A model improvement method, device, equipment and medium

Abstract

The application provides a model improvement method, including: acquiring a satellite multi-angle observation data set; inputting the multi-angle observation data set into an RTLSR model; inputting the multi-angle observation data set into an RTLSRS model; based on a first weight coefficient value of each pixel and a second weight coefficient value of each pixel, establishing a relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model through a random forest multi-output regression algorithm to acquire an expression of the second weight coefficient with respect to the first weight coefficient; and bringing the expression of the second weight coefficient with respect to the first weight coefficient into an expression of the RTLSRS model. The application establishes the relationship between the RTLSR model and the RTLSRS model, thereby obtaining the weight coefficients of the corresponding models, and simultaneously obtaining a new RTLSRS model, so that the target product is improved and the precision of the target product is improved.

Description

Technical Field

[0001] This invention belongs to the field of spatial information technology, specifically relating to a model improvement method, apparatus, device, and medium. Background Technology

[0002] Snow cover, as a crucial land cover type within the Earth-atmosphere system, has a wide distribution range and exhibits distinct interannual and seasonal variations. As a key indicator of climate change, snow cover is extremely sensitive to changes in surface radiation balance and temperature. Its high reflectivity and large coverage area make it a critical factor in the radiation energy balance of the Earth-atmosphere system. Therefore, accurately characterizing the reflectivity of snow cover is of great significance for understanding global energy radiation balance, climate change, and the carbon cycle.

[0003] Currently, MODIS-MCD43A products are widely used in quantitative remote sensing. However, the operational algorithm for MCD43A products (i.e., the kernel-driven RTLSR model) ignores the differences between snow reflectance characteristics and vegetation-soil reflectance characteristics, leading to significant errors in MCD43A product production. To better describe the anisotropic characteristics of snow cover, a kernel-driven RTLSRS model was developed, but it is difficult to directly use this model to improve the accuracy of MCD43A products. Summary of the Invention

[0004] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide the weight coefficient values ​​of each pixel based on the RTLSR model and the RTLSRS model. By establishing the relationship between the weight coefficient expressions of the RTLSR model and the RTLSRS model, the expression of the weight coefficients of the RTLSRS model with respect to the weight coefficients of the RTLSR model is obtained. At the same time, the expression is substituted into the RTLSRS model, thereby improving the accuracy of the target product, improving the snow reflection characteristics and albedo of the target product, and solving the problem of large errors generated during the production of the target product.

[0005] To achieve the above and other related objectives, the present invention provides a method for model improvement and product improvement, comprising: acquiring a satellite multi-angle observation dataset; inputting the multi-angle observation dataset into an RTLSR model to obtain a first weight coefficient value for each pixel retrieved by the RTLSR model; inputting the multi-angle observation dataset into an RTLSRS model to obtain a second weight coefficient value for each pixel retrieved by the RTLSRS model; based on the first weight coefficient value and the second weight coefficient value for each pixel, establishing a relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model using a random forest multi-output regression algorithm to obtain an expression for the second weight coefficient with respect to the first weight coefficient; and substituting the expression for the second weight coefficient with respect to the first weight coefficient into the expression of the RTLSRS model to obtain a new RTLSRS model.

[0006] In an optional embodiment of the present invention, in the step of inputting the multi-angle observation dataset into the RTLSR model to obtain the first weight coefficient value of each pixel inverted by the RTLSR model, the first weight coefficient value includes the weight coefficient values ​​of the isotropic kernel, the volume scattering kernel and the geometric optics kernel of the kernel-driven RTLSR model.

[0007] In an optional embodiment of the present invention, in the step of inputting the multi-angle observation dataset into the RTLSRS model to obtain the second weight coefficient value of each pixel inverted by the RTLSRS model, the second weight coefficient includes weight coefficient values ​​representing the isotropic kernel, the volume scattering kernel, the geometric optical kernel, and the snow kernel of the kernel-driven RTLSRS model.

[0008] In an optional embodiment of the present invention, in the step of inputting the multi-angle observation dataset into the RTLSR model to obtain the first weight coefficient value of each pixel inverted by the RTLSR model, the RTLSR model expression is as follows:

[0009]

[0010] Where R(θ) s θ v θ represents the surface reflectance of a natural feature at wavelength λ, where φ and λ represent the surface reflectance of the feature at wavelength λ. s θ v φ represents the solar zenith angle, the observed zenith angle, and the relative azimuth angle. f iso (λ) f vol (λ) and f geo (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSR model, respectively.vol (θ s θ v ,φλ) and geometrical optical kernel K geo (θ s θ v The weighting coefficients of , φ, λ).

[0011] In an optional embodiment of the present invention, in the step of inputting the multi-angle observation dataset into the RTLSRS model to obtain the second weight coefficient value of each pixel inverted by the RTLSRS model, the RTLSRS model expression is as follows:

[0012]

[0013] Where R(θ) s θ v θ represents the surface reflectance of a natural feature at wavelength λ, where φ and λ represent the surface reflectance of the feature at wavelength λ. s θ v φ represents the solar zenith angle, the observed zenith angle, and the relative azimuth angle. f ' iso (λ) f ' vol (λ) f ' geo (λ) and f ' snw (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSRS model, respectively. vol (θ s θ v , φ, λ), geometrical optical kernel K geo (θ s θ v (φ, λ) and snow core K snw (θ s θ v The weighting coefficients of , φ, λ).

[0014] In an optional embodiment of the present invention, in the step of establishing the relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model based on the first weight coefficient value and the second weight coefficient value of each pixel using the random forest multi-output regression algorithm, to obtain the expression of the second weight coefficient with respect to the first weight coefficient, the expression is as follows:

[0015]

[0016]

[0017]

[0018]

[0019] Where F represents the multi-output random forest algorithm, θ s Indicates the zenith angle of the sun. f iso (λ) f vol (λ) and f geo (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSR model, respectively. vol (θ s θ v , φ, λ) and geometrical optical kernel K geo (θ s θ v The weighting coefficients of φ and λ. f ' iso (λ) f ' vol (λ) f ' geo (λ) and f ' snw (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSRS model, respectively. vol (θ s θ v , φ, λ), geometrical optical kernel K geo (θ s θ v (φ, λ) and snow core K snw (θ s θ v The weighting coefficients of , φ, λ).

[0020] To achieve the above and other related objectives, the present invention also provides a product comprising: a new RTLSRS model; a target product; the new RTLSRS model including a model improvement method, the model improvement method comprising: acquiring a satellite multi-angle observation dataset; inputting the multi-angle observation dataset into the RTLSR model to obtain a first weight coefficient value for each pixel inverted by the RTLSR model; inputting the multi-angle observation dataset into the RTLSRS model to obtain a second weight coefficient value for each pixel inverted by the RTLSRS model; based on the first weight coefficient value and the second weight coefficient value for each pixel, establishing a relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model using a random forest multi-output regression algorithm to obtain an expression for the second weight coefficient with respect to the first weight coefficient; and substituting the expression for the second weight coefficient with respect to the first weight coefficient into the expression of the RTLSRS model to obtain a new RTLSRS model.

[0021] The new RTLSRS model is then incorporated into the target product to obtain the new target product.

[0022] To achieve the above and other related objectives, the present invention also provides a model improvement apparatus comprising: a dataset acquisition module for acquiring a satellite multi-angle observation dataset; a first weight coefficient value module for inputting the multi-angle observation dataset into an RTLSR model to obtain a first weight coefficient value for each pixel inverted by the RTLSR model; a second weight coefficient value module for inputting the multi-angle observation dataset into the RTLSRS model to obtain a second weight coefficient value for each pixel inverted by the RTLSRS model; an expression acquisition module for establishing a relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model based on the first weight coefficient value and the second weight coefficient value of each pixel using a random forest multi-output regression algorithm to obtain an expression for the second weight coefficient with respect to the first weight coefficient; and a model acquisition module for substituting the expression for the second weight coefficient with respect to the first weight coefficient into the expression of the RTLSRS model to obtain a new RTLSRS model.

[0023] The technical advantage of this invention lies in providing a model improvement method. Based on the weight coefficient values ​​of each pixel in the RTLSR model and the weight coefficient values ​​of each pixel in the RTLSRS model, the method establishes a relationship between the weight coefficients of the RTLSR model and the weight coefficients of the RTLSRS model to obtain an expression for the weight coefficients of the RTLSR model with respect to the weight coefficients of the RTLSRS model. This improves the RTLSRS model. Simultaneously, the expression for the weight coefficients of the RTLSRS model with respect to the weight coefficients of the RTLSR model is substituted into the RTLSRS model to achieve improvement and enhancement of the target product. Attached Figure Description

[0024] Figure 1 Application scenario diagram of the model improvement method proposed in this invention;

[0025] Figure 2 Flowchart of the model improvement method proposed in this invention;

[0026] Figure 3 A detailed flowchart of the model improvement method proposed in this invention;

[0027] Figure 4 The improved albedo data and the effect diagram of site albedo proposed in this invention;

[0028] Figure 5 The improved albedo data and the effect diagram of site albedo proposed in this invention;

[0029] Figure 6 A functional module diagram of a product proposed in this invention;

[0030] Figure 7 Functional block diagram of the model improvement device proposed in this invention;

[0031] Figure 8 The structural block diagram of the electronic device proposed in this invention. Detailed Implementation

[0032] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0033] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the illustrations only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0034] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0035] The Moderate-resolution Imaging Spectroradiometer (MODIS) is a large space-based remote sensing instrument developed by NASA to understand global climate change and the impact of human activities on climate. This instrument captures data in 36 mutually registered spectral bands, covering the visible to infrared spectrum. It provides Earth surface observation data every 1–2 days. It is designed to provide dynamic measurements of large-scale global data, including changes in cloud cover, Earth's energy radiation, ocean and land variations, and low-altitude atmospheric processes. MODIS Basic Parameters: Orbit Altitude: 705km, Scan Rate: 20.3rpm, Coverage Area: 2330 km, Size: 1.0x1.6x1.0m, Weight: 228.7kg, Power: 162.5W, Data Rate: 10.6Mbps (peak daytime); 6.1Mbps (orbital average), Quantization: 12 bits, Spatial Resolution: 250m (bands 1-2); 500m (bands 3-7); 1000m (bands 8-36); Design Life: 6 years.

[0036] The MODIS MCD43A series products are now widely used in quantitative remote sensing. Since 2000, MODIS instruments on the Terra and Aqua platforms have been offering operational MCD43A series products, including BRDF model parameter products (MCD43A1), product quality (MCD43A2), albedo products (MCD43A3), and zenith reflectance (Nadir BRDF A-adjusted Reflectance, NBAR) products (MCD43A4) (Schaaf et al., 2002). Compared to MODIS V005 products, the MODIS V006 daily MCD43A products with a spatial resolution of 500 meters offer improved temporal daily resolution (Wang et al., 2018).

[0037] Figure 1 The application scenario diagram of the model method provided in the embodiments of the present invention shows that the solar zenith angle, observed zenith angle, relative azimuth angle, and reflectance values ​​of each band are obtained through the Polarization and Directionality of the Earth's Reflectances (POLDER) sensor. The solar zenith angle, observed zenith angle, relative azimuth angle, and reflectance values ​​of each band are input into the RTLSR model and the RTLSRS model, respectively, to obtain the first weight coefficient value and the second weight coefficient value of the RTLSR model and the RTLSRS model inversion. Then, based on the first weight coefficient value and the second weight coefficient value, the relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model is established through the forest random multi-output regression algorithm to obtain the expression of the second weight coefficient with respect to the first weight coefficient. The expression of the second weight coefficient with respect to the first weight coefficient is substituted into the RTLSRS model to obtain a new RTLSRS model. The new RTLSRS model is put into the MODIS MCD43A product to improve and enhance the product.

[0038] In other application scenarios, improvements to the MODIS MCD43A product can be made according to the actual situation, and the embodiments of the present invention do not limit this.

[0039] The electronic device can be any electronic product that can interact with the user, such as a personal computer, tablet computer, smartphone, personal digital assistant (PDA), interactive network television (IPTV), smart wearable device, etc.

[0040] The electronic device may also include network devices and / or user devices. The network devices include, but are not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0041] The networks in which the electronic devices are located include, but are not limited to, the Internet, wide area networks, metropolitan area networks, local area networks, and virtual private networks (VPNs).

[0042] Figure 2 This is a flowchart illustrating a model improvement method provided by an embodiment of the present invention. The method addresses the issue of improving the accuracy of the MODIS MCD43A product in snow-covered areas. It proposes a method for improving the MODIS MCD43A product based on the weight coefficients of each pixel in the RTLSR model and the RTLSRS model. This method establishes a relationship between the weight coefficients of the RTLSR model and the RTLSRS model, thereby improving the product's accuracy. This method can be applied to... Figure 1 The implementation environment shown is intended to illustrate a method that can also be applied to other instance implementation environments and executed by devices in those environments. This embodiment does not limit the implementation environment to which the method is applicable.

[0043] like Figure 2 As shown, one model improvement method in this embodiment includes:

[0044] Step S21: Acquire satellite multi-angle observation dataset. It should be noted that, in one specific embodiment, the satellite multi-angle observation dataset is acquired through a POLDER sensor, and the multi-angle observation data acquired by the POLDER is converted into MODIS multi-angle observation data through band conversion. Specifically, the POLDER sensor includes a linear CCD camera, a wide-field-of-view electro-optics device, and a filter measurement instrument with a rotating wheel. It includes nine channels from the blue band (0.443µm) to the near-infrared band (1.020µm), with polarization observations at 0.490, 0.670, and 0.865µm. The bandwidth of the spectral bands is between 20 and 40 nm. The POLDER has a field of view of ±51° along the orbit and ±43° outside the orbit, allowing observation of ground targets from up to 15 angles. Its unique multi-angle observation data is highly suitable for probing the physical properties of clouds and aerosols.

[0045] In one specific embodiment, the satellite multi-angle observation data includes the solar zenith angle, observed zenith angle, relative azimuth angle, and reflectivity data for six wavelengths (wavelengths of 490 nm, 565 nm, 670 nm, 765 nm, 865 nm, and 1020 nm). Specifically, the satellite multi-angle observation data can be directly downloaded from existing websites; the download address is: https: / / doi.pangaea.de / 10.1594 / PANGAEA.864090 .

[0046] Step S22: Input the multi-angle observation dataset into the RTLSR model to obtain the first weight coefficient value of each pixel inverted by the RTLSR model. It should be noted that the first weight coefficient value includes the weight coefficient values ​​of the isotropic kernel, volume scattering kernel, and geometric optics kernel of the kernel-driven RTLSR model.

[0047] In one specific embodiment, the RTLSR model expression is as follows:

[0048]

[0049] Where R(θ) s θ v θ represents the surface reflectance of a natural feature at wavelength λ, where φ and λ represent the surface reflectance of the feature at wavelength λ. s θ v φ represents the solar zenith angle, the observed zenith angle, and the relative azimuth angle, respectively; fiso(λ), fvol(λ), and fgeo(λ) represent the isotropic kernel, the volume scattering kernel K, and the volume scattering kernel K in the kernel-driven RTLSR model, respectively. vol (θ s θ v , φ, λ) and geometrical optical kernel K geo (θ s θ v The weighting coefficients of , φ, λ).

[0050] Step S23: Input the multi-angle observation dataset into the RTLSRS model to obtain the second weight coefficient value of each pixel inverted by the RTLSRS model. It should be noted that the second weight coefficient value includes the weight coefficient values ​​of the isotropic kernel, volume scattering kernel, geometric optics kernel, and snow kernel of the kernel-driven RTLSR model.

[0051] In one specific embodiment, the RTLSRS model expression is as follows:

[0052]

[0053] Where R(θ) s θ v θ represents the surface reflectance of a natural feature at wavelength λ, where φ and λ represent the surface reflectance of the feature at wavelength λ. s θv φ represents the solar zenith angle, the observed zenith angle, and the relative azimuth angle. f ' iso (λ) f ' vol (λ) f ' geo (λ) and f ' snw (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSRS model, respectively. vol (θ s θ v , φ, λ), geometrical optical kernel K geo (θ s θ v (φ, λ) and snow core K snw (θ s θ v The weighting coefficients of , φ, λ).

[0054] Step S24: Based on the first weight coefficient value and the second weight coefficient value of each pixel, establish the relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model using the Random Forest Multi-Output Regression algorithm, to obtain the expression of the second weight coefficient with respect to the first weight coefficient. It should be noted that the Random Forest Multi-Output Regression algorithm trains a random forest, and since it is a regression task, the average value of the samples in the leaf nodes is taken as the predicted value. Of course, when building the random forest, valuable features are preferentially selected, and then important features are obtained through the random forest, namely the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model in this invention, thereby obtaining the expression of the second weight coefficient with respect to the first weight coefficient.

[0055] In one specific embodiment, the expression for the second weighting coefficient with respect to the first weighting coefficient is as follows:

[0056]

[0057]

[0058]

[0059]

[0060] Where F represents the multi-output random forest algorithm, θ s Indicates the zenith angle of the sun. f iso (λ) f vol (λ) and fgeo (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSR model, respectively. vol (θ s θ v , φ, λ) and geometrical optical kernel K geo (θ s θ v The weighting coefficients of φ and λ. f ' iso (λ) f ' vol (λ) f ' geo (λ) and f ' snw (λ) represents the isotropic kernel and the bulk scattering kernel K in the kernel-driven RTLSRS model, respectively. vol (θ s θ v , φ, λ), geometrical optical kernel K geo (θ s θ v (φ, λ) and snow core K snw (θ s θ v The weighting coefficients of , φ, λ).

[0061] Step S25: Substitute the expression of the second weight coefficient with respect to the first weight coefficient into the expression of the RTLSRS model to obtain the new RTLSRS model.

[0062] Figure 3 This is a flowchart illustrating the model improvement method provided in an embodiment of the present invention. The technical solution of the present invention will be described below with reference to a specific embodiment:

[0063] Multi-angle observation data acquired by the POLDER sensor is converted into MODIS multi-angle data (solar zenith angle, observed zenith angle, and relative azimuth angle) via band conversion. These data are then input into the RTLSR and RTLSRS models, respectively, to obtain the first and second weight coefficient values ​​derived from the RTLSR and RTLSRS models. Based on these values, a forest random multi-output regression algorithm is used to establish the relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model, yielding an expression for the second weight coefficient with respect to the first weight coefficient. This expression is then substituted into the RTLSRS model to obtain a new RTLSRS model. This new RTLSRS model is then applied to the MODIS MCD43A product to correct the site albedo of the MODIS MCD43A product, and the correction effect is obtained.

[0064] like Figure 4-5 As shown, in one specific embodiment, a correction method is used to correct the improved target product and the original target product to obtain a comparison result diagram before and after correction. Figure 4 It is the albedo data before correction and the site albedo. Figure 5 This refers to the corrected albedo data and the site albedo. Figure 4 It can be seen that the albedo data before correction significantly underestimates the station albedo, resulting in poor overall fitting accuracy. Compared with the albedo data before correction, the data from the original data is significantly lower. Figure 5 It can be seen that the corrected albedo data is significantly improved compared with the site albedo, the root mean square error of the fit is reduced by 10.8%, the bias is reduced by 56.7%, and the coefficient of determination is also significantly improved.

[0065] like Figure 6 As shown, this embodiment proposes a product including: a target product 61; a new RTLSRS model 62; the new RTLSRS model 62 includes a model improvement method, which includes: acquiring a satellite multi-angle observation dataset; inputting the multi-angle observation dataset into the RTLSR model to obtain a first weight coefficient value for each pixel inverted by the RTLSR model; inputting the multi-angle observation dataset into the RTLSRS model to obtain a second weight coefficient value for each pixel inverted by the RTLSRS model; based on the first weight coefficient value and the second weight coefficient value for each pixel, establishing a relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model using a random forest multi-output regression algorithm to obtain an expression for the second weight coefficient with respect to the first weight coefficient; and substituting the expression for the second weight coefficient with respect to the first weight coefficient into the expression of the RTLSRS model to obtain the new RTLSRS model 62. Please refer to the above for details.

[0066] The new RTLSR model 62 is placed into the target product 61 to obtain the new target product.

[0067] In summary, the model improvement method provided by this invention can solve the problems of existing technologies neglecting the differences between snow reflectance characteristics and vegetation-soil reflectance characteristics, and can also solve the problems of large errors caused by existing technologies in the production of MCD43A products. At the same time, it is difficult to improve the accuracy of MCD43A products using the kernel-driven RTLSRS model while improving snow reflectance characteristics and albedo.

[0068] like Figure 7As shown, the present invention also provides a model improvement system comprising: a dataset acquisition module 71, a first weight coefficient value module 72, a second weight coefficient value module 73, an expression acquisition module 74, and a model acquisition module 75. The dataset acquisition module 71 acquires a satellite multi-angle observation dataset; the first weight coefficient value module 72 inputs the multi-angle observation dataset into an RTLSR model to obtain a first weight coefficient value for each pixel inverted by the RTLSR model; the second weight coefficient value module 73 inputs the multi-angle observation dataset into the RTLSRS model to obtain a second weight coefficient value for each pixel inverted by the RTLSRS model; the expression acquisition module 74, based on the first weight coefficient value and the second weight coefficient value of each pixel, establishes a relationship between the first weight coefficient of the RTLSR model and the second weight coefficient of the RTLSRS model using a random forest multi-output regression algorithm to obtain an expression for the second weight coefficient with respect to the first weight coefficient; and the model acquisition module 75 substitutes the expression for the second weight coefficient with respect to the first weight coefficient into the expression of the RTLSRS model to obtain a new RTLSRS model.

[0069] It should be noted that the embodiments provided above are as follows: Figure 7 The model improvement device shown is based on the same concept as the model improvement method provided in the above embodiments. The specific ways in which each module and unit performs its operations have been described in detail in the method embodiments and will not be repeated here. In practical applications, the model improvement device provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0070] Embodiments of the present invention also provide an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to implement the model improvement methods provided in the above embodiments.

[0071] Figure 8 A schematic diagram of a computer system suitable for implementing embodiments of the present invention is shown. It should be noted that... Figure 8 The computer system 800 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0072] like Figure 8As shown, the computer system 800 includes a Central Processing Unit (CPU) 601, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 802 or programs loaded from storage portion 806 into Random Access Memory (RAM) 803, such as performing the methods described in the above embodiments. The RAM 803 also stores various programs and data required for system operation. The CPU 801, ROM 802, and RAM 803 are interconnected via a bus 804. An Input / Output (I / O) interface 805 is also connected to the bus 804.

[0073] The following components are connected to I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 806 including a hard disk, etc.; and a communication section 808 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 808 performs communication processing via a network such as the Internet. A drive 810 is also connected to I / O interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 810 as needed so that computer programs read from it can be installed into storage section 806 as needed.

[0074] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 808, and / or installed from removable medium 811. When the computer program is executed by central processing unit (CPU) 801, it performs various functions defined in the system of the present invention.

[0075] It should be noted that the computer-readable medium shown in the embodiments of the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0076] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0077] The units described in the embodiments of the present invention can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0078] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer's processor, causes the computer to perform the aforementioned model improvement method. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not incorporated into the electronic device.

[0079] Another aspect of the present invention provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the model improvement methods provided in the various embodiments described above.

[0080] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A model improvement method characterized by, The method comprises the following steps: acquiring a satellite multi-angle observation data set; inputting the multi-angle observation data set into an RTLSR model to acquire first weight coefficient values of each pixel inverted by the RTLSR model; inputting the multi-angle observation data set into an RTLSRS model to acquire second weight coefficient values of each pixel inverted by the RTLSRS model; based on the first weight coefficient values of each pixel and the second weight coefficient values of each pixel, establishing a relationship between the first weight coefficients of the RTLSR model and the second weight coefficients of the RTLSRS model through a random forest multi-output regression algorithm to acquire an expression of the second weight coefficients with respect to the first weight coefficients; inputting the expression of the second weight coefficients with respect to the first weight coefficients into an expression of the RTLSRS model to acquire a new RTLSRS model; wherein the expression of the second weight coefficients with respect to the first weight coefficients is as follows: where F denotes a multi-output random forest algorithm, θ s denotes the solar zenith angle, f iso (λ), f vol (λ) and f geo (λ) denote the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ) and the geometric optical kernel K geo (θ s , θ v , φ, λ) of the kernel-driven RTLSRS model, respectively. f ’ iso (λ), f ’ vol (λ), f ’ geo (λ) and f ’ snw (λ) denote the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ), the geometric optical kernel K geo (θ s , θ v , φ, λ) and the snow kernel K snw (θ s , θ v , φ, λ) of the kernel-driven RTLSRS model, respectively.

2. The model improvement method according to claim 1, characterized by, In the step of inputting the multi-angle observation data set into the RTLSR model to acquire the first weight coefficient values of each pixel inverted by the RTLSR model, the first weight coefficient values comprise weight coefficient values of an isotropic kernel, a volume scattering kernel and a geometric optical kernel of a kernel-driven RTLSR model.

3. The model improvement method according to claim 1, characterized by, In the step of inputting the multi-angle observation data set into the RTLSRS model to acquire the second weight coefficient values of each pixel inverted by the RTLSRS model, the second weight coefficients comprise weight coefficient values of an isotropic kernel, a volume scattering kernel, a geometric optical kernel and a snow kernel of a kernel-driven RTLSRS model.

4. The model improvement method according to claim 1, characterized by, In the step of inputting the multi-angle observation data set into the RTLSR model to acquire the first weight coefficient values of each pixel inverted by the RTLSR model, the expression of the RTLSR model is as follows: where R(θ s , θ v , φ, λ) represents the surface reflectance of natural features at wavelength λ, θ s , θ v and φ represent the solar zenith angle, the viewing zenith angle and the relative azimuth angle, f iso (λ), f vol (λ) and f geo (λ) represent the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ) and the geometric optics kernel K geo (θ s , θ v , φ, λ) of the kernel-driven RTLSR model, respectively.

5. The model refinement method of claim 4, wherein, In the step of inputting the multi-angle observation data set into the RTLSRS model to acquire the second weight coefficient values of each pixel inverted by the RTLSRS model, the expression of the RTLSRS model is as follows: where R(θ s , θ v , φ, λ) represents the surface reflectance of natural terrain at wavelength λ, θ s , θ v and φ represent the solar zenith angle, the observation zenith angle and the relative azimuth angle, f ’ iso (λ), f ’ vol (λ), f ’ geo (λ) and f ’ snw (λ) represent the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ), the geometric optics kernel K geo (θ s , θ v , φ, λ) and the snow kernel K snw (θ s , θ v , φ, λ) of the kernel-driven RTLSRS model, respectively.

6. A product characterized by, The method comprises the following steps: a new RTLSRS model; a target product; the new RTLSRS model comprises a model improvement method, and the model improvement method comprises the following steps: acquiring a satellite multi-angle observation data set; inputting the multi-angle observation data set into an RTLSR model to acquire first weight coefficient values of each pixel inverted by the RTLSR model; inputting the multi-angle observation data set into an RTLSRS model to acquire second weight coefficient values of each pixel inverted by the RTLSRS model; based on the first weight coefficient values of each pixel and the second weight coefficient values of each pixel, establishing a relationship between the first weight coefficients of the RTLSR model and the second weight coefficients of the RTLSRS model through a random forest multi-output regression algorithm to acquire an expression of the second weight coefficients with respect to the first weight coefficients; and inputting the expression of the second weight coefficients with respect to the first weight coefficients into an expression of the RTLSRS model to acquire a new RTLSRS model. Put the new RTLSRS model into the target product to obtain a new target product; The expression of the second weight coefficient with respect to the first weight coefficient is as follows: where F denotes a multi-output random forest algorithm, θ s denotes the solar zenith angle, f iso (λ), f vol (λ) and f geo (λ) denote the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ) and the geometric optical kernel K geo (θ s , θ v , φ, λ) of the kernel-driven RTLSRS model, respectively, f iso (λ), f vol (λ), f geo (λ) and f snw (λ) denote the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ), the geometric optical kernel K geo (θ s , θ v , φ, λ) and the snow kernel K snw (θ s , θ v , φ, λ) of the kernel-driven RTLSRS model, respectively.​​​​ 7. A model refinement system characterized by, The method comprises the following steps: An acquisition data set module acquires a satellite multi-angle observation data set; A first weight coefficient value module inputs the multi-angle observation data set into an RTLSR model to acquire a first weight coefficient value of each pixel inverted by the RTLSR model; A second weight coefficient value module inputs the multi-angle observation data set into an RTLSRS model to acquire a second weight coefficient value of each pixel inverted by the RTLSRS model; An acquisition expression module establishes a relationship between a first weight coefficient of the RTLSR model and a second weight coefficient of the RTLSRS model based on the first weight coefficient value of each pixel and the second weight coefficient value of each pixel through a random forest multi-output regression algorithm to acquire an expression of the second weight coefficient with respect to the first weight coefficient; An acquisition model module brings the expression of the second weight coefficient with respect to the first weight coefficient into an expression of the RTLSRS model to acquire a new RTLSRS model; The expression of the second weight coefficient with respect to the first weight coefficient is as follows: where F denotes a multi-output random forest algorithm, θ s denotes the solar zenith angle, f iso (λ), f vol (λ) and f geo (λ) denote the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ) and the geometric optical kernel K geo (θ s , θ v , φ, λ), f ’ iso (λ), f ’ vol (λ), f ’ geo (λ) and f ’ snw (λ) denote the weight coefficients of the isotropic kernel, the volume scattering kernel K vol (θ s , θ v , φ, λ), the geometric optical kernel K geo (θ s , θ v , φ, λ) and the snow kernel K snw (θ s , θ v , φ, λ).

8. An electronic device, comprising: The electronic device comprises: One or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, cause the electronic device to implement the model improvement method according to any one of claims 1-5.

9. A computer-readable storage medium, characterized in that, A computer program is stored thereon, which, when executed by a processor of a computer, causes the computer to execute the model improvement method according to any one of claims 1-5.

Images