A star sub-pixel centroid extraction method for discrete plane array remote sensing instrument

Abstract

The application discloses a star sub-pixel centroid extraction method for discrete plane array remote sensing instruments, and comprises the following steps: data simulation to obtain a simulation image; discrete plane array remote sensing instrument star observation image processing and splicing; star trajectory detection; star motion trajectory fitting to obtain east-west and south-north motion functions of star motion; and star sub-pixel centroid position extraction. The application can more accurately obtain the position of star centroid, and can obtain good star sub-pixel centroid position extraction effect for star images of different qualities and different star brightnesses.

Description

Technical Field

[0001] The invention relates to a star sub-pixel centroid extraction method for a discrete array remote sensing instrument, and belongs to the technical field of satellite remote sensing. Background Art

[0002] High-precision navigation, positioning, and registration for three-axis stabilized geostationary orbit satellites rely on the positions of stellar centroids observed by in-orbit remote sensing instruments. During in-orbit observations, satellite remote sensing instruments place high demands on instrument pointing accuracy. Star positions are crucial information for confirming instrument pointing, and their positions relative to the observer can be determined by observing a region of the sky with the instrument. Furthermore, instrument pointing can be corrected based on the deviation between the observed and theoretical positions of stars. Satellite remote sensing instruments can be used to observe regions of space and obtain continuous stellar observation images. High-precision sub-pixel stellar centroid positions can be used to determine the error between actual and theoretical positions. Therefore, research on obtaining high-precision sub-pixel stellar centroid positions has extremely high application value.

[0003] Traditional stellar centroid extraction methods are primarily designed for star sensors, which are more sensitive to stars. Stars are often unfocused, with a single star occupying multiple pixels, making them suitable for stellar centroid extraction. However, Earth observation remote sensing instruments, such as those on the US GOES satellite and my country's Fengyun-4 satellite, often use a focused observation mode, with a single star occupying only a single pixel. This makes it impossible to directly use traditional stellar centroid extraction methods to obtain sub-pixel positions of stellar centroids.

[0004] Zhang et al. investigated algorithms for extracting stellar centroids from images of Earth observation remote sensing instruments. They discussed stellar centroid extraction methods for linear arrays with 8×4 and 32×4 imagers, as well as for surface arrays with 256×330 imagers. The detectors in these devices were densely packed, resulting in continuous imaging between pixels. However, in some devices, the detectors are distributed discretely. Stellar centroid extraction methods for dense planar arrays are not applicable to discrete optical detectors, making linear array-based methods more valuable. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a star sub-pixel centroid extraction method for discrete array remote sensing instruments.

[0006] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0007] A method for extracting sub-pixel centroids of stars for discrete array remote sensing instruments comprises the following steps:

[0008] S1: Simulate the star observation data of the discrete array remote sensing instrument to obtain a simulated image;

[0009] S2: Processing and stitching simulated images of stars observed by discrete array remote sensing instruments;

[0010] S3: star trajectory detection;

[0011] S4: Fitting the star motion trajectory to obtain the east-west and north-south motion functions of the star motion;

[0012] S5: Extract the sub-pixel centroid position of the star.

[0013] Preferably, step S2 includes the following sub-steps:

[0014] S21: prune the mean of each pixel in the simulation image and remove fixed noise;

[0015] S22: stitching the simulation images according to the time series.

[0016] Preferably, the specific steps of processing and splicing the simulated images include:

[0017] S221: Flatten each frame of the simulation image, rearrange the pixels of the image, and transform the image with a length of H and a width of W into an image with a length of H×W and a width of 1;

[0018] S222: The flattened images are stitched together according to the time sequence.

[0019] Preferably, step S3 includes the following sub-steps:

[0020] S31: Screening foreground, for each pixel, the pixel with a value higher than the average value of all frames in the time series is considered as the foreground;

[0021] S32: Determine the star track area, calculate the number of foregrounds in each area by sliding window, and take the single area with the highest foreground number as the star track detection result.

[0022] Preferably, the specific steps for obtaining the east-west motion function of the star motion are:

[0023] S411: Establish the function of the east-west coordinate x and time t of the star motion;

[0024] S412: Fitting the function obtained in step S411 with a Gaussian function to obtain the moment when the star passes through the center of mass of the imager;

[0025] S413: further fitting the result obtained in step S412 to obtain the east-west motion function of the star;

[0026] Wherein, in step S412, the peak point of the fitted Gaussian function is the moment when the star passes through the center of mass of the imager.

[0027] Preferably, the east-west motion function of the star motion is expressed as:

[0028] x=b x +k x t

[0029] Among them, k x is the slope of the east-west motion obtained by fitting, b x is the intercept of the east-west motion obtained by fitting, and x is the sub-pixel coordinate of the star in the east-west direction on the phase plane when the star is at time t.

[0030] Preferably, the specific steps for obtaining the north-south motion function of the star motion are:

[0031] S421: Establish the function of the north-south coordinate y and time t of the star motion;

[0032] S422: Calculate the position of each frame in the star simulation by the centroid method, calculate the sub-pixel centroid position of the star, fit the function obtained in step S421, and obtain the north-south motion function of the star.

[0033] Preferably, the step S422 performs fitting three times; after the first and second fittings are completed, 10% of the image frames with the largest deviation from the result are removed.

[0034] Preferably, the north-south motion function of the star motion is expressed as:

[0035] y=b y +k y t

[0036] Among them, k y is the slope of the north-south movement obtained by fitting, b y is the intercept of the north-south motion obtained by fitting, and y is the sub-pixel coordinate of the star in the north-south direction on the phase plane when the star is at time t.

[0037] Preferably, in step S5, given any time t, the sub-pixel centroid position of the star at the target time is obtained, and the calculation formula is as follows:

[0038] (x(t),y(t))=(b x +k x t,b y +k y t).

[0039] Compared with the existing technology, the present invention can obtain the position of the star's center of mass more accurately, and can achieve good star sub-pixel center of mass position extraction effect for star images of different masses and different star brightness. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A flowchart of a method for extracting sub-pixel centroids of stars for discrete array remote sensing instruments provided by the present invention;

[0041] Figure 2 A schematic diagram of a detector array in an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of an east-west fitting curve in an embodiment of the present invention;

[0043] Figure 4 Schematic diagram of the north-south fitting results in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The technical content of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown, an embodiment of the present invention provides a method for extracting sub-pixel centroids of stars for a discrete array remote sensing instrument, comprising the following steps:

[0046] S1: Simulate the star observation data of the discrete array remote sensing instrument to obtain a simulated image.

[0047] Reliable star observation images are obtained by simulating the actual imaging conditions of satellite remote sensing instruments.

[0048] like Figure 2 As shown, the specific setting method of the simulation scene is:

[0049] The size of the detector array is set to 8×8, the distance between each detector is the same as the size of the detector (this parameter can be set flexibly), and the observation field of view of the satellite remote sensing instrument is set to 14μrad.

[0050] During the simulation, it is assumed that only one star passes through the detector array during the star observation period, and that the star passes through the detector array at a uniform speed and horizontally.

[0051] It should be noted that in order to simulate the installation bias of satellite remote sensing instruments, a random tilt in the range of (-3, 3) needs to be added to the trajectory of the star.

[0052] S2: Process and stitch the simulated images of star observations by discrete array remote sensing instruments.

[0053] Among them, the specific steps of simulation image processing and splicing are:

[0054] S21: Delete the mean value of each pixel in the simulation image and remove fixed noise.

[0055] Since random noise has the characteristic of a mean of 0 and there are enough frames in the sequence image, the noise of each pixel is removed by subtracting the mean of each pixel in the time series, and the influence of fixed noise is also removed.

[0056] S22: stitching the simulation images according to the time series.

[0057] In discrete array remote sensing instruments, there are large gaps between pixels. While the brightness of a single pixel changes continuously as a star passes through it, the brightness between two laterally adjacent pixels does not. In other words, a star does not immediately reach the area of ​​the next pixel after leaving the previous one. Therefore, processing methods similar to those used for dense arrays are inappropriate.

[0058] Specifically, in the embodiment of the present invention, the following method is used to achieve splicing:

[0059] S221: Flatten each frame of the simulation image and rearrange the image pixels, transforming the image with a length of H and a width of W into an image with a length of H×W and a width of 1, where the first column of images is from the 1st to the Hth pixels, the second column of images is from the (H+1)th to the 2Hth pixels, and so on. For the simulation case in the use case, each frame of the image is expanded into a 64-dimensional image vector, where the 1st to 8th dimensions correspond to the first column of the original image, the 9th to 16th dimensions correspond to the second column of the original image, and so on.

[0060] S222: The flattened images are stitched together in time sequence. In the stitched image, since the process of a star passing through a single pixel is continuous, the process of a star passing through a pixel will form a straight line on the stitched image.

[0061] S3: Star trajectory detection.

[0062] The specific detection methods are:

[0063] S31: Screening foreground, for each pixel, the pixel with a value higher than the average value of all frames in the time series is considered to be the foreground.

[0064] The purpose of foreground screening is to find pixels that are suspected to be stars. Since stars may be relatively low in brightness, the response in the star trajectory is darker and easily mixed with random noise in the image, which has a series of impacts on the star detection process. Therefore, the present invention chooses to lower the foreground screening threshold here to increase the probability of detecting stars.

[0065] S32: Determine the star track area, calculate the number of foregrounds in each area by sliding window, and take the single area with the highest foreground number as the star track detection result.

[0066] Also consider the case of low-brightness stars. This situation may cause discontinuous star trails, but the number of foregrounds in the star region must be greater than that in the region without stars. Therefore, star trails are detected by this method.

[0067] Since it is assumed that only one star passes through the sensor during step S1 , the single region with the highest number of foreground stars is the star trail detection result.

[0068] S4: Fit the star motion trajectory to obtain the east-west and north-south motion functions of the star.

[0069] In one embodiment of the present invention, the specific steps of obtaining the east-west motion function of a star are as follows:

[0070] S411: Establish the function of the east-west coordinate x and time t of the star motion.

[0071] Because the east-west motion of a star is continuous and the point spread function (PSF) of a star is a Gaussian function, a function of the east-west coordinate x and time t can be established.

[0072] S412: Fit the function obtained in step S411 by a Gaussian function to obtain the moment when the star passes through the center of mass of the imager.

[0073] like Figure 3 As shown, the function obtained in step S411 is fitted with a Gaussian function to obtain the moment when the star's centroid passes through the center of each column of pixels. The peak point of the fitted Gaussian function is the moment when the star passes through the imager's centroid.

[0074] S413: Further fitting is performed on the result obtained in step S412 to obtain the east-west motion function of the star motion.

[0075] The time t when the star passes through the center point of each column of pixels is calculated according to step S412 mid , and the east-west coordinate x of the pixel center point mid is also known, so according to xmid and t mid The eastward motion function of the star motion is obtained by fitting, and the expression is as follows:

[0076] x = b x +k x t

[0077] wherein k x is the slope of the eastward motion obtained by fitting, b x is the intercept of the eastward motion obtained by fitting, and x is the eastward sub-pixel coordinate of the star on the phase plane at time t.

[0078] In an embodiment of the present application, the specific steps for obtaining the north-south motion function of the star motion are as follows:

[0079] S421: Establishing the function of the north-south coordinate y of the star motion and time t.

[0080] S422: Calculating the position of each frame in the star simulation by the centroid method, calculating the sub-pixel centroid position of the star, and fitting the function obtained in step S421 to obtain the north-south motion function of the star motion.

[0081] The north-south trajectory fitting is to obtain the function of the north-south coordinate and time, to calculate the position of each frame in the star by the centroid method, and to calculate the sub-pixel centroid position of the star, thereby completing the fitting of the function. Because the star motion is approximately linear motion, a straight line is selected for the function of the north-south coordinate and time.

[0082] In order to further reduce the influence of noise, three times of fitting are required. After the first and second fitting are completed, 10% of the image frames with the largest deviation from the result are removed. The purpose is to obtain a higher-precision motion function between the north-south coordinate of the star and time.

[0083] As shown in FIG. 4, the gray dots in the figure represent the sub-pixel centroid positions of the star in the north-south direction at different times, the light gray straight line is the north-south coordinate-time function of the star obtained by the fitting method, and the black straight line is the true motion trajectory of the star. Figure 4

[0084] After three times of fitting, the north-south motion function of the star motion is obtained, and the expression is as follows:

[0085] y = b y +k y t

[0086] wherein k y is the slope of the north-south motion obtained by fitting, b y ​is the intercept of the north-south motion obtained by fitting, and y is the sub-pixel coordinate of the star in the north-south direction on the phase plane when the star is at time t.

[0087] S5: Extract the sub-pixel centroid position of the star.

[0088] By obtaining the east-west and north-south coordinates of the star's motion as a function of time, the east-west and north-south coordinates of the star at each moment are calculated, and based on the east-west and north-south coordinates, the sub-pixel center of mass position of the star in each frame is predicted.

[0089] Given any time t, the sub-pixel centroid position of the star at the target time is obtained, and the calculation formula is as follows:

[0090] (x(t),y(t))=(b x +k x t,b y +k y t).

[0091] It should be noted that the star sub-pixel centroid extraction method provided by the present invention has been verified by experiments with different point spread function radii, different noise standard deviation (noise_std) intensities, and different star magnitude brightnesses. The average centroid error is used to quantitatively describe the accuracy of the star centroid algorithm. The average centroid error includes the x-direction error, the y-direction error, and the average distance error, which respectively measure the east-west distance, north-south distance, and Euclidean distance between the predicted position and the actual position.

[0092] Among them, the lower the average centroid error, the higher the extraction accuracy of the method.

[0093] As shown in the following table, the present invention sets different point spread function radii:

[0094] PSF radius Average x-direction error Average y-direction error Average distance error 0.5 0.025 0.384 0.385 1 0.036 0.257 0.260 1.5 0.045 0.233 0.237

[0095] It can be seen from the above table that under different point spread function radii, the star sub-pixel centroid extraction method provided by the present invention can achieve good results, and as the radius increases, the star centroid positioning accuracy increases.

[0096] As shown in the following table, the present invention sets different noise standard deviation intensities:

[0097] Noise standard deviation Average x-direction error Average y-direction error Average distance error 0 0.032 0.153 0.156 10 0.042 0.221 0.225 20 0.058 0.307 0.312

[0098] It can be seen from the above table that the star sub-pixel centroid extraction method provided by the present invention can achieve good results for star images of different masses.

[0099] As shown in the following table, the present invention sets different magnitude brightness:

[0100] Magnitude Average x-direction error Average y-direction error Mean square error 7 0.057 0.364 0.368 6 0.043 0.312 0.315 5 0.042 0.308 0.311 4 0.021 0.167 0.168

[0101] It can be seen from the above table that the star sub-pixel centroid extraction method provided by the present invention can adapt to different star brightness and obtain effective star centroid positions.

[0102] It should be noted that the above embodiments are only examples, and the technical solutions of the various embodiments can be combined and are all within the protection scope of the present invention.

[0103] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0104] The above describes in detail the method for extracting stellar sub-pixel centroids for discrete array remote sensing instruments provided by the present invention. However, any obvious modification to this method without departing from the essence of the present invention would constitute an infringement of the present invention's patent rights and would incur corresponding legal liability.

Claims

1. A method for extracting sub-pixel centroids of stars for discrete array remote sensing instruments, characterized in that: include: S1: Simulate the star observation data of discrete array remote sensing instruments to obtain simulated images; S2: Processing and stitching simulated images of stars observed by discrete array remote sensing instruments; S3: star trajectory detection; S4: Fitting the star motion trajectory to obtain the east-west and north-south motion functions of the star motion; S5: Extract the sub-pixel centroid position of the star; The specific steps of obtaining the east-west motion function of the star motion are: S411: Establish the function of the east-west coordinate x and time t of the star motion; S412: Fitting the function obtained in step S411 with a Gaussian function to obtain the moment when the star passes through the center of mass of the imager; S413: further fitting the result obtained in step S412 to obtain the east-west motion function of the star; Wherein, in step S412, the peak point of the fitted Gaussian function is the moment when the star passes through the center of mass of the imager; The east-west motion function of the star motion is expressed as: in, is the slope of the east-west movement obtained by fitting, is the intercept of the east-west motion obtained by fitting, For the star at the moment Its sub-pixel coordinates in the east-west direction on the phase plane; The specific steps of obtaining the north-south motion function of the star motion are: S421: Establish the function of the north-south coordinate y and time t of the star motion; S422: Calculate the position of each frame in the star simulation using the centroid method, calculate the sub-pixel centroid position of the star, and fit the function obtained in step S421 to obtain the north-south motion function of the star; The north-south motion function of the star motion is expressed as: in, is the slope of the north-south movement obtained by fitting, is the intercept of the north-south motion obtained by fitting, For the star at the moment Its sub-pixel coordinates in the north-south direction on the phase plane.

2. The method according to claim 1, wherein The step S2 further includes: S21: prune the mean of each pixel in the simulation image and remove fixed noise; S22: stitching the simulation images according to the time series.

3. The method according to claim 2, wherein The specific steps of the process of processing and splicing the simulation images include: S221: Flatten each frame of the simulation image, rearrange the pixels of the image, and transform the image with a length of H and a width of W into an image with a length of H×W and a width of 1; S222: The flattened images are stitched together according to the time sequence.

4. The method according to claim 1, wherein The step S3 further includes: S31: Screening foreground, for each pixel, the pixel with a value higher than the average value of all frames in the time series is considered as the foreground; S32: Determine the star track area, calculate the number of foregrounds in each area by sliding window, and take the single area with the highest foreground number as the star track detection result.

5. The method according to claim 1, wherein In step S422 , three fittings are performed; after the first and second fittings are completed, 10% of the image frames with the largest deviation from the result are removed.

6. The method according to claim 1, wherein In step S5, given any time t, the sub-pixel centroid position of the star at the target time is obtained, and the calculation formula is as follows: 。

Images