Natural resource remote sensing intelligent interpretation sample database construction method

Abstract

The invention relates to the field of data processing, and relates to a natural resource remote sensing intelligent interpretation sample database construction method, which comprises the following steps: setting a classification standard of natural resource remote sensing interpretation sample data; the classification standard comprises a large class, a first-level class and a second-level class; the large classes comprise a natural land cover class, an artificial target class, a change detection class and a crop class; collecting image sample data, and making the image sample data into an image sample set according to the classification standard; setting a database concept structure, a database logic structure and a database table structure, and generating a database; preprocessing the image sample set; and storing the preprocessed image sample set into a corresponding database table. By the adoption of the method, the built remote sensing interpretation sample database is complete in sample type, and a larger number of sample data sets can be stored; in addition, the robustness of the remote sensing interpretation model can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and more specifically, to a method for constructing a natural resource remote sensing intelligent interpretation sample database. Background Art

[0002] In today's society, with the development of science and technology, the collection and application of remote sensing images have become increasingly important in many fields. In the field of environmental monitoring and protection, remote sensing image data can be used for climate change monitoring, pollution monitoring and ecosystem protection; in climate change monitoring, remote sensing images can record information such as the spectrum and temperature of the earth's surface, providing scientists with key data on changes in the earth system, which is helpful for formulating environmental policies and responding to climate change challenges. In pollution monitoring, remote sensing images can be used to monitor environmental problems such as air pollution, water pollution, and soil pollution, providing strong data support for environmental protection. In ecosystem protection, remote sensing images can also be used to monitor ecological parameters such as forest coverage and grassland conditions, providing important information for protecting ecosystems and wildlife habitats. In the field of agriculture and agricultural management, remote sensing image data can be used for crop growth monitoring, soil moisture monitoring, and yield estimation. In the field of natural resource management, remote sensing images can be used for resource exploration and resource protection.

[0003] In order to facilitate the use of remote sensing images, it is necessary to use remote sensing interpretation models to interpret remote sensing images, obtain remote sensing interpretation sample data, and store them in the established database for easy access. However, the database used to store remote sensing interpretation sample data in the prior art has the problem of insufficient diversity, and the preset sample types are difficult to fully cover the diversity of regional or global surface environments. When encountering new categories (open set problems), the interpretation system of the remote sensing image interpretation model cannot be flexibly expanded, resulting in misjudgment; in addition, the existing database storage space for remote sensing interpretation sample data is small, and it is difficult to store a huge number of sample data sets. Furthermore, the lack of geographic location and time attributes weakens the robustness of the remote sensing interpretation model. Summary of the invention

[0004] In order to solve the technical problems existing in the above-mentioned existing remote sensing interpretation sample database, the present invention provides solutions in the following aspects.

[0005] In a first aspect, the present invention provides a method for constructing a natural resource remote sensing intelligent interpretation sample database, comprising:

[0006] Setting classification standards for natural resource remote sensing interpretation sample data; the classification standards include major categories, primary categories and secondary categories; the major categories include: natural surface cover category, artificial target category, change detection category and crop category;

[0007] Collect image sample data and make it into an image sample set according to the classification standard; the image sample set includes several groups of sample patches and sample set metadata, and the sample set metadata is used to record the correspondence between the entire group of image sample sets and the corresponding image information and classification information; a group of sample patches includes image patches, label patches and patch metadata, the image patches refer to the image data of regular size after cutting, the label patches refer to the rasterized sample category label data corresponding to the image patches, and the patch metadata is used to record the sample patch attribute information, and the sample patch attribute information includes the name of a single group of sample patch data and the effective proportion of the sample;

[0008] Create a database, including: setting up the database conceptual structure, database logical structure and database table structure, and generating the database;

[0009] Preprocessing the image sample set includes: organizing the sample data according to the data storage specification and relevant format requirements, and performing consistency conversion on the contents of the image sample set to be stored, including the naming of the sample data to be stored, field structure adjustment and metadata item information filling operation;

[0010] The preprocessed image sample set is stored in the corresponding database table.

[0011] Preferably, the first-level categories belonging to the natural surface cover class include: cultivated land, gardens, woodlands, grasslands, waters and unused land; the first-level categories belonging to the artificial target class include: buildings, public service facilities, transportation, industrial and energy facilities and typical target elements; the first-level categories belonging to the change detection class include new / lost fill, new / lost structures, new / lost cultivated land, new / lost roads, new / lost water surfaces, new / lost ditches, new / lost gardens, new / lost woodlands, new / lost open-pit mining and new / lost wetlands; the first-level categories belonging to the crop class include: food crops, cash crops, vegetables and fruits.

[0012] Preferably, the secondary categories belonging to the cultivated land category include: paddy fields, dry land and greenhouses; the secondary categories belonging to the forest land category include: trees, shrubs and bamboo forests; the secondary categories belonging to the grassland category include: artificial grassland and other grasslands; the secondary categories belonging to the water area category include: water surface, ditch and inland tidal flat; the secondary categories belonging to the unused land category include: bare rock gravel and gravel; the secondary categories belonging to the building category include: residential and industrial plants; the secondary categories belonging to the public service facilities category include: open-air stadiums, basketball courts, tennis courts and hardened surfaces; the secondary categories belonging to the industrial The second-level categories of the energy facilities category include: wind power generation facilities, photovoltaic power generation facilities, dams, chimneys, substations, towers and oil storage tanks; the second-level categories of the typical target elements category include: greenhouses, cars and airplanes; the second-level categories of the transportation category include: roads and railways; the second-level categories of the food crops category include: wheat, corn, rice, peanuts, millet, sorghum, sweet potatoes and soybeans; the second-level categories of the cash crops category include: garlic, peppers, tobacco leaves, onions, watermelons, beans, carrots, cabbage, asparagus, yam, rehmannia and ginger.

[0013] Preferably, the content of the small piece metadata includes: sample small piece number, sample size and effective proportion; the content of the sample set metadata includes: sample information, image information and area information.

[0014] Preferably, the sample information includes: name, image clip, label clip, sample clip metadata, sample size, major category code, major category name, first-level category code, first-level category name, second-level category code, second-level category name, sample time and spatial reference; the image information includes: image type, image resolution, image satellite and image sensor; the regional information includes: province, city, county and terrain.

[0015] Preferably, the method for designing the conceptual structure of the database includes:

[0016] Classify and extract sample data of intelligent interpretation of various remote sensing images;

[0017] Determine the interpreted sample entity in the local application, the attributes of the sample and the corresponding identification number;

[0018] Determine the connections and types between entities in remote sensing image intelligent interpretation samples.

[0019] Preferably, the database table includes an image sample table, an image sample set table, a classification system table and a sample area table.

[0020] Preferably, the image sample table includes a sample data table, a sample metadata table and a sample image attribute table, the sample data table is used to record the sample ID, sample image data, vector data and range data ID; the sample metadata table is used to record various metadata information of the sample, and the metadata information includes the sample classification system, sample type, sample resolution, sample image attribute ID, sample spatial range, sample area ID, sample production information; the sample image attribute table is used to record various attribute information of samples of different image types and record the sample metadata ID.

[0021] Preferably, the image sample set table includes a sample set data table, a sample set small piece data table, a sample set small piece metadata table, a sample set metadata table and a sample set image attribute table; the information recorded in the sample set data table includes a sample set ID, a sample set name and the number of samples in the sample set; the information recorded in the sample set small piece data table includes: a single small piece sample ID, a sample set ID, a sample number, sample image data and a label data ID; the information recorded in the sample set small piece metadata table includes: a single small piece sample ID, a sample size, a category list and an effective proportion; the sample set metadata table is used to record various metadata information of the sample, including: a sample set classification system, a sample type, a sample resolution, a sample set image attribute ID, a sample set spatial range, a sample area ID and a sample set production information; the information recorded in the sample set image attribute table includes: a sample set metadata ID, an image attribute unique identifier, a remote sensing image type, the number of bands contained in the image, the composition information of the image bands, the image pixel storage type, the image pixel bit depth, the image shooting payload platform, the satellite platform for image acquisition and the sensor type for image acquisition.

[0022] Preferably, the classification system table includes: a major category data table, a first-level category data table and a second-level category data table; the information recorded in the major category data table includes: category ID, category code, category name, category description and color value; the information recorded in the first-level category data table and the second-level category data table includes: category ID, category code, category name, category description, parent category ID and color value.

[0023] The beneficial effects of the present invention are as follows: by adopting the natural resource remote sensing intelligent interpretation sample database construction method of the present invention, the sample types of the constructed remote sensing interpretation sample database are relatively complete, the structural design of the database table is relatively simple and reasonable, and a larger number of sample data sets can be stored; in addition, when using image sample data to produce an image sample set, geographic location attributes and time attributes are added to the content of the sample set, thereby improving the robustness of the remote sensing interpretation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1 is a flowchart schematically showing a method for constructing a natural resource remote sensing intelligent interpretation sample database according to an embodiment of the present invention;

[0026] Figure 2 is a schematic diagram schematically showing a classification coding specification according to an embodiment of the present invention;

[0027] Figure 3 is a schematic diagram schematically showing the organization of image sample data according to an embodiment of the present invention;

[0028] Figure 4 is a schematic diagram schematically showing the organization of an image sample set according to an embodiment of the present invention;

[0029] Figure 5 is a schematic diagram schematically showing the overall logical structure of a database according to an embodiment of the present invention;

[0030] Figure 6 is a schematic diagram schematically showing the logical structure of an image sample library according to an embodiment of the present invention;

[0031] Figure 7 is a schematic diagram schematically showing the logical structure of an image sample library according to an embodiment of the present invention;

[0032] Figure 8 is a schematic diagram schematically showing a conceptual structure of image sample data according to an embodiment of the present invention;

[0033] Fig. 9 FIG. 1 is a schematic diagram schematically showing the conceptual structure of image sample set data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0035] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0036] Example of method for constructing a natural resource remote sensing intelligent interpretation sample database:

[0037] like Figure 1 As shown, the method for constructing a natural resource remote sensing intelligent interpretation sample database of the present invention comprises:

[0038] S101, setting classification standards for remote sensing interpretation sample data, specifically: setting classification standards for natural resource remote sensing interpretation sample data; the classification standards include major categories, primary categories and secondary categories; the major categories include: natural surface cover category, artificial target category, change detection category and crop category;

[0039] When formulating the sample classification system standard, we should not only pay attention to the actual situation of the ground objects, but also take into account the spectral information of different ground objects, and use physical, chemical and biological indicators in a comprehensive manner. We should fully consider the application of all business departments involved to ensure the feasibility and applicability of the sample classification standard, and the design of the classification system should be scalable.

[0040] The classification codes for major categories, first-level categories, and second-level categories are composed of two-digit, four-digit, and six-digit hierarchical codes, respectively, reflecting the level and subordinate relationship of each category. Its structure is as follows Figure 2 As shown. Among them, the major categories are the major categories that can be distinguished by remote sensing, and are set with two-digit codes. The natural surface cover category is coded as 01, the artificial target category is coded as 02, the change detection category is 03, and the crop category is 04. The first-level category is the basic sample element category, which is set as a four-digit code. The second-level category is the detailed sample element category, which is set as a six-digit code.

[0041] Each major category includes multiple first-level categories, and each first-level category includes multiple second-level categories. In this embodiment, the sample major category classification content is shown in Table 1. The natural surface cover category includes 6 first-level categories and 14 second-level categories; the artificial target category includes 5 first-level categories and 16 second-level categories; the change detection category includes 12 first-level categories and 19 second-level categories; the crop category includes 2 first-level categories and 20 second-level categories.

[0042] Table 1

[0043] Serial number Category Name Number of first-class Number of secondary categories 1 Land cover 6 14 2 Artificial Target 5 16 3 Change Detection 12 19 4 crop 2 20

[0044] In this embodiment, the first-level category belonging to the natural surface cover category includes: cultivated land, garden, forest, grassland, water and unused land; the second-level category belonging to the cultivated land category includes: paddy field, dry land and greenhouse; the second-level category belonging to the forest category includes: trees, shrubs and bamboo forest; the second-level category belonging to the grassland category includes: artificial grassland and other grassland; the second-level category belonging to the water category includes: water surface, ditch and inland tidal flat; the second-level category belonging to the unused land category includes: bare rock gravel and gravel; greenhouse includes greenhouse and vegetable greenhouse; garden includes other orchards, such as tea garden and vineyard; water surface includes: river, lake, reservoir, pond and other water surface; ditch includes: main ditch, branch ditch, general ditch. The classification codes of each category under the natural surface cover category are shown in Table 2.

[0045] Table 2

[0046]

[0047] The first-level categories belonging to the artificial target class include: buildings, public service facilities, transportation, industrial and energy facilities, and typical target elements.

[0048] The second-level categories belonging to the building category include: residential buildings and industrial plants; the second-level categories belonging to the public service facility category include: open-air stadiums, basketball courts, tennis courts, and hardened surfaces; the second-level categories belonging to the industrial and energy facility category include: wind power generation facilities, photovoltaic power generation facilities, dams, chimneys, substations, towers, and oil storage tanks; the second-level categories belonging to the typical target element category include: greenhouses, cars, and airplanes; the second-level categories belonging to the transportation category include: roads and railways. The classification codes of each category of the artificial target class are shown in Table 3.

[0049] Table 3

[0050]

[0051] The first-level categories belonging to the change detection class include new / lost fill, new / lost buildings, new / lost cultivated land, new / lost roads, new / lost water surfaces, new / lost ditches, new / lost gardens, new / lost forest land, new / lost open-pit mining and new / lost wetlands; the classification codes of each first-level category and second-level category of the change detection class, as well as the second-level categories contained in each first-level category are shown in Table 4.

[0052] Table 4

[0053]

[0054]

[0055] The first-level categories belonging to the crop category include: food crops, cash crops, vegetables and fruits. Among them, the second-level categories belonging to the food crop category include: wheat, corn, rice, peanuts, millet, sorghum, sweet potato and soybean, and the second-level categories belonging to the cash crop category include: garlic, pepper, tobacco, green onion, watermelon, beans, carrots, cabbage, asparagus, yam, rehmannia and ginger. The classification codes of each first-level category and second-level category included in the crop category are shown in Table 5.

[0056] Table 5

[0057]

[0058]

[0059] S102, collecting image sample data and making an image sample set, specifically: collecting image sample data, and making it into an image sample set according to the classification standard; the image sample set includes several groups of sample patches and sample set metadata, and the sample set metadata is used to record the correspondence between the entire group of image sample sets and the corresponding image information and classification information; a group of sample patches includes image patches, label patches and patch metadata, the image patches refer to the image data of regular size after cutting, the label patches refer to the rasterized sample category label data corresponding to the image patches, and the patch metadata is used to record the sample patch attribute information, and the sample patch attribute information includes the name of a single group of sample patch data and the effective proportion of samples;

[0060] Image sample data consists of image data, sample annotation data, image sample valid range, and sample metadata. Image data is a remote sensing image data entity. Natural surface cover class, artificial target feature class, and other types of land features are represented by a single remote sensing image. Surface change features are represented by two periods of remote sensing images. The image data format is .tif format. Sample annotation data is vector data collected based on image data, and the general format is .shp. The image sample valid range is the matching annotated area vector data. The sample valid range is optional. If it does not exist, the default is to collect all samples within the image range. The valid range file format is .shp. Sample metadata is a file that describes its corresponding relationship and basic information of the image sample for data library management. The metadata file is in .xml format.

[0061] The data of the image sample data is based on the image name. The sample vector is named "image name + _sample", the valid range is named "image name + _region", and the sample metadata is named "image name + _meta". The change detection sample images are suffixed with '_before' and '_after' to distinguish the images before and after. The image sample data is organized as follows: Figure 3 shown.

[0062] The contents of the sample metadata are shown in Table 6.

[0063] Table 6

[0064]

[0065]

[0066] The sample collection method should adopt advanced design concepts, make full use of advanced technical means, and adopt stable and reliable software and hardware environment, technical design and standards. The sample data coordinates use the "2000 National Geodetic Coordinate System" (latitude and longitude), and the elevation uses the "1985 National Elevation Datum". The framing method and numbering of the standard framing data used should comply with the provisions of GB / T 13989. The date when the sample data is collected should use the Gregorian calendar era, and the time should use Beijing time. The format of the date and time attribute values ​​should be correct. The time measurement of various types of data should be accurate and reliable, and meet the requirements of data currency.

[0067] The definition and attribute value filling of the attribute items of the sample collection results data should comply with the specified requirements, and the attribute values ​​of the edge elements should be consistent. The organization, format and naming of the sample data should comply with the specified requirements, and the data files should be complete and correct. All types of samples required in the specification should be complete without omissions or redundancies. When making samples, the uniform distribution of training samples across the province and within the time range should also be considered.

[0068] The metadata information should be filled in in accordance with the specified requirements. The metadata items are divided into three types according to "constraints / conditions": "M" is a mandatory item; "C" is a conditional mandatory item; "O" is an optional item. The metadata items should be filled in truthfully, accurately and completely. When there is no value for a metadata item, "None" should be filled in. The metadata file uses a plain text format described by .xml, and the metadata item tags in the .xml file use the English name of the metadata item.

[0069] In this embodiment, the content of the small piece metadata includes: sample small piece number, sample size and effective proportion; as shown in Table 7, the content of the sample set metadata includes: sample information, image information and regional information. The sample information includes: name, image small piece, label small piece, sample small piece metadata, sample size, major category code, major category name, first-level category code, first-level category name, second-level category code, second-level category name, sample time and spatial reference; the image information includes: image type, image resolution, image satellite and image sensor; the regional information includes: province, city, county and terrain.

[0070] Table 7

[0071]

[0072]

[0073]

[0074] The image sample set needs to be organized and stored in the form of folders. Image patches, label patches, and patch metadata are stored in different folders with the same name. Surface change image patches are stored in different folders in the image patch folders according to the time period. The organization form of the image sample set is as follows: Figure 4 shown.

[0075] The content of the small piece metadata is shown in Table 8.

[0076] Table 8

[0077]

[0078] S103, creating a database, specifically including: setting a database conceptual structure, a database logical structure and a database table structure, and generating a database;

[0079] The purpose of setting up a database conceptual structure is to classify the various data managed by the database, comprehensively determine the entities in the database management system, the attributes of the entities, the codes that identify the entities, and the types of connections between entities, and establish an abstract conceptual data model.

[0080] The database conceptual design should adopt a bottom-up conceptual design method. First, abstract the data to form a local conceptual design, and then integrate the local design to form an overall conceptual model view. The conceptual model should reflect the information structure, information flow, mutual constraints and associations between information, and user requirements for information storage, query and application, and should not rely on a specific database management system.

[0081] The database logical structure needs to be designed based on the database conceptual structure, clarify the organization of remote sensing image data and metadata information, and establish a logical structure that conforms to the data model. The design process of the database logical structure includes:

[0082] A. Data Association and Organization

[0083] The optimized data structure and organization method should be used to establish logical associations between database entities to reduce data redundancy. The intelligent interpretation sample database can organize sample data according to the following organizational principles.

[0084] (1) Sample data organization: The database is divided into different databases according to sample types, and different sample type databases are organized separately according to sample entity data and metadata;

[0085] (2) Entity data organization: organized and stored in the form of files;

[0086] (3) Metadata organization: Use database table organization to establish associations with corresponding sample entity data, and realize the association between metadata and entities through sample IDs.

[0087] B. Generate database logical structure

[0088] The interpretation sample database is logically divided into image sample library, image sample set library, and object sample library. Each sample sub-library consists of sample entity data and metadata information. The overall logical structure of the database is as follows: Figure 5 shown.

[0089] The image sample library consists of entity image sample data and sample metadata. Entity data includes sample images and sample vectors. If the sample image is a change monitoring sample, it should include two images from the previous and next periods, and the sample vector is vector data; metadata, as sample description data, includes sample attributes, sample classification information, image information, spatial location information, sample area information, and production information. The logical structure of the image sample library is as follows: Figure 6 shown.

[0090] The image sample set library consists of entity small-piece sample data, small-piece set metadata and small-piece metadata. Entity small-piece sample data includes image small pieces and label small pieces. Small-piece set metadata records various attribute information of the sample set, including sample attributes, sample classification information, image information, spatial location information, sample area information and production information. Small-piece metadata records the independent attribute information of each small-piece sample. The logical structure of the image sample set library is as follows: Figure 7 shown.

[0091] The database table structure is an important basis for the storage management, rendering display, query retrieval, and download application of the collected sample set results data in the sample library. A good design can ensure fast and effective data acquisition. In this embodiment, the database table includes an image sample table, an image sample set table, a classification system table, and a sample area table.

[0092] The image sample table includes a sample data table, a sample metadata table and a sample image attribute table. The sample data table is used to record sample ID, sample image data, vector data and range data ID; the sample data table is shown in Table 9.

[0093] Table 9

[0094]

[0095] The sample metadata table is used to record various metadata information of the sample, including sample classification system, sample type, sample resolution, sample image attribute ID, sample space range, sample area ID, and sample production information; the sample image attribute table is used to record various attribute information of samples of different image types and record the sample metadata ID. The table structure of the sample metadata table is shown in Table 10.

[0096] Table 10

[0097]

[0098]

[0099] The sample image attribute table is used to record the attribute information of samples of different image types and record the sample metadata ID. The sample metadata image information part is linked by recording the image attribute unique ID in the sample metadata table. The specific table structure of the sample image attribute table is shown in Table 11.

[0100] Table 11

[0101]

[0102] The image sample set table includes a sample set data table, a sample set patch data table, a sample set patch metadata table, a sample set metadata table and a sample set image attribute table.

[0103] The information recorded in the sample set data table includes information such as the sample set ID, the sample set name, and the number of samples in the sample set; the table structure of the sample set data table is shown in Table 12.

[0104] Table 12

[0105]

[0106] The sample set small piece data table is used to record information including: single small piece sample ID, sample set ID, sample number, sample image data and label data ID; the table structure of the sample set small piece data table is shown in Table 13.

[0107] Table 13

[0108]

[0109] The sample set small piece metadata table is used to record information including: single small piece sample ID, sample size, category list and effective proportion; the table structure of the sample set small piece metadata table is shown in Table 14.

[0110] Table 14

[0111]

[0112]

[0113] The sample set metadata table is used to record various metadata information of the sample, including: sample set classification system, sample type, sample resolution, sample set image attribute ID, sample set spatial range, sample area ID and sample set production information; the table structure of the sample set metadata table is shown in Table 15.

[0114] Table 15

[0115]

[0116] The sample set image attribute table is used to record information including: sample set metadata ID, image attribute unique identifier, remote sensing image type, number of bands contained in the image, composition information of image bands, image pixel storage type, image pixel depth, image shooting payload platform, satellite platform for image acquisition, and sensor type for image acquisition. The table structure of the sample set image attribute table is shown in Table 16.

[0117] Table 16

[0118]

[0119] The classification system table includes: a major category data table, a first-level category data table and a second-level category data table; the major category data table is used to record information including: category ID, category code, category name, category description and color value; the table structure of the major category data table is shown in Table 17.

[0120] Table 17

[0121]

[0122] The information recorded in the primary class data table and the secondary class data table includes: category ID, category code, category name, category description, parent category ID and color value. The table structure of the primary class data table is shown in Table 18. The table structure of the secondary class data table is shown in Table 19.

[0123] Table 18

[0124]

[0125] Table 19

[0126]

[0127] The sample area table is used to record information such as sample area ID, sample area code, sample area type, terrain, climate, and hydrological characteristics. The specific table structure is shown in Table 20.

[0128] Table 20

[0129]

[0130] S104, preprocessing the image sample set, the preprocessing includes: arranging the sample data according to the data storage specification and relevant format requirements, and performing consistency conversion on the contents of the image sample set to be stored, including the naming of the sample data to be stored, field structure adjustment and metadata item information filling operation;

[0131] S105: Store the preprocessed image sample set into a corresponding database table.

[0132] Before storing the preprocessed image sample set in the corresponding database table, the image sample set needs to be quality checked before storage. Specifically, according to the GBT41149-2021 standard, general checks are carried out on spatial reference system, logical consistency, accuracy, integrity, topology and representation, image quality, and metadata. The specific inspection can be carried out by batch inspection with auxiliary script programs and manual interactive review and verification schemes until the samples meet the inspection requirements.

[0133] When entering the database, qualified samples are stored in the corresponding storage space of the database according to the specific sample library organization form. The specific entry process is mainly carried out by auxiliary scripts, supplemented by manual addition. In addition, log files are created and saved for operations before and after entry to avoid erroneous operations affecting database security.

[0134] After the image sample set is stored in the database, a post-storage inspection can be carried out to review the data after storage, including review of the data storage path, data integrity check, and consistency check of image data and label files. In addition, a corresponding daily maintenance script program is established to ensure the correctness of the database samples.

[0135] By adopting the natural resource remote sensing intelligent interpretation sample database construction method of the present invention, the sample types of the constructed remote sensing interpretation sample database are relatively complete, the structural design of the database table is relatively simple and reasonable, and a larger number of sample data sets can be stored; in addition, when using image sample data to produce an image sample set, geographic location attributes and time attributes are added to the content of the sample set, thereby improving the robustness of the remote sensing interpretation model.

[0136] A sample database with complete types and huge quantity suitable for natural resource remote sensing interpretation sample data can be constructed, thereby making it easier to store the collected natural resource remote sensing interpretation sample data.

[0137] In one embodiment, the method for designing the conceptual structure of the database includes:

[0138] S201, classify and extract sample data of intelligent interpretation of various remote sensing images;

[0139] S202, determining the interpreted sample entity in the local application, the attributes of the sample and the corresponding identification number;

[0140] S203, determining the connections and types between entities of the remote sensing image intelligent interpretation sample.

[0141] The designed database conceptual structure includes image sample data conceptual structure and image sample set data conceptual structure.

[0142] For image sample data resources, the main content includes three parts: the data entity file of the image sample, the metadata file associated with the data, and the attribute information and spatial information of the data. The designed conceptual structure of image sample data is as follows: Figure 8 shown.

[0143] For the image sample set data resource, the main content includes four parts: the data entity file of the image sample set, the metadata file associated with the data set, the metadata file associated with the small piece of data, and the attribute information and spatial information of the data. The designed conceptual structure of the image sample set data is as follows: Fig. 9 shown.

[0144] In one embodiment, the database table further includes an administrative division table, which is composed of a province data table, a city data table, and a county data table.

[0145] The provincial data table is used to record information such as provincial ID, administrative code, and provincial name. The specific table structure is shown in Table 21.

[0146] Table 21

[0147]

[0148] The city data table is used to record information such as city ID, administrative code, city name, and province ID. The specific table structure is shown in Table 22.

[0149] Table 22

[0150]

[0151] The county data table is used to record information such as county ID, administrative code, county name, and city ID. The specific table structure is shown in Table 23.

[0152] Table 23

[0153]

[0154] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0155] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for constructing a natural resource remote sensing intelligent interpretation sample database, characterized in that: include: Set classification standards for natural resource remote sensing interpretation sample data; The classification standard includes major categories, primary categories and secondary categories; the major categories include: natural land cover category, artificial target category, change detection category and crop category; Collect image sample data and make it into an image sample set according to the classification standard; the image sample set includes several groups of sample patches and sample set metadata, and the sample set metadata is used to record the correspondence between the entire group of image sample sets and the corresponding image information and classification information; a group of sample patches includes image patches, label patches and patch metadata, the image patches refer to the image data of regular size after cutting, the label patches refer to the rasterized sample category label data corresponding to the image patches, and the patch metadata is used to record the sample patch attribute information, and the sample patch attribute information includes the name of a single group of sample patch data and the effective proportion of the sample; Create a database, including: setting up the database conceptual structure, database logical structure and database table structure, and generating the database; Preprocessing the image sample set includes: organizing the sample data according to the data storage specification and relevant format requirements, and performing consistency conversion on the contents of the image sample set to be stored, including the naming of the sample data to be stored, field structure adjustment and metadata item information filling operation; The preprocessed image sample set is stored in the corresponding database table.

2. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 1, characterized in that: The first-level categories belonging to the natural surface cover category include: cultivated land, gardens, woodlands, grasslands, waters and unused land; the first-level categories belonging to the artificial target category include: buildings, public service facilities, transportation, industrial and energy facilities and typical target elements; the first-level categories belonging to the change detection category include new / lost fill, new / lost structures, new / lost cultivated land, new / lost roads, new / lost water surfaces, new / lost ditches, new / lost gardens, new / lost forest land, new / lost open-pit mining and new / lost wetlands; the first-level categories belonging to the crop category include: food crops, cash crops, vegetables and fruits.

3. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 2, characterized in that: The second-level categories of cultivated land include paddy fields, dry land and greenhouses; the second-level categories of forest land include trees, shrubs and bamboo forests; the second-level categories of grassland include artificial grassland and other grasslands; the second-level categories of water areas include water surfaces, ditches and inland beaches; the second-level categories of unused land include bare rock gravel and gravel; the second-level categories of buildings include residential buildings and industrial plants; the second-level categories of public service facilities include open-air stadiums, basketball courts, tennis courts and hardened surfaces; the second-level categories of industrial and energy The secondary categories of the power facility category include: wind power facilities, photovoltaic power generation facilities, dams, chimneys, substations, towers and oil storage tanks; the secondary categories of the typical target element category include: greenhouses, cars and airplanes; the secondary categories of the transportation category include: roads and railways; the secondary categories of the food crop category include: wheat, corn, rice, peanuts, millet, sorghum, sweet potatoes and soybeans; the secondary categories of the cash crop category include: garlic, peppers, tobacco leaves, onions, watermelons, beans, carrots, cabbage, asparagus, yam, rehmannia and ginger.

4. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 1, characterized in that: The content of the small piece metadata includes: sample small piece number, sample size and effective proportion; the content of the sample set metadata includes: sample information, image information and area information.

5. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 4, characterized in that: The sample information includes: name, image clip, label clip, sample clip metadata, sample size, major category code, major category name, first-level category code, first-level category name, second-level category code, second-level category name, sample time and spatial reference; the image information includes: image type, image resolution, image satellite and image sensor; the regional information includes: province, city, county and terrain.

6. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 1, characterized in that: The design method of the conceptual structure of the database includes: Classify and extract sample data of intelligent interpretation of various remote sensing images; Determine the interpreted sample entity in the local application, the attributes of the sample and the corresponding identification number; Determine the connections and types between entities in remote sensing image intelligent interpretation samples.

7. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to any one of claims 1 to 6, characterized in that: The database tables include an image sample table, an image sample set table, a classification system table, and a sample area table.

8. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 7, characterized in that: The image sample table includes a sample data table, a sample metadata table and a sample image attribute table. The sample data table is used to record the sample ID, sample image data, vector data and range data ID; the sample metadata table is used to record various metadata information of the sample, and the metadata information includes the sample classification system, sample type, sample resolution, sample image attribute ID, sample spatial range, sample area ID, and sample production information; the sample image attribute table is used to record various attribute information of samples of different image types and record the sample metadata ID.

9. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 7, characterized in that: The image sample set table includes a sample set data table, a sample set small piece data table, a sample set small piece metadata table, a sample set metadata table and a sample set image attribute table; the information recorded in the sample set data table includes a sample set ID, a sample set name and the number of samples in the sample set; The sample set small piece data table is used to record information including: single small piece sample ID, sample set ID, sample number, sample image data and label data ID; the sample set small piece metadata table is used to record information including: single small piece sample ID, sample size, category list and effective proportion; the sample set metadata table is used to record various metadata information of the sample, including: sample set classification system, sample type, sample resolution, sample set image attribute ID, sample set spatial range, sample area ID and sample set production information; the sample set image attribute table is used to record information including: sample set metadata ID, image attribute unique identifier, remote sensing image type, number of bands contained in the image, composition information of image bands, image pixel storage type, image pixel bit depth, image shooting payload platform, satellite platform for image acquisition and sensor type for image acquisition.

10. The method for constructing a natural resource remote sensing intelligent interpretation sample database according to claim 7, characterized in that: The classification system table includes: a major category data table, a first-level category data table and a second-level category data table; the major category data table is used to record information including: category ID, category code, category name, category description and color value; the first-level category data table and the second-level category data table are used to record information including: category ID, category code, category name, category description, parent category ID and color value.

Images