A potassium salt mine area land identification method and device based on remote sensing technology for Southeast Asia
Patent Information
- Application Number
- CN202611214989.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-12
- Publication Date
- 2026-10-30
AI Technical Summary
[0021]由于依赖单一高分卫星影像进行环境要素信息提取，会降低提取精度。因此，本发明在基于国产高分卫星数据的基础上，会同时结合多源数据进行辅助判断，以便提高信息提取精度。本发明借助GF-1、GF1-BCD、GF-2、GF-6、GF-7等全波段影像数据，结合DEM 数据、地质背景资料、矿产资源分类资料等多源异构数据，选择合适的机器学习分类器，对钾盐矿矿区用地信息进行提取和精度验证。
Smart Images
Figure CN123016231A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of land use identification technology, and in particular to a method and equipment for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia. Background Technology
[0002] Southeast Asia, located on the western Pacific tectonic-magmatic activity belt, is one of the most tectonically active regions in the Earth's lithosphere since the Mesozoic and Cenozoic eras, characterized by intense tectonic-magmatic activity. It is also one of the world's few regions with abundant mineral resources, boasting numerous world-class deposits. Adjacent to my country's Tibet, Yunnan, Guangxi, Guangdong, and Hainan provinces (autonomous regions), Southeast Asia shares similarities in regional metallogenic geological conditions and metallogenic characteristics, while also exhibiting certain differences and complementarities. Its advantageous mineral resources are well-complementary to my country's mineral resource structure. Among these, potash is a significant and advantageous mineral resource in Southeast Asia, holding a prominent position globally.
[0003] Southeast Asia is rich in potash resources, mainly distributed in Cretaceous saline strata around the Khorat Plateau. Potash deposits occur in large Mesozoic intracontinental sedimentary basins, with ore bodies occurring in the Upper Cretaceous Mahasharakan Formation. Utilizing domestic satellite data to automatically identify potash mining areas is beneficial for the management of mining activities and comprehensive mine governance, and also provides decision-making support for my country's international cooperation on mineral resources.
[0004] Based on the above background, this invention proposes a method and equipment for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia. Summary of the Invention
[0005] To overcome the problems existing in the prior art, this invention provides a method and equipment for identifying land use in potash mines based on remote sensing technology in Southeast Asia. It can realize the automated, refined and efficient extraction of land use information in mines, and provide decision-making reference for my country's international mining cooperation.
[0006] To achieve the above objectives, the present invention provides the following solution:
[0007] A method for identifying land use in potash mine areas based on remote sensing technology in Southeast Asia includes the following steps:
[0008] Step 1, Remote Sensing Data Acquisition: Acquire multi-source high-resolution satellite remote sensing images of the target area;
[0009] Step 2, Remote Sensing Interpretation and Verification: By interpreting the multi-source high-resolution satellite remote sensing images, investigate the environmental conditions of potash mines; using multiple periods of the multi-source high-resolution satellite remote sensing images, conduct an investigation into the land use and mining situation and resource and environmental changes in potash mine areas; and verify the remote sensing interpretation and investigation results; obtain the remote sensing interpretation and verification results.
[0010] Step 3: Constructing the sample library and selecting features: Based on the remote sensing interpretation and verification results, assign values to each unit to construct a sample library of potash mine land use based on multi-source high-resolution satellite remote sensing images; establish classification models for spectral features, texture features, and spatial features;
[0011] Step 4: Automatic identification of land use information based on remote sensing images: Based on the spectral curve characteristics of different land features in different bands, and utilizing spectral, spatial, and temporal features, land feature elements are extracted from remote sensing images using neural networks to obtain the automatic identification results of potash mine land use information.
[0012] Furthermore, in step 1, after acquiring multi-source high-resolution satellite remote sensing images of the target location, the acquired multi-source high-resolution satellite remote sensing images are preprocessed to obtain fused and enhanced data.
[0013] Furthermore, in step 1, the acquired multi-source high-resolution satellite remote sensing images are preprocessed, specifically including: registration, orthorectification, fusion, mosaicking, and enhancement processing of the acquired multi-source high-resolution satellite remote sensing images.
[0014] Furthermore, in step 2, the interpretation method is human-computer interaction, combining relevant regional background information and common interpretation markers for potash mines to infer and judge different land features and image characteristics on remote sensing images to extract the land use of potash mine areas.
[0015] Furthermore, in step 2, the land use of the potash mine area includes brine ponds, processing plants, mine buildings and others; it also includes identifying the natural geographical environment and socio-economic environment of the potash mine area, as well as the location, quantity, mining status and others of the mining.
[0016] Further, step 3 includes: cropping and dividing the preprocessed multi-source high-resolution satellite remote sensing image into blocks based on the remote sensing interpretation and verification results; extracting image features from the block-based single-band images and training them; identifying typical land features obtained from the training to form interpretation markers; and collecting and managing these markers as samples.
[0017] Further, step 4 includes: using neural networks for classification and edge optimization to improve recognition accuracy, constructing a potash mine land cover recognition model based on multi-source high-resolution satellite remote sensing images, wherein the deep learning images contain models and knowledge for land cover classification; and training the constructed potash mine land cover recognition model based on multi-source high-resolution satellite remote sensing images under the constraint of the target loss function.
[0018] Furthermore, step 4 also includes: cropping and segmenting the preprocessed multi-source high-resolution remote sensing images and inputting them into a trained potash mine land cover identification model based on multi-source high-resolution satellite remote sensing images. The identified typical land covers will be used for surveys and dynamic monitoring.
[0019] The present invention also provides a computing device, including a processor and a memory, wherein the memory is used to execute storage instructions; and the processor is used to execute the aforementioned method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia.
[0020] Compared with the prior art, the beneficial technical effects of the present invention are as follows:
[0021] Relying on a single high-resolution satellite image for environmental element information extraction reduces extraction accuracy. Therefore, this invention, based on domestic high-resolution satellite data, simultaneously incorporates multi-source data for auxiliary judgment to improve information extraction accuracy. This invention utilizes full-band imagery data such as GF-1, GF1-BCD, GF-2, GF-6, and GF-7, combined with multi-source heterogeneous data including DEM data, geological background information, and mineral resource classification data, and selects an appropriate machine learning classifier to extract and verify the accuracy of land use information in potash mining areas.
[0022] This invention aims to identify the mining area and environmental elements of potash mines. Based on domestic high-resolution satellite data and using remote sensing technology, it focuses on the northeastern region of Thailand in Southeast Asia, extracting information on brine ponds, processing plants, mining buildings, and other three types of potash mine land use markers.
[0023] This invention focuses on northeastern Thailand in Southeast Asia, specifically potash mines. Based on domestically produced high-resolution remote sensing data and deep learning technology, it establishes a technical method for land use identification and information extraction in potash mine areas across Southeast Asia. Furthermore, based on relevant data, the method was validated in northeastern Thailand, demonstrating good accuracy. This provides a decision-making reference for my country's international mining cooperation.
[0024] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1A schematic diagram of the process for the extraction method based on open-pit mines provided by this invention;
[0027] Figure 2 This invention provides a method for identifying land use in potash mine areas in Southeast Asia based on remote sensing technology. Detailed Implementation
[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0029] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0030] like Figures 1 to 2 As shown in the figure, this embodiment provides a method for identifying land use in potash mines based on remote sensing technology in Southeast Asia. This method can achieve automated, refined, and highly efficient extraction of land use information in mining areas.
[0031] This method includes the following steps:
[0032] Step 1: Remote sensing data acquisition: Acquire multi-source high-resolution satellite remote sensing images of the target area.
[0033] After acquiring multi-source high-resolution satellite remote sensing images of the target location, the acquired multi-source high-resolution satellite remote sensing images are preprocessed to obtain fused and enhanced data.
[0034] The data preprocessing includes registration, orthorectification, fusion, mosaicking, and enhancement of the acquired multi-source high-resolution satellite remote sensing images.
[0035] Step 2, Remote Sensing Interpretation and Verification: By interpreting the multi-source high-resolution satellite remote sensing images, investigate the environmental conditions of potash mines; using multiple periods of the multi-source high-resolution satellite remote sensing images, conduct an investigation into the land use and mining situation and resource and environmental changes in potash mine areas; and verify the remote sensing interpretation and investigation results; obtain the remote sensing interpretation and verification results.
[0036] The interpretation method involves human-computer interaction, combining relevant regional background information and common interpretation markers for potash mines to infer and judge different land features and image characteristics on remote sensing images to extract land use in potash mine areas.
[0037] The land use of potash mines includes brine ponds, processing plants, and mining buildings.
[0038] It also includes identifying the natural geographical environment and socio-economic environment of potash mining areas, as well as the location, quantity, and mining status of mining operations.
[0039] Step 3: Constructing the sample library and selecting features: Based on the remote sensing interpretation and verification results, assign values to each unit to construct a sample library of potash mine land use based on multi-source high-resolution satellite remote sensing images; establish classification models for spectral features, texture features, and spatial features.
[0040] Specifically, based on the remote sensing interpretation and verification results, the preprocessed multi-source high-resolution satellite remote sensing images are cropped and divided into blocks. Image features are extracted from the single-band images after block division and trained. Typical land features obtained from the training are identified to form interpretation markers, which are then collected and managed as samples.
[0041] Step 4: Automatic identification of land use information based on remote sensing images: Based on the spectral curve characteristics of different land features in different bands, and utilizing spectral, spatial, and temporal features, land feature elements are extracted from remote sensing images using neural networks to obtain the automatic identification results of potash mine land use information.
[0042] Specifically, by utilizing the feature learning capabilities of deep learning models, a land feature identification model for potash mine areas based on multi-source high-resolution satellite remote sensing imagery is constructed. The deep learning imagery contains models and knowledge for land category identification. Under the constraint of the target loss function, the constructed land feature identification model for potash mine areas based on multi-source high-resolution satellite remote sensing imagery is trained.
[0043] The preprocessed multi-source high-resolution remote sensing images are cropped and segmented, and then input into a trained potash mine land cover identification model based on multi-source high-resolution satellite remote sensing images. The identified typical land covers will be used for surveys and dynamic monitoring.
[0044] The specific applications of the method of the present invention are as follows:
[0045] Step 1: Remote sensing data acquisition:
[0046] Acquire multi-source high-resolution satellite remote sensing imagery of the target area. From 2024 to 2025, 265 scenes of high-quality, low-cloud-coverage domestic high-resolution satellite data covering the target area were selected. These include 211 scenes with 2m / 8m resolution (GF-1, GF1-BCD, GF6) and 54 scenes with 0.8m / 2m resolution (GF-2).
[0047] The selected satellite imagery undergoes preprocessing such as registration, orthorectification, fusion, mosaicking, and enhancement to ensure consistent geographic coordinates and minimal errors in the multi-source high-resolution satellite data, which is beneficial for the subsequent construction of the sample library. The next step involves directly inputting the preprocessed multi-source high-resolution satellite data.
[0048] Step 2, Remote Sensing Interpretation and Verification:
[0049] Using GF-1 and GF-6 data with a spatial resolution of 2m, a survey of the geographical environment of potash mines was conducted; using multiple periods of remote sensing imagery, a survey of land use and mining activities and changes in resources and the environment in potash mine areas was conducted; finally, the results of the remote sensing interpretation survey were verified.
[0050] The interpretation method primarily involves human-computer interaction. Based on the different image characteristics of various land features and combined with the image characteristics of domestic potash mines, a universal interpretation marker for potash mines was established. By combining relevant regional background data and interpretation markers, the land use of different potash mine areas on remote sensing images was inferred and determined, mainly including brine ponds, processing plants, mining buildings, and others.
[0051] In high-resolution remote sensing images, the most important feature of potash mines is the brine pools, which are regularly shaped, tightly packed, and green in color, with different shades of green depending on the processing stage of each brine pool.
[0052] Processing plants are typically located near brine pools and consist of a group of mining buildings with blue or bright-colored roofs. They also contain storage tanks, transport pipelines, and processing equipment used for flotation, crystallization, and solid-liquid separation.
[0053] In addition to processing plants, potash mining areas also include other buildings used for offices or other purposes, as well as roads and parking lots.
[0054] Step 3: Constructing the sample library and feature selection:
[0055] Based on the remote sensing interpretation and verification results, brine pools, processing plants, and mining buildings in the preprocessed multi-source high-resolution satellite imagery were identified and labeled. Labeling included land use category and extent. The images were then cropped and divided into training, testing, and validation sets to construct a land use sample library for potash mine areas based on multi-source high-resolution remote sensing imagery.
[0056] Step 4: Automatic identification of land use information based on remote sensing imagery:
[0057] By leveraging the feature learning capabilities of deep learning models, a land cover identification model for potash mine areas is constructed based on multi-source high-resolution remote sensing imagery. The deep learning imagery contains models and knowledge for land cover classification. The constructed identification model is then trained under the constraint of a target loss function.
[0058] The preprocessed multi-source high-resolution remote sensing images are cropped and segmented, and then input into a trained land use identification model for potash mine areas in Southeast Asia based on remote sensing technology. The resulting typical land feature classifications will be used for surveys and dynamic monitoring.
[0059] The extraction results were verified for accuracy. Ultimately, the overall verification rate reached 93.52%, and the overall verification accuracy rate was 80.38%, demonstrating the practicality and accuracy of the method used in this invention.
[0060] Furthermore, step 4 specifically adopts a classification method based on deep learning. The overall technical process is centered on "data-driven + automated iteration", covering six key links: "data preprocessing → sample construction → model training → classification reasoning → result optimization → verification output". This enables end-to-end automated production from raw remote sensing data to full-element classification results, adapting to the full-element identification needs of multiple scenarios such as land use, ecological monitoring, and urban management.
[0061] It should be noted that the application of deep learning in remote sensing image classification essentially simulates the human visual cognitive process through multi-layer neural networks (such as CNN, Transformer, etc.), automatically extracting feature layers from the bottom to the top from raw remote sensing data (multispectral, hyperspectral, SAR, etc.).
[0062] • Low-level features: pixel-level spectral response, brightness differences, and simple textures (such as edges and stripes);
[0063] • Mid-level features: Shape combination and spatial relationship at the target component level (such as building outline and vegetation canopy structure).
[0064] • High-level features: semantic-level category associations (such as "farmland + irrigation facilities = arable land" and "roads + buildings = urban land").
[0065] This end-to-end feature learning model completely eliminates the limitations of "manually designed feature generators"—it does not rely on domain experts to manually select spectral indices (such as NDVI) or texture operators (such as GLCM). Instead, it automatically generates the feature representation that best suits the classification task through data-driven methods, making it particularly suitable for complex scenes of "different spectra for the same object and the same spectrum for different objects" in high-resolution remote sensing images.
[0066] This invention employs a deep learning-based classification method to achieve:
[0067] Fully automated: After inputting the original image and a small number of initial labeled samples, the deep learning method automatically completes feature extraction, classifier training, and result output by iteratively adjusting parameters for different regions and different sensor data through a cycle of "feature selection - threshold adjustment - result correction". That is, from data preprocessing to result output, only initial parameter settings are required, without manual intervention, making it suitable for large-scale, long-term remote sensing monitoring projects (such as national land use dynamic update and global vegetation cover change assessment).
[0068] Highly efficient sample generation: Relying on prior products, massive training samples are automatically generated through a process of "georegistration-pixel-level matching-automatic annotation". Combined with pseudo-labels and data augmentation, this solves the industry pain point of "scarcity of labeled samples" in remote sensing classification (manually annotating a 10km×10km high-resolution image takes several days). On the other hand, by utilizing the spatial coverage integrity of prior products, it ensures the uniform distribution of samples in different landform types (mountains, plains, wetlands, etc.) and different climate zones, avoiding classification bias caused by sample bias (e.g., a model trained only with urban samples cannot accurately identify rural land use).
[0069] Highly adaptable: Supports multi-source data (optical / SAR / LiDAR), multi-scale imagery, and multi-scene full-element classification, and can be flexibly adapted to different computing power and business needs through model selection;
[0070] High-precision classification: Deep learning models' ability to capture deep semantic features makes them stand out in complex scene classification. For example, challenging problems such as "mixed pixels of vegetation and bare soil" in high-resolution images, "distinguishing between small buildings and sheds," and "identifying water bodies and shadows in SAR images" can be solved more accurately by utilizing the local receptive field of CNNs and the global attention mechanism of Transformers, achieving more precise boundary segmentation and category determination. Practical data shows that, under the same data source conditions, the overall accuracy (OA) of deep learning classification is typically 10%-20% higher than that of traditional SVMs and maximum likelihood methods, with a Kappa coefficient improvement of more than 0.15.
[0071] In addition, the present invention also provides a computing device, including a processor and a memory. The memory is used to execute storage instructions; the processor is used to execute a land use identification method for potash mine areas based on remote sensing technology, targeting Southeast Asia.
[0072] Therefore, this invention provides a method and equipment for identifying potash mine land use in Southeast Asia based on remote sensing technology. Focusing on northeastern Thailand in Southeast Asia and potash mines as the research object, this invention establishes a technical method for identifying potash mine land use and extracting information based on domestically produced high-resolution remote sensing data and deep learning technology. Furthermore, based on relevant data, the method for identifying potash mine land use in Southeast Asia based on remote sensing technology was verified in northeastern Thailand, showing good accuracy. This provides a decision-making reference for my country's international mining cooperation.
[0073] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.
Claims
1. A method for identifying land use in potash mine areas based on remote sensing technology in Southeast Asia, characterized in that, Includes the following steps: Step 1, Remote Sensing Data Acquisition: Acquire multi-source high-resolution satellite remote sensing images of the target area; Step 2, Remote Sensing Interpretation and Verification: By interpreting the multi-source high-resolution satellite remote sensing images, investigate the environmental conditions of potash mines; using multiple periods of the multi-source high-resolution satellite remote sensing images, conduct an investigation into the land use and mining situation and resource and environmental changes in potash mine areas; and verify the remote sensing interpretation and investigation results; obtain the remote sensing interpretation and verification results. Step 3: Constructing the sample library and selecting features: Based on the remote sensing interpretation and verification results, assign values to each unit to construct a sample library of potash mine land use based on multi-source high-resolution satellite remote sensing images; establish classification models for spectral features, texture features, and spatial features; Step 4: Automatic identification of land use information based on remote sensing images: Based on the spectral curve characteristics of different land features in different bands, and utilizing spectral, spatial, and temporal features, land feature elements are extracted from remote sensing images using neural networks to obtain the automatic identification results of potash mine land use information.
2. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 1, characterized in that, In step 1, after acquiring multi-source high-resolution satellite remote sensing images of the target location, the acquired multi-source high-resolution satellite remote sensing images are preprocessed to obtain fused and enhanced data.
3. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 2, characterized in that, In step 1, the acquired multi-source high-resolution satellite remote sensing images are preprocessed, specifically including: registration, orthorectification, fusion, mosaicking, and enhancement processing of the acquired multi-source high-resolution satellite remote sensing images.
4. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 1, characterized in that, In step 2, the interpretation method is human-computer interaction. By combining relevant regional background information and common interpretation markers for potash mines, we can infer and judge different land features and image characteristics on the remote sensing images to extract the land use of the potash mine area.
5. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 4, characterized in that, Step 2 includes the land use of the potash mine area, including brine ponds, processing plants, mine buildings and others; it also includes identifying the natural geographical environment and socio-economic environment of the potash mine area, as well as the location, quantity, mining status and others of the mining.
6. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 2, characterized in that, Step 3 includes: based on the remote sensing interpretation and verification results, cropping and dividing the preprocessed multi-source high-resolution satellite remote sensing images into blocks, extracting image features from the block-based single-band images, training them, identifying typical land features obtained from the training to form interpretation markers, and using them as samples for collection and management.
7. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 1, characterized in that, Step 4 includes: utilizing the feature learning capabilities of deep learning models to construct a potash mine land cover identification model based on multi-source high-resolution satellite remote sensing images, wherein the deep learning images contain models and knowledge for land cover identification; and training the constructed potash mine land cover identification model based on multi-source high-resolution satellite remote sensing images under the constraint of the target loss function.
8. The method for identifying land use in potash mine areas based on remote sensing technology for Southeast Asia according to claim 7, characterized in that, Step 4 also includes: cropping and dividing the preprocessed multi-source high-resolution remote sensing images into blocks, and inputting them into the trained potash mine land cover identification model based on multi-source high-resolution satellite remote sensing images. The identified typical land covers will be used for surveys and dynamic monitoring.
9. A computing device, characterized in that, It includes a processor and a memory, the memory being used to execute storage instructions; the processor being used to execute the remote sensing-based potash mine land use identification method for Southeast Asia as described in any one of claims 1-8.