Remote sensing image generation method and system based on SAR and optical image fusion

Abstract

The present invention discloses a remote sensing image generation method and system based on SAR and optical image fusion, which relates to the field of image processing technology. The method includes an image monitoring module, a data acquisition module, an edge computing module and a central server. SAR images and optical images are acquired through the image monitoring module, and image data and remote sensing data are acquired through the data acquisition module. A data processing model is then constructed to perform image analysis. The edge computing module preliminarily analyzes and evaluates the quality of the source image and sequentially analyzes the image texture features and image color features. The central server then performs image fusion and feature correction and generates remote sensing images. From the perspective of SAR and optical image fusion, the advantages of the stability of SAR images under poor atmospheric conditions and the high spatial resolution of optical images are complemented, thereby improving the generalization ability of the system in different application scenarios. Real-time monitoring requirements are met through edge computing, ensuring the data accuracy and generation stability of remote sensing images.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a remote sensing image generation method and system based on SAR and optical image fusion. Background Art

[0002] Remote sensing satellites provide a wealth of remote sensing image data with varying spatial resolutions and spectra. These massive amounts of remote sensing images play a significant role in military, natural disaster monitoring, surveying and mapping, aerospace, and other fields. However, due to the varying imaging principles of different sensor types and the influence of factors such as the natural environment, a single sensor image often cannot fully capture all the information in a scene.

[0003] SAR (Synthetic Aperture Radar) is a type of remote sensing data that uses radar technology to observe the ground. Optical imaging technology uses optical devices such as cameras to capture light, convert it into electrical or digital signals, and then remotely sense and visualize.

[0004] However, existing remote sensing image generation technologies suffer from incomplete data, insufficient accuracy, and unstable image generation caused by a single sensor, as well as the resulting lack of system generalization capabilities. For example, the spatial resolution of SAR images is easily limited. In complex terrain areas such as mountainous and hilly areas, SAR imaging may exhibit shadows and overlapping images. Optical images are easily restricted by weather and lighting conditions. Image quality may be affected in cloud cover, rain, snow, at night, or in low light conditions.

[0005] In view of the above technical defects, a solution is now proposed. Summary of the Invention

[0006] The purpose of the present invention is to solve the problems of incomplete remote sensing data, insufficient accuracy and unstable image generation caused by a single sensor in remote sensing monitoring in the prior art, as well as the resulting defect of insufficient system generalization ability.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] The remote sensing image generation method based on SAR and optical image fusion includes the following steps:

[0009] S1, acquiring SAR images and optical images through an image monitoring module: the image monitoring module includes a SAR radar sensor and an optical sensor;

[0010] S2, the data acquisition module acquires image data and remote sensing data from the SAR image and the optical image, wherein the image data includes band parameters and imaging parameters, the band parameters include the SAR band frequency and the optical band wavelength, and the imaging parameters include spatial resolution and temporal resolution; remote sensing data A is then extracted from the SAR image, and remote sensing data B is extracted from the optical image, and remote sensing data B includes the chromaticity values ​​of the pixels in the optical image;

[0011] In step S3, the edge computing module performs a preliminary analysis of the image data and remote sensing data: it evaluates the source image quality through preliminary analysis of the image data, and then analyzes the image texture features and image color features using remote sensing data A and remote sensing data B in turn;

[0012] S4, the central server performs image fusion and feature correction and generates remote sensing images: image texture and color fusion is performed through remote sensing data A and remote sensing data B, and feature correction is performed based on the source image quality to obtain remote sensing feature data and perform deep fusion, thereby generating and outputting remote sensing images.

[0013] Furthermore, the image data and remote sensing data are analyzed and processed by constructing a data processing model, which includes an image quality assessment sub-model, a remote sensing data analysis sub-model, a remote sensing feature fusion sub-model, and a remote sensing image generation sub-model;

[0014] The image quality assessment sub-model preliminarily analyzes the image data and assesses the source image quality;

[0015] The remote sensing data A and remote sensing data B are classified and processed by the remote sensing data analysis sub-model to obtain the texture features of the SAR image and the color features of the optical image;

[0016] The remote sensing data A and remote sensing data B are image-fused through the remote sensing feature fusion sub-model, texture features and color features are classified and configured, and the image correction index is obtained by combining the source image quality. The image correction index is used to perform feature correction to obtain remote sensing feature data.

[0017] The remote sensing feature data is deeply integrated through the remote sensing image generation sub-model to generate and output remote sensing images.

[0018] Furthermore, the collection and labeling process of image data and remote sensing data is as follows:

[0019] Band parameters include SAR band frequency Fa and optical band wavelength Wb;

[0020] Imaging parameters include spatial resolution and temporal resolution; the spatial resolution and temporal resolution of SAR images are marked as SRa and TRa respectively; the spatial resolution and temporal resolution of optical images are marked as SRb and TRb respectively;

[0021] The remote sensing data A includes the grayscale values ​​of the pixels in the SAR image. Any pixel in the SAR image is marked as i, and the grayscale value of pixel i is marked as Gi;

[0022] The remote sensing data B includes the chromaticity values ​​of the pixels of the optical image. Any pixel of the optical image is marked as j, and the chromaticity value of the pixel j is marked as Cj.

[0023] Furthermore, the specific process of the image quality assessment sub-model is as follows:

[0024] The image quality assessment sub-model preliminarily analyzes the image data and assesses the source image quality;

[0025] S3-101, obtain the SAR image quality assessment coefficient QUsar‌ by combining the SAR band frequency Fa, the SAR image spatial resolution SRa and the temporal resolution TRa;

[0026] The source image quality of the SAR image is evaluated by setting the evaluation interval of the quality assessment coefficient QUsar and performing interval comparison;

[0027] S3-102, obtain the quality evaluation coefficient QUopt of the optical image by combining the optical band wavelength Wb, the spatial resolution SRb and the temporal resolution TRb of the optical image;

[0028] The source image quality of the optical image is evaluated by setting the evaluation interval of the quality evaluation coefficient QUopt and performing interval comparison.

[0029] Furthermore, the specific process of the remote sensing data analysis sub-model is as follows:

[0030] Classify remote sensing data A and remote sensing data B through remote sensing data analysis sub-model;

[0031] S3-201, analyze and process remote sensing data A to obtain texture features of SAR images;

[0032] The SAR image is divided into m1 local regions, the number of pixels in any local region I is marked as N1, and the gray level co-occurrence matrix is ​​constructed through the gray value Gi of N1 pixels in the local region I;

[0033] Mark any pixel pair in the gray-level co-occurrence matrix as (x, y), where x is the row number of the gray-level co-occurrence matrix and y is the column number of the gray-level co-occurrence matrix, representing the pixel pair with gray levels (x) and (y). Mark the probability of the pixel pair (x, y) as P(x, y);

[0034] The uniformity, contrast and fineness of pixel pairs are evaluated through the gray level co-occurrence matrix, and then the texture feature evaluation coefficient TEX of the SAR image is obtained;

[0035] S3-202, analyzing and processing remote sensing data B to obtain color features of the optical image;

[0036] The optical image is divided into m2 local regions, the number of pixels in any local region J is marked as N2, and the chromaticity co-occurrence matrix is ​​constructed through the chromaticity values ​​Cj of the N2 pixels in the local region J;

[0037] Mark any pixel pair in the chromatic co-occurrence matrix as (p, q), where p is the row number of the chromatic co-occurrence matrix and q is the column number of the chromatic co-occurrence matrix, representing the pixel pair with chromaticity values ​​(p) and (q), and mark the probability of the pixel pair (p, q) as P(p, q);

[0038] The richness, contrast, and saturation of pixel pairs are evaluated through the chromatic co-occurrence matrix, and the color feature evaluation coefficient COL of the optical image is obtained.

[0039] Furthermore, the specific process of the remote sensing feature fusion sub-model is as follows:

[0040] Perform image fusion of remote sensing data A and remote sensing data B through remote sensing feature fusion sub-model;

[0041] S4-101, classify and configure texture features and color features;

[0042] A risk threshold R1 of the texture feature evaluation coefficient TEX of the SAR image is set, and the texture feature status of the SAR image is evaluated by threshold comparison. When the texture feature evaluation coefficient TEX of the SAR image is lower than the risk threshold R1, the texture feature status of the SAR image is determined to be poor, and the remote sensing data A is configured and corrected;

[0043] A risk threshold R2 is set for the color feature evaluation coefficient COL of the optical image, and the color feature status of the optical image is evaluated by threshold comparison. When the color feature evaluation coefficient COL of the optical image is lower than the risk threshold R2, the color feature status of the optical image is determined to be poor, and the remote sensing data B is configured and corrected.

[0044] S4-102, obtaining an image correction index and then performing feature correction;

[0045] The remote sensing data A is configured and corrected by the SAR image quality assessment coefficient QUsar and the texture feature assessment coefficient TEX to obtain the image correction coefficient No. 1. ;

[0046] The remote sensing data B is configured and corrected through the image correction index CRopt of the optical image and the color feature evaluation coefficient COL to obtain the second image correction coefficient ;

[0047] S4-103, obtaining remote sensing feature data by modifying the feature data;

[0048] Establish a feature correction model, input the source image E and its pixel parameters Re;

[0049] Mark the total number of pixels in the source image E as Ne, mark any pixel as e, and obtain the neighboring pixels of pixel e. Preset the number of neighboring pixels of any pixel e as Nu, and mark any neighboring pixel as f;

[0050] Substitute the source image E into the spatial coordinate system and obtain the coordinates of each pixel;

[0051] Obtain the spatial domain weight ws(e, f) between pixel e and pixel f through the distance between the coordinate De of pixel e and the coordinate Df of neighboring pixel f;

[0052] Then, the parameter domain weight wr(e, f) between pixel e and pixel f is obtained by the difference between the parameter Re of pixel e and the parameter value Rf of the neighboring pixel f.

[0053] The comprehensive weight w(e, f) of pixel e is obtained by combining the spatial domain weight ws(e, f) between pixel e and Nu neighboring pixels f and the parameter domain weight wr(e, f);

[0054] Set the standard interval Qw of the comprehensive weight w(e,f). When the comprehensive weight w(e,f) of pixel e is within the standard interval Qw, the pixel e is considered normal and no processing is performed on the pixel e. When the comprehensive weight w(e,f) of pixel e is lower or higher than the standard interval Qw, the pixel e is considered abnormal and the parameter Re of the pixel e is corrected using the image correction coefficient.

[0055] The feature correction model outputs the corrected source image E and marks it as the feature image H.

[0056] Furthermore, the specific process of the remote sensing image generation sub-model is as follows:

[0057] Input the SAR image into the feature correction model to obtain the abnormal pixels of the SAR image and perform grayscale correction. The image correction coefficient Select the No. 1 image correction coefficient based on actual conditions , output the corrected SAR image and mark it as the SAR feature image;

[0058] Input the optical image into the feature correction model to obtain the abnormal pixels of the optical image and perform chromaticity correction and image correction coefficient Select the No. 2 image correction coefficient based on actual conditions , output the corrected optical image and mark it as an optical feature image;

[0059] Combine SAR characteristic images with optical characteristic images to comprehensively generate and output remote sensing images;

[0060] The grayscale data of the SAR feature image and the colorimetric data of the optical feature image are integrated and marked as remote sensing feature data. The remote sensing feature data includes the feature parameter matrix of the remote sensing image. The feature parameter matrix includes the remote sensing feature data vectors of all pixel points. , where GiR is the corrected remote sensing grayscale parameter and CjR is the corrected remote sensing chromaticity parameter.

[0061] A remote sensing image generation system based on SAR and optical image fusion includes an image monitoring module, a data acquisition module, an edge computing module and a central server. The image monitoring module, the data acquisition module, the edge computing module and the central server are communicatively connected, wherein the edge computing module includes an image analysis submodule and a remote sensing analysis submodule, and the central server includes a feature correction submodule and a depth analysis submodule. The system applies the above-mentioned remote sensing image generation method based on SAR and optical image fusion.

[0062] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0063] The present invention obtains SAR images and optical images through an image monitoring module, and obtains image data and remote sensing data through a data acquisition module, then constructs a data processing model for image analysis, preliminarily analyzes and evaluates the source image quality through an edge computing module, and sequentially analyzes the image texture features and image color features, and then performs image fusion and feature correction through a central server to generate remote sensing images. From the perspective of SAR and optical image fusion, the stability of SAR images under poor atmospheric conditions and the high spatial resolution of optical images are complemented by each other, thereby improving the generalization ability of the system in different application scenarios, meeting real-time monitoring requirements through edge computing, and ensuring the data accuracy and generation stability of remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A schematic diagram showing the steps of the method flow of the present invention is shown;

[0065] Figure 2 A schematic flow chart of a data processing model of the present invention is shown;

[0066] Figure 3 A connection diagram of the system modules of the present invention is shown. DETAILED DESCRIPTION

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0068] Example 1:

[0069] like Figure 1-3 As shown in FIG, the remote sensing image generation method based on SAR and optical image fusion includes the following steps:

[0070] S1, acquiring SAR images and optical images through an image monitoring module: the image monitoring module includes a SAR radar sensor and an optical sensor;

[0071] The SAR radar sensor transmits electromagnetic waves and receives echoes to record the SAR image reflected by the surface; optical sensors that use visible light, infrared and other electromagnetic waves to obtain optical images;

[0072] SAR (Synthetic Aperture Radar) images: SAR uses the principles of microwave signal emission and reflection to form high-resolution images through synthetic aperture technology. SAR can capture images not only on sunny days, but also at night or in adverse weather conditions such as cloud cover and haze. However, in complex terrain areas such as mountainous and hilly areas, SAR imaging quality may be affected, resulting in shadows and overlapping images, reducing image accuracy and readability.

[0073] Optical imaging: Utilizes electromagnetic waves such as visible light and infrared rays to form images. Optical imaging relies on favorable atmospheric conditions and is therefore limited in environments such as clouds, rain, and snow.

[0074] S2, the data acquisition module acquires image data and remote sensing data from the SAR image and the optical image, wherein the image data includes band parameters and imaging parameters, the band parameters include the SAR band frequency and the optical band wavelength, and the imaging parameters include spatial resolution and temporal resolution; remote sensing data A is then extracted from the SAR image, and remote sensing data B is extracted from the optical image, and remote sensing data B includes the chromaticity values ​​of the pixels in the optical image;

[0075] The collection and marking process of image data and remote sensing data is as follows:

[0076] Band parameters include SAR band frequency Fa and optical band wavelength Wb;

[0077] Imaging parameters include spatial resolution and temporal resolution; the spatial resolution and temporal resolution of SAR images are marked as SRa and TRa respectively; the spatial resolution and temporal resolution of optical images are marked as SRb and TRb respectively;

[0078] The remote sensing data A includes the grayscale values ​​of the pixels in the SAR image. Any pixel in the SAR image is marked as i, and the grayscale value of pixel i is marked as Gi;

[0079] The remote sensing data B includes the chromaticity values ​​of the pixels of the optical image. Any pixel of the optical image is marked as j, and the chromaticity value of pixel j is marked as Cj.

[0080] Grayscale value Gi: Grayscale refers to the depth of an image or color. It is used to represent the brightness level of an image. It is usually used to represent transition colors from black to white. It is calculated by measuring the equality of RGB values ​​or by using existing image processing tools to identify the grayscale of pixels.

[0081] Chromaticity value Cj: Chromaticity refers to the basic properties of color, excluding brightness, and reflects the hue and saturation of the color. The color matching function is calculated by measuring the hue and saturation of the RGB color, or the chromaticity of the pixel can be identified using existing image processing tools.

[0082] Image data and remote sensing data are analyzed and processed by building a data processing model, which includes an image quality assessment sub-model, a remote sensing data analysis sub-model, a remote sensing feature fusion sub-model, and a remote sensing image generation sub-model;

[0083] In step S3, the edge computing module performs a preliminary analysis of the image data and remote sensing data: it evaluates the source image quality through preliminary analysis of the image data, and then analyzes the image texture features and image color features using remote sensing data A and remote sensing data B in turn;

[0084] S3-1, preliminarily analyzing image data through the image quality assessment sub-model to assess the source image quality;

[0085] The specific process of the image quality assessment sub-model is as follows:

[0086] The image quality assessment sub-model preliminarily analyzes the image data and assesses the source image quality;

[0087] In step S3-101, the quality assessment coefficient QUsar of the SAR image is obtained by combining the SAR band frequency Fa, the spatial resolution SRa, and the temporal resolution TRa of the SAR image:

[0088] ;

[0089] in, 、 and are the weight coefficients of SAR band frequency Fa, spatial resolution SRa and temporal resolution TRa of SAR image, respectively, and 、 and are all greater than 0. When the SAR band frequency Fa, the spatial resolution SRa, and the temporal resolution TRa of the SAR image are higher, the SAR image quality assessment coefficient QUsar is higher, and the source image quality of the SAR image is better. A higher carrier frequency means that the SAR image has a higher resolution and can provide clear image details. High resolution is crucial for identifying small targets, conducting refined management, and improving target recognition accuracy.

[0090] The source image quality of the SAR image is evaluated by setting the evaluation interval of the quality assessment coefficient QUsar and performing interval comparison;

[0091] S3-102, by combining the optical band wavelength Wb, the spatial resolution SRb and the temporal resolution TRb of the optical image, obtain the quality assessment coefficient QUopt of the optical image:

[0092] ;

[0093] in, 、 and are the weight coefficients of the optical band wavelength Wb, the spatial resolution SRb and the temporal resolution TRb of the optical image, respectively, and 、 and are all greater than 0; when the optical band wavelength Wb, the spatial resolution SRb and the temporal resolution TRb of the optical image are higher, the optical image quality evaluation coefficient QUopt is higher, and the source image quality of the evaluated light image is better; the longer the wavelength, the stronger the penetration ability, and the low-frequency band has stronger penetration ability, which is suitable for penetrating vegetation, soil or observing underground targets;‌

[0094] The source image quality of the optical image is evaluated by setting the evaluation interval of the quality evaluation coefficient QUopt and performing interval comparison;

[0095] S3-2, classifying and processing remote sensing data A and remote sensing data B through the remote sensing data analysis sub-model, thereby obtaining texture features of the SAR image and color features of the optical image;

[0096] The specific process of the remote sensing data analysis sub-model is as follows:

[0097] Classify remote sensing data A and remote sensing data B through remote sensing data analysis sub-model;

[0098] S3-201, analyze and process remote sensing data A to obtain texture features of SAR images;

[0099] The SAR image is divided into m1 local regions, the number of pixels in any local region I is marked as N1, and the gray level co-occurrence matrix is ​​constructed through the gray value Gi of N1 pixels in the local region I;

[0100] Mark any pixel pair in the gray-level co-occurrence matrix as (x, y), where x is the row number of the gray-level co-occurrence matrix and y is the column number of the gray-level co-occurrence matrix, representing the pixel pair with gray levels (x) and (y). Mark the probability of the pixel pair (x, y) as P(x, y);

[0101] The texture feature evaluation coefficient TEX of the SAR image is obtained through the gray level co-occurrence matrix:

[0102]

[0103] Among them, n1 is the number of pixel pairs in the gray-level co-occurrence matrix, and the uniformity of n1 pixel pairs in the gray-level co-occurrence matrix is ​​obtained by traversing , contrast and fineness , thereby comprehensively evaluating the texture feature state of the SAR image. When the uniformity, contrast and fineness of the pixel pairs are higher, the texture feature evaluation coefficient TEX of the SAR image is higher, and the texture feature state of the SAR image is better.

[0104] S3-202, analyzing and processing remote sensing data B to obtain color features of the optical image;

[0105] The optical image is divided into m2 local regions, the number of pixels in any local region J is marked as N2, and the chromaticity co-occurrence matrix is ​​constructed through the chromaticity values ​​Cj of the N2 pixels in the local region J;

[0106] Mark any pixel pair in the chromatic co-occurrence matrix as (p, q), where p is the row number of the chromatic co-occurrence matrix and q is the column number of the chromatic co-occurrence matrix, representing the pixel pair with chromaticity values ​​(p) and (q), and mark the probability of the pixel pair (p, q) as P(p, q);

[0107] The color feature evaluation coefficient COL of the optical image is obtained through the chromatic co-occurrence matrix:

[0108]

[0109] Among them, n2 is the number of pixel pairs in the gray-level co-occurrence matrix, Cave refers to the average value of the chromaticity value Cj of N2 pixels in the local area J, and the richness of n2 pixel pairs in the gray-level co-occurrence matrix is ​​obtained by traversing , contrast , saturation , thereby comprehensively evaluating the color feature state of the optical image. When the richness, contrast, and saturation of the pixel pairs are higher, the color feature evaluation coefficient COL of the optical image is higher, and the color feature state of the optical image is better evaluated;

[0110] S4, the central server performs image fusion and feature correction to generate remote sensing images: image texture and color fusion is performed through remote sensing data A and remote sensing data B, and feature correction is performed based on the quality of the source images to obtain remote sensing feature data and perform deep fusion to generate and output remote sensing images;

[0111] S4-1, remote sensing data A and remote sensing data B are image-fused using the remote sensing feature fusion sub-model, texture features and color features are classified and configured, and image correction indexes are obtained based on the source image quality. Feature correction is performed using the image correction indexes to obtain remote sensing feature data;

[0112] The specific process of the remote sensing feature fusion sub-model is as follows:

[0113] Perform image fusion of remote sensing data A and remote sensing data B through remote sensing feature fusion sub-model;

[0114] S4-101, classify and configure texture features and color features;

[0115] A risk threshold R1 of the texture feature evaluation coefficient TEX of the SAR image is set, and the texture feature status of the SAR image is evaluated by threshold comparison. When the texture feature evaluation coefficient TEX of the SAR image is lower than the risk threshold R1, the texture feature status of the SAR image is determined to be poor, and the remote sensing data A is configured and corrected;

[0116] A risk threshold R2 is set for the color feature evaluation coefficient COL of the optical image, and the color feature status of the optical image is evaluated by threshold comparison. When the color feature evaluation coefficient COL of the optical image is lower than the risk threshold R2, the color feature status of the optical image is determined to be poor, and the remote sensing data B is configured and corrected.

[0117] S4-102, obtaining an image correction index and then performing feature correction;

[0118] The remote sensing data A is configured and corrected by the SAR image quality assessment coefficient QUsar and the texture feature assessment coefficient TEX to obtain the image correction coefficient No. 1. : ,in, is the conversion coefficient of SAR image, It refers to combining the quality assessment coefficient QUsar with the texture feature assessment coefficient TEX, and then converting it into the No. 1 image correction coefficient The preset constant of The preset value is in the interval (0,1). When the quality evaluation coefficient QUsar and texture feature evaluation coefficient TEX of the SAR image are higher, the image correction coefficient The lower;

[0119] The remote sensing data B is configured and corrected through the image correction index CRopt of the optical image and the color feature evaluation coefficient COL to obtain the second image correction coefficient : ,in, is the conversion coefficient of the optical image, the conversion coefficient It refers to combining the image correction index CRopt with the color feature evaluation coefficient COL and then converting it into the second image correction coefficient The preset constant of The preset value is in the interval (0,1). When the image correction index CRopt and the color feature evaluation coefficient COL of the optical image are higher, the second image correction coefficient The lower;

[0120] S4-103, obtaining remote sensing feature data by modifying the feature data;

[0121] Establish a feature correction model, input the source image E and its pixel parameters Re;

[0122] Mark the total number of pixels in the source image E as Ne, mark any pixel as e, and obtain the neighboring pixels of pixel e. Preset the number of neighboring pixels of any pixel e as Nu, and mark any neighboring pixel as f;

[0123] Substitute the source image E into the spatial coordinate system and obtain the coordinates of each pixel;

[0124] The spatial domain weight ws(e, f) between pixel e and pixel f is obtained by the distance between the coordinate De of pixel e and the coordinate Df of the neighboring pixel f: ;in, It refers to the Euclidean distance between the coordinate De of the pixel point e and the coordinate Df of the neighboring pixel point f; when the distance The higher it is, the higher the spatial domain weight ws(e, f) between pixel e and pixel f is;

[0125] Then, the parameter domain weight wr(e, f) between pixel e and pixel f is obtained by the difference between the parameter Re of pixel e and the parameter value Rf of the neighboring pixel f: ;in, Refers to the difference between the parameter Re of pixel e and the parameter value Rf of neighboring pixel f; when the difference The higher it is, the higher the parameter domain weight wr(e, f) between pixel e and pixel f is;

[0126] The comprehensive weight w(e, f) of pixel e is obtained by combining the spatial domain weight ws(e, f) between pixel e and Nu neighboring pixels f and the parameter domain weight wr(e, f):

[0127] ;

[0128] Set the standard interval Qw of the comprehensive weight w(e,f). When the comprehensive weight w(e,f) of pixel e is within the standard interval Qw, the pixel e is considered normal and no processing is performed on the pixel e. When the comprehensive weight w(e,f) of pixel e is lower or higher than the standard interval Qw, the pixel e is considered abnormal, indicating that the overall difference between pixel e and its neighboring pixels is too small or too large, and the parameter Re of pixel e is corrected.

[0129] The specific correction process is to compare the depth of pixel e with the neighboring pixels:

[0130] like If it is greater than 0, it means that the parameter Re of pixel e is too large relative to that of its neighboring pixels. The parameter Re of pixel e is reduced, and the corrected parameter of pixel e is marked as RCe-: ,in, is the image correction coefficient;

[0131] like If it is less than 0, it means that the parameter Re of pixel e is smaller than that of its neighboring pixels. The parameter Re of pixel e is increased, and the corrected parameter of pixel e is marked as RCe+: ,in, is the image correction coefficient;

[0132] The feature correction model outputs the corrected source image E and marks it as the feature image H;

[0133] S4-2, deep fusion of remote sensing feature data through remote sensing image generation sub-model, thereby generating and outputting remote sensing images;

[0134] The specific process of the remote sensing image generation sub-model is as follows:

[0135] Input the SAR image into the feature correction model to obtain the abnormal pixels of the SAR image and perform grayscale correction. The image correction coefficient Select the No. 1 image correction coefficient based on actual conditions , output the corrected SAR image and mark it as the SAR feature image;

[0136] Input the optical image into the feature correction model to obtain the abnormal pixels of the optical image and perform chromaticity correction and image correction coefficient Select the No. 2 image correction coefficient based on actual conditions , output the corrected optical image and mark it as an optical feature image;

[0137] Combine SAR characteristic images with optical characteristic images to comprehensively generate and output remote sensing images;

[0138] The grayscale data of the SAR feature image and the colorimetric data of the optical feature image are integrated and marked as remote sensing feature data. The remote sensing feature data includes the feature parameter matrix of the remote sensing image. The feature parameter matrix includes the remote sensing feature data vectors of all pixel points. , where GiR is the corrected remote sensing grayscale parameter and CjR is the corrected remote sensing chromaticity parameter.

[0139] A remote sensing image generation system based on SAR and optical image fusion includes an image monitoring module, a data acquisition module, an edge computing module, and a central server. The image monitoring module, the data acquisition module, the edge computing module, and the central server are communicatively connected. The edge computing module includes an image analysis submodule and a remote sensing analysis submodule. The central server includes a feature correction submodule and a depth analysis submodule. The system applies the above-mentioned remote sensing image generation method based on SAR and optical image fusion.

[0140] The image monitoring module acquires SAR images and optical images through SAR radar sensors and optical sensors;

[0141] The data acquisition module acquires image data and remote sensing data through SAR images and optical images;

[0142] The edge computing module performs preliminary analysis on image data and remote sensing data;

[0143] The central server performs image fusion and feature correction and generates remote sensing images;

[0144] The system analyzes and processes image data and remote sensing data by constructing a data processing model. The data processing model includes an image quality assessment sub-model, a remote sensing data analysis sub-model, a remote sensing feature fusion sub-model and a remote sensing picture generation sub-model. Among them: the image analysis sub-module applies the image quality assessment sub-model; the remote sensing analysis sub-module applies the remote sensing data analysis sub-model; the feature correction sub-module applies the remote sensing feature fusion sub-model; and the depth analysis sub-module applies the remote sensing picture generation sub-model.

[0145] In summary, the present invention obtains SAR images and optical images through the image monitoring module, obtains image data and remote sensing data through the data acquisition module, and then constructs a data processing model for analysis. The edge computing module preliminarily analyzes and evaluates the source image quality and analyzes the image texture features and image color features in turn. The central server then performs image fusion and feature correction and generates remote sensing pictures. From the perspective of SAR and optical image fusion, the stability of SAR images under poor atmospheric conditions and the high spatial resolution of optical images complement each other, improve the generalization ability of the model in different application scenarios, meet real-time monitoring needs through edge computing, and ensure the data accuracy and generation stability of remote sensing pictures.

[0146] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technical personnel in this field for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0147] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by those skilled in the art according to actual conditions.

[0148] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A remote sensing image generation method based on SAR and optical image fusion, characterized by: The following steps are involved: S1, acquiring SAR images and optical images through an image monitoring module: the image monitoring module includes a SAR radar sensor and an optical sensor; S2, the data acquisition module acquires image data and remote sensing data from the SAR image and the optical image, wherein the image data includes band parameters and imaging parameters, the band parameters include the SAR band frequency and the optical band wavelength, and the imaging parameters include spatial resolution and temporal resolution; remote sensing data A is then extracted from the SAR image, and remote sensing data B is extracted from the optical image, and remote sensing data B includes the chromaticity values ​​of the pixels in the optical image; In step S3, the edge computing module performs a preliminary analysis of the image data and remote sensing data: it evaluates the source image quality through preliminary analysis of the image data, and then analyzes the image texture features and image color features using remote sensing data A and remote sensing data B in turn; S4, the central server performs image fusion and feature correction to generate remote sensing images: image texture and color fusion is performed through remote sensing data A and remote sensing data B, and feature correction is performed based on the quality of the source images to obtain remote sensing feature data and perform deep fusion to generate and output remote sensing images; Image data and remote sensing data are analyzed and processed by building a data processing model, which includes a remote sensing feature fusion sub-model and a remote sensing image generation sub-model; The remote sensing data A and remote sensing data B are image-fused through the remote sensing feature fusion sub-model, texture features and color features are classified and configured, and the image correction index is obtained by combining the source image quality. The image correction index is used to perform feature correction to obtain remote sensing feature data. The remote sensing feature data is deeply integrated through the remote sensing image generation sub-model to generate and output remote sensing images; The specific process of the remote sensing feature fusion sub-model is as follows: Perform image fusion of remote sensing data A and remote sensing data B through remote sensing feature fusion sub-model; S4-101, classify and configure texture features and color features; A risk threshold R1 of the texture feature evaluation coefficient TEX of the SAR image is set, and the texture feature status of the SAR image is evaluated by threshold comparison. When the texture feature evaluation coefficient TEX of the SAR image is lower than the risk threshold R1, the texture feature status of the SAR image is determined to be poor, and the remote sensing data A is configured and corrected; A risk threshold R2 is set for the color feature evaluation coefficient COL of the optical image, and the color feature status of the optical image is evaluated by threshold comparison. When the color feature evaluation coefficient COL of the optical image is lower than the risk threshold R2, the color feature status of the optical image is determined to be poor, and the remote sensing data B is configured and corrected. S4-102, obtaining an image correction index and then performing feature correction; The remote sensing data A is configured and corrected by the SAR image quality assessment coefficient QUsar and the texture feature assessment coefficient TEX to obtain the image correction coefficient No.

1. ; The remote sensing data B is configured and corrected through the image correction index CRopt of the optical image and the color feature evaluation coefficient COL to obtain the second image correction coefficient ; S4-103, obtaining remote sensing feature data by modifying the feature data; Establish a feature correction model, input the source image E and its pixel parameters Re; Mark the total number of pixels in the source image E as Ne, mark any pixel as e, and obtain the neighboring pixels of pixel e. Preset the number of neighboring pixels of any pixel e as Nu, and mark any neighboring pixel as f; Substitute the source image E into the spatial coordinate system and obtain the coordinates of each pixel; Obtain the spatial domain weight ws(e, f) between pixel e and pixel f through the distance between the coordinate De of pixel e and the coordinate Df of neighboring pixel f; Then, the parameter domain weight wr(e, f) between pixel e and pixel f is obtained by the difference between the parameter Re of pixel e and the parameter value Rf of the neighboring pixel f. The comprehensive weight w(e, f) of pixel e is obtained by combining the spatial domain weight ws(e, f) between pixel e and Nu neighboring pixels f and the parameter domain weight wr(e, f); Set the standard interval Qw of the comprehensive weight w(e,f). When the comprehensive weight w(e,f) of pixel e is within the standard interval Qw, the pixel e is considered normal and no processing is performed on the pixel e. When the comprehensive weight w(e,f) of pixel e is lower or higher than the standard interval Qw, the pixel e is considered abnormal and the parameter Re of the pixel e is corrected using the image correction coefficient. The feature correction model outputs the corrected source image E and marks it as the feature image H; The specific process of the remote sensing image generation sub-model is as follows: Input the SAR image into the feature correction model to obtain the abnormal pixels of the SAR image and perform grayscale correction. The image correction coefficient Select the image correction coefficient No. 1 according to the actual situation , output the corrected SAR image and mark it as the SAR feature image; The optical image is input into the feature correction model to obtain the abnormal pixels of the optical image and perform chromaticity correction. The image correction coefficient is The second image correction coefficient is selected based on the actual situation , output the corrected optical image and mark it as an optical feature image; Combine SAR characteristic images with optical characteristic images to comprehensively generate and output remote sensing images; The grayscale data of the SAR feature image and the colorimetric data of the optical feature image are integrated and marked as remote sensing feature data. The remote sensing feature data includes the feature parameter matrix of the remote sensing image. The feature parameter matrix includes the remote sensing feature data vectors of all pixel points. , where GiR is the corrected remote sensing grayscale parameter and CjR is the corrected remote sensing chromaticity parameter.

2. The remote sensing image generation method based on SAR and optical image fusion according to claim 1, characterized in that: The data processing model also includes an image quality assessment sub-model and a remote sensing data analysis sub-model; The image quality assessment sub-model preliminarily analyzes the image data and assesses the source image quality; The remote sensing data A and remote sensing data B are classified and processed by the remote sensing data analysis sub-model to obtain the texture features of the SAR image and the color features of the optical image.

3. The remote sensing image generation method based on SAR and optical image fusion according to claim 2, characterized in that: The collection and marking process of image data and remote sensing data is as follows: Band parameters include SAR band frequency Fa and optical band wavelength Wb; Imaging parameters include spatial resolution and temporal resolution; the spatial resolution and temporal resolution of SAR images are marked as SRa and TRa respectively; the spatial resolution and temporal resolution of optical images are marked as SRb and TRb respectively; The remote sensing data A includes the grayscale values ​​of the pixels in the SAR image. Any pixel in the SAR image is marked as i, and the grayscale value of pixel i is marked as Gi; The remote sensing data B includes the chromaticity values ​​of the pixels of the optical image. Any pixel of the optical image is marked as j, and the chromaticity value of the pixel j is marked as Cj.

4. The remote sensing image generation method based on SAR and optical image fusion according to claim 3, characterized in that: The specific process of the image quality assessment sub-model is as follows: The image quality assessment sub-model preliminarily analyzes the image data and assesses the source image quality; S3-101, obtain the SAR image quality assessment coefficient QUsar‌ by combining the SAR band frequency Fa, the SAR image spatial resolution SRa and the temporal resolution TRa; The source image quality of the SAR image is evaluated by setting the evaluation interval of the quality assessment coefficient QUsar and performing interval comparison; S3-102, obtain the quality evaluation coefficient QUopt of the optical image by combining the optical band wavelength Wb, the spatial resolution SRb and the temporal resolution TRb of the optical image; The source image quality of the optical image is evaluated by setting the evaluation interval of the quality evaluation coefficient QUopt and performing interval comparison.

5. The remote sensing image generation method based on SAR and optical image fusion according to claim 4, characterized in that: The specific process of the remote sensing data analysis sub-model is as follows: Classify remote sensing data A and remote sensing data B through remote sensing data analysis sub-model; S3-201, analyze and process remote sensing data A to obtain texture features of SAR images; The SAR image is divided into m1 local regions, the number of pixels in any local region I is marked as N1, and the gray level co-occurrence matrix is ​​constructed through the gray value Gi of N1 pixels in the local region I; Mark any pixel pair in the gray-level co-occurrence matrix as (x, y), where x is the row number of the gray-level co-occurrence matrix and y is the column number of the gray-level co-occurrence matrix, representing the pixel pair with gray levels (x) and (y). Mark the probability of the pixel pair (x, y) as P(x, y); The uniformity, contrast and fineness of pixel pairs are evaluated through the gray level co-occurrence matrix, and then the texture feature evaluation coefficient TEX of the SAR image is obtained; S3-202, analyzing and processing remote sensing data B to obtain color features of the optical image; The optical image is divided into m2 local regions, the number of pixels in any local region J is marked as N2, and the chromaticity co-occurrence matrix is ​​constructed through the chromaticity values ​​Cj of the N2 pixels in the local region J; Mark any pixel pair in the chromatic co-occurrence matrix as (p, q), where p is the row number of the chromatic co-occurrence matrix and q is the column number of the chromatic co-occurrence matrix, representing the pixel pair with chromaticity values ​​(p) and (q), and mark the probability of the pixel pair (p, q) as P(p, q); The richness, contrast, and saturation of pixel pairs are evaluated through the chromatic co-occurrence matrix, and the color feature evaluation coefficient COL of the optical image is obtained.

6. A remote sensing image generation system based on SAR and optical image fusion, characterized by: The system includes an image monitoring module, a data acquisition module, an edge computing module and a central server, which are communicatively connected to each other. The edge computing module includes an image analysis submodule and a remote sensing analysis submodule, and the central server includes a feature correction submodule and a depth analysis submodule. The system applies the remote sensing image generation method based on SAR and optical image fusion as described in any one of claims 1 to 5 above.

Images