Method and device for satellite image sub - frame cutting and RPC parameter generation

Abstract

The present invention provides a method and device for satellite image sub - frame cutting and RPC parameter generation, which relates to the field of satellite image cutting and includes: sub - frame cutting the original satellite image to obtain all the cut images; inputting the cut images into a preset classification model to obtain the image recognition results output by the preset classification model; determining the feature of elements to be excluded according to the application scenario of the cut images, and excluding the image recognition results with the feature of elements to be excluded from the cut images to obtain the images after exclusion, where the application scenarios include natural environment scenarios, urban construction scenarios, and agricultural and land use scenarios; calculating RPC parameters according to the application scenario, and using the RPC parameters to perform geometric correction on the images after exclusion to determine the corrected images. The present invention can flexibly determine the feature of excluded elements and perform geometric correction according to the requirements of different application scenarios, providing an image data processing solution for fields such as natural environment, urban construction, and agricultural land use.

Description

Technical Field

[0001] The present invention relates to the field of satellite image cropping, and in particular to a method and device for satellite image sub-frame cropping and RPC parameter generation. Background Art

[0002] When the existing satellite image cropping methods are applied to satellite images with actual usage scenarios, it is impossible to determine the degree of association between the feature characteristics in the actual application scenario and the cropped images, so a large number of redundant feature characteristics need to be retained, thereby reducing the data processing efficiency.

[0003] When performing geometric calibration on sub-frame images, polynomial coefficients (Rational Polynomial Coefficients, RPC) are usually used. However, if the influence of terrain on image geometry is considered, the calculation process of polynomial coefficients is usually particularly complex. When dealing with each cropped image after satellite image cropping, if different cropped images will be used for different application scenarios, it is usually impossible to calculate the polynomial coefficients separately for different application scenarios, so that the finally obtained polynomial coefficients do not take into account different application scenarios, that is, the polynomial coefficients used for geometric calibration are not accurate enough.

[0004] Currently, there is no technical solution that can solve the above technical problems, and there is no method and device for satellite image sub-frame cropping and RPC parameter generation. Summary of the Invention

[0005] The present invention provides a method and device for satellite image sub-frame cropping and RPC parameter generation. When sub-frame cropping the original satellite image, for different application scenarios corresponding to different cropped images, redundant feature elimination and RPC parameter calculation are respectively performed, so as to improve the data processing efficiency while making the corrected image more accurate.

[0006] In a first aspect, the present invention provides a method for satellite image sub-frame cropping and RPC parameter generation, including:

[0007] Sub-frame crop the original satellite image to obtain all cropped images;

[0008] For each cropped image, input the cropped image into a preset classification model to obtain an image recognition result output by the preset classification model;

[0009] According to the application scenario of the cropped image, determine the element features to be eliminated, and eliminate the image recognition results with the element features to be eliminated from the cropped image to obtain an image after elimination. The application scenarios include natural environment scenarios, urban construction scenarios, and agricultural and land use scenarios;

[0010] According to the application scenario, calculate the RPC parameters corresponding to the image after rejection, and geometrically correct the image after rejection by using the RPC parameters to determine the corrected image;

[0011] The preset classification model is determined after being trained according to all image samples and the labeled tags corresponding to different feature characteristics in each image sample, and the labeled tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts.

[0012] According to the satellite image sub - frame cutting and RPC parameter generation method provided by the present invention, the determination of the feature characteristics of elements to be rejected according to the application scenario of the image after cutting includes:

[0013] When the application scenario is a natural environment scenario, determine that the feature characteristics of elements to be rejected are houses, roads, and bridges;

[0014] When the application scenario is an urban construction scenario, determine that the feature characteristics of elements to be rejected are lakes, rivers, reservoirs, forests, grasslands, and farmlands;

[0015] When the application scenario is an agricultural and land - use scenario, determine that the feature characteristics of elements to be rejected are houses, roads, bridges, bare soil, rocks, and deserts.

[0016] According to the satellite image sub - frame cutting and RPC parameter generation method provided by the present invention, when the application scenario is a natural environment scenario, the calculation of the RPC parameters corresponding to the image after rejection according to the application scenario includes:

[0017] Register the original digital elevation model data according to the geographical coordinates and longitude - latitude information of the image after cutting to obtain the target digital elevation model data corresponding to the image after cutting;

[0018] Perform terrain correction on the image after cutting by using the target digital elevation model data to obtain the corrected image;

[0019] Construct a first RPC model according to the platform position, platform attitude, and pixel size corresponding to the corrected image, and solve the RPC parameters corresponding to the image after rejection based on the first RPC model.

[0020] According to the satellite image sub - frame cutting and RPC parameter generation method provided by the present invention, the construction of the first RPC model according to the platform position, platform attitude, and pixel size corresponding to the corrected image includes:

[0021]

[0022] Wherein, (X im, Y im , Z im (X, Y) are the pixel coordinates of the platform position in the image coordinate system considering the terrain, (X, Y) are the geographical coordinates in the geographical coordinate system, (Roll, Pitch, Yaw) is the platform attitude, (Sx, Sy) is the pixel size, and functions f and g are polynomial functions.

[0023] According to the satellite image mosaicking and cutting and RPC parameter generation method provided by the present invention, solving the RPC parameters corresponding to the image after rejection based on the first RPC model includes:

[0024] Solving all the first polynomial coefficients in function f by the least squares method, solving all the second polynomial coefficients in function g, and determining all the first polynomial coefficients and all the second polynomial coefficients as the RPC parameters corresponding to the image after rejection.

[0025] According to the satellite image mosaicking and cutting and RPC parameter generation method provided by the present invention, when the application scenario is an urban construction scenario, calculating the RPC parameters corresponding to the image after rejection according to the application scenario includes: constructing a second RPC model according to the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the image after rejection, and solving the RPC parameters corresponding to the image after rejection based on the second RPC model;

[0026] Constructing the second RPC model according to the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the image after rejection includes:

[0027]

[0028] Wherein, (X im , Y im ) are the pixel coordinates of the platform position in the image coordinate system without considering the terrain, (X, Y) are the geographical coordinates in the geographical coordinate system, (Roll, Pitch, Yaw) is the platform attitude, (Sx, Sy) is the pixel size, BE is the building edge information Building_Edges, UF is the urban feature information Urban_Features, and functions f and g are polynomial functions.

[0029] According to the method for generating satellite image mosaicking and cutting and RPC parameters provided by the present invention, in the case where the application scenario is the agricultural and land use scenario, calculating the RPC parameters corresponding to the image after elimination according to the application scenario includes: constructing a third RPC model based on the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the image after elimination, and solving the RPC parameters corresponding to the image after elimination based on the third RPC model;

[0030] Constructing the third RPC model based on the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the image after elimination includes:

[0031]

[0032] wherein, (X im , Y im ) are the pixel coordinates of the platform position in the image coordinate system without considering the terrain, (X, Y) are the geographical coordinates in the geographical coordinate system, (Roll, Pitch, Yaw) are the platform attitude, (Sx, Sy) are the pixel size, VI is the vegetation index information Vegetation_Index, FF is the farmland feature information Farmland_Features, and the functions f and g are polynomial functions.

[0033] According to the method for generating satellite image mosaicking and cutting and RPC parameters provided by the present invention, after determining the corrected image, the method further includes:

[0034] Traversing all the original satellite images to determine the corrected image and the RPC parameters corresponding to each original satellite image;

[0035] Outputting all the corrected images and the RPC parameters corresponding to each corrected image to a preset storage unit to implement geometric information processing of different corrected images according to different application scenarios.

[0036] According to the method for generating satellite image mosaicking and cutting and RPC parameters provided by the present invention, mosaicking and cutting the original satellite images to obtain all the cut images includes:

[0037] Loading the original satellite image using GIS software, and determining the cutting range from the original satellite image according to the geographical coordinates and longitude and latitude information;

[0038] Mosaicking and cutting the original satellite image in the cutting range to obtain all the cut images;

[0039] The geographical coordinates and the longitude and latitude information are determined after overlaying a preset geographical entity as a layer on the original satellite image.

[0040] In a second aspect, a satellite image sub - frame cutting and RPC parameter generation device is provided, including:

[0041] An acquisition unit, configured to perform sub - frame cutting on the original satellite image to obtain all the cut images;

[0042] An input unit, configured to input each of the cut images into a preset classification model for each cut image, and obtain the image recognition result output by the preset classification model;

[0043] An elimination unit, configured to determine the feature of elements to be eliminated according to the application scenario of the cut image, and eliminate the image recognition results with the feature of elements to be eliminated from the cut image to obtain the eliminated image, where the application scenario includes natural environment scenarios, urban construction scenarios, and agricultural and land - use scenarios;

[0044] A determination unit, configured to calculate the RPC parameters corresponding to the eliminated image according to the application scenario, perform geometric correction on the eliminated image by using the RPC parameters, and determine the corrected image;

[0045] The preset classification model is determined after being trained according to all image samples and the labeled tags corresponding to different element features in each image sample, and the labeled tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts.

[0046] In the present invention, by inputting the cut image into a preset classification model, accurate recognition results of different element features are obtained, aiming to cut and identify elements in a specific scenario from the original satellite image, eliminate and geometrically correct the image, so as to obtain image data that meets the requirements of a specific application scenario. Then, in combination with the application scenario again, the RPC parameters corresponding to the eliminated image are calculated, and geometric correction is performed on the image, thereby more precisely converting the image pixel coordinates into geographic coordinates, ensuring the spatial accuracy and consistency of the image data. The present invention can flexibly determine the element features to be eliminated and perform geometric correction according to the requirements of different application scenarios, providing customized image data processing solutions for different fields such as natural environment, urban construction, and agricultural land use. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following - described drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1It is one of the schematic flowcharts of the satellite image sub - frame cutting and RPC parameter generation method provided by the present invention;

[0049] Figure 2 It is the second of the schematic flowcharts of the satellite image sub - frame cutting and RPC parameter generation method provided by the present invention;

[0050] Figure 3 It is the schematic structural diagram of the satellite image sub - frame cutting and RPC parameter generation device provided by the present invention;

[0051] Figure 4 It is the schematic structural diagram of the electronic device provided by the present invention. Detailed implementation manners

[0052] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0053] Figure 1 It is one of the schematic flowcharts of the satellite image sub - frame cutting and RPC parameter generation method provided by the present invention. The satellite image sub - frame cutting and RPC parameter generation method includes:

[0054] Step 101: Sub - frame cut the original satellite image to obtain all the cut images;

[0055] Step 102: For each cut image, input the cut image into a preset classification model to obtain the image recognition result output by the preset classification model;

[0056] Step 103: According to the application scenario of the cut image, determine the feature of the element to be excluded, and exclude the image recognition result with the feature of the element to be excluded from the cut image to obtain the image after exclusion. The application scenarios include natural environment scenarios, urban construction scenarios, and agricultural and land - use scenarios;

[0057] Step 104: According to the application scenario, calculate the RPC parameters corresponding to the image after exclusion, and perform geometric correction on the image after exclusion using the RPC parameters to determine the corrected image;

[0058] The preset classification model is determined after being trained according to all image samples and the marked labels corresponding to different element features in each image sample. The marked labels include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts.

[0059] In step 101, the original satellite image is divided and cropped to obtain all the cropped images, including: loading the original satellite image using GIS software, determining the cropping range from the original satellite image according to the geographic coordinates and longitude and latitude information; dividing and cropping the original satellite image within the cropping range to obtain all the cropped images; the geographic coordinates and the longitude and latitude information are determined after overlaying a preset geographic entity as a layer on the original satellite image.

[0060] Optionally, the present invention uses a Geographic Information System (GIS) tool to load the original satellite image and geographic information data, determines the cropping range by drawing a vector boundary or a polygon range, accurately determines the boundary of the cropping range according to the geographic coordinates and longitude and latitude information, ensures that the cropped image corresponds exactly to the geographic entity, and can also refer to existing geographic data, topographic maps or satellite images to assist in determining the cropping range to ensure that the cropped image matches the surrounding environment.

[0061] By using GIS software or professional remote sensing image processing software, the automatic cropping function can be used to remove redundant pixel data and reduce redundant information; image processing algorithms, such as edge detection, feature extraction, etc., can be applied to identify and remove unnecessary data and retain key information; according to the project requirements and analysis purposes, the cropping range can be carefully selected to remove unnecessary areas and information and reduce data redundancy.

[0062] Optionally, the present invention can also perform multi-threaded cropping on the original satellite image to improve the processing efficiency, adopt parallel processing technology to process multiple images simultaneously, speed up the cropping speed, and while increasing the cropping speed, ensure that the cropped image corresponds exactly to the actual geographical location, avoid information deviation and inconsistency, remove redundant information, reduce the data volume, and improve the data processing and storage efficiency.

[0063] In step 102, the present invention determines the preset classification model after training according to all the image samples and the marker tags corresponding to different element features in each image sample. Among them, different element features are marked in the form of manual marking. Specifically, the marker tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts, so that after inputting the cropped image into the preset classification model, the image recognition result output by the preset classification model is obtained.

[0064] Specifically, the present invention can use a Support Vector Machine (SVM) to classify remote sensing images through a preset classification model to determine the image recognition result, thereby removing the corresponding redundant information.

[0065] In step 103, determining the feature of the element to be excluded according to the application scenario of the cropped image includes: when the application scenario is a natural environment scenario, determining the feature of the element to be excluded as houses, roads, and bridges; when the application scenario is an urban construction scenario, determining the feature of the element to be excluded as lakes, rivers, reservoirs, forests, grasslands, and farmlands; when the application scenario is an agriculture and land use scenario, determining the feature of the element to be excluded as houses, roads, bridges, bare soil, rocks, and deserts.

[0066] In an optional embodiment, if the application scenario corresponding to the cropped image is a natural environment scenario, and the image recognition results output by the preset classification model include houses and roads, then the image recognition results with the feature of the element to be excluded are removed from the cropped image, that is, houses and roads are removed from the cropped image to obtain the post-removal image; in another optional embodiment, if the application scenario corresponding to the cropped image is an urban construction scenario, and the image recognition results output by the preset classification model include grasslands and farmlands, then the image recognition results with the feature of the element to be excluded are removed from the cropped image, that is, grasslands and farmlands are removed from the cropped image to obtain the post-removal image; in another optional embodiment, if the application scenario corresponding to the cropped image is an agriculture and land use scenario, and the image recognition results output by the preset classification model include bridges, bare soil, and deserts, then the image recognition results with the feature of the element to be excluded are removed from the cropped image, that is, bridges, bare soil, and deserts are removed from the cropped image to obtain the post-removal image.

[0067] In step 104, the present invention combines the application scenario again to calculate the RPC parameters corresponding to the post-removal image, that is, for different application scenarios, different methods are used to calculate the post-removal image, so as to obtain more accurate RPC parameters. The present invention uses a terrain-independent RPC parameter conversion method: this method does not consider the influence of terrain on the image geometry, and is more focused on the accuracy of geometric correction. According to the RPC parameters, geometric correction is performed on the post-removal image to adjust the geometric shape of the post-removal image to conform to the geographic coordinate system and geographic coordinate information.

[0068] Optionally, according to the corrected image and RPC parameters, the geographic positioning accuracy of the image is further adjusted to ensure the accurate geographical position of the image. It can be registered with a geographic information system or other geographic data to further improve the geographic positioning accuracy of the image. After completing geometric correction and geographic positioning, the present invention can also verify the corrected image to ensure the accuracy of the geometric shape and geographical location.

[0069] Optionally, after determining the corrected image, the method further includes:

[0070] Traverse all the original satellite images to determine the corrected image and the RPC parameters corresponding to each of the original satellite images;

[0071] Output all the corrected images and the RPC parameters corresponding to each corrected image to a preset storage unit, so as to implement geometric information processing of different corrected images according to different application scenarios.

[0072] Optionally, the present invention traverses all the original satellite images, uses the above steps to determine the corrected image and the RPC parameters corresponding to each of the original satellite images, and then outputs all the corrected images after geometric correction and geolocation processing and the RPC parameters corresponding to each corrected image. These information will be used for subsequent geospatial information processing and applications, and can be pre-stored in a preset storage unit for subsequent call at any time.

[0073] The present invention inputs the cropped image into a preset classification model to obtain accurate recognition results of different feature characteristics, aims to crop and identify elements in a specific scenario from the original satellite image, and performs rejection and geometric correction on the image to obtain image data that meets the requirements of a specific application scenario. Then, in combination with the application scenario again, calculates the RPC parameters corresponding to the rejected image, performs geometric correction on the image, and further accurately converts the image pixel coordinates into geographic coordinates to ensure the spatial accuracy and consistency of the image data. The present invention can flexibly determine the rejection of feature characteristics and perform geometric correction according to the requirements of different application scenarios, providing customized image data processing solutions for different fields such as natural environment, urban construction, and agricultural land use.

[0074] Figure 2 It is the second flowchart of the satellite image sub - frame cropping and RPC parameter generation method provided by the present invention. When the application scenario is a natural environment scenario, calculating the RPC parameters corresponding to the rejected image according to the application scenario includes:

[0075] Step 201: Register the original digital elevation model data according to the geographic coordinates and longitude - latitude information of the cropped image to obtain the target digital elevation model data corresponding to the cropped image;

[0076] Step 202: Perform terrain correction on the cropped image using the target digital elevation model data to obtain the corrected image;

[0077] Step 203: Construct a first RPC model based on the platform position, platform attitude, and pixel size corresponding to the corrected image, and solve the RPC parameters corresponding to the rejected image based on the first RPC model.

[0078] In step 201, it is necessary to obtain the geographic coordinate information of the cropped image, which usually includes the longitude and latitude coordinates of the upper left corner and the lower right corner of the image, as well as information such as the resolution of the image. The original digital elevation model data can be interpolated or resampled to make it have the same spatial resolution and coverage as the image. Using the geographic coordinate information of the cropped image, the original digital elevation model data is aligned with the cropped image to ensure that the original digital elevation model data and the cropped image are in the same coordinate system and have a consistent spatial reference framework, and then the target digital elevation model data corresponding to the cropped image is obtained.

[0079] In step 202, before calculating the RPC parameters, the influence of the terrain needs to be considered in the natural environment scene. The digital elevation model provides surface elevation information. The calculation of the RPC parameters needs to combine the terrain data to ensure that the geometric shape of the image conforms to the actual surface situation. The present invention uses the target digital elevation model data to extract the terrain features of the image, such as terrain height, slope, aspect, etc. These terrain features can help understand the influence of the terrain on the relationship between the pixel coordinates and the geographic coordinates of the image and provide a basis for terrain adjustment. According to the terrain information provided by the target digital elevation model data, the elevation correction of the cropped image can be performed. By comparing the elevation information of the pixels of the cropped image with the target digital elevation model data, the elevation values of the image pixels are adjusted to make them more conform to the actual surface elevation situation, and finally the corrected image is obtained.

[0080] In step 203, constructing the first RPC model according to the platform position, platform attitude, and pixel size corresponding to the corrected image includes:

[0081]

[0082] Among them, (X im , Y im , Z im ) are the pixel coordinates of the platform position in the image coordinate system considering the terrain, (X, Y) are the geographic coordinates in the geographic coordinate system, (Roll, Pitch, Yaw) are the platform attitudes, (Sx, Sy) are the pixel sizes, and the functions f and g are polynomial functions.

[0083] Optionally, based on information such as platform position, platform attitude, and pixel size, the RPC parameters of the image can be calculated. The RPC model is a mathematical model used for geometric correction and geolocation, which describes the relationship between pixel coordinates and geographic coordinates through polynomial coefficients. In a natural environment scenario, the influence of terrain needs to be considered particularly. Therefore, the RPC parameter calculation method based on the digital elevation model is commonly used. By combining the digital elevation data with the corrected image, the RPC parameters can be calculated more accurately.

[0084] Optionally, solving the RPC parameters corresponding to the cropped image based on the first RPC model includes:

[0085] Solving all the first polynomial coefficients in the function f using the least squares method, solving all the second polynomial coefficients in the function g, and determining all the first polynomial coefficients and all the second polynomial coefficients as the RPC parameters corresponding to the cropped image.

[0086] Optionally, in the case where the application scenario is an urban construction scenario, calculating the RPC parameters corresponding to the cropped image according to the application scenario includes: constructing a second RPC model based on the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the cropped image, and solving the RPC parameters corresponding to the cropped image based on the second RPC model;

[0087] Constructing the second RPC model based on the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the cropped image includes:

[0088]

[0089] where (X im , Y im ) are the pixel coordinates of the platform position in the image coordinate system without considering terrain, (X, Y) are the geographic coordinates in the geographic coordinate system, (Roll, Pitch, Yaw) are the platform attitude, (Sx, Sy) are the pixel size, BE is the building edge information Building_Edges, UF is the urban feature information Urban_Features, and the functions f and g are polynomial functions.

[0090] Optionally, in the urban construction scenario, it is usually necessary to consider the impacts of buildings, roads, etc. These artificial objects will affect the geometric transformation of the image. Therefore, these factors need to be considered when calculating the RPC parameters. The present invention adopts an RPC parameter calculation method based on building edge extraction, and combines urban planning and road network urban features to calculate the RPC parameters. Specifically, the present invention first uses remote sensing image processing software or computer vision algorithms to extract the edges of buildings in the image, and extract the geometric information of the buildings, including the outlines, corner points, etc. of the buildings. Then, urban feature information related to urban planning data and road network data is obtained, including urban planning maps, road network maps, building distribution maps, etc., which are used to combine with the building edge information to construct a second RPC model. In the urban construction scenario, by combining the building edges and urban feature information, the RPC parameters can be calculated more accurately, considering the impacts of urban features such as urban planning and road network on the image.

[0091] Optionally, in the case where the application scenario is the agriculture and land use scenario, calculating the RPC parameters corresponding to the image after rejection according to the application scenario includes: constructing a third RPC model based on the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the image after rejection, and solving the RPC parameters corresponding to the image after rejection based on the third RPC model;

[0092] Constructing the third RPC model based on the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the image after rejection includes:

[0093]

[0094] where (X im , Y im ) are the pixel coordinates of the platform position in the image coordinate system without considering the terrain, (X, Y) are the geographical coordinates in the geographical coordinate system, (Roll, Pitch, Yaw) are the platform attitude, (Sx, Sy) are the pixel sizes, VI is the vegetation index information Vegetation_Index, FF is the farmland feature information Farmland_Features, and the functions f and g are polynomial functions.

[0095] Optionally, in agricultural and land use scenarios, it is usually necessary to consider the influence of natural elements such as farmland and vegetation. These natural elements will affect the geometric transformation of the image. Therefore, these factors need to be considered when calculating the RPC parameters. The present invention first uses remote sensing image processing software or vegetation index calculation methods to extract the vegetation index in the image after elimination, such as the normalized difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), etc. The vegetation index can reflect the growth status of vegetation and affect the geometric transformation of the image. Then, agricultural and land use characteristic information related to farmland distribution data and land use type data is obtained. These data can include farmland distribution maps, land use type maps, vegetation coverage, etc. In agricultural and land use scenarios, by combining building edges and urban characteristic information, the RPC parameters can be calculated more accurately, considering the influence of natural elements such as farmland and vegetation on the image after elimination.

[0096] Figure 3 FIG. 4 is a schematic structural diagram of a satellite image sub-frame cutting and RPC parameter generation device provided by the present invention. The satellite image sub-frame cutting and RPC parameter generation device includes an acquisition unit 1 for sub-frame cutting the original satellite image to obtain all the cut images. The working principle of the acquisition unit 1 can refer to the foregoing step 101 and will not be elaborated here.

[0097] The satellite image sub-frame cutting and RPC parameter generation device further includes an input unit 2 for inputting each cut image into a preset classification model for each cut image to obtain an image recognition result output by the preset classification model. The working principle of the input unit 2 can refer to the foregoing step 102 and will not be elaborated here.

[0098] The satellite image sub-frame cutting and RPC parameter generation device further includes an elimination unit 3 for determining the characteristics of elements to be eliminated according to the application scenario of the cut image, and eliminating the image recognition results with the characteristics of the elements to be eliminated from the cut image to obtain an image after elimination. The application scenarios include natural environment scenarios, urban construction scenarios, and agricultural and land use scenarios. The working principle of the elimination unit 3 can refer to the foregoing step 103 and will not be elaborated here.

[0099] The satellite image sub-frame cutting and RPC parameter generation device further includes a determination unit 4 for calculating the RPC parameters corresponding to the image after elimination according to the application scenario, geometrically correcting the image after elimination using the RPC parameters, and determining a corrected image. The working principle of the determination unit 4 can refer to the foregoing step 104 and will not be elaborated here.

[0100] The preset classification model is determined after training based on all image samples and the labeled tags corresponding to different feature characteristics in each image sample. The labeled tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts.

[0101] By inputting the cropped image into the preset classification model, the present invention obtains accurate recognition results for different feature characteristics, aiming to crop and identify the features in the original satellite image under a specific scenario, and perform image rejection and geometric correction on the image to obtain image data that meets the requirements of a specific application scenario. Then, in combination with the application scenario again, the RPC parameters corresponding to the image after rejection are calculated, and geometric correction is performed on the image, so as to more accurately convert the image pixel coordinates into geographical coordinates, ensuring the spatial accuracy and consistency of the image data. The present invention can flexibly determine the rejected feature characteristics and perform geometric correction according to the requirements of different application scenarios, providing customized image data processing solutions for different fields such as natural environment, urban construction, and agricultural land use.

[0102] Figure 4 It is a schematic structural diagram of the electronic device provided by the present invention. As Figure 4 shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440. Among them, the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call the logical instructions in the memory 430 to execute the satellite image sub-frame cropping and RPC parameter generation method, which includes: sub-frame cropping the original satellite image to obtain all cropped images; for each cropped image, inputting the cropped image into the preset classification model to obtain the image recognition result output by the preset classification model; according to the application scenario of the cropped image, determining the feature characteristics to be rejected, and rejecting the image recognition results with the feature characteristics to be rejected from the cropped image to obtain the image after rejection. The application scenario includes natural environment scenarios, urban construction scenarios, and agricultural and land use scenarios; according to the application scenario, calculating the RPC parameters corresponding to the image after rejection, and using the RPC parameters to perform geometric correction on the image after rejection to determine the corrected image; the preset classification model is determined after training based on all image samples and the labeled tags corresponding to different feature characteristics in each image sample. The labeled tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts.

[0103] In addition, when the logical instructions in the above-mentioned memory 430 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0104] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a method for satellite image sub-framing and RPC parameter generation provided by the above-mentioned various methods. The method includes: sub-framing and cutting the original satellite image to obtain all the cut images; for each cut image, input the cut image into a preset classification model to obtain the image recognition result output by the preset classification model; according to the application scenario of the cut image, determine the feature of the element to be excluded, and exclude the image recognition result with the feature of the element to be excluded from the cut image to obtain the image after exclusion. The application scenario includes natural environment scenarios, urban construction scenarios, and agricultural and land use scenarios; according to the application scenario, calculate the RPC parameters corresponding to the image after exclusion, and use the RPC parameters to perform geometric correction on the image after exclusion to determine the corrected image; the preset classification model is determined after training according to all image samples and the marked labels corresponding to different element features in each image sample. The marked labels include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts.

[0105] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the satellite image sub-framing and cutting and RPC parameter generation methods provided by the above various methods. The method includes: sub-framing and cutting the original satellite image to obtain all the cut images; for each cut image, input the cut image into a preset classification model to obtain the image recognition result output by the preset classification model; according to the application scenario of the cut image, determine the feature of the element to be excluded, and exclude the image recognition result with the feature of the element to be excluded from the cut image to obtain the image after exclusion. The application scenarios include natural environment scenarios, urban construction scenarios, and agricultural and land use scenarios; according to the application scenario, calculate the RPC parameters corresponding to the image after exclusion, and use the RPC parameters to perform geometric correction on the image after exclusion to determine the corrected image; the preset classification model is determined after being trained according to all image samples and the marked labels corresponding to different element features in each image sample, and the marked labels include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soils, rocks, and deserts.

[0106] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0107] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solutions, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for satellite image sub - frame cutting and RPC parameter generation, characterized in that, Including: Performing sub - image cutting on the original satellite image to obtain all the cut images; For each cut image, inputting the cut image into a preset classification model to obtain the image recognition result output by the preset classification model; According to the application scenario of the cut image, determining the feature of the element to be removed, and removing the image recognition result with the feature of the element to be removed from the cut image to obtain the image after removal. The application scenarios include natural environment scenarios, urban construction scenarios, and agricultural and land - use scenarios; According to the application scenario, calculating the RPC parameters corresponding to the image after removal, and performing geometric correction on the image after removal using the RPC parameters to determine the corrected image; The preset classification model is determined after training based on all image samples and the labeled tags corresponding to different element features in each image sample. The labeled tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soil, rocks, and deserts; When the application scenario is a natural environment scenario, the calculating the RPC parameters corresponding to the image after removal according to the application scenario includes: Registering the original digital elevation model data according to the geographic coordinates and longitude - latitude information of the cut image to obtain the target digital elevation model data corresponding to the cut image; Performing terrain correction on the cut image using the target digital elevation model data to obtain the corrected image; Constructing a first RPC model according to the platform position, platform attitude, and pixel size corresponding to the corrected image, and solving the RPC parameters corresponding to the image after removal based on the first RPC model; When the application scenario is an urban construction scenario, the calculating the RPC parameters corresponding to the image after removal according to the application scenario includes: constructing a second RPC model according to the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the image after removal, and solving the RPC parameters corresponding to the image after removal based on the second RPC model; When the application scenario is an agricultural and land - use scenario, the calculating the RPC parameters corresponding to the image after removal according to the application scenario includes: constructing a third RPC model according to the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the image after removal, and solving the RPC parameters corresponding to the image after removal based on the third RPC model.

2. The method for satellite image sub - frame cutting and RPC parameter generation according to claim 1, wherein, The determining the feature of the element to be removed according to the application scenario of the cut image includes: When the application scenario is a natural environment scenario, determining the feature of the element to be removed as houses, roads, and bridges; When the application scenario is an urban construction scenario, determining the feature of the element to be removed as lakes, rivers, reservoirs, forests, grasslands, and farmlands; When the application scenario is an agricultural and land - use scenario, determining the feature of the element to be removed as houses, roads, bridges, bare soil, rocks, and deserts.

3. The method for satellite image sub - frame cutting and RPC parameter generation according to claim 1, wherein, Constructing a first RPC model based on the platform position, platform attitude, and pixel size corresponding to the corrected image includes: Among them, (X im , Y im , Z im ) are the pixel coordinates of the platform position in the image coordinate system considering the terrain, (X, Y) are the geographical coordinates in the geographical coordinate system, (Roll, Pitch, Yaw) are the platform attitudes, (Sx, Sy) are the pixel sizes, and the functions f and g are polynomial functions.

4. The method for satellite image sub - frame cutting and RPC parameter generation according to claim 1, characterized in that, Solving the RPC parameters corresponding to the image after rejection based on the first RPC model includes: Solving all the first polynomial coefficients in the function f and all the second polynomial coefficients in the function g using the least squares method, and determining all the first polynomial coefficients and all the second polynomial coefficients as the RPC parameters corresponding to the image after rejection.

5. The method for satellite image mosaicking and RPC parameter generation according to claim 1, wherein Constructing a second RPC model based on the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the image after rejection includes: Among them, (X im , Y im ) is the pixel coordinate of the platform position in the image coordinate system without considering the terrain, (X, Y) is the geographic coordinate in the geographic coordinate system, (Roll, Pitch, Yaw) is the platform attitude, (Sx, Sy) is the pixel size, BE is the building edge information Building_Edges, UF is the urban feature information Urban_Features, and the functions f and g are polynomial functions.

6. The method for satellite image sub-frame cutting and RPC parameter generation according to claim 1, wherein Constructing a third RPC model based on the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the image after rejection includes: Among them, (X im , Y im ) is the pixel coordinate of the platform position in the image coordinate system without considering the terrain, (X, Y) is the geographic coordinate in the geographic coordinate system, (Roll, Pitch, Yaw) is the platform attitude, (Sx, Sy) is the pixel size, VI is the vegetation index information Vegetation_Index, FF is the farmland feature information Farmland_Features, and the functions f and g are polynomial functions.

7. The method for satellite image sub - frame cutting and RPC parameter generation according to claim 1, wherein After determining the corrected image, the method further includes: Traversing all the original satellite images to determine the corrected image and RPC parameters corresponding to each original satellite image; Outputting all the corrected images and the RPC parameters corresponding to each corrected image to a preset storage unit to implement geometric information processing of different corrected images according to different application scenarios.

8. The method for satellite image sub - frame cutting and RPC parameter generation according to claim 1, wherein, Mosaicking and cutting the original satellite images to obtain all the cut images includes: Loading the original satellite images using GIS software, and determining the cutting range from the original satellite images according to the geographic coordinates and longitude and latitude information; Mosaicking and cutting the original satellite images within the cutting range to obtain all the cut images; The geographic coordinates and the longitude and latitude information are determined after overlaying a preset geographic entity as a layer on the original satellite images.

9. A satellite image sub - frame cutting and RPC parameter generation device, characterized in that, Includes: An acquisition unit for mosaicking and cutting the original satellite images to obtain all the cut images; An input unit for inputting each cut image into a preset classification model for each cut image to obtain an image recognition result output by the preset classification model; A rejection unit for determining the feature of the element to be rejected according to the application scenario of the cut image, and rejecting the image recognition result with the feature of the element to be rejected from the cut image to obtain the image after rejection, where the application scenario includes natural environment scenario, urban construction scenario, and agricultural and land use scenario; A determination unit for calculating the RPC parameters corresponding to the image after rejection according to the application scenario, geometrically correcting the image after rejection using the RPC parameters, and determining the corrected image; The preset classification model is determined after training according to all the image samples and the labeled tags corresponding to different element features in each image sample, and the labeled tags include lakes, rivers, reservoirs, forests, grasslands, farmlands, houses, roads, bridges, bare soils, rocks, and deserts; In the case where the application scenario is a natural environment scenario, calculating the RPC parameters corresponding to the image after rejection according to the application scenario includes: Register the original digital elevation model data according to the geographic coordinates and latitude and longitude information of the cropped image, and obtain the target digital elevation model data corresponding to the cropped image; Perform terrain correction on the cropped image using the target digital elevation model data to obtain the corrected image; Construct a first RPC model based on the platform position, platform attitude, and pixel size corresponding to the corrected image, and solve the RPC parameters corresponding to the removed image based on the first RPC model; When the application scenario is an urban construction scenario, calculating the RPC parameters corresponding to the removed image according to the application scenario includes: constructing a second RPC model based on the platform position, platform attitude, pixel size, building edge information, and urban feature information corresponding to the removed image, and solving the RPC parameters corresponding to the removed image based on the second RPC model; When the application scenario is an agriculture and land use scenario, calculating the RPC parameters corresponding to the removed image according to the application scenario includes: constructing a third RPC model based on the platform position, platform attitude, pixel size, vegetation index information, and farmland feature information corresponding to the removed image, and solving the RPC parameters corresponding to the removed image based on the third RPC model.

Images