Multispectral remote sensing image data preprocessing method and system
Through multispectral remote sensing image data preprocessing methods, including noise reduction, geometric correction and atmospheric correction, the noise and distortion problems in image processing are solved, efficient and accurate image preprocessing is achieved, and image quality and registration accuracy are improved.
Patent Information
- Application Number
- CN202411569877.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2024-12-03
AI Technical Summary
There are problems in the multi-spectral remote sensing image processing that the data volume is large and complex, the noise processing is not accurate enough, which affects the image quality and subsequent analysis accuracy. In the prior art, it is difficult to retain image details during denoising, and the image registration accuracy is low.
Using multi-spectral remote sensing image data preprocessing methods, including noise reduction processing, geometric correction, radiation calibration and atmospheric correction, high-precision geometric correction algorithm eliminates image distortion, combined with intelligent image registration technology, automatically identify feature points for precise matching Accurate and get preprocessed images by cropping vector boundaries.
Effectively remove noise, improve image quality, ensure accurate position of land objects, significantly improve image registration accuracy and efficiency, reduce manual intervention, and provide higher quality image data support.
Smart Images

Figure CN119067878A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of image preprocessing, and in particular relates to a multi-spectral remote sensing image data preprocessing method and system. Background Art
[0002] With the development of remote sensing technology and the widespread application of hyperspectral remote sensing data, efficient processing and accurate analysis of multispectral images have become increasingly important. Multispectral cameras can capture the reflection or radiation information of target objects in different spectral bands. These spectral information contains rich features and attributes of objects, which is of great significance for environmental monitoring, agricultural assessment, urban planning and other fields.
[0003] At present, multispectral remote sensing image processing faces two major challenges: first, the data volume is large and complex, requiring efficient preprocessing methods to improve data quality; second, the spectral feature extraction in multispectral images is not accurate enough, making it difficult to fully explore the correlation between spectral segments.
[0004] Since the original remote sensing images are often affected by a variety of factors, the image quality is reduced, which directly affects the accuracy and effect of subsequent image analysis and application. These influencing factors mainly include atmospheric scattering and absorption, sensor noise, changes in lighting conditions, geometric distortion, etc. Atmospheric scattering and absorption will change the intensity of the spectral signal, sensor noise will increase the random error of the image, changes in lighting conditions will affect the brightness and contrast of the image, and geometric distortion will cause inaccurate spatial position of the image. In order to eliminate the impact of these factors on the image quality of multispectral cameras and improve the accuracy and efficiency of image processing, remote sensing data preprocessing has become an indispensable link.
[0005] Common preprocessing methods include noise filtering, image enhancement, and spatial domain to frequency domain conversion. Noise filtering methods can use mean filtering, median filtering, etc. to eliminate noise interference in the image. Image enhancement methods can adjust the contrast and brightness of the image to make it more suitable for subsequent processing and analysis. Spatial domain to frequency domain conversion methods can use Fourier transform and other techniques to convert the image from the spatial domain to the frequency domain, filter and enhance in the frequency domain, and then convert back to the spatial domain.
[0006] However, when the above-mentioned existing technologies are directly applied to the remote sensing field for noise reduction processing, they often cannot retain image details well while removing noise, affecting the subsequent processing quality of the image after noise reduction; at the same time, manual intervention is still required in the automatic identification and processing of noise, missing values, and duplicate values in the data, which increases the workload and the possibility of errors.
[0007] In addition, when performing image registration in the remote sensing field, the method of selecting feature points is often not universally applicable, and there is a lack of more detailed geometric correction methods for remote sensing images. It is impossible to provide an accurate and reliable image basis for subsequent image processing. There are certain deficiencies in the precise registration of images, and the registration accuracy is low. Summary of the invention
[0008] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a multispectral remote sensing image data preprocessing method and system, which can efficiently remove noise in multispectral camera images, and utilize a high-precision geometric correction algorithm to eliminate multispectral image distortion caused by factors such as sensor viewing angle and earth curvature. Accurate geometric correction can ensure that the position of ground objects in the image is accurate, providing a reliable foundation for subsequent image registration, fusion and other processing steps.
[0009] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: A first aspect of the present invention provides a multispectral remote sensing image data preprocessing method.
[0010] The multispectral remote sensing image data preprocessing method comprises the following steps: Acquire multispectral remote sensing images and perform noise reduction processing to obtain images to be corrected; Determine the reference image, select corresponding ground control points from the image to be corrected and the reference image, perform geometric correction on the image to be corrected based on the selected ground control points, perform image resampling on the geometrically corrected image, and complete image registration; Perform radiometric calibration and atmospheric correction on the registered images; Obtain a full-color image, and fuse the image after radiometric calibration and atmospheric correction with the full-color image to obtain a fused image; The fused image is cropped using the vector boundary to obtain the preprocessed multispectral remote sensing image.
[0011] A second aspect of the present invention provides a multispectral remote sensing image data preprocessing system.
[0012] Multispectral remote sensing image data preprocessing system, including: The noise reduction module is configured to: acquire a multispectral remote sensing image and perform noise reduction processing to obtain an image to be corrected; The geometric correction module is configured to: determine a reference image, select corresponding ground control points from the image to be corrected and the reference image, perform geometric correction on the image to be corrected based on the selected ground control points, perform image resampling on the geometrically corrected image, and complete image registration; The atmospheric correction module is configured to: perform radiometric calibration and atmospheric correction on the registered image; The fusion module is configured to: acquire a full-color image, fuse the image after radiometric calibration and atmospheric correction with the full-color image, and obtain a fused image; The cropping module is configured to: crop the fused image using the vector boundary to obtain a preprocessed multispectral remote sensing image. One or more of the above technical solutions have the following beneficial effects: (1) The present invention provides a multispectral remote sensing image data preprocessing method and system. The noise reduction process uses advanced denoising algorithms to efficiently remove noise in multispectral camera images. These algorithms can effectively eliminate image noise caused by factors such as weather changes and sensor noise, and improve image quality. Compared with traditional denoising methods, these algorithms can more thoroughly remove noise while maintaining image details, providing higher quality image data for subsequent processing.
[0013] (2) The present invention uses a high-precision geometric correction algorithm to eliminate multispectral image distortion caused by factors such as sensor viewing angle and earth curvature. Accurate geometric correction can ensure the accurate position of ground objects in the image, providing a reliable basis for subsequent image registration, fusion and other processing steps.
[0014] (3) The image registration of the present invention adopts intelligent image registration technology based on feature matching, which can automatically identify and match feature points (such as boundaries, intersections of linear objects, regional contours, etc.) in multispectral camera images to achieve accurate image registration. Intelligent image registration can significantly improve registration accuracy and efficiency, reduce manual intervention, and reduce processing costs.
[0015] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0017] Figure 1 This is a flow chart of the method of the first embodiment.
[0018] Figure 2 It is a system structure diagram of the second embodiment. DETAILED DESCRIPTION
[0019] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0020] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.
[0021] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0022] Embodiment 1 The present invention proposes a remote sensing data preprocessing method, which includes several links, such as noise elimination, geometric correction (geographic positioning, geometric precision correction, image registration, orthorectification, etc.), radiation correction, atmospheric correction, image fusion, image mosaic and image clipping. In the geometric correction link, the panchromatic image is geometrically corrected by selecting control points from standard data, the multispectral image is registered with the panchromatic image as the reference image, the multispectral and panchromatic images are fused, and the fusion results are clipped using vector boundaries, and finally a multispectral image with geographic coordinates and high resolution is obtained.
[0023] This method can provide high-quality data support for subsequent image analysis and applications. It aims to achieve efficient and accurate processing of multispectral camera images and improve the application value of remote sensing data by optimizing the preprocessing process and combining advanced algorithms and technologies.
[0024] like Figure 1 As shown, the multispectral remote sensing image data preprocessing method disclosed in this embodiment includes the following steps: Acquire multispectral remote sensing images and perform noise reduction processing to obtain images to be corrected; Determine the reference image, select corresponding ground control points from the image to be corrected and the reference image, perform geometric correction on the image to be corrected based on the selected ground control points, perform image resampling on the geometrically corrected image, and complete image registration; Perform radiometric calibration and atmospheric correction on the registered images; Obtain a full-color image, and fuse the image after radiometric calibration and atmospheric correction with the full-color image to obtain a fused image; The fused image is cropped using the vector boundary to obtain the preprocessed multispectral remote sensing image.
[0025] In some embodiments, the multispectral remote sensing image is subjected to noise reduction processing, specifically comprising: According to the noise type and intensity in multispectral remote sensing images, choose the appropriate noise reduction method: For the peaks or bright spots formed by periodic noise, use bandpass or slot filtering to eliminate them; For the spike noise that is not parallel to the scanning direction, Fourier transform is used for filtering; For the stripes parallel to the scanning direction and the bad lines irrelevant to the radiation signal, Fourier transform and low-pass filtering are used to eliminate or weaken them.
[0026] In some embodiments, after performing noise reduction processing on the multispectral remote sensing image, the method further includes: Determine the area, characteristics and influence of thin clouds and shadows in multispectral remote sensing images, use filtering technology to reduce the noise and blur caused by thin clouds, and use methods including ratio method, interpolation method and model method to remove the shadow area to obtain the image to be corrected.
[0027] In some embodiments, corresponding ground control points are selected from the image to be corrected and the reference image, and the specific selection method includes: The ground control points have the same positions on the image to be corrected and the reference image and have obvious features and clear point marks, including road intersections and river confluences; The ground objects on the ground control points do not change over time; The ground control points are evenly distributed in the entire image and a certain number is guaranteed; The ground control points can be selected by taking a topographic map as a reference, or obtained by field GPS measurement, or obtained from a calibrated image.
[0028] In some embodiments, geometric correction is performed on the image to be corrected based on the selected ground control points, specifically including: Determine the coordinate transformation relationship between the corresponding ground control points in the reference image and the image to be corrected, and obtain a geometric correction model; Based on the geometric correction model, the pixel coordinates in the image to be corrected are transformed into the corrected coordinate system to obtain the new coordinates of the pixels in the image to be corrected; Based on the coordinate position of the ground control points in the reference image, it is determined whether the new coordinates of the ground control points in the image to be corrected are within the set error range. If so, the geometric correction is completed. If not, the ground control points are readjusted, and an accurate geometric correction model is iteratively obtained and applied to the entire image to be corrected.
[0029] In some embodiments, image resampling is performed on the geometrically corrected image to complete image registration, specifically including: By interpolation calculation, according to the values of the original adjacent pixels in the image to be corrected, the pixel values at the non-integer coordinates in the geometrically corrected image are estimated to complete the image resampling; The interpolation calculation method includes the nearest neighbor method, the bilinear interpolation method and the cubic convolution method.
[0030] In some embodiments, radiometric calibration and atmospheric correction are performed on the registered image, specifically including: Convert the brightness grayscale value of the registered image into an absolute radiation brightness value, eliminate the radiation value difference of the unchanged ground objects, and complete the radiation calibration; According to the preset radiation brightness value and atmospheric parameters, the real reflectivity of surface objects is inverted to complete the atmospheric correction.
[0031] In some embodiments, before the fused image is cropped using the vector boundary, the fused image is further mosaicked, specifically including: Obtain the image to be mosaicked, and perform geometric registration on the fused image and the image to be mosaicked so that they are in the same spatial coordinates; Color matching is performed on the fused image after geometric registration and the image to be mosaicked, so that the images at different phases are coordinated and consistent in color. Determine the overlapping area, select a stitching line connecting the two images in the overlapping area, and complete the image mosaic; Alternatively, the fused image is cropped using the vector boundary, specifically: Image cropping is performed according to administrative boundaries or natural boundaries to remove areas outside the study area.
[0032] In some embodiments, the process further includes performing orthorectification on the registered image, specifically including: Obtain orthophoto reference data, and perform orthophoto correction on the registered image based on the orthophoto reference data to eliminate image distortion caused by terrain effects and sensor viewing angle factors; The orthophoto reference data includes orthophoto images, vector data and terrain elevation models.
[0033] Next, the above technical solution of this embodiment will be explained in detail.
[0034] The multispectral remote sensing image data preprocessing method provided in this embodiment may specifically include the following steps: Step 1: Obtain a multispectral remote sensing image and perform a preliminary analysis of the image feature information, select appropriate noise reduction, thin cloud, and shadow processing methods, and obtain a multispectral remote sensing image after noise reduction, i.e., the image to be corrected; Step 2: Determine the reference image, select ground control points from the image to be corrected and the reference image, and perform geometric precision correction, image registration, and orthorectification according to the geometric deformation of the image to be corrected; Step 3: Obtain a full-color image, fuse the image processed in the above steps into a high-resolution multispectral image, i.e., a fused image, and complete image mosaicking and cropping of the fused image; Step 4: Eliminate sensor response differences and atmospheric effects through radiation correction, thereby extracting more accurate remote sensing information.
[0035] Due to factors such as sensors, weather, and solar altitude angle, various types of noise, thin clouds, and mountain shadows may appear in some acquired remote sensing images, which will not only reduce the visual effect of the image, but may also affect subsequent data analysis and processing, so they must be weakened or eliminated.
[0036] Furthermore, in step 1, a multispectral remote sensing image is obtained and a preliminary analysis of the image feature information is performed, and appropriate noise reduction, thin cloud, and shadow processing methods are selected, specifically: S1: Perform necessary preprocessing on the original image (i.e., the acquired multispectral remote sensing image), analyze the noise type and intensity in the image, and select a suitable noise reduction method; S2: Remove periodic and sharp noise, as well as bad lines and bands using filter-based methods or Fourier transform; S3: Determine the area, characteristics and impact of thin clouds and shadows, and select appropriate methods to deal with thin clouds and shadows; S4: Post-process and evaluate the quality of the image (evaluation indicators may include image clarity, contrast, color reproduction, etc.) to ensure that the processed image meets the needs of subsequent analysis and application.
[0037] Furthermore, the above process specifically includes: Periodic noise is usually superimposed on the original image, forming a periodic interference pattern with different amplitudes, frequencies and phases. It forms a series of peaks or bright spots, which are most prominent at certain spatial frequency positions and can be eliminated by bandpass or slot filtering. To eliminate peak noise, especially those that are not parallel to the scanning direction, Fourier transform filtering is used.
[0038] The stripes parallel to the scanning direction that appear in the remote sensing image, as well as some strip noise that is unrelated to the radiation signal, are bad lines, which can be eliminated or weakened by Fourier transform and low-pass filtering.
[0039] The image is further processed using filtering technology to reduce the noise and blur caused by thin clouds, smooth the image, and reduce the contrast between the thin cloud area and the surrounding surface, making it easier to identify and remove. In some cases, multi-source remote sensing data can be fused to take advantage of the advantages of different data sources to make up for the shortcomings of a single data source in the thin cloud coverage area.
[0040] For the detected shadow area, appropriate methods are used to remove it, including ratio method, interpolation method and model method. The ratio method estimates the surface reflectance under the shadow by comparing the spectral characteristics of the shadow area with the surrounding non-shadow area; the interpolation method estimates the pixel value of the shadow area using the values of the surrounding pixels; the model method predicts and removes shadows by establishing a mathematical model between the surface reflectance and factors such as the solar altitude angle and terrain.
[0041] After removing the shadows, the image needs to be further repaired to fill the holes or uneven areas caused by the shadow removal.
[0042] Furthermore, in the above step 2: The factors that cause image geometric deformation are generally divided into two categories, including systematic and non-systematic: Systematicity is usually caused by the sensor itself, has regularity and is predictable, and can be corrected using the sensor model; Non-systematic geometric deformation is irregular and may be caused by instability in the height and attitude of the sensor platform itself, or by changes in the earth's curvature, air refraction, and terrain.
[0043] In order to correct the geometric deformation, in the above step 2, ground control points are selected, and the image to be corrected is subjected to geometric precision correction, image registration and orthorectification according to the image geometric deformation.
[0044] It can be understood that the above-mentioned reference image refers to an image used as a reference or standard in the geometric correction process. It is usually an image known to be free of distortion or with little distortion, and is used to help determine and correct the geometric position deviation of other images. The reference image can be an image generated by ground-based measured data, a high-resolution satellite image, or a strictly calibrated map data.
[0045] The image to be corrected is subjected to geometric precision correction, image registration and orthorectification, specifically including the following steps: S1: Select GCPs (ground control points) from the image to be corrected and the reference image. The points can be selected based on the topographic map, obtained by field GPS measurement, or obtained from the corrected image. S2: Use the reference geographic coordinates to calculate the transformation function and establish a geometric correction model; S3: After calculating the model, image resampling is performed and image registration is completed; S4: Obtain ortho reference data to perform ortho correction, perform error checking and control on the results, and eliminate image distortion caused by factors such as terrain effects and sensor viewing angle.
[0046] Furthermore, ground control points are selected, and the ground control points must meet the following characteristics: (1) GCPs have obvious and clear point marks on the image, such as road intersections, river intersections, etc. (2) The features on the ground control points do not change over time; (3) GCPs are evenly distributed throughout the image and a certain number of them must be guaranteed. Different correction models have different requirements for the number of control points.
[0047] A rigorous physical model can be established for the auxiliary data provided by the satellite, which only requires 9 control points; for the rational polynomial model, generally no less than 30 control points are required for each scene, and the number of control points can be appropriately increased in difficult areas; the geometric polynomial model will be determined according to the terrain conditions, and it requires more control points than the above models, usually around 30-50 control points per scene, especially in mountainous areas, where the number of control points should be appropriately increased.
[0048] The specific process of establishing the geometric correction model: After the ground control points are determined, a reasonable coordinate transformation function (i.e., geometric correction model) is selected using the pixel coordinates (x, y) on the image to be corrected and their reference geographic coordinates (X, Y) on the reference image, and the transformation function is calculated. Based on the obtained geometric correction model, the pixel coordinates in the image to be corrected are transformed into the corrected coordinate system to obtain the new coordinates of the pixels in the image to be corrected, that is, the coordinates of the control points in the image to be corrected are recalculated based on the calculated transformation function to obtain (X', Y'); The error of the control point, i.e., RMS, is calculated using the reference geographic coordinates (X, Y) on the reference image and the coordinates (X', Y') of the control point in the recalculated image to be corrected. If the RMS is too large, the control points need to be readjusted and the above process is repeated. Finally, an accurate transformation function is obtained and applied to the entire image.
[0049] The specific process of image resampling: After using the precise transformation function to act on the entire image to be corrected, since the distribution of the relocated pixels in the original image (i.e., the image to be corrected) is uneven, it is necessary to resample the original image according to certain rules, perform interpolation calculations on the brightness values, and establish a new image matrix.
[0050] Commonly used interpolation methods include the nearest neighbor method, bilinear interpolation method, and cubic convolution interpolation method. It is generally believed that the nearest neighbor method is conducive to maintaining the grayscale of the original image, but it has a greater damage to the geometric structure of the image; although the latter two methods approximate the pixel values, they largely retain the original geometric structure of the image, such as road networks, water systems, and boundaries of land objects.
[0051] The above process is further explained as follows: The purpose of geometric correction is to eliminate geometric distortion in the image and make the image in a standard geometric position. Image registration is to align two or more images into the same coordinate system or space through geometric transformation to achieve the correspondence between images and the fusion of information.
[0052] In the process of geometric correction, it is first necessary to establish a geometric correction model based on the ground control points, and transform the pixel coordinates in the distorted image (image to be corrected) into the corrected coordinate system. However, since the transformed coordinates are often not integers, that is, they are not at the pixel center of the original image (image to be corrected), it is necessary to recalculate the pixel values at these non-integer coordinates through resampling.
[0053] There are many methods for image resampling, such as the nearest neighbor method, bilinear interpolation, and cubic convolution. These methods estimate the pixel values at non-integer coordinates based on the values of adjacent pixels in the original image (the image to be corrected) through interpolation calculation, thereby obtaining the corrected image.
[0054] Although image resampling is not performed directly for image registration, it is an important part of the geometric correction process and provides a more accurate and reliable image basis for subsequent image registration.
[0055] Perform orthorectification. The specific process includes: First, orthophotos or vector data and terrain elevation models (DEM) are needed as orthophoto reference data, and then the orthophoto reference data and DEM data are used to perform geometric correction and projection transformation on the multispectral data. The correction process should consider factors such as the satellite's viewing angle, the curvature of the earth, and terrain changes. After the orthophoto correction is completed, the results need to be carefully checked and the error controlled, usually requiring the error to be controlled within a certain range (such as within 2 pixels).
[0056] In step 3, the panchromatic image is a black and white image of the entire visible light wavelength range (generally defined as 0.4μm-0.7μm) acquired by the remote sensor. It is a mixed image of this wavelength range, generally a grayscale image. Panchromatic images and multispectral images are parallel and both belong to remote sensing images.
[0057] The images after radiometric calibration and atmospheric correction are fused into high-resolution multispectral images to obtain fused images, and the fused images are mosaicked and cropped, including the following steps: S1: Image fusion based on Fourier transform or decomposition method to generate a high-resolution multispectral image to provide more comprehensive information; S2: Geometrically align the mosaic image and the fused image so that they are in the same spatial coordinate system, and then use a certain method to color match adjacent images so that the images at different phases are consistent in color. S3: determining the overlapping area, selecting a stitching line connecting the two images in the overlapping area, completing the image mosaicking, and then performing color uniformity processing to ensure the color balance of the mosaicked image; S4: Crop the image according to administrative boundaries or natural boundaries to remove areas outside the study area.
[0058] It can be understood that in this embodiment, the image after radiation calibration and atmospheric correction is fused with the full-color image based on the Fourier transform or decomposition method to obtain a fused image.
[0059] Furthermore, since the total radiation brightness of the ground target finally measured by the sensor is not a reflection of the true reflectivity of the surface, it also includes radiation errors caused by atmospheric absorption, scattering, etc., which will affect the quality of the image and the accuracy of subsequent analysis. In order to reduce the radiation error, the above step 4 of this embodiment eliminates the corresponding differences of the sensors and the influence of the atmosphere through radiation correction, thereby performing more accurate remote sensing information extraction, including the following steps: S1: Radiometric calibration, converting the brightness grayscale value of the image into an absolute radiometric brightness value, eliminating the difference in the radiation value of the unchanged objects, and obtaining the necessary parameters; S2: After all parameters are set, the atmospheric correction operation is performed to invert the true reflectivity of the surface objects according to the input radiance value and atmospheric parameters; S3: Extract remote sensing information and identify the type, nature, spatial position, shape, size and other attributes of the object based on the features of the object on the remote sensing image; S4: Through visual interpretation, remote sensing images are interpreted manually, and the scope of target objects on the remote sensing images is manually outlined to achieve the purpose of information extraction.
[0060] Embodiment 2 This embodiment discloses a multispectral remote sensing image data preprocessing system.
[0061] like Figure 2 As shown, the multispectral remote sensing image data preprocessing system includes: The noise reduction module is configured to: acquire a multispectral remote sensing image and perform noise reduction processing to obtain an image to be corrected; The geometric correction module is configured to: determine a reference image, select corresponding ground control points from the image to be corrected and the reference image, perform geometric correction on the image to be corrected based on the selected ground control points, perform image resampling on the geometrically corrected image, and complete image registration; The atmospheric correction module is configured to: perform radiometric calibration and atmospheric correction on the registered image; The fusion module is configured to: acquire a full-color image, fuse the image after radiometric calibration and atmospheric correction with the full-color image, and obtain a fused image; The cropping module is configured to: crop the fused image using the vector boundary to obtain a preprocessed multispectral remote sensing image.
[0062] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.
[0063] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. A multispectral remote sensing image data preprocessing method, characterized in that: The following steps are involved: Acquire multispectral remote sensing images and perform noise reduction processing to obtain images to be corrected; Determine the reference image, select corresponding ground control points from the image to be corrected and the reference image, perform geometric correction on the image to be corrected based on the selected ground control points, perform image resampling on the geometrically corrected image, and complete image registration; Perform radiometric calibration and atmospheric correction on the registered images; Obtain a full-color image, and fuse the image after radiometric calibration and atmospheric correction with the full-color image to obtain a fused image; The fused image is cropped using the vector boundary to obtain the preprocessed multispectral remote sensing image.
2. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: The multispectral remote sensing image is subjected to noise reduction processing, including: According to the noise type and intensity in multispectral remote sensing images, choose the appropriate noise reduction method: For the peaks or bright spots formed by periodic noise, use bandpass or slot filtering to eliminate them; For the spike noise that is not parallel to the scanning direction, Fourier transform is used for filtering; For the stripes parallel to the scanning direction and the bad lines irrelevant to the radiation signal, Fourier transform and low-pass filtering are used to eliminate or weaken them.
3. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: After the multispectral remote sensing image is subjected to noise reduction, it also includes: Determine the area, characteristics and influence of thin clouds and shadows in multispectral remote sensing images, use filtering technology to reduce the noise and blur caused by thin clouds, and use methods including ratio method, interpolation method and model method to remove the shadow area to obtain the image to be corrected.
4. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: Select corresponding ground control points from the image to be corrected and the reference image. The specific selection methods include: The ground control points have the same positions on the image to be corrected and the reference image and have obvious features and clear point marks, including road intersections and river confluences; The ground objects on the ground control points do not change over time; The ground control points are evenly distributed in the entire image and a certain number is guaranteed; The ground control points can be selected by taking a topographic map as a reference, or obtained by field GPS measurement, or obtained from a calibrated image.
5. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: Based on the selected ground control points, the image to be corrected is geometrically corrected, including: Determine the coordinate transformation relationship between the corresponding ground control points in the reference image and the image to be corrected, and obtain a geometric correction model; Based on the geometric correction model, the pixel coordinates in the image to be corrected are transformed into the corrected coordinate system to obtain the new coordinates of the pixels in the image to be corrected; Based on the coordinate position of the ground control points in the reference image, it is determined whether the new coordinates of the ground control points in the image to be corrected are within the set error range. If so, the geometric correction is completed. If not, the ground control points are readjusted, and an accurate geometric correction model is iteratively obtained and applied to the entire image to be corrected.
6. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: Image resampling is performed on the geometrically corrected image to complete image registration, including: By interpolation calculation, according to the values of the original adjacent pixels in the image to be corrected, the pixel values at the non-integer coordinates in the geometrically corrected image are estimated to complete the image resampling; The interpolation calculation method includes the nearest neighbor method, the bilinear interpolation method and the cubic convolution method.
7. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: Perform radiometric calibration and atmospheric correction on the registered images, including: Convert the brightness grayscale value of the registered image into an absolute radiation brightness value, eliminate the radiation value difference of the unchanged ground objects, and complete the radiation calibration; According to the preset radiation brightness value and atmospheric parameters, the real reflectivity of surface objects is inverted to complete the atmospheric correction.
8. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: Before the fused image is cropped using the vector boundary, the fused image is mosaicked, including: Obtain the image to be mosaicked, and perform geometric registration on the fused image and the image to be mosaicked so that they are in the same spatial coordinates; Color matching is performed on the fused image after geometric registration and the image to be mosaicked, so that the images at different phases are coordinated and consistent in color. Determine the overlapping area, select a stitching line connecting the two images in the overlapping area, and complete the image mosaic; Alternatively, the fused image is cropped using the vector boundary, specifically: Image cropping is performed according to administrative boundaries or natural boundaries to remove areas outside the study area.
9. The multispectral remote sensing image data preprocessing method according to claim 1, characterized in that: It also includes orthorectification of the registered images, including: Obtain orthophoto reference data, and perform orthophoto correction on the registered image based on the orthophoto reference data to eliminate image distortion caused by terrain effects and sensor viewing angle factors; The orthophoto reference data includes orthophoto images, vector data and terrain elevation models.
10. A multispectral remote sensing image data preprocessing system, characterized in that: include: The noise reduction module is configured to: acquire a multispectral remote sensing image and perform noise reduction processing to obtain an image to be corrected; The geometric correction module is configured to: determine a reference image, select corresponding ground control points from the image to be corrected and the reference image, perform geometric correction on the image to be corrected based on the selected ground control points, perform image resampling on the geometrically corrected image, and complete image registration; The atmospheric correction module is configured to: perform radiometric calibration and atmospheric correction on the registered image; The fusion module is configured to: acquire a full-color image, fuse the image after radiometric calibration and atmospheric correction with the full-color image, and obtain a fused image; The cropping module is configured to: crop the fused image using the vector boundary to obtain a preprocessed multispectral remote sensing image.
Citation Information
Patent Citations
Low-spatial-resolution multisource remote sensing image space consistency correction method
CN103413272A
Remote sensing image processing method based on geographic information system
CN110060218A
Geometric fine correction method suitable for high-resolution remote sensing image
CN113592744A
Satellite remote sensing image fusion method
CN113902661A
Remote sensing image processing method and system for natural resource management
CN115311556A
Cited By
Ecological hydrological simulation method based on machine learning model
CN119476055A
Method and system for improving geometric precision correction of historical geoscience map
CN120219252A
Terrain change monitoring system based on image analysis
CN120298398A
A terrain change monitoring system based on image analysis
CN120298398B
Multi-spectral remote sensing image strip radiation consistency correction method and system
CN120495088A
