Natural resource element change detection method based on deep learning
Through the detection method of changes in natural resource elements based on deep learning, combined with multi-source heterogeneous data and knowledge graphs, the problems of low detection accuracy and large labor in the existing technology are solved, and higher detection accuracy and depth of understanding are achieved. Adapt to the dynamic changes of natural resources.
Patent Information
- Application Number
- CN202411072017.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art has poor detection accuracy in natural resource change detection and requires a large amount of manual identification of labor, which cannot effectively improve detection accuracy and reduce labor.
The natural resource element change detection method based on deep learning is adopted, and through structured knowledge representation and powerful pattern recognition capabilities, multi-source heterogeneous data and knowledge graphs are used to improve the accuracy and depth of detection.
It enhances feature representation and interpretability, optimizes data utilization, realizes in-depth causal analysis and dynamic updates of changes in natural resources, and improves detection accuracy and robustness.
Smart Images

Figure CN119003664A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of natural resource element detection, and in particular to a natural resource element change detection method based on deep learning. Background Art
[0002] Natural resources broadly cover all kinds of materials, energy and conditions that exist naturally on the earth and can provide humans with the necessary survival and development. They can be mainly summarized into the following categories:
[0003] 1. Land resources: including cultivated land, forest land, grassland, construction land, etc., which are the basis for human settlement, agricultural planting, forestry development and other activities.
[0004] 2. Water resources: including surface water (such as rivers, lakes, reservoirs), groundwater, atmospheric water (such as rain and snow), glacier water, etc., are the key to life maintenance and economic and social development.
[0005] 3. Biological resources: refers to animal and plant resources and the ecosystems they constitute, such as forest resources, fishery resources, wild animal and plant resources, etc., which provide food, medicine, materials and ecological services.
[0006] 4. Mineral resources: including metal minerals (such as iron, copper, and aluminum), non-metallic minerals (such as coal, oil, natural gas, salt, and graphite), and other types of rocks and minerals, which are the main sources of industrial production and energy.
[0007] 5. Climate resources: such as solar energy, wind energy, precipitation, temperature, etc., provide conditions for agricultural production, energy development, climate regulation, etc.
[0008] 6. Marine resources: including marine life, seabed minerals, seawater chemical resources and marine energy such as tidal energy and wave energy, which are precious wealth that has not been fully developed on the earth.
[0009] 7. Air resources: Clean air is a basic requirement for human health and ecosystem balance.
[0010] 8. Renewable resources: such as solar energy, wind energy, hydropower, etc. These resources can be continuously regenerated if used reasonably.
[0011] 9. Non-renewable resources: such as oil, natural gas, coal and most minerals, which, once consumed, are difficult to replenish on a human time scale.
[0012] 10. Natural landscape and cultural resources: Nature reserves, scenic spots, historical and cultural sites, etc., although not natural resources in the traditional sense, have important value in tourism, education and scientific research. The above natural resource elements are interrelated and together constitute the natural environment system of the earth, which is crucial to the sustainable development of human society.
[0013] However, there is currently no perfect detection method for natural resource change detection, which means that when conducting natural resource change detection in a certain area, a large amount of manual labor is required to identify changes in natural resource elements, and the detection accuracy is not ideal. Therefore, when conducting natural resource change detection on natural resources, how to improve detection accuracy and reduce the large amount of manual labor to identify changes in natural resource elements is an urgent problem to be solved. Summary of the invention
[0014] To solve the above problems, the present invention provides a natural resource element change detection method based on deep learning, which improves the accuracy and understanding depth of natural resource change detection through structured knowledge representation and powerful pattern recognition capabilities, and overcomes the limitations of a single technology.
[0015] To achieve the above object, the present invention provides a natural resource element change detection method based on deep learning, comprising the following steps:
[0016] S1, collect multi-source heterogeneous data and pre-process them;
[0017] S2. Use deep learning models to extract features from multi-source heterogeneous data and train deep learning models to identify and classify natural resource elements;
[0018] S3, constructing a knowledge graph containing natural resource elements, attributes and relationships based on the recognition results of step S2, to provide additional semantic information for the deep learning model;
[0019] S4, embedding the knowledge in the knowledge graph obtained in step S3 into the deep learning model described in step S2 as a constraint condition, to obtain a deep learning model embedded in the knowledge graph;
[0020] S5. Collect multi-source heterogeneous data of the area to be detected, input it into the deep learning model embedded in the knowledge graph described in step S4 after preprocessing, and output the detection results.
[0021] Preferably, the multi-source heterogeneous data described in step S1 includes remote sensing images, meteorological data, terrain data, and historical change records;
[0022] The preprocessing described in step S1 includes correction, registration and segmentation.
[0023] Preferably, step S3 specifically includes the following steps:
[0024] S31. Determine the target domain, application scenarios, and problems that the knowledge graph is expected to solve, and define the ontology of the knowledge graph, where the ontology of the knowledge graph includes natural resource categories, attributes, and relationships;
[0025] S32, using a deep learning model to identify natural resource entities from multi-source heterogeneous data, and then classifying the identified entities into predefined categories;
[0026] S33, extract relations and attributes:
[0027] Extract relations: annotate relations between entities through dependency analysis;
[0028] Attribute extraction: extract key features and attributes of each entity;
[0029] S34. Build a knowledge graph:
[0030] S341, Generate triples: Integrate entities, relations, and attributes into triples, i.e., subject-relation-object;
[0031] S342. Storage graph: Use a graph database or relational database to store triples and build a knowledge graph.
[0032] Preferably, step S4 specifically includes the following steps:
[0033] S41. Knowledge graph preprocessing: converting entities and relations in the knowledge graph into numerical vector representations;
[0034] S42, Dataset division: divide the triples in the knowledge graph into training set, validation set and test set for subsequent model training and evaluation;
[0035] S43. Build a deep learning model embedded in the knowledge graph:
[0036] S431, select GraphSAGE as the entity and relationship embedding model;
[0037] S432, Initialize embedding vector: Randomly initialize low-dimensional vector representation for each entity and relationship in the knowledge graph;
[0038] S433, Knowledge Graph Embedding Training:
[0039] S4331. Define loss function: Use Margin-based Loss or Pairwise Loss as the loss function for knowledge graph embedding to optimize the embedding vector;
[0040] S4332, Training: Use the back-propagation algorithm to update the embedding vector through the gradient descent optimization method, minimize the loss function, and enable the deep learning model embedded in the knowledge graph to learn the intrinsic structure and semantics of the knowledge graph;
[0041] S434, Integrate knowledge graphs into deep learning models:
[0042] S4341, Feature Fusion: The knowledge graph embedding vector is used as an additional feature and input into the deep learning model together with the original input features;
[0043] S4342, Model architecture adjustment: Adjust the deep learning model architecture according to the goal;
[0044] S4343, Joint training: Train the deep learning model and the knowledge graph embedding model simultaneously, and update the parameters of the deep learning model and the knowledge graph embedding model through the back propagation algorithm;
[0045] S4344, Evaluation and Verification:
[0046] Evaluation indicators: Select accuracy, recall, F1 score, and MRR as evaluation indicators to evaluate the performance of the model on the validation set and test set;
[0047] Model verification: Verify whether the error between the output result of the deep learning model embedded in the knowledge graph and the true value meets the set error threshold requirement.
[0048] Preferably, step S41 specifically includes the following steps:
[0049] S411. Based on the manual definition of natural resource elements in multi-source heterogeneous data, the manual definition information is used to construct the connection relationship between natural resource elements, attributes and relationships, where the relationship includes adjacent relationship, overlapping relationship, dependent relationship and penetrating relationship:
[0050] S412. Determine the types of natural resource elements that need to be identified and classified by observing remote sensing images;
[0051] S413. Manually classify each natural resource element, associate it with the entities in the knowledge graph, and assign a unique identifier to each entity;
[0052] S414, using a graph theory algorithm to convert the connection relationship into an undirected graph, wherein each node represents a natural resource element, and each edge represents a connection relationship between two natural resource elements;
[0053] S415. Traverse the undirected graph and store the relationship information between each group of natural resource elements in the knowledge graph.
[0054] The present invention has the following beneficial effects:
[0055] 1. Enhanced feature representation: Deep learning can automatically extract complex spatiotemporal features from remote sensing images, while knowledge graphs can provide prior knowledge and entity relationships about natural resource elements, such as the seasonal changes of different vegetation types and the impact of geological structure on water resource distribution. Combining the two, these domain knowledge can be incorporated into the model training process to enhance the model's ability to recognize specific change patterns;
[0056] 2. Improve interpretability: Deep learning models are often difficult to explain due to their black box characteristics. The addition of knowledge graphs can guide model learning through structured knowledge, making the model decision process more transparent. For example, graph reasoning can be used to verify whether the changes identified by the deep learning model conform to known natural laws or expert experience.
[0057] 3. Optimize data utilization: Knowledge graphs can help integrate multi-source heterogeneous data, such as meteorological data, terrain data, historical change records, etc., and input them into deep learning models together with remote sensing image data to form richer contextual information and improve the accuracy and robustness of change detection;
[0058] 4. Analysis of causes of change: By associating change detection results with possible driving factors, such as climate change and human activities, through knowledge graphs, we can achieve in-depth causal analysis of natural resource changes, rather than just identifying surface phenomena.
[0059] 5. Dynamic update and adaptation: As an updateable knowledge storage, the knowledge graph can incorporate new research results or monitoring data in real time, enabling the deep learning model to adjust and optimize according to the latest knowledge and better adapt to the dynamic changes of natural resource elements over time.
[0060] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] Figure 1 This is a flow chart of a natural resource element change detection method based on deep learning of the present invention. DETAILED DESCRIPTION
[0062] In order to make the purpose, technical scheme and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention are further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions.
[0063] It should be noted that the terms "include", "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.
[0064] Like reference numerals and letters denote similar items in the following drawings, and thus, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.
[0065] like Figure 1 As shown, a natural resource element change detection method based on deep learning includes the following steps:
[0066] S1, collect multi-source heterogeneous data and pre-process them;
[0067] The multi-source heterogeneous data described in step S1 includes remote sensing images, meteorological data, terrain data, and historical change records;
[0068] The preprocessing described in step S1 includes correction, registration and segmentation.
[0069] S2. Use deep learning models to extract features from multi-source heterogeneous data and train deep learning models to identify and classify natural resource elements;
[0070] S3, constructing a knowledge graph containing natural resource elements, attributes and relationships based on the recognition results of step S2, to provide additional semantic information for the deep learning model;
[0071] Step S3 specifically includes the following steps:
[0072] S31. Determine the target domain, application scenarios, and problems that the knowledge graph is expected to solve, and define the ontology of the knowledge graph, where the ontology of the knowledge graph includes natural resource categories, attributes, and relationships;
[0073] S32, using a deep learning model to identify natural resource entities from multi-source heterogeneous data, and then classifying the identified entities into predefined categories;
[0074] S33, extract relations and attributes:
[0075] Extract relations: annotate relations between entities through dependency analysis;
[0076] Attribute extraction: extract key features and attributes of each entity;
[0077] S34. Build a knowledge graph:
[0078] S341, Generate triples: Integrate entities, relations, and attributes into triples, i.e., subject-relation-object;
[0079] S342. Storage graph: Use a graph database or a relational database to store triples and build a knowledge graph.
[0080] S4, embedding the knowledge in the knowledge graph obtained in step S3 into the deep learning model described in step S2 as a constraint condition, to obtain a deep learning model embedded in the knowledge graph;
[0081] Step S4 specifically includes the following steps:
[0082] S41. Knowledge graph preprocessing: converting entities and relations in the knowledge graph into numerical vector representations;
[0083] S42, Dataset division: divide the triples in the knowledge graph into training set, validation set and test set for subsequent model training and evaluation;
[0084] S43. Build a deep learning model embedded in the knowledge graph:
[0085] S431, select GraphSAGE as the entity and relationship embedding model;
[0086] S432, Initialize embedding vector: Randomly initialize low-dimensional vector representation for each entity and relationship in the knowledge graph;
[0087] S433, Knowledge Graph Embedding Training:
[0088] S4331. Define loss function: Use Margin-based Loss or Pairwise Loss as the loss function for knowledge graph embedding to optimize the embedding vector;
[0089] S4332, Training: Use the back-propagation algorithm to update the embedding vector through the gradient descent optimization method, minimize the loss function, and enable the deep learning model embedded in the knowledge graph to learn the intrinsic structure and semantics of the knowledge graph;
[0090] S434, Integrate knowledge graphs into deep learning models:
[0091] S4341, Feature Fusion: The knowledge graph embedding vector is used as an additional feature and input into the deep learning model together with the original input features;
[0092] S4342, Model architecture adjustment: Adjust the deep learning model architecture according to the goal;
[0093] S4343, Joint training: Train the deep learning model and the knowledge graph embedding model simultaneously, and update the parameters of the deep learning model and the knowledge graph embedding model through the back propagation algorithm;
[0094] S4344, Evaluation and Verification:
[0095] Evaluation indicators: Select accuracy, recall, F1 score, and MRR as evaluation indicators to evaluate the performance of the model on the validation set and test set;
[0096] Model verification: Verify whether the error between the output result of the deep learning model embedded in the knowledge graph and the true value meets the set error threshold requirement.
[0097] Preferably, step S41 specifically includes the following steps:
[0098] S411. Based on the manual definition of natural resource elements in multi-source heterogeneous data, the manual definition information is used to construct the connection relationship between natural resource elements, attributes and relationships, where the relationship includes adjacent relationship, overlapping relationship, dependent relationship and penetrating relationship:
[0099] S412. Determine the types of natural resource elements that need to be identified and classified by observing remote sensing images;
[0100] S413. Manually classify each natural resource element, associate it with the entities in the knowledge graph, and assign a unique identifier to each entity;
[0101] S414, using a graph theory algorithm to convert the connection relationship into an undirected graph, wherein each node represents a natural resource element, and each edge represents a connection relationship between two natural resource elements;
[0102] S415. Traverse the undirected graph and store the relationship information between each group of natural resource elements in the knowledge graph.
[0103] S5. Collect multi-source heterogeneous data of the area to be detected, input it into the deep learning model embedded in the knowledge graph described in step S4 after preprocessing, and output the detection results.
[0104] Therefore, the present invention adopts the above-mentioned natural resource element change detection method based on deep learning, which improves the accuracy and understanding depth of natural resource change detection through structured knowledge representation and powerful pattern recognition capabilities, and overcomes the limitations of a single technology.
[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.
Claims
1. A natural resource element change detection method based on deep learning, characterized in that: The following steps are involved: S1, collect multi-source heterogeneous data and pre-process them; S2. Use deep learning models to extract features from multi-source heterogeneous data and train deep learning models to identify and classify natural resource elements; S3, constructing a knowledge graph containing natural resource elements, attributes and relationships based on the recognition results of step S2, to provide additional semantic information for the deep learning model; S4, embedding the knowledge in the knowledge graph obtained in step S3 into the deep learning model described in step S2 as a constraint condition, to obtain a deep learning model embedded in the knowledge graph; S5. Collect multi-source heterogeneous data of the area to be detected, input it into the deep learning model embedded in the knowledge graph described in step S4 after preprocessing, and output the detection results.
2. The method for detecting changes in natural resource elements based on deep learning according to claim 1, characterized in that: The multi-source heterogeneous data described in step S1 includes remote sensing images, meteorological data, terrain data, and historical change records; The preprocessing described in step S1 includes correction, registration and segmentation.
3. The method for detecting changes in natural resource elements based on deep learning according to claim 2, characterized in that: Step S3 specifically includes the following steps: S31. Determine the target domain, application scenarios, and problems that the knowledge graph is expected to solve, and define the ontology of the knowledge graph, where the ontology of the knowledge graph includes natural resource categories, attributes, and relationships; S32, using a deep learning model to identify natural resource entities from multi-source heterogeneous data, and then classifying the identified entities into predefined categories; S33, extract relations and attributes: Extract relations: annotate the relations between entities through dependency analysis; Attribute extraction: extract key features and attributes of each entity; S34. Build a knowledge graph: S341, Generate triples: Integrate entities, relations, and attributes into triples, i.e., subject-relation-object; S342. Storage graph: Use a graph database or a relational database to store triples and build a knowledge graph.
4. The method for detecting changes in natural resource elements based on deep learning according to claim 3, characterized in that: Step S4 The specific steps include: S41. Knowledge graph preprocessing: converting entities and relations in the knowledge graph into numerical vector representations; S42, Dataset division: divide the triples in the knowledge graph into training set, validation set and test set for subsequent model training and evaluation; S43. Build a deep learning model embedded in the knowledge graph: S431, select GraphSAGE as the entity and relationship embedding model; S432, Initialize embedding vector: Randomly initialize low-dimensional vector representation for each entity and relationship in the knowledge graph; S433, Knowledge Graph Embedding Training: S4331. Define loss function: Use Margin-based Loss or Pairwise Loss as the loss function for knowledge graph embedding to optimize the embedding vector; S4332, Training: Use the back-propagation algorithm to update the embedding vector through the gradient descent optimization method, minimize the loss function, and enable the deep learning model embedded in the knowledge graph to learn the intrinsic structure and semantics of the knowledge graph; S434, Integrate knowledge graphs into deep learning models: S4341, Feature Fusion: The knowledge graph embedding vector is used as an additional feature and input into the deep learning model together with the original input features; S4342, Model architecture adjustment: Adjust the deep learning model architecture according to the goal; S4343, Joint training: Train the deep learning model and the knowledge graph embedding model simultaneously, and update the parameters of the deep learning model and the knowledge graph embedding model through the back propagation algorithm; S4344, Evaluation and Verification: Evaluation indicators: Select accuracy, recall, F1 score, and MRR as evaluation indicators to evaluate the performance of the model on the validation set and test set; Model verification: Verify whether the error between the output result of the deep learning model embedded in the knowledge graph and the true value meets the set error threshold requirement.
5. The method for detecting changes in natural resource elements based on deep learning according to claim 4, characterized in that: Step S41 specifically includes the following steps: S411. Based on the manual definition of natural resource elements in multi-source heterogeneous data, the manual definition information is used to construct the connection relationship between natural resource elements, attributes and relationships, where the relationship includes adjacent relationship, overlapping relationship, dependent relationship and penetrating relationship: S412. Determine the types of natural resource elements that need to be identified and classified by observing remote sensing images; S413. Manually classify each natural resource element, associate it with the entities in the knowledge graph, and assign a unique identifier to each entity; S414, using a graph theory algorithm to convert the connection relationship into an undirected graph, wherein each node represents a natural resource element, and each edge represents a connection relationship between two natural resource elements; S415. Traverse the undirected graph and store the relationship information between each group of natural resource elements in the knowledge graph.
Citation Information
Patent Citations
Public security event-oriented multi-source heterogeneous data knowledge graph construction method
CN110598005A
Deep learning-based natural resources element change detection method
CN113362286A
Urban construction and natural resource archive knowledge graph construction method
CN116089628A
Wind turbine generator fault prediction method fusing knowledge graph and deep learning
CN116187508A
Network security dynamic early warning system and method based on knowledge graph, and storage medium
CN117834282A
Cited By
Knowledge graph ontology constrained vegetation remote sensing classification method
CN119580103A
A vegetation remote sensing classification method constrained by knowledge graph ontology
CN119580103B
Natural resource change pattern spot purification method, device and equipment based on knowledge graph
CN120472306A
Natural resource data intelligent acquisition method
CN121658911A
