A polar-orbiting meteorological satellite image display system and display method

Abstract

The present invention discloses a polar-orbiting meteorological satellite image display system and display method, a data acquisition unit: used to acquire images taken by an imaging device carried by a polar-orbiting meteorological satellite, divide the global image area, and perform separate shooting according to the divided areas, and the present invention relates to the field of remote sensing imaging technology. The polar-orbiting meteorological satellite image display system and display method acquires local influence planning through the data acquisition unit, reduces the time for subsequent image processing, and the image optimization processing unit has a good image enhancement effect by constructing a Retinex‑Net model, can retain more image detail information, and cooperates with dynamic histogram equalization and homomorphic filtering to optimize the image, which can reduce the degree of distortion of image optimization, and then cooperates with the image splicing correction unit to fuse the image, ensuring the integrity of the global image.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing image technology, and in particular to a polar-orbiting meteorological satellite image display system and a display method. Background Art

[0002] Polar-orbiting meteorological satellites have the ability to obtain remote sensing images covering the entire world, but the imaging equipment carried by polar-orbiting meteorological satellites can only observe local areas of the earth's surface at the same time. To generate global images, it is necessary to stitch together multiple images, multi-track images, or multi-scene images obtained over a period of time. At this time, the images are not observed at the same time, and are affected by changes in the angle of sunlight, etc., and the image quality is inconsistent.

[0003] However, when existing polar-orbiting meteorological satellite images are fused, a disordered shooting and acquisition method is adopted, which makes it difficult to fuse subsequent images. When adjacent images are directly fused, due to different image qualities, the fused color difference is visible and the quality is poor. Directly adjusting the illumination will cause distortion of image detail information. Therefore, the present invention proposes a polar-orbiting meteorological satellite image display system and display method to solve the above-mentioned problems. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the deficiencies in the prior art, the present invention provides a polar-orbiting meteorological satellite image display system and display method, which solve the problems mentioned in the above background technology.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A polar-orbiting meteorological satellite image display system, comprising:

[0008] Data acquisition unit: used to acquire images taken by the imaging equipment carried by the polar-orbiting meteorological satellite, divide the global image area, take separate photos according to the divided areas, and send the information to the image optimization processing unit, wherein there are overlapping parts in the images of two adjacent divided areas;

[0009] Image optimization processing unit: used to receive the information sent by the data acquisition unit, then build a Ret i nex neural network model, convert the original image from RGB space to YCrCb space, use the network model to process the Y component alone, only use guided filtering to process the Cr and Cb components, use a multi-scale average algorithm to adaptively enhance the low-light image, and send the information to the image stitching correction unit;

[0010] Image stitching correction unit: used to receive the information sent by the image optimization processing unit, then fuse the adjacent images so that the adjacent images overlap at the intersection, remove the redundant image parts, and send the information to the background color selection unit;

[0011] Background color selection unit: used to receive the information sent by the image stitching and correction unit, then analyze the image information in different areas, and select a background color with obvious distinction in the image, wherein the initial background color of the image is white.

[0012] As an improved technical solution, the specific method of performing separate shooting according to the divided areas in the data acquisition unit is:

[0013] S1: First, mark the global image area as W, then divide the W area and mark it as S j , j = 1, 2, ..., n, j is the number of the divided area, where the reference calculation formula is: And, S j The shape is rectangular, according to S j The shape of the rectangle is marked as AS. j ,BS j , CS j and DS j ;

[0014] S2: Then S j The four edges AS j ,BS j , CS j and DS j The corresponding adjacent regions are labeled XS1, XS2, XS3, and XS4. j When it is a side figure, 2 or 3 sides exist in adjacent regions, and the side is located in the adjacent region. j When it is an internal figure, the four sides exist in adjacent areas, and the side edges are located in the adjacent areas.

[0015] As an improved technical solution, the specific expression for adaptively enhancing the low-light image in the image optimization processing unit is:

[0016]

[0017] In the formula, R(x, y) represents the enhanced image, M(x, y) represents the enhanced reflection image of the Y component, and K(x, y) represents the enhanced illumination image of the Y component. Represents element-by-element product operation, C r (x, y) represents the enhanced C r Quantity, C b (x, y) represents the enhanced Cb Quantity;

[0018] The expressions of M(x, y) and K(x, y) are:

[0019] M(x, y) = RTN_D θ (Y(x,y))

[0020] K(x, y) = RTN_E α (Y(x,y))

[0021] Where, RTN_D θ Represents the decomposed sub-network model based on Ret i nex, RTN_E α (Y(x, y)) represents the Retinex-based enhancement subnetwork model, θ and α represent learnable network parameters, and Y(x, y) represents the Y component of the original input image.

[0022] As an improved technical solution, the processing method of the Retinex neural network model in the image optimization processing unit to enhance the reflectance map of the Y component is:

[0023] The enhanced result of the Y component illumination map is connected pixel by pixel with the Y component reflection map that has not been enhanced as a self-attention, that is, the channels are connected, and then input into the enhancement sub-network for enhancement.

[0024] As an improved technical solution, the image stitching correction unit performs fusion processing on adjacent images so that the adjacent images overlap at the intersection. The specific method is:

[0025] P1: First, two points are determined in two adjacent images in sequence. The two points of the two images are marked as A, B and Among them, the two points of the image are selected as the two end points of the image, and the positions of the two points on the two images are the same. At this time, the overlapping sides of the two images are placed together, and AB and are two line segments of equal length;

[0026] P2: Then A, B, and The rectangular area formed is cropped. At this time, point A and Point B and The points overlap, and the fusion of the two adjacent images is completed. The above operations P1 and P2 are repeated for the images on other adjacent sides.

[0027] As an improved technical solution, the two points of the image in P1 are selected as the two end points of the image, and the specific method of determining the positions of the two points on the two images to be the same is:

[0028] After confirming the two endpoints A and B of one of the edges, it is necessary to check whether the adjacent images A and B on the other two sides are located therein. When the two endpoints A and B are both located in the adjacent images on the other two sides, the endpoints A and B can be determined. Otherwise, the two endpoints A and B need to be re-determined.

[0029] As an improved technical solution, the background color selection unit analyzes the image information in different areas, and selects the background color with obvious distinction in the image in the following specific manner:

[0030] M1: First divide the background color and mark it as R j , j = 1, 2, ..., n, j is the type number of the background color, and then the image colors in different areas are marked as P j , where P j ∈R j , when P j When there is no white area in the image, the background color is selected as white;

[0031] M2: Then when P j When there is a white area in the j Does it include R j All the colors in the j =R j When R j The least one background color appears in the j ≠R j When R j A background color that does not appear in the .

[0032] As an improved technical solution, R is selected in M2 j At least one background color appears in the selection and R j If a background color that does not appear in the selection is selected, if multiple choices appear, it will be selected randomly.

[0033] As an improved technical solution, the method further includes a data storage unit, which is used to store data in the data acquisition unit, the image optimization processing unit, the image stitching correction unit and the background color selection unit.

[0034] The present invention also discloses a polar-orbiting meteorological satellite image display method, which specifically comprises the following steps:

[0035] Step 1: Obtain images taken by imaging equipment carried by polar-orbiting meteorological satellites, divide the global image area, and take separate photos according to the divided areas;

[0036] Step 2: Construct a Ret i nex neural network model, transform the original image from RGB space to YCrCb space, use the network model to process the Y component alone, only use guided filtering to process the Cr and Cb components, and use a multi-scale average algorithm to adaptively enhance the low-light image;

[0037] Step 3: Fuse adjacent images so that they overlap at the intersection and remove redundant image parts;

[0038] Step 4: Analyze the image information in different areas and select a background color with obvious distinction in the image.

[0039] (III) Beneficial effects

[0040] The present invention provides a polar-orbiting meteorological satellite image display system and display method. Compared with the prior art, the present invention has the following beneficial effects:

[0041] The polar-orbiting meteorological satellite image display system and method acquire local image planning through a data acquisition unit, thereby reducing the time for subsequent image processing. The image optimization processing unit has a good image enhancement effect by constructing a Ret i nex-Net model, which can retain more image detail information, and cooperates with dynamic histogram equalization and homomorphic filtering to optimize the image, which can reduce the degree of distortion of image optimization. Then, the image can be fused in cooperation with an image splicing correction unit, thereby ensuring the integrity of the global image. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The block diagram of the polar-orbiting meteorological satellite image display system of the present invention;

[0043] Figure 2 The present invention is a flow chart of the polar-orbiting meteorological satellite image display method. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0045] See also Figure 1-Figure 2 , the embodiment of the present invention provides six technical solutions:

[0046] Embodiment 1: A polar-orbiting meteorological satellite image display system, comprising:

[0047] Data acquisition unit: used to acquire images taken by the imaging device carried by the polar-orbiting meteorological satellite, divide the global image area, take separate photos according to the divided areas, and send the information to the image optimization processing unit, wherein there is an overlapped portion between the images of two adjacent divided areas; the specific method of taking separate photos according to the divided areas in the data acquisition unit is as follows: S1: first mark the global image area as W, then divide the W area and mark it as S j , j = 1, 2, ..., n, j is the number of the divided area, where the reference calculation formula is: And, S j The shape is rectangular, according to S j The shape of the rectangle is marked as AS. j ,BS j , CS j and DS j ; S2: Then S j The four edges AS j ,BS j , CS j and DS j The corresponding adjacent regions are labeled XS1, XS2, XS3, and XS4. j When it is a side figure, 2 or 3 sides exist in adjacent regions, and the side is located in the adjacent region. j When it is an internal figure, the four sides exist in adjacent areas, and the side edges are located in the adjacent areas.

[0048] Image optimization processing unit: used to receive the information sent by the data acquisition unit, then build a Ret i nex neural network model, transform the original image from RGB space to YCrCb space, use the network model to process the Y component alone, and only use guided filtering to process the Cr and Cb components. Adopt a multi-scale average algorithm to adaptively enhance the low-light image, and send the information to the image stitching correction unit; the specific expression for adaptively enhancing the low-light image in the image optimization processing unit is:

[0049]

[0050] In the formula, R(x, y) represents the enhanced image, M(x, y) represents the enhanced reflection image of the Y component, and K(x, y) represents the enhanced illumination image of the Y component. Represents element-by-element product operation, C r (x, y) represents the enhanced C r Quantity, C b (x, y) represents the enhanced C b Quantity;

[0051] The expressions of M(x, y) and K(x, y) are:

[0052] M(x, y) = RTN_D θ (Y(x,y))

[0053] K(x, y) = RTN_E α (Y(x,y))

[0054] Where, RTN_D θ Represents the decomposed sub-network model based on Ret i nex, RTN_E α (Y(x, y)) represents the enhanced sub-network model based on Retinex, θ and α represent learnable network parameters, and Y(x, y) represents the Y component of the original input image. The processing method of the Retinex neural network model in the image optimization processing unit for enhancing the reflectance map of the Y component is:

[0055] The enhanced result of the Y component illumination map is connected pixel by pixel with the Y component reflection map that has not been enhanced as a self-attention, that is, the channels are connected, and then input into the enhancement sub-network for enhancement.

[0056] The Ret i nex neural network is divided into a decomposition sub-network and an enhancement sub-network. Both sub-networks are based on the standard 4-layer U-net. The down-sampling parts of the decomposition sub-network are connected with 3×3 convolution kernels using residual connections to form residual blocks; 3×3 void convolutions with a void rate of 2 are used to replace all the down-sampling convolution kernels of the enhancement sub-network; a multi-feature fusion enhancement module is added to all the up-sampling layers of the enhancement sub-network; and a non-local self-supervision module is inserted before the first down-sampling void convolution of the enhancement sub-network. The parameters of the pooling layer in the sub-network are adjusted to parameters adapted to single-channel input.

[0057] Image stitching correction unit: used for receiving information sent by the image optimization processing unit, and then fusing adjacent images so that adjacent images overlap at the intersection, and removing redundant image parts, and sending information to the background color selection unit; the specific method of fusing adjacent images in the image stitching correction unit so that adjacent images overlap at the intersection is:

[0058] P1: First, two points are determined in two adjacent images in sequence. The two points of the two images are marked as A, B and Among them, the two points of the image are selected as the two end points of the image, and the positions of the two points on the two images are the same. At this time, the overlapping sides of the two images are placed together, and AB and are two line segments of the same length; P2: Then A, B, and The rectangular area formed is cropped. At this time, point A and Point B and The points overlap, and the two adjacent images are fused. The above operations P1 and P2 are repeated for the other adjacent images. The two points in P1 are selected as the two endpoints of the image, and the specific method for determining the same position of the two points on the two images is:

[0059] After confirming the two endpoints A and B of one of the edges, it is necessary to check whether the adjacent images A and B on the other two sides are located therein. When the two endpoints A and B are both located in the adjacent images on the other two sides, the endpoints A and B can be determined. Otherwise, the two endpoints A and B need to be re-determined.

[0060] Embodiment 2: This embodiment is based on Embodiment 1, and the difference from Embodiment 1 is that it also includes a background color selection unit: used to receive information sent by the image stitching correction unit, and then analyze the image information in different areas, and select a background color with obvious distinction in the image, wherein the initial background color of the image is white. The specific method of analyzing the image information in different areas and selecting a background color with obvious distinction in the image in the background color selection unit is as follows: M1: First, divide the background color and mark it as R j , j = 1, 2, ..., n, j is the type number of the background color, and then the image colors in different areas are marked as P j , where P j ∈R j , when P j When there is no white area in the image, the background color is selected as white;

[0061] M2: Then when P j When there is a white area in the j Does it include R j All the colors in the j =R j When R j The least one background color appears in the j ≠R j When R j A background color that does not appear in M2. Select R j At least one background color appears in the selection and R j If a background color that does not appear in the selection is selected, if multiple choices appear, it will be selected randomly.

[0062] Embodiment 3: It also includes a data storage unit, which is used to store data in the data acquisition unit, the image optimization processing unit, the image stitching correction unit and the background color selection unit.

[0063] Embodiment 4: Compared with Embodiment 1, Embodiment 2 and Embodiment 3, the technical solution of this embodiment is to combine the solutions of Embodiment 1, Embodiment 2 and Embodiment 3 for implementation.

[0064] Embodiment 5: A method for displaying polar-orbiting meteorological satellite images, comprising the following steps:

[0065] Step 1: Obtain images taken by imaging equipment carried by polar-orbiting meteorological satellites, divide the global image area, and take separate photos according to the divided areas;

[0066] Step 2: Construct a Ret i nex neural network model, transform the original image from RGB space to YCrCb space, use the network model to process the Y component alone, and only use guided filtering to process the Cr and Cb components. Use a multi-scale average algorithm to adaptively enhance the low-light image.

[0067] Step 3: Fuse adjacent images so that they overlap at the intersection and remove redundant image parts;

[0068] Step 4: Analyze the image information in different areas and select a background color with obvious distinction in the image.

[0069] Embodiment 6: A polar-orbiting meteorological satellite image display system, comprising:

[0070] Data acquisition unit: used to acquire images taken by the imaging equipment carried by the polar-orbiting meteorological satellite, divide the global image area, take separate photos according to the divided areas, and send the information to the image optimization processing unit, wherein there are overlapping parts in the images of two adjacent divided areas;

[0071] Image optimization processing unit: used to receive the information sent by the data acquisition unit, then build a Ret i nex neural network model, transform the original image from RGB space to YCrCb space, use the network model to process the Y component alone, and only use guided filtering to process the Cr and Cb components. Adopt a multi-scale average algorithm to adaptively enhance the low-light image, and send the information to the image stitching correction unit;

[0072] Image stitching correction unit: used to receive the information sent by the image optimization processing unit, then fuse the adjacent images so that the adjacent images overlap at the intersection, remove the redundant image parts, and send the information to the background color selection unit;

[0073] Background color selection unit: used to receive the information sent by the image stitching and correction unit, then analyze the image information in different areas, and select a background color with obvious distinction in the image, wherein the initial background color of the image is white.

[0074] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0075] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0076] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A polar-orbiting meteorological satellite image display system, characterized in that: include: Data acquisition unit: used to acquire images taken by the imaging equipment carried by the polar-orbiting meteorological satellite, divide the global image area, take separate photos according to the divided areas, and send the information to the image optimization processing unit, wherein there are overlapping parts in the images of two adjacent divided areas; Image optimization processing unit: used to receive the information sent by the data acquisition unit, then build a Retinex neural network model, convert the original image from the RGB space to the YCrCb space, use the network model to process the Y component alone, only use guided filtering to process the Cr and Cb components, use a multi-scale average algorithm to adaptively enhance the low-light image, and send the information to the image stitching correction unit; Image stitching correction unit: used to receive the information sent by the image optimization processing unit, and then fuse the adjacent images to make the adjacent images overlap at the intersection, remove the redundant image parts, and send the information to the background color selection unit; Background color selection unit: used to receive the information sent by the image stitching and correction unit, and then analyze the image information in different areas, and select a background color with obvious distinction in the image, wherein the initial background color of the image is white; The specific expression for adaptively enhancing the low-light image in the image optimization processing unit is: In the formula, R(x, y) represents the enhanced image, M(x, y) represents the enhanced reflection image of the Y component, and K(x, y) represents the enhanced illumination image of the Y component. Represents element-by-element product operation, C r (x, y) represents the enhanced C r Quantity, C b (x, y) represents the enhanced C b Quantity; The expressions of M(x, y) and K(x, y) are: M(x,y)=RTN_D θ (Y(x,y)) K(x,y)=RTN_E α (Y(x,y)) Where, RTN_D θ Represents the Retinex-based decomposition sub-network model, RTN_E α (Y(x, y)) represents the enhanced subnetwork model based on Retinex, θ and α represent learnable network parameters, and Y(x, y) represents the Y component of the original input image; The specific method of fusing adjacent images in the image stitching correction unit so that adjacent images overlap at the intersection is: P1: First, two points are determined in two adjacent images, and the two points of the two images are marked as A, B and The two points of the image are selected as the two end points of the image, and the positions of the two points on the two images are the same. At this time, the overlapping sides of the two images are placed together, and AB and are two line segments of equal length; P2: Then A, B, and The rectangular area formed is cropped. At this time, point A and Point B and The points overlap, and the fusion of the two adjacent images is completed. The above operations P1 and P2 are repeated for the images on other adjacent sides.

2. A polar-orbiting meteorological satellite image display system according to claim 1, characterized in that: The specific method of performing separate shooting according to the divided areas in the data acquisition unit is: S1: First, mark the global image area as W, then divide the W area and mark it as S j , j = 1, 2, ..., n, j is the number of the divided area, where the reference calculation formula is: And, S j The shape is rectangular, according to S j The shape of the rectangle is marked as AS. j ,BS j , CS j and DS j ; S2: Then S j The four edges AS j ,BS j , CS j and DS j The corresponding adjacent regions are labeled XS1, XS2, XS3, and XS4. j When it is a side figure, 2 or 3 sides exist in adjacent regions, and the side is located in the adjacent region. j When it is an internal figure, the four sides exist in adjacent areas, and the side edges are located in the adjacent areas.

3. The polar-orbiting meteorological satellite image display system according to claim 1, characterized in that: The processing method of the Retinex neural network model in the image optimization processing unit for enhancing the reflectance image of the Y component is: The enhanced result of the Y component illumination map is used as a self-attention and connected to the Y component reflection map that has not been enhanced by pixels, that is, the channels are connected, and then input into the enhancement sub-network for enhancement.

4. The polar-orbiting meteorological satellite image display system according to claim 1, characterized in that: The specific method of selecting the two points of the image in P1 as the two end points of the image and determining the same position of the two points on the two images is: After confirming the two endpoints A and B of one of the edges, we need to check whether the adjacent images A and B on the other two sides are located in it. When the two endpoints A and B are both located in adjacent images on the other two sides, the endpoints A and B can be determined. Otherwise, the two endpoints A and B need to be re-determined.

5. The polar-orbiting meteorological satellite image display system according to claim 1, characterized in that: The background color selection unit analyzes the image information in different areas, and selects the background color with obvious difference in the image in the following specific manner: M1: First divide the background color and mark it as R j , j = 1, 2, ..., n, j is the type number of the background color, and then the image colors in different areas are marked as P j , where P j ∈R j , when P j When there is no white area in the image, the background color is selected as white; M2: Then when P j When there is a white area in the j Does it include R j All the colors in the j =R j When R j The least one background color appears in the j ≠R j When R j A background color that does not appear in the .

6. A polar-orbiting meteorological satellite image display system according to claim 5, characterized in that: The M2 selects R j At least one background color appears in the selection and R j If a background color that does not appear in the selection is selected, if multiple choices appear, it will be selected randomly.

7. The polar-orbiting meteorological satellite image display system according to claim 1, characterized in that: It also includes a data storage unit, which is used to store data in the data acquisition unit, the image optimization processing unit, the image stitching correction unit and the background color selection unit.

8. A polar-orbiting meteorological satellite image display system implementing any one of claims 1 to 7, characterized in that: The display method specifically includes the following steps: Step 1: Obtain images taken by imaging equipment carried by polar-orbiting meteorological satellites, divide the global image area, and take separate photos according to the divided areas; Step 2: Construct a Retinex neural network model, transform the original image from RGB space to YCrCb space, use the network model to process the Y component alone, use guided filtering to process the Cr and Cb components only, and use a multi-scale average algorithm to adaptively enhance the low-light image; Step 3: Fuse adjacent images so that they overlap at the intersection and remove redundant image parts; Step 4: Analyze the image information in different areas and select a background color with obvious distinction in the image.

Images