Land resource utilization identification method and system, storage medium and equipment

Abstract

The invention discloses a land resource utilization identification method and system, a storage medium and equipment. The method comprises the steps of performing classification training on remote sensing image sample data according to land class codes to generate a remote sensing image model; performing classification training on the field proof photo sample data according to the land class codes to generate a field photo model; generating the business data into a spatial knowledge graph database according to the spatial position relationship; gridding processing is carried out on a to-be-identified photo, remote sensing image identification aiming at a remote sensing image model, proof photo identification aiming at a field photo model and vector data identification processing aiming at a space knowledge graph database are respectively carried out on each piece of grid data, and confidence is set; and returning a final recognition result according to the remote sensing photo recognition result, the proof photo recognition result and the vector data recognition result in combination with the parameter confidence. The land category is recognized through three recognition dimensions, and the accuracy is improved according to a multi-dimension mutual verification mode, so that the land utilization recognition precision is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, system, storage medium, and device for identifying land resource utilization. Background Technology

[0002] Land surveys are a major national survey of conditions and strength. Land use classification, as a key component of land surveys, is an important means of verifying and clarifying land resources. The purpose of classifying all land features is to comprehensively ascertain the current land use status nationwide, obtain accurate basic land data, improve land survey, monitoring, and statistical systems, strengthen the socialized services of land resource information, and meet the needs of economic and social development and land resource management. However, traditional land use identification methods rely heavily on manual judgment, which consumes the vast majority of land survey time, significantly reducing the efficiency of land use identification. Therefore, how to rapidly improve the efficiency of land use identification has become a crucial issue in land resource surveys. Summary of the Invention

[0003] This invention addresses the shortcomings of existing technologies by providing a method for identifying land resource utilization.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] A method for identifying land resource use includes the following steps:

[0006] Remote sensing image sample data is classified and trained according to land use codes to generate remote sensing image models;

[0007] The field evidence photo sample data is classified and trained to generate a field photo model based on the land category code;

[0008] Generate a spatial knowledge graph library from business data based on spatial location relationships;

[0009] The photos to be identified are processed into grids. For each grid, remote sensing image recognition is performed on the remote sensing image model, evidence photo recognition is performed on the field photo model, and vector data recognition is performed on the spatial knowledge graph database. Confidence levels are set, and the final recognition result is returned based on the remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with the parameter confidence level.

[0010] As a preferred method, a method for training and generating remote sensing image models by classifying remote sensing image sample data according to land cover codes includes:

[0011] Remote sensing image sample data is classified and cropped according to land use coding: the coordinate strings of land use patches are sequentially generated into vector data, and the vector data are overlaid and cropped to generate patch image format;

[0012] Set the grid parameter m*n, determine whether the pixel count of the remote sensing image is greater than k times m*n, and if it is, perform slider land class cropping on the image of the patch format based on m*n pixels, and output the slider land class patch;

[0013] If k is less than or equal to a multiple of m*n, then no trimming is needed.

[0014] As a preferred method, the method of training and generating remote sensing image models by classifying remote sensing image sample data according to land cover coding also includes...

[0015] If the effective area of ​​any slider land feature is less than a preset percentage of the slider area, it is fused with the adjacent slider features and then trained for recognition.

[0016] If the effective area of ​​any slider land feature is greater than a preset percentage of the slider area, then it will be trained and recognized separately.

[0017] As a preferred method, the method of training and generating a field photo model from field evidence photo sample data based on land use coding includes:

[0018] Calculate the shooting range of the photo based on the five elements of field photography;

[0019] Based on the cropping parameters of the remote sensing image, the four boundaries of the block where the field photo is located are determined, and each block is trained and identified separately.

[0020] As a preferred method, the final identification result is returned based on the identification results of remote sensing photos, evidence photos, and vector data, combined with parameter confidence scores. This includes:

[0021] If the remote sensing image recognition result of the photo to be identified is consistent with the vector data recognition result, then the land category level 1 is returned;

[0022] If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are consistent, then return to level 2 land category;

[0023] If the identification result of the evidence photo is consistent with the identification result of the vector data, then return the land category as Level 1 or Level 2;

[0024] If the remote sensing image recognition result matches the evidence photo data recognition result, then return to Level 1 land category.

[0025] As a preferred option, if the remote sensing image recognition result, the vector data recognition result, and the evidence photo recognition result are all different, the evidence photo recognition result will be returned; if the confidence level is lower than the threshold, the vector data recognition result will be returned.

[0026] As a preferred method, the final identification result is returned based on the identification results of remote sensing photos, evidence photos, and vector data, combined with parameter confidence scores. This includes:

[0027] If the remote sensing image recognition result of the photo to be identified is consistent with the vector data recognition result, then the Level 1 land category is returned. If the confidence level of the evidence photo recognition is lower than the threshold, then the vector data recognition result is returned; otherwise, the evidence photo recognition result is returned.

[0028] If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are consistent, then the land category level 2 is returned, and the vector data recognition result is also returned.

[0029] If the identification result of the evidence photo is consistent with the identification result of the vector data, then the land category of level 1 or level 2 is returned; if the confidence level of the evidence photo identification is lower than the threshold, then the vector data identification result is returned; otherwise, the evidence photo identification result is returned.

[0030] If the remote sensing image recognition result is consistent with the evidence photo data recognition result, then return the land category level 1, and return the result with higher confidence in the evidence photo or image photo recognition.

[0031] Furthermore, a land resource use identification system is proposed, comprising the following structure:

[0032] The model training unit is used to train remote sensing image sample data to generate remote sensing image models based on land use codes; to train field evidence photo sample data to generate field photo models based on land use codes; and to generate a spatial knowledge graph library based on business data according to spatial location relationships.

[0033] The photo recognition unit is used to process the photos to be recognized into grids, and to perform remote sensing image recognition for remote sensing image models, evidence photo recognition for field photo models, and vector data recognition for spatial knowledge graph databases for each grid data. Confidence levels are set, and recognition processing is performed based on the remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with the parameter confidence levels.

[0034] The result output unit is used to output the results identified by the photo recognition unit.

[0035] The beneficial effects of this invention are:

[0036] This invention proposes a land use identification method. This scheme trains a remote sensing image model, an evidence photo model, and a spatial knowledge graph database. It identifies land use categories through three identification dimensions, performs comprehensive judgment based on interpretation rules, and increases accuracy by cross-verifying multiple dimensions, thereby improving the precision of land use identification. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0038] Figure 1 This is a flowchart of a land resource utilization identification method. Detailed Implementation

[0039] The present invention will be further described in detail below with reference to the embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following embodiments.

[0040] A method for identifying land resource use includes the following steps:

[0041] Remote sensing image sample data is classified and trained according to land use codes to generate remote sensing image models;

[0042] The field evidence photo sample data is classified and trained to generate an evidence photo model based on the land category code;

[0043] Generate a spatial knowledge graph library from business data based on spatial location relationships;

[0044] The photos to be identified are processed into grids. For each grid, remote sensing image recognition is performed for the remote sensing image model, evidence photo recognition is performed for the evidence photo model, and vector data recognition is performed for the spatial knowledge graph database. Confidence levels are set, and the final recognition result is returned based on the remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with the parameter confidence level.

[0045] This scheme trains remote sensing image models, evidence photo models, and spatial knowledge graph databases, identifies land use categories through three recognition dimensions, and conducts comprehensive analysis based on interpretation rules, thereby improving the accuracy of land use identification.

[0046] As a preferred solution, the identification process uses AI intelligent identification technology for land category identification. Specifically, the Inception v3 convolutional neural network is used for AI land category identification.

[0047] As a preferred approach, a method for training and generating remote sensing image models from remote sensing image sample data based on land cover coding classification includes:

[0048] Remote sensing image sample data is classified and cropped according to land use coding: the coordinate strings of land use patches are sequentially generated into vector data, and the vector data are overlaid and cropped to generate patch image format;

[0049] Specifically, the coordinate strings of land parcels are sequentially generated into vector data (e.g., .shp format) and raster data (e.g., .tif format) by stringing them together into WKT data. The vector and raster data are then overlaid and cropped to generate a parcel image format (e.g., .png format).

[0050] As a preferred approach, the vector data is expanded by 10-15% to form a buffer, creating a greater contrast between the current land type and other surrounding land types, thus enhancing recognizability. The raster data and the expanded vector data are then overlaid and cropped to generate a patch image format.

[0051] Set the pixel value grid parameter m*n, determine whether the pixel count of the remote sensing image is greater than k times m*n, and if it is, perform slider land class cropping on the image of the patch format based on m*n pixels, and output the slider land class patch;

[0052] If k is less than or equal to a multiple of m*n, then no trimming is needed. Here, m and n can be the same value. In this scheme, k is taken as 1.5.

[0053] Furthermore, if the effective area of ​​any slider land feature is less than a preset percentage of the slider area, it is fused with the adjacent slider features and then trained for recognition.

[0054] If the effective area of ​​any slider land feature is greater than a preset percentage of the slider area, then it will be trained and recognized separately.

[0055] In this scheme, the preset value is 40%. If the effective area of ​​any slider land feature is less than 40% of the slider area, it will be merged with the adjacent slider features and then trained for recognition.

[0056] If the effective area of ​​any slider land feature is greater than 40% of the slider area, then it is trained and identified separately.

[0057] The remote sensing image model is output based on the above method.

[0058] As a preferred approach, methods for training and generating field photo models from field evidence photo sample data based on land use coding include:

[0059] Calculate the shooting range of the photo based on the five elements of field photography;

[0060] Based on the grid parameters m*n of the land cover cropping of the remote sensing image, the four boundaries of the block where the field evidence photo is located are determined. Then, each block is trained and identified separately to form an evidence photo model.

[0061] The five elements are: heading angle, pitch angle, roll angle, shooting altitude, and shooting coordinates. The actual shooting range of the photo is calculated based on these shooting parameters.

[0062] As a preferred approach, the knowledge graph database is generated by using land parcel data from the Third National Land Survey in this embodiment, which identifies the specific land parcel type for each land parcel.

[0063] First, assign a color to each land category for the business data. Then, rasterize the business data vector layer into an image format. Slice the image format business data into 6 to 21 levels according to the Web Mercator rule. Store each level of data in a local db file for subsequent comparison and recognition with the photo to be identified.

[0064] As a preferred approach, a method for returning the final identification result based on remote sensing image recognition results, evidence image recognition results, and vector data recognition results, combined with parameter confidence scores, includes selecting the land use type based on different weights of the three dimensions:

[0065] If the remote sensing image recognition result of the photo to be identified is consistent with the vector data recognition result, then the land category level 1 is returned;

[0066] If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are consistent, then return to level 2 land category;

[0067] If the identification result of the evidence photo is consistent with the identification result of the vector data, then return the land category as Level 1 or Level 2;

[0068] If the remote sensing image recognition result matches the evidence photo data recognition result, then return to Level 1 land category.

[0069] Specifically, if the remote sensing image recognition result of the photo to be identified is consistent with the vector data recognition result, then the level 1 land category is returned; if the confidence level of the evidence photo recognition is lower than the threshold, then the vector data recognition result is returned; otherwise, the evidence photo recognition result is returned.

[0070] If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are consistent, then the land category level 2 is returned, and the vector data recognition result is also returned.

[0071] If the identification result of the evidence photo is consistent with the identification result of the vector data, then the land category of level 1 or level 2 is returned; if the confidence level of the evidence photo identification is lower than the threshold, then the vector data identification result is returned; otherwise, the evidence photo identification result is returned.

[0072] If the remote sensing image recognition result is consistent with the evidence photo data recognition result, then return the land category level 1, and return the result with higher confidence in the evidence photo or image photo recognition.

[0073] If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are all different, the evidence photo recognition result will be returned. If the confidence level is lower than the threshold, the vector data recognition result will be returned.

[0074] The table below illustrates a specific analytical strategy:

[0075]

[0076] A land resource use identification system includes the following structure:

[0077] The model training unit is used to train remote sensing image sample data to generate remote sensing image models based on land use codes; to train evidence photo models based on field evidence photo sample data based on land use codes; and to generate spatial knowledge graphs based on business data according to spatial location relationships.

[0078] The photo recognition unit is used to process the photos to be recognized into grids, and to perform remote sensing image recognition for remote sensing image models, evidence photo recognition for field photo models, and vector data recognition for spatial knowledge graph databases for each grid data. Confidence levels are set, and recognition processing is performed based on the remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with the parameter confidence levels.

[0079] The result output unit is used to output the results identified by the photo recognition unit.

[0080] The photo recognition unit also includes:

[0081] The remote sensing image recognition unit is used to compare and identify the gridded photo to be recognized with the remote sensing image model.

[0082] The evidence photo recognition unit is used to identify and compare the gridded photo to be recognized with the evidence photo model.

[0083] The vector data recognition unit is used to compare and identify the gridded photo to be recognized with the spatial knowledge graph database.

[0084] The model training unit also includes:

[0085] The remote sensing image model training unit is used to train and generate remote sensing image models by classifying and classifying remote sensing image sample data according to land use codes.

[0086] The evidence photo model is generated by training field evidence photo sample data according to land category coding.

[0087] Spatial knowledge graph library is used to generate spatial knowledge graph libraries based on the spatial location relationships of business data.

[0088] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules or units is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed.

[0089] The units may or may not be physically separate. The components shown as units can be one or more physical units, meaning they can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0090] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0091] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, essentially, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] In addition, this solution should also include a computer device, including a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, characterized in that the program instructions are loaded and executed by the processor to implement the steps of the land resource utilization identification method.

[0093] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A land resource utilization recognition method characterized by comprising: Includes the following steps: Remote sensing image sample data is classified and trained according to land use codes to generate remote sensing image models; The field evidence photo sample data is classified and trained to generate a field photo model based on the land category code; Generate a spatial knowledge graph library from business data based on spatial location relationships; The photos to be identified are processed into grids. For each grid, remote sensing image recognition is performed on the remote sensing image model, evidence photo recognition is performed on the field photo model, and vector data recognition is performed on the spatial knowledge graph database. Confidence levels are set, and the final recognition result is returned based on the remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with the parameter confidence level.

2. The land resource utilization identification method according to claim 1, characterized in that, Methods for training remote sensing image models based on land cover coding classification of remote sensing image sample data include: Remote sensing image sample data is classified and cropped according to land use coding: the coordinate strings of land use patches are sequentially generated into vector data, and the vector data are overlaid and cropped to generate patch image format; Set the grid parameter m*n, determine whether the pixel count of the remote sensing image is greater than k times m*n, and if it is, perform slider land class cropping on the image of the patch format based on m*n pixels, and output the slider land class patch; If k is less than or equal to a multiple of m*n, then no trimming is needed.

3. The method of claim 2, wherein Methods for training remote sensing image models based on land cover coding classification of remote sensing image sample data also include If the effective area of ​​any slider land feature is less than a preset percentage of the slider area, it is fused with the adjacent slider features and then trained for recognition. If the effective area of ​​any slider land feature is greater than a preset percentage of the slider area, then it will be trained and recognized separately.

4. The land resource utilization identification method according to claim 2, characterized in that, Methods for training and generating field photo models from field evidence photo sample data based on land use coding include: Calculate the shooting range of the photo based on the five elements of field photography; Based on the cropping parameters of the remote sensing image, the four boundaries of the field photo block are determined, and each block is trained and identified separately.

5. The land resource utilization identification method according to claim 1, characterized in that, Methods for returning the final recognition result based on remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with parameter confidence scores, include: If the remote sensing image recognition result of the photo to be identified is consistent with the vector data recognition result, then the land category level 1 is returned; If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are consistent, then return to level 2 land category; If the identification result of the evidence photo is consistent with the identification result of the vector data, then return the land category as Level 1 or Level 2; If the remote sensing image recognition result matches the evidence photo data recognition result, then return to Level 1 land category.

6. The land resource utilization identification method according to claim 1, characterized in that, If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are all different, the evidence photo recognition result will be returned. If the confidence level is lower than the threshold, the vector data recognition result will be returned.

7. The land resource utilization identification method according to claim 1, characterized in that, Methods for returning the final recognition result based on remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with parameter confidence scores, include: If the remote sensing image recognition result of the photo to be identified is consistent with the vector data recognition result, then the Level 1 land category is returned. If the confidence level of the evidence photo recognition is lower than the threshold, then the vector data recognition result is returned; otherwise, the evidence photo recognition result is returned. If the remote sensing image recognition result, vector data recognition result, and evidence photo recognition result are consistent, then the land category level 2 is returned, and the vector data recognition result is also returned. If the identification result of the evidence photo is consistent with the identification result of the vector data, then the land category of level 1 or level 2 is returned; if the confidence level of the evidence photo identification is lower than the threshold, then the vector data identification result is returned; otherwise, the evidence photo identification result is returned. If the remote sensing image recognition result is consistent with the evidence photo data recognition result, then return the land category level 1, and return the result with higher confidence in the evidence photo or image photo recognition.

8. A land resource use identification system, characterized in that, Includes the following structure: The model training unit is used to train remote sensing image sample data to generate remote sensing image models based on land use codes; to train field evidence photo sample data to generate field photo models based on land use codes; and to generate a spatial knowledge graph library based on business data according to spatial location relationships. The photo recognition unit is used to process the photos to be recognized into grids, and to perform remote sensing image recognition for remote sensing image models, evidence photo recognition for field photo models, and vector data recognition for spatial knowledge graph databases for each grid data. Confidence levels are set, and recognition processing is performed based on the remote sensing photo recognition results, evidence photo recognition results, and vector data recognition results, combined with the parameter confidence levels. The result output unit is used to output the results identified by the photo recognition unit.

9. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, it controls the device containing the storage medium to perform the land resource utilization identification method according to any one of claims 1 to 7.

10. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, characterized in that, when the program instructions are loaded and executed by the processor, they implement the steps of the land resource utilization identification method according to any one of claims 1 to 7.

Images