multi-mode multi-source space-time vector products is provided, comprising:
[0041] S101. Preprocess the multi-source space-time vector data to obtain the data of each residential plot in the multi-source space-time vector data, wherein the preprocessing includes data cleaning processing, data conversion processing and data standardization processing.
[0042] The crowdsourced residential vector data was cleaned and standardized, including removing duplicate records, correcting errors, and filling missing values, and a new data set was constructed for comparison with the original data set. The original data was cleaned, standardized, and converted to eliminate inconsistencies and errors; the coordinates and coordinate systems of each data were unified to ensure that the same plot was roughly located in a similar position.
[0043] In addition, data preprocessing can also be designed to include normalized attribute fields to facilitate subsequent feature extraction and analysis.
[0044] S102, extracting multi-dimensional spatiotemporal features of each residential plot data, wherein the multi-dimensional spatiotemporal features include time features, space features, and metadata features.
[0045] It mainly constructs the characteristics of the three dimensions of time, space, and metadata of each block of data. Point, line, and surface features represent different geographic information elements. For example, point features usually represent single location markers, such as connection points, road signs, traffic lights, traffic signs, and road marking lights. These elements are displayed separately as points on the map. Line features represent geographic entities with a certain length, such as roads, ferry features, regional boundaries, and road marking lights. They are displayed as lines on the map to indicate the direction and path of the geographic entities. Surface features represent areas with a certain area, such as regions and land cover lights. They are stored as discrete point strings and displayed as closed graphics when rendered to indicate the scope and area of geographic entities.
[0046] For temporal features, the main information collected includes the update timestamp of the data, the planning, construction, maintenance, and update time information of the represented features; for spatial features, it includes intuitive geometric attributes, such as the area and boundary shape complexity of the building, as well as more complex spatial features, such as location information, directionality, topological relationship between features, corner information, the order of points on the boundary of surface elements, adjacent elements, and feature point sets (visual feature points of surface elements or vertices with larger boundary curvature are used as feature points); metadata features include relevant information of the data source, including relevant information of the data source, credibility of the data source, unique code of the plot, etc.
[0047] Through feature extraction, a feature library of residential vector data is constructed, denoted as T, and the feature library of original data is denoted as B. This provides rich input data for subsequent change detection.
[0048] S103, calculating the similarity between each residential plot based on the multi-dimensional spatiotemporal features, and using the similarity to match and align the residential plots in different time series and / or different data sources.
[0049] Based on the extracted features, the direction similarity, position similarity, topological similarity, overlapping area similarity, shape similarity, etc. between each residential plot are calculated, and the final total similarity is calculated in combination with the credibility of the data source. The total similarity is used to match the data of the same plot of the same residential area in different time series and from different sources.
[0050] Specifically, the similarity between residential plots is calculated based on multi-dimensional spatiotemporal features, which can be designed to include:
[0051] Setting feature matching indicators for judging whether residential land plot elements need to be matched and updated, the feature matching indicators include direction matching, position matching, topology matching, overlapping area matching and shape matching;
[0052] The matching degree of each indicator between residential plots in different time series and / or different data sources is calculated in the order of direction, position, topology, overlapping area and shape, and the total matching degree of residential plots in different data sources is obtained by weighted summation.
[0053] The residential plot with the highest total matching degree is selected, and the total matching degree of the residential plot is compared with the preset threshold. If the total matching degree is greater than the preset threshold, the residential plots of different time series and / or different data sources currently matched are matched and aligned, and the matching degree of each feature matching indicator is recorded.
[0054] By setting indicators of different dimensions, we can determine whether the two sets of features are matched and updated. In the feature matching stage, we select any vector surface in T and compare it with the vector data in the original feature library B, and calculate the matching degree of the two sets of features from different dimensions. The detailed steps are as follows:
[0055] (1) In order to reduce the amount of calculation, the extended buffer method is used to obtain matching candidate objects.
[0056] (2) Calculate the matching degree of each indicator in the order of direction, position, topology, overlapping area, and shape, and store each matching degree in a collection as an input parameter for the subsequent machine learning model. The meaning and calculation formula of each indicator matching degree are as follows:
[0057] 1) Direction matching: Direction matching is defined as the degree of similarity between the two residential objects to be matched in terms of overall orientation. The absolute value of the angle difference between the diagonal lines connected to the lower left corner vertex of the circumscribed rectangle is used as a tool to measure the direction matching. dir for
[0058] S dir =1-sin(|θ|)
[0059] 2) Position matching: The centroid of the settlement outline polygon is used to represent the location of the settlement, and the specific coordinates of the centroid are determined by the template accumulation method. Then the position matching of two settlement objects can be represented by the Euclidean distance of the centroid. Assuming that the maximum tolerable centroid distance is L, the range of the centroid distance l is [0, L], and the position matching degree S is defined as loc for
[0060] S loc =(Ll) / L
[0061] 3) Topological matching: The matching of two residential plots is mainly expressed by calculating the Hausdorff distance matching of the two plots. First, the boundary points of the two surface elements are extracted to form point sets A and B. For each point ai in set A, calculate its distance to all points in point set B, record the minimum value as Lai, and record the maximum value in Lai as LA; conversely, perform the same operation on B to obtain LB. The larger value of LA and LB can be recorded as the Hausdorff distance LH of the two plots. The maximum tolerance distance LL can be set, then the topological matching degree S of the two plots top for
[0062] S top =(LL-LH) / LL
[0063] 4) Overlap area matching: Assume T a ,B bare two face elements in the T and B datasets respectively, then the matching degree S between the two face elements sim The calculation formula is as follows:
[0064]
[0065] 5) Shape matching: Shape is the most intuitive and accurate means for human eyes to distinguish objects. This method divides the polygon into n equal segments, takes the distance from each equal segment point to the centroid as the value of the shape description function, and uses a matching starting point P0 to any point P on the boundary. i The arc length l i As the parameter of the shape description function, the absolute distance between the shape description functions is used to represent the shape matching, then the shape phase S f The calculation formula is
[0066]
[0067] 6) Total matching degree: The weighted sum of each matching degree is used as the total matching degree T of the boundary points in different data sets, where ω X The weight of each matching degree, R is the data credibility.
[0068] T=(ω dir S dir +ω loc S loc +ω top S top +ω sim S sim +ω f S f )×R
[0069] (3) Calculate the total matching degree of the remaining matching objects, and compare the object with the highest matching degree with the threshold of the total matching degree. If it is greater than the threshold, it is determined to be a matching object and the matching work is completed. At the same time, the similarity of each indicator is recorded.
[0070] (4) Repeat (1)-(3) to complete the matching of the target image to the source image.
[0071] S104. Use the surface feature similarity of residential plot data and the matching alignment results of residential plots as training data, use the training data to train a specified heterogeneous machine learning model, and build an integrated learning model based on the heterogeneous machine learning model and using a machine ensemble learning method to obtain the final matching output through a voting mechanism or a stacking method in machine ensemble learning.
[0072] Specifically, the specified heterogeneous machine learning model is trained using training data, which can be designed to include:
[0073] Constructing a heterogeneous machine learning model, wherein the heterogeneous machine learning model includes: a decision tree, a SVM, and a BP neural network;
[0074] The matching degree of each feature matching index and the multi-dimensional spatiotemporal characteristics of residential plots are used as the input of each heterogeneous machine learning model, and the heterogeneous machine learning model is iteratively trained to learn the weight of each feature matching index, the threshold of a single feature matching index, and the total matching judgment matching threshold;
[0075] Until the model converges, obtain the converged weight matrix and store it.
[0076] The feature matching index cut-off thresholds of each heterogeneous machine learning model can be recorded in machine ensemble learning, and the total matching judgment matching threshold can be obtained based on the feature matching index cut-off thresholds; the residential ground feature matching results can be graded according to the size of the total matching degree and the total matching judgment matching threshold.
[0077] Construct different machine learning models, and use the corrected matching results and the matching degree of each indicator as the training data input of the model. The model iteratively learns the weight of each indicator, the cut-off threshold of a single indicator, and the overall matching judgment matching threshold. Until a converged weight matrix is obtained, the weight matrix is finally stored according to the matching scenario.
[0078] To ensure the accuracy of the test results, cartographers checked the matching results, focusing on checking and evaluating the residential objects with a total matching value in the range of (0.7, 0.9), and correcting any mismatches and missed matches.
[0079] By combining the prediction results of multiple basic learning models, the integrated model can reduce the risk of overfitting of a single model and improve the ability to identify complex change patterns. In addition, integrated learning can also optimize the final prediction output through voting mechanisms or stacking methods.
[0080] Integrate different machine learning models to build an integrated learning model, learn and record the cut-off threshold of each model, and form the final overall matching judgment matching threshold. The matching results are graded according to the size of the overall matching degree. For example, assuming that the total matching degree is greater than 0.8, it can be considered a match, then it can be divided into 4 levels (0.9, 1) as a reliable matching interval; (0.8, 0.9) as a judgment matching interval; (0.7, 0.8) as a judgment mismatch interval; (0, 0.7) as a determined mismatch interval.
[0081] S105. Use the integrated learning model to perform matching detection on the residential area data to be detected. If the matching detection result exceeds the residential area element update threshold, the residential area data to be detected is updated.
[0082] An integrated learning model is used to construct an update threshold for residential surface features. This threshold defines whether the feature needs to be updated. Once the model detection result exceeds this threshold, the system will automatically trigger the update process to ensure the timeliness and accuracy of the data set. Object update operations can be divided into creation, deletion, geometric modification, and attribute modification. For newly added or disappeared objects, the creation or deletion operation can be used directly; for objects with changes in geometric shape or attributes, geometric modification or attribute modification is performed. In the case of object merging, decomposition, and aggregation, the original object can be deleted and a matching target object can be created for processing.
[0083] Furthermore, based on the above method, an embodiment of the present invention also provides a system for updating settlement information in multi-source spatiotemporal vector products under multiple modes, comprising: a data processing module, a feature extraction module, a matching alignment module, an integrated learning module and a target matching module, wherein:
[0084] A data processing module is used to pre-process the multi-source spatiotemporal vector data to obtain the data of each residential area in the multi-source spatiotemporal vector data, wherein the pre-processing includes data cleaning processing, data conversion processing and data standardization processing;
[0085] A feature extraction module is used to extract multi-dimensional spatiotemporal features of each residential plot data, wherein the multi-dimensional spatiotemporal features include time features, spatial features and metadata features;
[0086] A matching and alignment module is used to calculate the similarity between each residential area plot based on multi-dimensional spatiotemporal features, and use the similarity to match and align residential areas plots in different time series and / or different data sources;
[0087] An integrated learning module is used to use the surface feature similarity of residential plot data and the matching alignment results of residential plots as training data, use the training data to train a specified heterogeneous machine learning model, and build an integrated learning model based on the heterogeneous machine learning model and using a machine integrated learning method to obtain the final matching output through a voting mechanism or a stacking method in the machine integrated learning;
[0088] The target matching module is used to use the integrated learning model to match the residential data to be detected. If the matching detection result exceeds the residential feature update threshold, the residential data to be detected will be updated.
[0089] Unless otherwise specifically stated, the relative steps, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the present invention.
[0090] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0091] The units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation is not considered to be beyond the scope of the present invention.
[0092] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits, and accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present invention is not limited to any specific form of combination of hardware and software.
[0093] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims
1. A method for updating settlement information in multi-source spatiotemporal vector products under multiple modes, characterized in that: Include: Preprocessing the multi-source spatiotemporal vector data to obtain the data of each residential area in the multi-source spatiotemporal vector data, wherein the preprocessing includes data cleaning processing, data conversion processing and data standardization processing; Extracting multi-dimensional spatiotemporal features of each residential plot data, wherein the multi-dimensional spatiotemporal features include time features, spatial features, and metadata features; Calculate the similarity between each residential area plot based on multi-dimensional spatiotemporal characteristics, and use the similarity to match and align residential areas in different time series and / or different data sources; The surface feature similarity of residential plot data and the matching alignment results of residential plots are used as training data, and the specified heterogeneous machine learning model is trained using the training data. An integrated learning model is constructed based on the heterogeneous machine learning model and using a machine integrated learning method, so as to obtain the final matching output through a voting mechanism or a stacking method in the machine integrated learning. The integrated learning model is used to match the residential data to be detected. If the matching detection result exceeds the residential feature update threshold, the residential data to be detected will be updated.
2. The method for updating settlement information in multi-mode multi-source space-time vector products according to claim 1, characterized in that: Preprocessing of multi-source spatiotemporal vector data, including: Perform data cleaning on crowd-source spatiotemporal vector data to remove invalid and erroneous data; The vector data after data cleaning is standardized, and the standardization of each vector data is ensured by unifying the data format and coordinate system.
3. The method for updating settlement information in multi-mode multi-source space-time vector products according to claim 1, characterized in that: Extract the multi-dimensional spatiotemporal features of each residential area data, including: Extracting relevant time features of residential land plot data, wherein the relevant time features include land plot update timestamp, land plot feature planning time and land plot construction time; Extracting relevant spatial features of residential plot data, the relevant spatial features include plot building area, building boundary shape, plot location, plot direction, topological relationship of features, feature corner information and order of points on the boundary of surface elements; Extract relevant metadata features of residential land parcel data, wherein the relevant metadata features include land parcel data source, land parcel data source credibility and land parcel code.
4. The method for updating settlement information in multi-mode multi-source spatio-temporal vector products according to claim 1, characterized in that: The similarity between each residential area is calculated based on multi-dimensional spatiotemporal features, including: Setting feature matching indicators for judging whether residential land plot elements need to be matched and updated, the feature matching indicators include direction matching, position matching, topology matching, overlapping area matching and shape matching; The matching degree of each indicator between residential plots in different time series and / or different data sources is calculated in the order of direction, position, topology, overlapping area and shape, and the total matching degree of residential plots in different data sources is obtained by weighted summation.
5. The method for updating settlement information in multi-mode multi-source space-time vector products according to claim 4, characterized in that: Use similarity to match and align residential plots from different time series and / or different data sources, including: The residential plot with the highest total matching degree is selected, and the total matching degree of the residential plot is compared with the preset threshold. If the total matching degree is greater than the preset threshold, the residential plots of different time series and / or different data sources currently matched are matched and aligned, and the matching degree of each feature matching indicator is recorded.
6. The method for updating settlement information in multi-mode multi-source spatio-temporal vector products according to claim 5, characterized in that: Use training data to train specified heterogeneous machine learning models, including: Constructing a heterogeneous machine learning model, wherein the heterogeneous machine learning model includes: a decision tree, a SVM, and a BP neural network; The matching degree of each feature matching index and the multi-dimensional spatiotemporal characteristics of residential plots are used as the input of each heterogeneous machine learning model, and the heterogeneous machine learning model is iteratively trained to learn the weight of each feature matching index, the threshold of a single feature matching index, and the total matching judgment matching threshold; Until the model converges, obtain the converged weight matrix and store it.
7. The method for updating settlement information in multi-mode multi-source space-time vector products according to claim 6, characterized in that: In machine ensemble learning, the final matching output is obtained through a voting mechanism or a stacking method, which also includes: In machine ensemble learning, the feature matching index cut-off thresholds of each heterogeneous machine learning model are recorded, and the total matching judgment matching threshold is obtained based on the feature matching index cut-off thresholds; The residential ground feature matching results are graded according to the size of the total matching degree and the total matching judgment matching threshold.
8. A system for updating settlement information in multi-source spatiotemporal vector products under multiple modes, characterized in that: It includes: data processing module, feature extraction module, matching alignment module, integrated learning module and target matching module, among which, A data processing module is used to pre-process the multi-source spatiotemporal vector data to obtain the data of each residential area in the multi-source spatiotemporal vector data, wherein the pre-processing includes data cleaning processing, data conversion processing and data standardization processing; A feature extraction module is used to extract multi-dimensional spatiotemporal features of each residential plot data, wherein the multi-dimensional spatiotemporal features include time features, spatial features and metadata features; A matching and alignment module is used to calculate the similarity between each residential area plot based on multi-dimensional spatiotemporal features, and use the similarity to match and align residential areas plots in different time series and / or different data sources; An integrated learning module is used to use the surface feature similarity of residential plot data and the matching alignment results of residential plots as training data, use the training data to train a specified heterogeneous machine learning model, and build an integrated learning model based on the heterogeneous machine learning model and using a machine integrated learning method to obtain the final matching output through a voting mechanism or a stacking method in the machine integrated learning; The target matching module is used to use the integrated learning model to match the residential data to be detected. If the matching detection result exceeds the residential feature update threshold, the residential data to be detected will be updated.
9. An electronic device, characterized in that: include: at least one processor, and a memory coupled to the at least one processor; The memory stores a computer program, and the computer program can be executed by the at least one processor to implement the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 7 can be implemented.