A model training method and system for intelligence analysis
Feature extraction and classification analysis of satellite images through high-dimensional mapping and local feature acquisition functions, solving the complex problem of satellite intelligence data processing and realizing the accurate classification and analysis of intelligence data.
Patent Information
- Application Number
- CN202411718782.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-28
- Publication Date
- 2025-05-02
- Estimated Expiration
- 2044-11-28
AI Technical Summary
Satellite intelligence data processing is complex and it is difficult to conduct effective intelligence analysis.
The high-dimensional feature value of the satellite image is obtained through the image feature high-dimensional mapping function, and the spatial context information is obtained based on the local feature acquisition function, the distance between the feature value and the feature center is calculated, and classification analysis is performed through the category allocation function.
Accurate classification and analysis of satellite image intelligence data is achieved, and the efficiency and accuracy of intelligence processing are improved.
Smart Images
Figure CN119229310B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of satellite intelligence analysis, and more specifically, relates to a model training method and system for intelligence analysis. Background Art
[0002] Satellite intelligence is the technology and method of using satellite equipment to observe and collect data on the ground to obtain various types of information such as military, meteorological, and environmental information. Satellite intelligence is widely used in national defense security, scientific research, disaster monitoring, environmental protection and other fields.
[0003] The main types of satellite intelligence include:
[0004] Optical imaging: Use visible and infrared light to obtain high-resolution ground images.
[0005] Radar imaging: Using synthetic aperture radar (SAR) technology, electromagnetic wave imaging can penetrate clouds and night conditions to provide ground images in all weather and all day.
[0006] Signal intelligence: monitoring and analysis of electromagnetic signals from the ground, air and sea, including communication signals, radar signals, etc.
[0007] Electronic intelligence: Collection and analysis of signals emitted by electronic equipment, such as operating frequency and signal strength of radar and communication equipment.
[0008] The advantages of satellite intelligence are that it can cover large areas, quickly obtain information, and has the characteristics of concealment and high resolution. However, satellite intelligence also faces challenges such as huge data volume and complex information processing. Summary of the invention
[0009] In order to solve the above technical problems, the present invention proposes a model training method for intelligence analysis, wherein the intelligence comes from satellite images, including:
[0010] Acquire the satellite image, set an image feature high-dimensional mapping function, and acquire a high-dimensional image feature value of the satellite image at each pixel;
[0011] Set a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image;
[0012] Randomly initialize the feature center of the category, set the distance measurement function, and calculate the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value;
[0013] A category assignment function is set, and the final distance to each category is calculated based on the distance, the color vector of the satellite image at each pixel, the illumination intensity, and the edge intensity, and the pixel points corresponding to the feature values of the high-dimensional image are divided into corresponding categories to complete the classification analysis.
[0014] Furthermore, the image feature high-dimensional mapping function includes:
[0015] ,
[0016] in, For satellite images In pixels The high-dimensional image feature value at is the ReLU function or the Sigmoid function, is the horizontal coordinate of the pixel point, is the vertical coordinate of the pixel, is the third weight, is the second weight, is the first weight, For satellite images In pixels The color vector at For satellite images In pixels The light intensity at For satellite images Length, For satellite images The width of is the first bias term, is the second bias term, is the third bias term, For satellite images In pixels The edge strength at .
[0017] Furthermore, the local feature acquisition function includes:
[0018] ,
[0019] in, For satellite images In pixels The spatial context information feature value of the neighborhood at is used to capture the pixel point The local eigenvalues of Neighborhood pixel The weight of For satellite images In the neighborhood pixels The high-dimensional image feature value at Pixel Neighborhood of is the horizontal coordinate of the neighborhood pixel, is the vertical coordinate of the neighborhood pixel.
[0020] Furthermore, the distance metric function includes:
[0021] ,
[0022] in, For satellite images In pixels The high-dimensional image eigenvalue at Characteristic center of the class The distance is the first weight of the distance, For the The center point of all pixels in the class in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the vertical direction, For the The center point of all pixels in the class in the vertical direction, is the second weight of the distance, For the The feature center of the spatial context of the class.
[0023] Furthermore, the category assignment function include:
[0024] ,
[0025] in, For the The color center of the class, For the The geometric feature center of the class, is the color weight, is the context weight, is the geometric weight.
[0026] The present invention also proposes a model training system for intelligence analysis, wherein the intelligence comes from satellite images, including:
[0027] A feature value acquisition module is used to acquire the satellite image, set an image feature high-dimensional mapping function, and acquire a high-dimensional image feature value of the satellite image at each pixel point;
[0028] A local feature acquisition module is used to set a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image;
[0029] A distance calculation module, used for randomly initializing the feature center of a category, setting a distance measurement function, and calculating the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value;
[0030] The classification module is used to set the category assignment function and calculate the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the illumination intensity and the edge intensity, and divide the pixel points corresponding to the feature value of the high-dimensional image into corresponding categories to complete the classification analysis.
[0031] Furthermore, the image feature high-dimensional mapping function includes:
[0032] ,
[0033] in, For satellite images In pixels The high-dimensional image feature value at is the ReLU function or the Sigmoid function, is the horizontal coordinate of the pixel point, is the vertical coordinate of the pixel, is the third weight, is the second weight, is the first weight, For satellite images In pixels The color vector at For satellite images In pixels The light intensity at For satellite images Length, For satellite images The width of is the first bias term, is the second bias term, is the third bias term, For satellite images In pixels The edge strength at .
[0034] Furthermore, the local feature acquisition function includes:
[0035] ,
[0036] in, For satellite images In pixels The spatial context information feature value of the neighborhood at is used to capture the pixel point The local eigenvalues of Neighborhood pixel The weight of For satellite images In the neighborhood pixels The high-dimensional image feature value at Pixel Neighborhood of is the horizontal coordinate of the neighborhood pixel, is the vertical coordinate of the neighborhood pixel.
[0037] Furthermore, the distance metric function includes:
[0038] ,
[0039] in, For satellite images In pixels The high-dimensional image eigenvalue at Characteristic center of the class The distance is the first weight of the distance, For the The center point of all pixels in the class in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the vertical direction, For the The center point of all pixels in the class in the vertical direction, is the second weight of the distance, For the The feature center of the spatial context of the class.
[0040] Furthermore, the category assignment function include:
[0041] ,
[0042] in, For the The color center of the class, For the The geometric feature center of the class, is the color weight, is the context weight, is the geometric weight.
[0043] Compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:
[0044] The present invention obtains the high-dimensional image feature value of the satellite image at each pixel point through an image feature high-dimensional mapping function; obtains the spatial context information feature value of the neighborhood of the satellite image at each pixel point by setting a local feature acquisition function; calculates the distance between the high-dimensional image feature value and the feature center by setting a distance measurement function; finally, calculates the final distance to each category by setting a category allocation function, and divides the pixel points corresponding to the high-dimensional image feature value into corresponding categories, so that the intelligence data on the satellite image can be accurately classified and analyzed according to the classification results. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 is a flow chart of the method of embodiment 1 of the present invention;
[0046] Figure 2 It is a system structure diagram of embodiment 2 of the present invention. DETAILED DESCRIPTION
[0047] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.
[0048] The method provided by the present invention can be implemented in the following terminal environment, and the terminal may include one or more of the following components: a processor, a storage medium, and a display screen. The storage medium stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method described in the following embodiment.
[0049] The processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts in the entire terminal, and executes various functions of the terminal and processes data by running or executing instructions, programs, code sets or instruction sets stored in the storage medium, and calling data stored in the storage medium.
[0050] The storage medium may include a random access memory (RAM) or a read-only memory (ROM). The storage medium may be used to store instructions, programs, codes, code sets or instructions.
[0051] The display is used to show the interactive cross-section of each application.
[0052] All subscripts in the formulas of the present invention are only used to distinguish parameters and have no actual meaning.
[0053] In addition, those skilled in the art can understand that the structure of the above terminal does not constitute a limitation on the terminal, and the terminal may include more or fewer components, or combine certain components, or arrange the components differently. For example, the terminal also includes components such as a radio frequency circuit, an input unit, a sensor, an audio circuit, and a power supply, which will not be described in detail here.
[0054] Example 1
[0055] like Figure 1 As shown, the present invention proposes a model training method for intelligence analysis, wherein the intelligence comes from satellite images, including:
[0056] Step 101, obtaining the satellite image, setting an image feature high-dimensional mapping function, and obtaining a high-dimensional image feature value at each pixel of the satellite image;
[0057] Specifically, the image feature high-dimensional mapping function includes:
[0058] ,
[0059] in, For satellite images In pixels The high-dimensional image feature value at is the ReLU function or the Sigmoid function, is the horizontal coordinate of the pixel point, is the vertical coordinate of the pixel, is the third weight, is the second weight, is the first weight, For satellite images In pixels The color vector at For satellite images In pixels The light intensity at For satellite images Length, For satellite images The width of is the first bias term, is the second bias term, is the third bias term, For satellite images In pixels The edge strength at .
[0060] Step 102, setting a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image;
[0061] Specifically, the local feature acquisition function includes:
[0062] ,
[0063] in, For satellite images In pixels The spatial context information feature value of the neighborhood at is used to capture the pixel point The local eigenvalues of Neighborhood pixel The weight of For satellite images In the neighborhood pixels The high-dimensional image feature value at Pixel Neighborhood of is the horizontal coordinate of the neighborhood pixel, is the vertical coordinate of the neighborhood pixel.
[0064] Step 103, randomly initialize the feature center of the category, set the distance measurement function, and calculate the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value;
[0065] Specifically, the distance metric function includes:
[0066] ,
[0067] in, For satellite images In pixels The high-dimensional image eigenvalue at Characteristic center of the class The distance is the first weight of the distance, For the The center point of all pixels in the class in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the vertical direction, For the The center point of all pixels in the class in the vertical direction, is the second weight of the distance, For the The feature center of the spatial context of the class.
[0068] Step 104, setting a category assignment function, and calculating the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the illumination intensity and the edge intensity, and dividing the pixel points corresponding to the high-dimensional image feature value into corresponding categories to complete the classification analysis.
[0069] Specifically, the category assignment function include:
[0070] ,
[0071] in, For the The color center of the class, For the The geometric feature center of the class, is the color weight, is the context weight, is the geometric weight.
[0072] Example 2
[0073] The embodiment of the present invention further provides a model training system for intelligence analysis, wherein the intelligence comes from satellite images, including:
[0074] A feature value acquisition module is used to acquire the satellite image, set an image feature high-dimensional mapping function, and acquire a high-dimensional image feature value of the satellite image at each pixel point;
[0075] Specifically, the image feature high-dimensional mapping function includes:
[0076] ,
[0077] in, For satellite images In pixels The high-dimensional image feature value at is the ReLU function or the Sigmoid function, is the horizontal coordinate of the pixel point, is the vertical coordinate of the pixel, is the third weight, is the second weight, is the first weight, For satellite images In pixels The color vector at For satellite images In pixels The light intensity at For satellite images Length, For satellite images The width of is the first bias term, is the second bias term, is the third bias term, For satellite images In pixels The edge strength at .
[0078] A local feature acquisition module is used to set a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image;
[0079] Specifically, the local feature acquisition function includes:
[0080] ,
[0081] in, For satellite images In pixels The spatial context information feature value of the neighborhood at is used to capture the pixel point The local eigenvalues of Neighborhood pixel The weight of For satellite images In the neighborhood pixels The high-dimensional image feature value at Pixel Neighborhood of is the horizontal coordinate of the neighborhood pixel, is the vertical coordinate of the neighborhood pixel.
[0082] A distance calculation module, used for randomly initializing the feature center of a category, setting a distance measurement function, and calculating the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value;
[0083] Specifically, the distance metric function includes:
[0084] ,
[0085] in, For satellite images In pixels The high-dimensional image eigenvalue at Characteristic center of the class The distance is the first weight of the distance, For the The center point of all pixels in the class in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the vertical direction, For the The center point of all pixels in the class in the vertical direction, is the second weight of the distance, For the The feature center of the spatial context of the class.
[0086] The classification module is used to set the category assignment function and calculate the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the illumination intensity and the edge intensity, and divide the pixel points corresponding to the feature value of the high-dimensional image into corresponding categories to complete the classification analysis.
[0087] Specifically, the category assignment function include:
[0088] ,
[0089] in, For the The color center of the class, For the The geometric feature center of the class, is the color weight, is the context weight, is the geometric weight.
[0090] Example 3
[0091] An embodiment of the present invention also proposes a storage medium storing a plurality of instructions, wherein the instructions are used to implement the model training method for intelligence analysis.
[0092] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.
[0093] Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: Step 101 acquires the satellite image, sets an image feature high-dimensional mapping function, and acquires a high-dimensional image feature value of the satellite image at each pixel;
[0094] Specifically, the image feature high-dimensional mapping function includes:
[0095] ,
[0096] in, For satellite images In pixels The high-dimensional image feature value at is the ReLU function or the Sigmoid function, is the horizontal coordinate of the pixel point, is the vertical coordinate of the pixel, is the third weight, is the second weight, is the first weight, For satellite images In pixels The color vector at For satellite images In pixels The light intensity at For satellite images Length, For satellite images The width of is the first bias term, is the second bias term, is the third bias term, For satellite images In pixels The edge strength at .
[0097] Step 102, setting a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image;
[0098] Specifically, the local feature acquisition function includes:
[0099] ,
[0100] in, For satellite images In pixels The spatial context information feature value of the neighborhood at is used to capture the pixel point The local eigenvalues of Neighborhood pixel The weight of For satellite images In the neighborhood pixels The high-dimensional image feature value at Pixel Neighborhood of is the horizontal coordinate of the neighborhood pixel, is the vertical coordinate of the neighborhood pixel.
[0101] Step 103, randomly initialize the feature center of the category, set the distance measurement function, and calculate the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value;
[0102] Specifically, the distance metric function includes:
[0103] ,
[0104] in, For satellite images In pixels The high-dimensional image eigenvalue at Characteristic center of the class The distance is the first weight of the distance, For the The center point of all pixels in the class in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the vertical direction, For the The center point of all pixels in the class in the vertical direction, is the second weight of the distance, For the The feature center of the spatial context of the class.
[0105] Step 104, setting a category assignment function, and calculating the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the illumination intensity and the edge intensity, and dividing the pixel points corresponding to the high-dimensional image feature value into corresponding categories to complete the classification analysis.
[0106] Specifically, the category assignment function include:
[0107] ,
[0108] in, For the The color center of the class, For the The geometric feature center of the class, is the color weight, is the context weight, is the geometric weight.
[0109] Example 4
[0110] An embodiment of the present invention also proposes an electronic device, including a processor and a storage medium connected to the processor, wherein the storage medium stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute a model training method for intelligence analysis.
[0111] Specifically, the electronic device of this embodiment may be a computer terminal, and the computer terminal may include: one or more processors, and a storage medium.
[0112] Among them, the storage medium can be used to store software programs and modules, such as a model training method for intelligence analysis in an embodiment of the present invention, corresponding program instructions / modules, and the processor executes various functional applications and data processing by running the software programs and modules stored in the storage medium, that is, realizing the above-mentioned model training method for intelligence analysis. The storage medium may include a high-speed random storage medium, and may also include a non-volatile storage medium, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state storage media. In some instances, the storage medium may further include a storage medium remotely arranged relative to the processor, and these remote storage media may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0113] The processor may call the information and application program stored in the storage medium through the transmission system to execute the steps: Step 101 acquires the satellite image, sets the image feature high-dimensional mapping function, and acquires the high-dimensional image feature value of the satellite image at each pixel point;
[0114] Specifically, the image feature high-dimensional mapping function includes:
[0115] ,
[0116] in, For satellite images In pixels The high-dimensional image feature value at is the ReLU function or the Sigmoid function, is the horizontal coordinate of the pixel point, is the vertical coordinate of the pixel, is the third weight, is the second weight, is the first weight, For satellite images In pixels The color vector at For satellite images In pixels The light intensity at For satellite images Length, For satellite images The width of is the first bias term, is the second bias term, is the third bias term, For satellite images In pixels The edge strength at .
[0117] Step 102, setting a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image;
[0118] Specifically, the local feature acquisition function includes:
[0119] ,
[0120] in, For satellite images In pixels The spatial context information feature value of the neighborhood at is used to capture the pixel point The local eigenvalues of Neighborhood pixel The weight of For satellite images In the neighborhood pixels The high-dimensional image feature value at Pixel Neighborhood of is the horizontal coordinate of the neighborhood pixel, is the vertical coordinate of the neighborhood pixel.
[0121] Step 103, randomly initialize the feature center of the category, set the distance measurement function, and calculate the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value;
[0122] Specifically, the distance metric function includes:
[0123] ,
[0124] in, For satellite images In pixels The high-dimensional image eigenvalue at Characteristic center of the class The distance is the first weight of the distance, For the The center point of all pixels in the class in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the horizontal direction, For satellite images In pixels The high-dimensional image feature value in the vertical direction, For the The center point of all pixels in the class in the vertical direction, is the second weight of the distance, For the The feature center of the spatial context of the class.
[0125] Step 104, setting a category assignment function, and calculating the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the illumination intensity and the edge intensity, and dividing the pixel points corresponding to the high-dimensional image feature value into corresponding categories to complete the classification analysis.
[0126] Specifically, the category assignment function include:
[0127] ,
[0128] in, For the The color center of the class, For the The geometric feature center of the class, is the color weight, is the context weight, is the geometric weight.
[0129] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.
[0130] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0131] In the several embodiments provided by the present invention, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the system embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.
[0132] 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 on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0133] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0134] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only storage medium (ROM, Read-Only Memory), random access storage medium (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.
[0135] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.
Claims
1. A model training method for intelligence analysis, wherein: The intelligence is derived from satellite images and is characterized by including: Acquire the satellite image, set an image feature high-dimensional mapping function, and acquire a high-dimensional image feature value of the satellite image at each pixel; Set a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image; The local feature acquisition function includes: Among them, ψ(I, x, y) is the spatial context information feature value of the neighborhood of the satellite image I at the pixel point (x, y), which is used to capture the local feature value of the pixel point (x, y), and w′ ij is the weight of the neighborhood pixel point (i, j), φ(I, i, j) is the high-dimensional image feature value of satellite image I at the neighborhood pixel point (i, j), N(x, y) is the neighborhood of the pixel point (x, y), i is the horizontal coordinate of the neighborhood pixel point, and j is the vertical coordinate of the neighborhood pixel point; Randomly initialize the feature center of the category, set the distance measurement function, and calculate the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value; Setting a category assignment function, and calculating the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the illumination intensity, and the edge intensity, and dividing the pixel points corresponding to the feature values of the high-dimensional image into corresponding categories to complete the classification analysis; The category assignment function J includes: Where H is the length of satellite image I, W is the width of satellite image I, D(φ(I, x, y), θ k ) is the high-dimensional image feature value of the satellite image I at the pixel point (x, y) and the feature center θ of the kth class k The distance, I′ x，y is the color vector of the satellite image I at the pixel (x, y), v k is the feature center of the spatial context of the kth class, G x，y is the edge strength of the satellite image I at the pixel (x, y), ω k is the color center of the kth class, χ k is the geometric feature center of the kth class, β is the color weight, λ′ is the context weight, and δ is the geometric weight.
2. A model training method for intelligence analysis as claimed in claim 1, characterized in that: The image feature high-dimensional mapping function includes: Where φ(I, x, y) is the high-dimensional image feature value of the satellite image I at the pixel point (x, y), σ is the ReLU function or the Sigmoid function, x is the horizontal coordinate of the pixel point, y is the vertical coordinate of the pixel point, W3 is the third weight, W2 is the second weight, W1 is the first weight, and I′ x，y is the color vector of the satellite image I at the pixel (x, y), L x，y is the illumination intensity of satellite image I at pixel point (x, y), H is the length of satellite image I, W is the width of satellite image I, b1 is the first bias term, b2 is the second bias term, b3 is the third bias term, G x，y is the edge strength of the satellite image I at the pixel point (x, y).
3. A model training method for intelligence analysis as claimed in claim 2, characterized in that: The distance metric function includes: Among them, D(φ(I, x, y), θ k ) is the high-dimensional image feature value of the satellite image I at the pixel point (x, y) and the feature center θ of the kth class k The distance, γ is the first weight of the distance, μ x，k is the center point of all pixels in the kth class in the horizontal direction, φ x (I, x, y) is the high-dimensional image feature value of satellite image I in the horizontal direction at pixel point (x, y), φ y (I, x, y) is the high-dimensional image feature value of satellite image I in the vertical direction at pixel point (x, y), μ y，k is the center point of all pixels in the kth class in the vertical direction, λ is the second weight of the distance, v k is the feature center of the spatial context of the kth class.
4. A model training system for intelligence analysis, wherein: The intelligence is derived from satellite images and is characterized by including: A feature value acquisition module is used to acquire the satellite image, set an image feature high-dimensional mapping function, and acquire a high-dimensional image feature value of the satellite image at each pixel point; A local feature acquisition module is used to set a local feature acquisition function to obtain the spatial context information feature value of the neighborhood of each pixel of the satellite image; The local feature acquisition function includes: Among them, ψ(I, x, y) is the spatial context information feature value of the neighborhood of the satellite image I at the pixel point (x, y), which is used to capture the local feature value of the pixel point (x, y), and w′ ij is the weight of the neighborhood pixel point (i, j), φ(I, i, j) is the high-dimensional image feature value of satellite image I at the neighborhood pixel point (i, j), N(x, y) is the neighborhood of the pixel point (x, y), i is the horizontal coordinate of the neighborhood pixel point, and j is the vertical coordinate of the neighborhood pixel point; A distance calculation module, used for randomly initializing the feature center of a category, setting a distance measurement function, and calculating the distance between the high-dimensional image feature value and the feature center according to the high-dimensional image feature value and the spatial context information feature value; A classification module is used to set a category assignment function, and calculate the final distance to each category based on the distance, the color vector of the satellite image at each pixel, the light intensity and the edge intensity, and divide the pixel points corresponding to the feature value of the high-dimensional image into corresponding categories to complete the classification analysis; The category assignment function J includes: Where H is the length of satellite image I, W is the width of satellite image I, D(φ(I, x, y), θ k ) is the high-dimensional image feature value of the satellite image I at the pixel point (x, y) and the feature center θ of the kth class k The distance, I′ x，y is the color vector of the satellite image I at the pixel (x, y), v k is the feature center of the spatial context of the kth class, G x，y is the edge strength of the satellite image I at the pixel (x, y), ω k is the color center of the kth class, χ k is the geometric feature center of the kth class, β is the color weight, λ′ is the context weight, and δ is the geometric weight.
5. A model training system for intelligence analysis as claimed in claim 4, characterized in that: The image feature high-dimensional mapping function includes: Where φ(I, x, y) is the high-dimensional image feature value of the satellite image I at the pixel point (x, y), σ is the ReLU function or the Sigmoid function, x is the horizontal coordinate of the pixel point, y is the vertical coordinate of the pixel point, W3 is the third weight, W2 is the second weight, W1 is the first weight, and I′ x，y is the color vector of the satellite image I at the pixel (x, y), L x，y is the illumination intensity of satellite image I at pixel point (x, y), H is the length of satellite image I, W is the width of satellite image I, b1 is the first bias term, b2 is the second bias term, b3 is the third bias term, G x，y is the edge strength of the satellite image I at the pixel point (x, y).
6. A model training system for intelligence analysis as claimed in claim 5, characterized in that: The distance metric function includes: Among them, D(φ(I, x, y), θ k ) is the high-dimensional image feature value of the satellite image I at the pixel point (x, y) and the feature center θ of the kth class k The distance, γ is the first weight of the distance, μ x，k is the center point of all pixels in the kth class in the horizontal direction, φ x (I, x, y) is the high-dimensional image feature value of satellite image I in the horizontal direction at pixel point (x, y), φ y (I, x, y) is the high-dimensional image feature value of satellite image I in the vertical direction at pixel point (x, y), μ y，k is the center point of all pixels in the kth class in the vertical direction, λ is the second weight of the distance, v k is the feature center of the spatial context of the kth class.
Citation Information
Patent Citations
Image classification method
CN112836629A
SAR image change detection algorithm based on bilateral weighted kernel graph segmentation
CN113487578A