A method for building an artificial intelligence model for geographic information integration

Abstract

The present invention discloses a method for constructing an artificial intelligence model for geographic information integration, which relates to the technical field of virtual model construction. By constructing a backbone network search space with different network architectures, and collecting corresponding sample data according to the type of geographic information to be integrated, the corresponding network architecture in the backbone network search space is trained in a targeted manner, thereby obtaining a network architecture capable of processing different types of data, realizing corresponding processing of different types of data contained in the geographic information, and improving the data compatibility of geographic information processing; when processing geographic information, the network architecture in the backbone network search space can be selectively activated according to the data type contained in the geographic information to be processed, thereby improving the adaptability between the network architecture and the geographic information, and further improving data processing efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual model construction, and in particular to a method for constructing an artificial intelligence model for geographic information integration. Background Art

[0002] In today's digital age, geographic information is becoming increasingly important and is widely used in a wide range of fields, including urban planning, disaster warning, and resource management. However, the sources of geographic information are complex and diverse, encompassing satellite remote sensing imagery, ground-based monitoring data, and mapping information. Data from these sources vary significantly in format, accuracy, and semantics, posing significant challenges to integrating geographic information.

[0003] Traditional geographic information integration methods rely primarily on manually formulated rules and algorithms, which are not only inefficient but also difficult to handle large-scale, high-dimensional, and complex geographic data. Furthermore, these methods lack a deep understanding of the inherent connections and semantics of different types of geographic information, making it impossible to fully tap the potential value of geographic data. Therefore, we propose a method for building an artificial intelligence model for geographic information integration. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for constructing an artificial intelligence model for geographic information integration.

[0005] The purpose of the present invention can be achieved by the following technical solution: A method for constructing an artificial intelligence model for geographic information integration, comprising:

[0006] Obtain geographic information integration requirements, collect corresponding sample data based on geographic information integration requirements, and build a backbone network search space that deploys several network architectures;

[0007] The collected sample data is divided into factors, and the corresponding sample sub-datasets are obtained according to the factor division results. The network architecture in the backbone network search space is trained using the obtained sample sub-datasets;

[0008] Establish an associated database for each trained network architecture and import the corresponding sample sub-datasets into the associated database;

[0009] Input the geographic information that needs to be integrated, and selectively activate the network architecture within the backbone network search space based on the input geographic information, and complete the extraction of geographic information features through the activated network architecture;

[0010] The extracted geographic information features are verified for accuracy, and the input geographic information is imported into the corresponding database to update the sample sub-dataset in the database.

[0011] Furthermore, the process of collecting corresponding sample data according to geographic information integration requirements and constructing a backbone network search space that deploys several network architectures includes:

[0012] The geographic information integration requirements include the data types, time series, spatial series and the correlation between different data of the geographic information expected to be involved;

[0013] Construct a backbone network search space, and arrange the required network architectures in the backbone network search space according to the requirements of geographic information integration, and initialize each network architecture;

[0014] According to the characteristics of each network architecture, the network architecture is associated with the data type, and a type of activation node corresponding to the network architecture is set. All network architectures with a type of activation node are recorded as feature extraction architectures;

[0015] According to the correlation between time series, spatial series and different data, the corresponding network architectures are associated respectively, and two types of activation nodes corresponding to the network architectures are set. The architecture with the two types of activation nodes is recorded as the correlation extraction architecture;

[0016] Collect geographic information sample data of the target area according to the needs of geographic information integration.

[0017] Furthermore, the collected sample data is divided into factors, and the process of obtaining corresponding sample sub-datasets according to the factor division results includes:

[0018] Divide the collected geographic information sample data according to type factors to obtain geographic information data of different data types;

[0019] Aggregate geographic information data of the same type as corresponding sample sub-datasets;

[0020] Set a corresponding type label for each sample sub-dataset and associate the type label with the sample sub-dataset;

[0021] Each sample sub-dataset is divided into corresponding training sets, validation sets and test sets.

[0022] Furthermore, the process of training the network architecture in the backbone network search space using the obtained sample sub-dataset includes:

[0023] Import the training set, validation set, and test set corresponding to each obtained sample sub-dataset into the backbone network search space;

[0024] Read the type labels associated with the sample sub-datasets corresponding to the imported training set, validation set, and test set;

[0025] activating a corresponding class of activation nodes according to the type label, and after activating the corresponding class of activation nodes, importing the training set, the validation set, and the test set into the feature extraction architecture corresponding to the class of activation nodes to train the feature extraction architecture;

[0026] After completing the training of the feature extraction architecture, the corresponding sample sub-dataset is input into the corresponding feature extraction architecture to extract the sample data features in the sample sub-dataset;

[0027] Summarize the sample data features extracted by each feature extraction architecture to obtain a sample feature set, and divide the sample feature set into corresponding training set, validation set, and test set;

[0028] Feature labels are set for the obtained sample feature sets, and the second-class activation nodes are activated according to the generated feature labels. After the second-class activation nodes are activated, the training set, validation set, and test set divided by the sample feature sets are input into the correlation extraction architecture associated with the second-class activation nodes to train the correlation extraction architecture.

[0029] Furthermore, the process of establishing a database associated with each trained network architecture and importing the corresponding sample sub-datasets into the associated database includes:

[0030] Constructing a database, and constructing a corresponding class of databases according to the type labels corresponding to each sample data subset within the database, and associating the constructed class of databases with the type labels;

[0031] Importing the sample data in the sample data subset into a corresponding type of database;

[0032] A corresponding second-class database is constructed in the database according to the feature labels corresponding to the sample feature set, and the sample data features in the sample feature set are imported into the second-class database.

[0033] Furthermore, the geographic information to be integrated is input, and the network architecture within the backbone network search space is selectively activated based on the input geographic information. The process of extracting geographic information features through the activated network architecture includes:

[0034] The geographic information to be integrated is recorded as information to be processed, the data type contained in the information to be processed is obtained, and a corresponding type label is generated according to the data type contained in the information to be processed;

[0035] Summarize the type labels to obtain a label set, and associate the label set with the information to be processed;

[0036] After inputting the information to be processed and the label set into the backbone network search space, the type label in the label set is read;

[0037] Activate a class of activation nodes corresponding to the read type labels in the backbone network search space, and call the feature extraction architecture corresponding to the activated class of activation nodes;

[0038] Extract features from the information to be processed through the called feature extraction architecture and obtain corresponding geographic information features;

[0039] After completing the feature extraction of all the information to be processed, the second-class activation nodes are activated, and the correlation extraction architecture corresponding to the second-class activation nodes is called;

[0040] The obtained geographic information features are input into the called relevance extraction framework, and the temporal features, spatial features and relevance features of the geographic information features are output through the relevance extraction framework.

[0041] Furthermore, the process of verifying the accuracy of the extracted geographic information features includes:

[0042] Randomly sample the obtained geographic information features to obtain corresponding sampling samples;

[0043] Compare the sampled data with the actual content of the information to be processed. If the comparison results are consistent, it means that the accuracy of the corresponding sampled data has passed. Otherwise, it has failed.

[0044] The proportion of sampling samples that pass the accuracy test to the total number of sampling samples;

[0045] Set a percentage threshold and compare the obtained percentage with the percentage threshold. If the percentage is higher than the percentage threshold, it means that the accuracy verification of the extracted geographic information features has passed, and a prompt message indicating successful integration is generated. Otherwise, it means that the accuracy verification has failed, and a prompt message indicating failed integration is generated.

[0046] Furthermore, the process of updating the sample sub-dataset in the database includes:

[0047] Obtaining geographic information corresponding to each geographic information feature and obtaining a type label corresponding to the geographic information;

[0048] According to the type label index, the corresponding database is found, and the geographic information is matched with the sample data stored in the database. If the geographic information is the same as the sample data stored in the database, it indicates a successful match, otherwise it indicates a failed match.

[0049] The geographic information and corresponding geographic information features that failed to be matched are summarized, and the technical staff will mark the summarized geographic information and geographic information features;

[0050] The labeling results include "recognition error" and "unrecognized";

[0051] The geographic information and geographic information features identified as "identification error" are eliminated, and the geographic information and geographic information features identified as "unidentified" are used as new sample data, and the new sample data are imported into the corresponding first-category database and second-category database, thereby updating the sample data subset in the database.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. By building a backbone network search space with different network architectures and collecting corresponding sample data according to the type of geographic information to be integrated, the corresponding network architecture in the backbone network search space is trained in a targeted manner, thereby obtaining a network architecture capable of processing different types of data. This enables the corresponding processing of different types of data contained in geographic information, and improves the data compatibility of geographic information processing;

[0054] 2. When processing geographic information, the network architecture within the backbone network search space can be selectively activated according to the type of data contained in the geographic information to be processed, thereby improving the adaptability between the network architecture and geographic information, and further improving data processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0056] Figure 1 is a flow chart of the present invention;

[0057] Figure 2 This is a schematic diagram of the backbone network search space structure of the present invention;

[0058] Figure 3 This is a flow chart of information processing to be processed according to the present invention. DETAILED DESCRIPTION

[0059] like Figure 1 As shown, a method for constructing an artificial intelligence model for geographic information integration includes:

[0060] Obtain geographic information integration requirements, collect corresponding sample data based on geographic information integration requirements, and build a backbone network search space that deploys several network architectures;

[0061] The collected sample data is divided into factors, and the corresponding sample sub-datasets are obtained according to the factor division results. The network architecture in the backbone network search space is trained using the obtained sample sub-datasets;

[0062] Establish an associated database for each trained network architecture and import the corresponding sample sub-datasets into the associated database;

[0063] Input the geographic information that needs to be integrated, and selectively activate the network architecture within the backbone network search space based on the input geographic information, and complete the extraction of geographic information features through the activated network architecture;

[0064] The extracted geographic information features are verified for accuracy, and the input geographic information is imported into the corresponding database to update the sample sub-dataset in the database.

[0065] like Figure 2 As shown, it is necessary to further explain that, in the specific implementation process, the process of collecting corresponding sample data according to the geographic information integration requirements and constructing a backbone network search space that deploys several network architectures includes:

[0066] The geographic information integration requirements include the data types, time series, spatial series and the correlation between different data of the geographic information expected to be involved;

[0067] Construct a backbone network search space, and arrange the required network architectures in the backbone network search space according to the requirements of geographic information integration, and initialize each network architecture;

[0068] According to the characteristics of each network architecture, the network architecture is associated with the data type, and a type of activation node corresponding to the network architecture is set. All network architectures with a type of activation node are recorded as feature extraction architectures;

[0069] According to the correlation between time series, spatial series and different data, the corresponding network architectures are associated respectively, and two types of activation nodes corresponding to the network architectures are set. The architecture with the two types of activation nodes is recorded as the correlation extraction architecture;

[0070] Collect geographic information sample data of the target area according to the needs of geographic information integration.

[0071] It should be further explained that, in the specific implementation process, the collected sample data is divided into factors, corresponding sample sub-datasets are obtained according to the factor division results, and the network architecture in the backbone network search space is trained using the obtained sample sub-datasets. The process includes:

[0072] Divide the collected geographic information sample data according to type factors to obtain geographic information data of different data types;

[0073] Aggregate geographic information data of the same type as corresponding sample sub-datasets;

[0074] Set a corresponding type label for each sample sub-dataset and associate the type label with the sample sub-dataset;

[0075] Divide each sample sub-dataset into corresponding training sets, validation sets, and test sets;

[0076] Import the training set, validation set, and test set corresponding to each obtained sample sub-dataset into the backbone network search space;

[0077] Read the type labels associated with the sample sub-datasets corresponding to the imported training set, validation set, and test set;

[0078] activating a corresponding class of activation nodes according to the type label, and after activating the corresponding class of activation nodes, importing the training set, the validation set, and the test set into the feature extraction architecture corresponding to the class of activation nodes to train the feature extraction architecture;

[0079] After completing the training of the feature extraction architecture, the corresponding sample sub-dataset is input into the corresponding feature extraction architecture to extract the sample data features in the sample sub-dataset;

[0080] Summarize the sample data features extracted by each feature extraction architecture to obtain a sample feature set, and divide the sample feature set into corresponding training set, validation set, and test set;

[0081] Setting feature labels for the obtained sample feature set, activating the second-class activation node according to the generated feature labels, and after activating the second-class activation node, inputting the training set, validation set, and test set divided by the sample feature set into the relevance extraction framework associated with the second-class activation node to train the relevance extraction framework;

[0082] It should be noted that the process of training the network architecture using the training set, validation set, and test set obtained from the sample data is a common technical means used by those skilled in the art, so it will not be described in detail.

[0083] It should be further explained that, in the specific implementation process, the process of establishing a related database for each trained network architecture and importing the corresponding sample sub-dataset into the related database includes:

[0084] Constructing a database, and constructing a corresponding class of databases according to the type labels corresponding to each sample data subset within the database, and associating the constructed class of databases with the type labels;

[0085] Importing the sample data in the sample data subset into a corresponding type of database;

[0086] A corresponding second-class database is constructed in the database according to the feature labels corresponding to the sample feature set, and the sample data features in the sample feature set are imported into the second-class database.

[0087] like Figure 3 As shown, it is necessary to further explain that, in the specific implementation process, the geographic information to be integrated is input, and the network architecture within the backbone network search space is selectively activated based on the input geographic information. The process of extracting geographic information features through the activated network architecture includes:

[0088] The geographic information to be integrated is recorded as information to be processed, the data type contained in the information to be processed is obtained, and a corresponding type label is generated according to the data type contained in the information to be processed;

[0089] Summarize the type labels to obtain a label set, and associate the label set with the information to be processed;

[0090] After inputting the information to be processed and the label set into the backbone network search space, the type label in the label set is read;

[0091] Activate a class of activation nodes corresponding to the read type labels in the backbone network search space, and call the feature extraction architecture corresponding to the activated class of activation nodes;

[0092] Extract features from the information to be processed through the called feature extraction architecture and obtain corresponding geographic information features;

[0093] After completing the feature extraction of all the information to be processed, the second-class activation nodes are activated, and the correlation extraction architecture corresponding to the second-class activation nodes is called;

[0094] The obtained geographic information features are input into the called relevance extraction framework, and the temporal features, spatial features and relevance features of the geographic information features are output through the relevance extraction framework.

[0095] It should be further explained that, in the specific implementation process, the process of verifying the accuracy of the extracted geographic information features includes:

[0096] Randomly sample the obtained geographic information features to obtain corresponding sampling samples;

[0097] Compare the sampled data with the actual content of the information to be processed. If the comparison results are consistent, it means that the accuracy of the corresponding sampled data has passed. Otherwise, it has failed.

[0098] The proportion of sampling samples that pass the accuracy test to the total number of sampling samples;

[0099] Set a percentage threshold and compare the obtained percentage with the percentage threshold. If the percentage is higher than the percentage threshold, it means that the accuracy verification of the extracted geographic information features has passed, and a prompt message indicating successful integration is generated. Otherwise, it means that the accuracy verification has failed, and a prompt message indicating failed integration is generated.

[0100] It should be further explained that, in the specific implementation process, the process of updating the sample sub-dataset in the database includes:

[0101] Obtaining geographic information corresponding to each geographic information feature and obtaining a type label corresponding to the geographic information;

[0102] According to the type label index, the corresponding database is found, and the geographic information is matched with the sample data stored in the database. If the geographic information is the same as the sample data stored in the database, it indicates a successful match, otherwise it indicates a failed match.

[0103] The geographic information and corresponding geographic information features that failed to be matched are summarized, and the technical staff will mark the summarized geographic information and geographic information features;

[0104] The labeling results include "recognition error" and "unrecognized";

[0105] The geographic information and geographic information features marked as "identification error" are eliminated, and the geographic information and geographic information features marked as "unidentified" are used as new sample data, and the new sample data are imported into the corresponding first-class database and second-class database, thereby updating the sample data subset in the database, making the amount of sample data corresponding to each network architecture larger and larger, thereby improving the performance of each network architecture.

[0106] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any modification or equivalent replacement of the above embodiments made according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A method for constructing an artificial intelligence model for geographic information integration, characterized in that: include: Obtain geographic information integration requirements, collect corresponding sample data based on geographic information integration requirements, and build a backbone network search space that deploys several network architectures; The collected sample data is divided into factors, and the corresponding sample sub-datasets are obtained according to the factor division results. The network architecture in the backbone network search space is trained using the obtained sample sub-datasets; Establish an associated database for each trained network architecture and import the corresponding sample sub-datasets into the associated database; Input the geographic information that needs to be integrated, and selectively activate the network architecture within the backbone network search space based on the input geographic information, and complete the extraction of geographic information features through the activated network architecture; Verify the accuracy of the extracted geographic information features and import the input geographic information into the corresponding database to update the sample sub-dataset in the database; The process of collecting corresponding sample data according to geographic information integration requirements and building a backbone network search space that deploys several network architectures includes: The geographic information integration requirements include the data types, time series, spatial series and the correlation between different data of the geographic information expected to be involved; Construct a backbone network search space, and arrange the required network architectures in the backbone network search space according to the requirements of geographic information integration, and initialize each network architecture; According to the characteristics of each network architecture, the network architecture is associated with the data type, and a type of activation node corresponding to the network architecture is set. All network architectures with a type of activation node are recorded as feature extraction architectures; According to the correlation between time series, spatial series and different data, the corresponding network architectures are associated respectively, and two types of activation nodes corresponding to the network architectures are set. The architecture with the two types of activation nodes is recorded as the correlation extraction architecture; Collect geographic information sample data of the target area according to the needs of geographic information integration.

2. The method for constructing an artificial intelligence model for geographic information integration according to claim 1, characterized in that: The collected sample data is divided into factors, and the process of obtaining the corresponding sample sub-datasets according to the factor division results includes: Divide the collected geographic information sample data according to type factors to obtain geographic information data of different data types; Aggregate geographic information data of the same type as corresponding sample sub-datasets; Set a corresponding type label for each sample sub-dataset and associate the type label with the sample sub-dataset; Each sample sub-dataset is divided into corresponding training sets, validation sets and test sets.

3. The method for constructing an artificial intelligence model for geographic information integration according to claim 2, characterized in that: The process of training the network architecture in the backbone network search space using the obtained sample sub-dataset includes: Import the training set, validation set, and test set corresponding to each obtained sample sub-dataset into the backbone network search space; Read the type labels associated with the sample sub-datasets corresponding to the imported training set, validation set, and test set; activating a corresponding class of activation nodes according to the type label, and after activating the corresponding class of activation nodes, importing the training set, the validation set, and the test set into the feature extraction architecture corresponding to the class of activation nodes to train the feature extraction architecture; After completing the training of the feature extraction architecture, the corresponding sample sub-dataset is input into the corresponding feature extraction architecture to extract the sample data features in the sample sub-dataset; Summarize the sample data features extracted by each feature extraction architecture to obtain a sample feature set, and divide the sample feature set into corresponding training set, validation set, and test set; Feature labels are set for the obtained sample feature sets, and the second-class activation nodes are activated according to the generated feature labels. After the second-class activation nodes are activated, the training set, validation set, and test set divided by the sample feature sets are input into the correlation extraction architecture associated with the second-class activation nodes to train the correlation extraction architecture.

4. The method for constructing an artificial intelligence model for geographic information integration according to claim 3, characterized in that: The process of establishing a database associated with each trained network architecture and importing the corresponding sample sub-dataset into the associated database includes: Constructing a database, and constructing a corresponding class of databases according to the type labels corresponding to each sample data subset within the database, and associating the constructed class of databases with the type labels; Importing the sample data in the sample data subset into a corresponding type of database; A corresponding second-class database is constructed in the database according to the feature labels corresponding to the sample feature set, and the sample data features in the sample feature set are imported into the second-class database.

5. The method for constructing an artificial intelligence model for geographic information integration according to claim 4, characterized in that: The geographic information to be integrated is input, and the network architecture within the backbone network search space is selectively activated based on the input geographic information. The process of extracting geographic information features through the activated network architecture includes: The geographic information to be integrated is recorded as information to be processed, the data type contained in the information to be processed is obtained, and a corresponding type label is generated according to the data type contained in the information to be processed; Summarize the type labels to obtain a label set, and associate the label set with the information to be processed; After inputting the information to be processed and the label set into the backbone network search space, the type label in the label set is read; Activate a class of activation nodes corresponding to the read type labels in the backbone network search space, and call the feature extraction architecture corresponding to the activated class of activation nodes; Extract features from the information to be processed through the called feature extraction architecture and obtain corresponding geographic information features; After completing the feature extraction of all the information to be processed, the second-class activation nodes are activated, and the correlation extraction architecture corresponding to the second-class activation nodes is called; The obtained geographic information features are input into the called relevance extraction framework, and the temporal features, spatial features and relevance features of the geographic information features are output through the relevance extraction framework.

6. The method for constructing an artificial intelligence model for geographic information integration according to claim 5, characterized in that: The process of verifying the accuracy of the extracted geographic information features includes: Randomly sample the obtained geographic information features to obtain corresponding sampling samples; Compare the sampled data with the actual content of the information to be processed. If the comparison results are consistent, it means that the accuracy of the corresponding sampled data has passed. Otherwise, it has failed. The proportion of sampling samples that pass the accuracy test to the total number of sampling samples; Set a percentage threshold and compare the obtained percentage with the percentage threshold. If the percentage is higher than the percentage threshold, it means that the accuracy verification of the extracted geographic information features has passed, and a prompt message indicating successful integration is generated. Otherwise, it means that the accuracy verification has failed, and a prompt message indicating failed integration is generated.

7. The method for constructing an artificial intelligence model for geographic information integration according to claim 6, characterized in that: The process of updating the sample sub-dataset in the database includes: Obtaining geographic information corresponding to each geographic information feature and obtaining a type label corresponding to the geographic information; According to the type label index, the corresponding database is found, and the geographic information is matched with the sample data stored in the database. If the geographic information is the same as the sample data stored in the database, it indicates a successful match, otherwise it indicates a failed match. The geographic information and corresponding geographic information features that failed to be matched are summarized, and the technical staff will mark the summarized geographic information and geographic information features; The labeling results include "recognition error" and "unrecognized"; The geographic information and geographic information features with the identification result of "recognition error" are eliminated, and the geographic information and geographic information features with the identification result of "unrecognized" are used as new sample data, and the new sample data are imported into the corresponding first-class database and second-class database, thereby updating the sample data subset in the database.

Images