A stray light correction method for city night light images based on 45-degree mirror imaging

Abstract

The present invention discloses a stray light correction method for city night light images based on 45-degree mirror imaging. The method comprises the following steps: according to the spatiotemporal regularity characteristics of stray light in the nighttime city light images of the low-light channel in the air domain and frequency domain, the image is restricted in the zenith angle range and pixel-dimensional image decomposition is performed, the neighborhood smoothing filter method and the first-order gradient method are used in the flight dimension to remove the stray light background in the city night light images, the threshold segmentation, domain smoothing filter and the first-order gradient method are used in the scanning dimension, and the separation of effective radiation targets and stray light background is achieved by restricting the solar zenith angle range in the image. The correction method effectively improves the calibration accuracy of the low-light channel, ensures data quality and data application efficiency, provides a correction algorithm for unified radiation recalibration of historical long-time series data sets, and lays a solid foundation for the study of long-sequence satellite data climate, human activities, etc.

Description

Technical Field

[0001] The invention belongs to the technical field of stray light processing of low-light-level images, and in particular relates to a stray light correction method for city night light images based on 45-degree mirror imaging. Background Art

[0002] With the development of optoelectronic technology, low-light imaging under low-light conditions and its application have become one of the research hotspots in recent years. Remote sensing application technology based on low-light channel imaging is on the fast track of accelerated development. At present, night remote sensing images have been applied to many fields, such as monitoring the radiation changes of visible light in land, atmosphere, ice and snow, and ocean, so as to analyze the information of climate and ecosystem; detecting and describing information such as urban light, ship movement, forest fire sources and specific targets to analyze human activities; dynamically monitoring volcanic activities, lightning, aurora and other unique low-light remote sensing applications, which have important application prospects and scientific research value, and gradually become the basic configuration of meteorological and marine environmental satellites.

[0003] At present, the low-light detection and imaging payloads carried by existing polar-orbiting satellites generally have the problem of stray light pollution. Their low-light channels are highly sensitive and easily polluted by external stray light. Multiple low-light satellite-borne equipment has obvious stray light problems in the Earth observation area, which seriously affects the quality of nighttime remote sensing data and restricts the depth and breadth of global quantitative application of polar-orbiting satellite data. In order to provide important and basic application research for the quantitative research of low-light detection and imaging payloads, and to give full play to the detection capabilities of the instruments in the visible and near-infrared spectrum at night and during the dawn and dusk periods, it is urgent and challenging to carry out research on accurate stray light correction algorithms for low-light detection and imaging payloads.

[0004] At present, the stray light removal method can be achieved through the preliminary stray light suppression design, software stray light calculation and image processing methods. The stray light correction method of polluted data usually adopts the image processing method, mainly using digital filtering method, histogram matching based on the statistical characteristics of image grayscale information, variational and partial differential equation method, moment matching and other methods. For example, reference [1] (Correction of solar pollution in FY-3C / VIRR mid-infrared image, Zhu Jibiao et al., 2021) and reference [2] (Identification and correction of solar pollution in the infrared channel of FY-3C / visible light infrared scanning radiometer, Xu Hanlie et al., 2015) give the infrared channel stray light correction method, but such methods are not suitable for the stray light image of the low-light channel. The low-light channel has a strong degree of pollution on the ground image and a large pollution range. Its pollution path and stray light source are complex and diverse. The imaging principle, pollution mechanism and target characteristics of low-light detection make the low-light channel different from the infrared channel stray light processing method. Summary of the invention

[0005] In order to overcome the shortcomings of the existing technology in correcting stray light pollution in low-light detection, the present invention provides a stray light correction method for city night light images based on 45-degree mirror imaging.

[0006] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0007] The present invention discloses a stray light correction method for a city night light map based on 45-degree mirror imaging. The low-light scanning radiometer for earth imaging adopts a 45-degree scanning mirror and a K mirror derotation technology, and uses N detectors to scan across tracks to obtain earth observation images. The number of sampling pixels in the scanning direction is M, and one scan can obtain N×M two-dimensional images. In the flight direction, the K mirror will have an AB surface effect, and the two reflective mirror surfaces of the K mirror, namely the A surface and the B surface, are alternately collected to form an earth scanning image. The method is characterized in that it includes the following steps:

[0008] Step 1): Perform single-track image synthesis on multiple Earth observation images of the low-light scanning radiometer, and use the solar zenith angle data corresponding to each imaging position to calculate the solar zenith angle greater than The coordinate position at that time is set to 0, and the image is decomposed into a nighttime city light image in a single-pixel single-scan state according to the imaging pixel and the A / B plane;

[0009] Step 2): Based on the single-pixel single-scan state image obtained in step 1), stray light correction processing of the flight direction image data is carried out.

[0010] 2.1): Use the neighborhood smoothing filtering method to perform low-pass filtering on a single column of image data, and obtain a threshold greater than The one-dimensional coordinate matrix y1 of the night lights and other object points at that time:

[0011] (1) (2)

[0012] (3)

[0013] Among them, y i is the coordinate position of a single column of image data, f(y i ) is the image single column data coordinate y i The value at is f(y i ) when the smoothing width is m1, g(y i ) is f(y i ) is the value after deducting the smoothing value, W is the number of values ​​in a single column of image data, y1 is represented by a binary one-dimensional matrix, 1 represents object point information, and 0 represents background information;

[0014] 2.2): For the same single column data as above, the first-order gradient method is used to extract the point information of objects such as night lights, and the threshold is greater than The one-dimensional coordinate matrix y2 of the object point:

[0015] (4)

[0016] (5)

[0017] Among them, h(y i ) is the coordinate y i and i-1 The difference in response values ​​between adjacent positions, y2, is represented by a binary one-dimensional matrix, where 1 represents object point information and 0 represents background information;

[0018] 2.3): Based on the neighborhood smoothing filter method and the first-order gradient method, the coordinate matrices y1 and y2 are obtained, and the one-dimensional coordinate matrix Y of the object point is obtained by using the union of the two coordinate matrices:

[0019] (6)

[0020] Among them, Y is represented by a binary number, 1 represents object point information, and 0 represents background information. The two-dimensional coordinate matrix corresponding to the object point information in the two-dimensional image is established through the one-dimensional coordinate matrix corresponding to all column data in the two-dimensional image, and the coordinate matrices of all single-pixel single-scan state images are synthesized into a two-dimensional coordinate matrix Z corresponding to the original image according to the original position of the pixel;

[0021] 2.4): For the image two-dimensional coordinate matrix Z, with the object point coordinates as the center, the object point coordinates are expanded in the flight dimension using formula (1), and the flight dimension smoothing width is set to m2 to obtain the two-dimensional coordinate matrix Z1. Then, the object point coordinates are expanded in the scanning dimension using formula (1), and the scanning dimension smoothing width is set to n2 to obtain the two-dimensional coordinate matrix Z2. The coordinate matrix Z3 corresponding to the image stray light background is:

[0022] (7)

[0023] For the original two-dimensional image, based on the stray light background coordinate matrix Z3, the first stray light background image under the original image coordinates is obtained by interpolation. The first stray light corrected image is:

[0024] The first corrected image = original image - the first stray light background image (8)

[0025] Step 3): For the first corrected image obtained in step 2), correct the influence of stray light on the observed image data in the scanning direction.

[0026] 3.1): Use the threshold method to obtain the number of rows of data in the scanning direction that is greater than The coordinate matrix x0 corresponding to the object point is:

[0027] (9)

[0028] in, is the scanning direction coordinate x in the first corrected image i The corresponding corrected value;

[0029] 3.2): Based on formula (1) (3), the neighborhood smoothing filter method is used to The data is used to extract the coordinates of the object points for the second time. Greater than The coordinate position matrix x1 is:

[0030] (10)

[0031] Among them, g(x i ) is f(x i ) is the one-dimensional matrix value after deducting the smooth background, and the smoothing width is set to n3.

[0032] 3.3): Based on formula (5), the first-order gradient method is used to The data is extracted for the third time to obtain the coordinates of the object points. The coordinate position matrix x2 of the threshold:

[0033] (11)

[0034] 3.4): The coordinate matrices x0, x1 and x2 obtained by the above three methods are used to calculate the coordinate position matrix X corresponding to the object point using the union of the three:

[0035] (12)

[0036] Among them, the coordinates in x0, x1, x2, and X are all represented by 1 and 0, 1 represents the object point, and 0 represents the background point. The one-dimensional coordinate matrix calculated by all row data in the two-dimensional image is used to establish the two-dimensional coordinate matrix P corresponding to the object point information in the two-dimensional image;

[0037] 3.5): For the image two-dimensional coordinate matrix P, with the object point coordinate as the center, the object point coordinate is expanded in the flight dimension using formula (1), and the flight dimension smoothing width is set to m3 to obtain the two-dimensional coordinate matrix P1. Then, the object point coordinate is expanded in the scanning dimension using formula (1), and the scanning dimension smoothing width is set to n3 to obtain the two-dimensional coordinate matrix P2. The coordinate matrix P3 corresponding to the stray light background of the object point is:

[0038] (13)

[0039] 3.6): Based on the first corrected image and the stray light background coordinates P3, the second stray light background image under the image coordinates is obtained by interpolation. The two-dimensional image after stray light correction is:

[0040] The second corrected image = the first corrected image - the second stray light background image (14)

[0041] Step 4): Based on the second stray light correction image obtained in step 3.6), the imaging area of ​​the night shadow area is limited by using the solar zenith angle. The values ​​of are all set to 0, and the image obtained is the night city light image corrected for stray light.

[0042] Furthermore, the target signal in the city night light image is composed of stray light, ground object moon reflection radiation and night light radiation. The stray light and ground object reflection background are used as low-frequency information. The imaging image removes the stray light information through the low-pass filtering method, and only retains the high-frequency information part such as the city night light target.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows: in view of the common problem of stray light in low-light detection imaging payloads, the invention of this method provides help and reference for improving the observation accuracy and data quality of low-light detection instruments, promotes the quantitative application level of low-light remote sensing data, helps to improve the understanding of the changing laws of remote sensing instruments in orbit, and provides valuable analysis experience for improving the quality of space payload development. The present invention provides a correction algorithm for reprocessing long-term series historical data of low-light detection payloads, establishes high-quality and high-precision global nighttime low-light observation data, and lays a solid foundation for long-sequence satellite data climate research. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0045] Figure 1 This is the process of 45-degree scanning mirror image decomposition and coordinate matrix conversion in the present invention.

[0046] Figure 2 This is a flow chart of the stray light correction method for city night light images in the present invention.

[0047] Figure 3 This is a comparison diagram of the city night light image before and after the stray light correction in the present invention.

[0048] Figure (1) is the original image of stray light pollution in the city night light image.

[0049] Figure (2) is a correction image of stray light pollution in urban night light images. DETAILED DESCRIPTION

[0050] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further described below with reference to the accompanying drawings and in combination with embodiments.

[0051] The present invention discloses a stray light correction method for a city night light image based on 45-degree mirror imaging, and the specific implementation steps include:

[0052] Step 1: Single-track image synthesis of multiple earth observation images of a low-light scanning radiometer, wherein the low-light scanning radiometer adopts a 45° scanning mirror and a K mirror derotation technology, such as Figure 1 As shown, the earth observation image is obtained by using the cross-track scanning method of 10 detectors. The sampling pixels in the scanning direction are 1536. A 10×1536 two-dimensional image can be obtained in one scan. In the flight direction, the K mirror will have an AB surface effect. The A surface and the B surface are alternately collected to form a ground scanning image. The solar zenith angle data corresponding to each imaging position are used to calculate the coordinate position when the solar zenith angle in the synthetic image is greater than 90°. By setting the signal code value corresponding to its coordinate position to 0, the image is decomposed according to the A / B surface and single pixel in the flight dimension, and is decomposed into 20 single-pixel night city light images in the single scanning state.

[0053] Step 2): According to the single pixel single scan state image obtained in step 1), Figure 2 As shown in FIG. 1 , the influence of stray light correction on the observed image data in the flight direction is shown.

[0054] 2.1): Use the neighborhood smoothing filtering method to perform low-pass filtering on a single column of image data, and obtain a threshold greater than The one-dimensional coordinate matrix y1 of the night lights and other object points at that time:

[0055] (1)

[0056] (2)

[0057] (3)

[0058] Among them, y i is the coordinate position of a single column of image data, f(y i ) is the image single column data coordinate y i The value at is f(y i ) when the smoothing width is m1, g(y i ) is f(y i) is the value after deducting the smoothing value, 1536 is the number of values ​​in a single column of image data, y1 is represented by a binary one-dimensional matrix, 1 represents object point information, and 0 represents background information;

[0059] 2.2): For the same single column data as above, the first-order gradient method is used to extract the point information of objects such as night lights, and the threshold is greater than The one-dimensional coordinate matrix y2 of the object point:

[0060] (4)

[0061] (5)

[0062] Among them, h(y i ) is the coordinate y i and i-1 The difference in response values ​​between adjacent positions, y2, is represented by a binary one-dimensional matrix, where 1 represents object point information and 0 represents background information;

[0063] 2.3): Based on the neighborhood smoothing filter method and the first-order gradient method, the coordinate matrices y1 and y2 are obtained, and the one-dimensional coordinate matrix Y of the object point is obtained by using the union of the two coordinate matrices:

[0064] (6)

[0065] Among them, Y is represented by a binary number, 1 represents object point information, and 0 represents background information. The two-dimensional coordinate matrix corresponding to the object point information in the two-dimensional image is established through the one-dimensional coordinate matrix corresponding to all column data in the two-dimensional image, and the coordinate matrices of all single-pixel single-scan state images are synthesized into a two-dimensional coordinate matrix Z corresponding to the original image according to the original position of the pixel;

[0066] 2.4): For the image two-dimensional coordinate matrix Z, with the object point coordinates as the center, the object point coordinates are expanded in the flight dimension using formula (1), and the flight dimension smoothing width is set to m2 to obtain the two-dimensional coordinate matrix Z1. Then, the object point coordinates are expanded in the scanning dimension using formula (1), and the scanning dimension smoothing width is set to n2 to obtain the two-dimensional coordinate matrix Z2. The coordinate matrix Z3 corresponding to the image stray light background is:

[0067] (7)

[0068] For the original two-dimensional image, based on the stray light background coordinate matrix Z3, the first stray light background image under the original image coordinates is obtained by interpolation. The first stray light corrected image is:

[0069] The first corrected image = original image - the first stray light background image (8)

[0070] Step 3): For the first corrected image obtained in step 2), Figure 2 As shown in FIG. 1 , the influence of stray light on the observed image data is corrected in the scanning direction.

[0071] 3.1): Use the threshold method to obtain the number of rows of data in the scanning direction that is greater than The coordinate matrix x0 corresponding to the object point is:

[0072] (9)

[0073] in, is the scanning direction coordinate x in the first corrected image i The corresponding corrected value;

[0074] 3.2): Based on formula (1) (3), the neighborhood smoothing filter method is used to The data is used to extract the coordinates of the object points for the second time. Greater than The coordinate position matrix x1 is:

[0075] (10)

[0076] Among them, g(x i ) is f(x i ) is the one-dimensional matrix value after deducting the smooth background, and the smoothing width is set to n3.

[0077] 3.3): Based on formula (5), the first-order gradient method is used to The data is extracted for the third time to obtain the coordinates of the object points. The coordinate position matrix x2 of the threshold:

[0078] (11)

[0079] 3.4): The coordinate matrices x0, x1 and x2 obtained by the above three methods are used to calculate the coordinate position matrix X corresponding to the object point using the union of the three:

[0080] (12)

[0081] Among them, the coordinates in x0, x1, x2, and X are all represented by 1 and 0, 1 represents the object point, and 0 represents the background point. The one-dimensional coordinate matrix calculated by all row data in the two-dimensional image is used to establish the two-dimensional coordinate matrix P corresponding to the object point information in the two-dimensional image;

[0082] 3.5): For the image two-dimensional coordinate matrix P, with the object point coordinate as the center, the object point coordinate is expanded in the flight dimension using formula (1), and the flight dimension smoothing width is set to m3 to obtain the two-dimensional coordinate matrix P1. Then, the object point coordinate is expanded in the scanning dimension using formula (1), and the scanning dimension smoothing width is set to n3 to obtain the two-dimensional coordinate matrix P2. The coordinate matrix P3 corresponding to the stray light background of the object point is:

[0083] (13)

[0084] 3.6): Based on the first corrected image and the stray light background coordinates P3, the second stray light background image under the image coordinates is obtained by interpolation. The two-dimensional image after stray light correction is:

[0085] The second corrected image = the first corrected image - the second stray light background image (14)

[0086] Step 4): The second corrected image of stray light obtained in step 3.6) is as follows: Figure 2 As shown in FIG. 1 , the imaging area of ​​the night shadow area is limited by using the solar zenith angle, and all values ​​with zenith angles greater than 105° are set to 0. The image obtained is the night city light image corrected for stray light.

[0087] In this embodiment, the target signal in the city night light image is composed of stray light, ground object moon reflection radiation and night light radiation. The stray light and ground object reflection background are used as low-frequency information. The ground image is removed by low-pass filtering method to remove the stray light information, and only the high-frequency information part such as the city night light target is retained. The processing result of the stray light correction method of the city night light image based on 45-degree mirror imaging is as follows: Figure 3 As shown, the processing results use the on-orbit imaging data of the low-light channel 10 elements and the 45-degree mirror scanning system.

[0088] Obviously, the above embodiments are only examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the protection scope of the invention.

Claims

1. A method for correcting stray light of a city night light image based on 45-degree mirror imaging, wherein the low-light scanning radiometer based on 45-degree mirror imaging for ground imaging adopts a 45° scanning mirror and a K mirror derotation technology, and uses N detectors to scan across tracks to obtain nighttime city light images. The number of sampling pixels in the scanning direction is M, and one scan can obtain an N×M two-dimensional image. In the flight direction, the K mirror will have an AB surface effect, and the two reflective mirror surfaces of the K mirror, namely the A surface and the B surface, are alternately collected to form a ground scanning image, characterized in that The correction method comprises the following steps: Step 1): Perform single-track image synthesis on multiple Earth observation images of the low-light scanning radiometer, and use the solar zenith angle data corresponding to each imaging position to calculate the solar zenith angle greater than The coordinate position at that time is set to 0, and the image is decomposed into a nighttime city light image in a single-pixel single-scan state according to the imaging pixel and the A / B plane; Step 2): According to the single-pixel single-scan state image obtained in step 1), the stray light correction processing of the flight direction image data is carried out. The specific steps are as follows: 2.1): Use the neighborhood smoothing filtering method to perform low-pass filtering on a single column of image data, and obtain a threshold greater than The one-dimensional coordinate matrix y1 of the object point at time: (1) (2) (3) Among them, y i is the coordinate position of a single column of image data, f(y i ) is the image single column data coordinate y i The value at is f(y i ) when the smoothing width is m1, g(y i ) is f(y i ) is the value after deducting the smoothing value, W is the number of values ​​in a single column of image data, y1 is represented by a binary one-dimensional matrix, 1 represents object point information, and 0 represents background information; object points include night lights; 2.2): For the same single column data as above, the first-order gradient method is used to extract the object point information, and the threshold is greater than The one-dimensional coordinate matrix y2 of the object point: (4) (5) Among them, h(y i ) is the coordinate y i and i-1 The difference in response values ​​between adjacent positions, y2, is represented by a binary one-dimensional matrix, where 1 represents object point information and 0 represents background information; 2.3): Based on the neighborhood smoothing filter method and the first-order gradient method, the coordinate matrices y1 and y2 are obtained, and the one-dimensional coordinate matrix Y of the object point is obtained by using the union of the two coordinate matrices: (6) Among them, Y is represented by a binary number, 1 represents object point information, and 0 represents background information. The two-dimensional coordinate matrix corresponding to the object point information in the two-dimensional image is established through the one-dimensional coordinate matrix corresponding to all column data in the two-dimensional image, and the coordinate matrices of all single-pixel single-scan state images are synthesized into a two-dimensional coordinate matrix Z corresponding to the original image according to the original position of the pixel; 2.4): For the image two-dimensional coordinate matrix Z, with the object point coordinates as the center, the object point coordinates are expanded in the flight dimension using formula (1), and the flight dimension smoothing width is set to m2 to obtain the two-dimensional coordinate matrix Z1. Then, the object point coordinates are expanded in the scanning dimension using formula (1), and the scanning dimension smoothing width is set to n2 to obtain the two-dimensional coordinate matrix Z2. The coordinate matrix Z3 corresponding to the image stray light background is: (7) For the original two-dimensional image, based on the stray light background coordinate matrix Z3, the first stray light background image under the original image coordinates is obtained by interpolation. The first stray light corrected image is: The first corrected image = original image - the first stray light background image; (8) Step 3): For the first corrected image obtained in step 2), the influence of stray light on the observed image data is corrected in the scanning direction. The specific steps are as follows: 3.1): Use the threshold method to obtain the number of rows of data in the scanning direction that is greater than The coordinate matrix x0 corresponding to the object point is: (9) in, is the scanning direction coordinate x in the first corrected image i The corresponding corrected value; 3.2): Based on formula (1) (3), the neighborhood smoothing filter method is used to The data is used to extract the coordinates of the object points for the second time. Greater than The coordinate position matrix x1 is: (10) Among them, g(x i ) is f(x i ) is the one-dimensional matrix value after deducting the smooth background, and the smoothing width is set to n3; 3.3): Based on formula (5), the first-order gradient method is used to The data is extracted for the third time to obtain the coordinates of the object points. The coordinate position matrix x2 of the threshold: (11) 3.4): The coordinate matrices x0, x1 and x2 obtained by the above three methods are used to calculate the coordinate position matrix X corresponding to the object point using the union of the three: (12) in, The coordinates in x0, x1, x2, and X are all represented by 1 and 0, 1 represents the object point, and 0 represents the background point. The one-dimensional coordinate matrix calculated by all the row data in the two-dimensional image is used to establish the two-dimensional coordinate matrix P corresponding to the object point information in the two-dimensional image; 3.5): For the image two-dimensional coordinate matrix P, with the object point coordinate as the center, the object point coordinate is expanded in the flight dimension using formula (1), and the flight dimension smoothing width is set to m3 to obtain the two-dimensional coordinate matrix P1. Then, the object point coordinate is expanded in the scanning dimension using formula (1), and the scanning dimension smoothing width is set to n3 to obtain the two-dimensional coordinate matrix P2. The coordinate matrix P3 corresponding to the stray light background of the object point is, ； (13) 3.6): Based on the first corrected image and the stray light background coordinates P3, the second stray light background image under the image coordinates is obtained by interpolation. The two-dimensional image after stray light correction is: The second corrected image = the first corrected image - the second stray light background image; (14) Step 4): Based on the second stray light correction image obtained in step 3.6), the imaging area of ​​the night shadow area is limited by using the solar zenith angle. The values ​​of are all set to 0, and the image obtained is the night city light image corrected for stray light.

Images