A method for extracting photometric data of spatial targets

Abstract

This invention relates to a method for extracting photometric data of space targets, addressing the problem of numerous noise points and singularities in photometric data extraction due to random interference. The invention employs a method in the two-dimensional spatial dimension of the image to eliminate the influence of background on the image in the frequency domain and to eliminate the influence of background sequences on light scattering intensity sequences in the frequency domain. Specifically, it includes five steps: image preparation, image region selection, two-dimensional image frequency domain background noise elimination, one-dimensional light intensity frequency domain background noise elimination, and magnitude data calculation. This invention effectively suppresses random noise and achieves high-precision extraction of target magnitude data, possessing significant theoretical and practical value for the acquisition and application of magnitude data, thus providing technical support for research on space target detection and identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of space target photometric feature measurement and inversion, and particularly relates to a space target photometric data extraction method. BACKGROUND

[0002] With the development of space technology, the number of artificial space targets is increasing year by year, and how to use the multi-dimensional characteristics of the target to realize the operation and maintenance and safety management of space assets has become a hot research direction. The photometric feature is one of the important physical characteristics of the space target, which can reflect the size, shape, category, surface material, attitude and running state of the target to a certain extent, and can be used for auxiliary identification of space targets. Using a camera to shoot a space target and performing photometric inversion calculation according to the obtained image is the most effective and widely used method to obtain the photometric characteristics of the space target.

[0003] At present, in the traditional calculation of target photometric information extraction method, the image area of the target is subtracted from the background area, and then the sum of the image difference value is used as the light scattering intensity of the target. The intensity ratio of the target to the standard reference star in the visible light band can be used to calculate the brightness magnitude value of the target. When the ground-based measurement equipment measures the photometric data of the artificial target, the photometric data of the target is weak and the distance is far, and the light scattering intensity extraction calculation is greatly affected by random interference on the imaging surface, resulting in many noise points and singular points of the photometric data. SUMMARY

[0004] Therefore, the present application provides a space target photometric data extraction method, which can effectively suppress random noise and realize high-precision target magnitude data extraction, has important theoretical significance and practical value for the acquisition and application of magnitude data, and further provides technical support for space target detection and identification research.

[0005] Specifically includes the following contents:

[0006] In the two-dimensional space dimension of the image, the method of eliminating the influence of the background on the image in the frequency domain and eliminating the influence of the background sequence on the light scattering intensity sequence in the frequency domain is adopted, which specifically includes five steps of image preparation, image region selection, two-dimensional image frequency domain background noise elimination, one-dimensional light intensity frequency domain background noise elimination and magnitude data solving.

[0007] The image region selection further includes: extracting optical image data frame by frame, selecting three regions in each image, which are target region, background reference region 1 and background reference region 2, the areas of the three regions are consistent, and the target region contains the space target image to be extracted.

[0008] The two-dimensional image frequency domain background noise elimination further includes:

[0009] Firstly, do two-dimensional Fourier transform for the target region, background reference region 1 and background reference region 2 of each frame respectively; get the two-dimensional amplitude distribution results and phase value distribution results in the Fourier transform results of the target region and background reference region 1; similarly get the two-dimensional amplitude distribution results in the Fourier transform results of background reference region 2;

[0010] Then, subtract the two-dimensional amplitude distribution results of reference region 2 from the two-dimensional amplitude distribution results of the target region to get the two-dimensional distribution results of the amplitude residual intensity of the target region; subtract the two-dimensional amplitude distribution results of reference region 2 from the two-dimensional amplitude results of reference region 1 to get the two-dimensional distribution results of the amplitude residual intensity of reference region 1;

[0011] Finally, multiply the two-dimensional distribution results of the amplitude residual intensity of the target region by the phase value distribution results of the target region, and sequentially perform inverse Fourier transform and amplitude calculation to get the two-dimensional distribution results of the denoised intensity value of the target region; multiply the two-dimensional distribution results of the amplitude residual intensity of reference region 1 by the phase value distribution results of background reference region 1, and sequentially perform inverse Fourier transform and amplitude calculation to get the two-dimensional distribution results of the denoised intensity value of background reference region 1.

[0012] The one-dimensional light intensity frequency domain background noise elimination further comprises:

[0013] Firstly, count the number of actual target covering pixels in the two-dimensional distribution results of the denoised intensity value of the target region, and calculate the sum of the denoised intensity values corresponding to these pixels, that is, get the light scattering intensity value of the target; randomly extract the same number of pixels as the number of actual target covering pixels in the two-dimensional distribution results of the denoised intensity value of background reference region 1, and calculate the sum of the denoised intensity values corresponding to these pixels, that is, get the light scattering intensity value of reference region 1;

[0014] Then, according to the light scattering intensity value of the target of each frame of optical image and the time corresponding to each frame of optical image, get the preliminary curve of the light scattering intensity value of the target versus time; according to the light scattering intensity value of reference region 1 of each frame of optical image and the time corresponding to each frame of optical image, get the preliminary curve of the light scattering intensity value of reference region 1 versus time;

[0015] Finally, do Fourier transform on the preliminary curve of the light scattering intensity value of the target and separate the amplitude and phase values, do Fourier transform on the preliminary curve of the light scattering intensity value of reference region 1 and separate the amplitude and phase values, then subtract the Fourier amplitude of the preliminary curve of the light scattering intensity value of reference region 1 from the Fourier amplitude of the preliminary curve of the light scattering intensity value of the target, multiply the result by the phase value of the preliminary curve of the light scattering intensity value of the target, and then perform inverse Fourier transform and amplitude calculation to get the final light scattering intensity curve E 目标 (t) of the target.

[0016] The star magnitude data solving further comprises:

[0017] The target final light scattering intensity curve E 目标 (t) is obtained by comparing the standard type luminosity with the equal brightness curve, i.e. the spatial target luminosity data.

[0018] The spatial target luminosity curve M 目标 (t) is represented as follows:

[0019] M 目标 (t) = M 恒 -2.5*log(E 目标 (t) / E 恒 )

[0020] Wherein, M 恒 represents the luminosity value of the reference star, and E 恒 represents the brightness value of the reference star.

[0021] Beneficial effects:

[0022] 1) The present application is a method for eliminating background noise in the frequency domain. In the process of eliminating background noise, secondary noise is not increased due to inconsistent noise distribution, the accuracy of data is improved without compromising data confidence.

[0023] 2) The present application adopts the method of eliminating the influence of background on image in the frequency domain and eliminating the influence of background sequence on light scattering intensity sequence in the frequency domain under the two-dimensional spatial dimension of image, which greatly improves the anti-interference ability in the process of luminosity data extraction.

[0024] 3) The present application has important theoretical significance and practical value for the acquisition and application of spatial target star magnitude data, and further provides technical support for spatial target detection and identification research. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The flow chart of star magnitude data extraction;

[0026] Figure 2 The schematic diagram of luminosity image data and the schematic diagram of target and background area selection position;

[0027] Fig. 3(a) is a two-dimensional distribution diagram of denoised intensity value of target area;

[0028] Fig. 3(b) is a two-dimensional distribution diagram of denoised intensity value of reference area 1;

[0029] Figure 4 The luminosity curve processing effect diagram of eliminating the influence of background in one-dimensional frequency domain; DETAILED DESCRIPTION

[0030] The application will be described in detail below with reference to the drawings and examples.

[0031] The application provides a method for extracting space target luminosity data, and the method flow is shown in Figure 1 The method specifically comprises five steps of image preparation, image area selection, two-dimensional image frequency domain background noise elimination, one-dimensional light intensity frequency domain background noise elimination and star magnitude data solving. Since some steps are further subdivided into multiple steps, the following will be described according to the detailed process:

[0032] Step 1, preparing data, arranging all luminosity images obtained by tracking measurement according to time, and corresponding to the orbit elements of the target:

[0033] Step 2, extracting optical image data frame by frame, selecting three areas in each frame of image, which are target area, background reference area 1 and background reference area 2, the three areas have the same size, and the target area contains the space target image to be extracted;

[0034] Step 3, performing two-dimensional Fourier transform on the target area, the background reference area 1 and the background reference area 2 of each frame respectively; obtaining the two-dimensional amplitude distribution result and the phase value distribution result in the Fourier transform result of the target area and the background reference area 1; and obtaining the two-dimensional amplitude distribution result in the Fourier transform result of the background reference area 2;

[0035] Step 4, subtracting the two-dimensional amplitude distribution result of the reference area 2 from the two-dimensional amplitude distribution result of the target area to obtain the two-dimensional distribution result of the residual intensity of the target area amplitude; subtracting the two-dimensional amplitude distribution result of the reference area 2 from the two-dimensional amplitude result of the reference area 1 to obtain the two-dimensional distribution result of the residual intensity of the reference area 1 amplitude;

[0036] Step 5, multiplying the two-dimensional distribution result of the residual intensity of the target area amplitude obtained in step 4 by the phase value distribution result of the target area obtained in step 3, and sequentially performing inverse Fourier transform and amplitude value operation to obtain the two-dimensional distribution result of the denoised intensity value of the target area; multiplying the two-dimensional distribution result of the residual intensity of the reference area 1 amplitude obtained in step 4 by the phase value distribution result of the background reference area 1 obtained in step 3, and sequentially performing inverse Fourier transform and amplitude value operation to obtain the two-dimensional distribution result of the denoised intensity value of the background reference area 1;

[0037] Step 6, counting the number of actual covered pixels of the target from the two-dimensional distribution result of the denoised intensity value of the target area, and calculating the sum of the denoised intensity values corresponding to the pixels, that is, obtaining the light scattering intensity value of the target; randomly extracting the same number of pixels as the number of actual covered pixels of the target from the two-dimensional distribution result of the denoised intensity value of the background reference area 1, and calculating the sum of the denoised intensity values corresponding to the pixels, that is, obtaining the light scattering intensity value of the reference area 1;

[0038] Step 7: Based on the target light scattering intensity value of each optical image frame and the corresponding time of each optical image frame, obtain a preliminary curve of the target light scattering intensity value versus time; based on the light scattering intensity value of reference region 1 of each photometric image frame and the corresponding time of each photometric image frame, obtain a preliminary curve of the light scattering intensity value of reference region 1 versus time.

[0039] Step 8: Perform a Fourier transform on the preliminary light scattering intensity curve of the target and separate the amplitude and phase values. Perform a Fourier transform on the preliminary light scattering intensity curve of reference region 1 and separate the amplitude and phase values. Then, subtract the Fourier amplitude of the preliminary light scattering intensity curve of reference region 1 from the Fourier amplitude of the preliminary light scattering intensity curve of the target. Multiply the result by the phase value of the preliminary light scattering intensity curve of the target, and then perform an inverse Fourier transform and amplitude calculation to obtain the final light scattering intensity curve of the target.

[0040] Step 9: Obtain the equal brightness curve, i.e., the spatial target luminance data, by comparing the final light scattering intensity curve of the target with the standard luminance curve.

[0041] Furthermore, in step 2, the target region selection method is as follows: In each frame of optical image, the approximate position of the target within the field of view is obtained based on the target's trajectory and tracking status. The target image is in the shape of a diffraction spot, and the maximum value P of the diffraction spot pixel amplitude is calculated. 斑 and the average pixel amplitude Av in the background area 背 According to P 斑 and Av 背 The value determines the segmentation threshold D between the target and the background. r Generally speaking, for images with a high signal-to-noise ratio, formula D can be used. r =0.25P 斑 +0.75Av 背 The minimum bounding rectangle of the target in each frame of the photometric image is obtained by finding the connected components. The target region image I is then delineated with the point of strongest target light spot as the center and dimensions at least five times the length and width of the minimum bounding rectangle. 目标r (M, N), where M represents the number of pixels horizontally in the region, N represents the number of pixels vertically in the region, L represents the length of the target region, and H represents the width of the target region;

[0042] Furthermore, in step 2, the selection method for background reference region 1 and background reference region 2 is as follows: Based on the target's position in the field of view, regions with similar imaging conditions are selected as reference regions as much as possible. If the target is at the center of the field of view, the center points of reference region 1 and reference region 2 are selected in the region close to the target, and the coverage area I of the two regions is determined by length L and width H. 参考1r (M, N) and I 参考2r(M, N); If the target is located at the edge of the field of view, then select the center points of reference region 1 and reference region 2 at positions symmetrical to the target with respect to the horizontal or vertical central axis of the image, and determine the coverage area I of the two regions with length L and width H. 参考1r (M, N) and I 参考2r (M, N), the two regions do not overlap.

[0043] Figure 2 This is a schematic diagram of an image obtained from photometric measurements. When selecting the target region, the approximate position of the target within the field of view can be obtained based on the target's trajectory and tracking status. The target image is in the shape of a diffraction spot. In this embodiment, the maximum value P of the diffraction spot amplitude... 斑 =13200, average amplitude of background region Av 背 =8790, according to P 斑 and Av 背 The threshold for segmenting the target from the background is set to 9893. The minimum bounding rectangle of the target is obtained by finding the connected components. The target region image I is then delineated with the strongest point of the target's light spot as the center and its dimensions being five times the length and width of the minimum bounding rectangle. 目标r (M, N), the target area is 40 in length and 40 in width. Figure 2 The white solid line box in the middle represents the target area.

[0044] Depend on Figure 2 If the target is located at a relatively edge of the field of view, then the center point of reference region 1 is selected at a position symmetrical to the target with respect to the horizontal central axis of the image, and the center point of reference region 2 is selected at a position symmetrical to reference region 1 with respect to the vertical central axis of the image. The coverage area I of the two regions is determined with a length of 40 and a width of 40. 参考1r (M, N) and I 参考2r (M, N), such as Figure 2 The area enclosed by the midpoint line and the dashed line. The center points of reference areas 1 and 2 can be interchanged without affecting the implementation of the method.

[0045] Furthermore, in step 3, the Fast Fourier Transform algorithm is used for the Fourier transform of the three rectangular regions, wherein the two-dimensional Fourier transform of the target region is expressed as: F 目标r (U, V) = FFT2(I) 目标r (M, N)),

[0046] The formula for extracting the amplitude of the target region is, A 目标r (U, V) = ABS(F) 目标r (U, V)),

[0047] The phase extraction formula is P 目标r (U, V) = PHASE(F) 目标r (U, V))

[0048] Similarly, the amplitude A of reference region 1 is obtained. 参考1r (U, V), phase P 参考1r (U, V) and the amplitude A of reference region 2 参考2r (U, V);

[0049] FFT2 represents two-dimensional Fourier transform, ABS represents amplitude extraction, and PHASE represents phase extraction.

[0050] Furthermore, in step 5, the formula for the two-dimensional distribution of the denoising intensity values ​​in the target region is as follows:

[0051] I 目标s (M, N) = ABS(iFFT2((A 目标r (U, V)-A 参考2r (U, V)).*P 目标r (U, V))));

[0052] The formula for the two-dimensional distribution of denoising intensity values ​​in reference region 1 is as follows:

[0053] I 参考1s (M, N) = ABS(iFFT2((A 参考1r (U, V)-A 参考2r (U, V)).*P 参考1r (U, V))));

[0054] Where iFFT2 represents the inverse Fourier transform.

[0055] Get I 目标s (M, N) and I 参考1s The results (M, N) are shown in Figure 3(a) and (b).

[0056] Furthermore, step 6 specifically includes extracting the denoised image I. 目标s The maximum value P of the (M, N) diffraction spot amplitude 斑s The average value Av of the background region 背s According to P 斑s and Av 背s Set threshold D s Generally speaking, for images with a high signal-to-noise ratio, formula D can be used. s =0.25P 斑s +0.75Av 背s Calculate and select image I 目标s All values ​​greater than D in (M, N) s The value of T s 1 point, and this T s Summing the amplitude values ​​at each point yields the target scattered light intensity value at the corresponding moment in the image frame; similarly, in I... 参考1s T is randomly selected from (M, N)s a point, and sum the amplitude of the T s points, i.e. the light scattering intensity value of the reference region 1 at the corresponding time of the frame image is obtained;

[0057] Further, step 7 specifically comprises processing the sequence images frame by frame, i.e. obtaining the preliminary light scattering intensity value curve E 目标s (t) about time. Similarly, the sequence images are processed frame by frame, i.e. obtaining the preliminary light scattering intensity value curve E 背景s (t) of the reference region 1 about time;

[0058] Step 8, Fourier transform is performed on the preliminary light scattering intensity value curve of the target and the preliminary light scattering intensity value curve of the reference region 1, and the amplitude and phase values are separated, then the Fourier amplitude of the preliminary light scattering intensity value curve of the target is subtracted from the Fourier amplitude of the preliminary light scattering intensity value curve of the reference region 1, the result is multiplied by the phase value of the preliminary light scattering intensity value curve of the target, and inverse Fourier transform and amplitude calculation are performed to obtain the final light scattering intensity curve of the target;

[0059] Further, step 8 is described as follows. Fourier transform is performed on E 目标s (t), and since E 目标s (t) is a one-dimensional curve, one-dimensional Fourier transform is used to obtain the Fourier transformed result value T 目标s (v), and the amplitude and phase values A 目标s (v) and P 目标s (v) are separated. Similarly, the Fourier transformed amplitude A 背景s (v) of E 背景s (t) is obtained, and the final light scattering intensity curve of the target is obtained, and the specific formula is as follows:

[0060] E 目标 (t) = ABS(iFFT((A 目标s (v) - A 背景s (v)).*P 目标s (v)))) ;

[0061] Further, the final light curve of the target is obtained, and is expressed as follows M 目标 (t) = M 恒 - 2.5*log(E 目标 (t) / E 恒 ), wherein M 目标 (t) represents the light value of the target, M 恒 represents the light value of the reference star, and E 恒 represents the brightness value of the reference star.

[0062] In the embodiment, the denoised image I目标s Maximum value P of the diffraction spot amplitude in (M, N) 斑s = 3260, average value Av of the background region amplitude 背s = 23, according to P 斑s and Av 背s Set threshold value D s = 832, sum of all 10 points greater than D 目标s in the denoised image I s (M, N) is the target light scattering intensity value at the corresponding moment, and the light scattering intensity curve E 目标s (t) is obtained by processing the sequence images frame by frame. 参考1s Similarly, 10 points are randomly extracted in I 背景s (M, N) and summed to obtain the background light intensity value at the corresponding moment, and the light scattering intensity curve E 目标s (t) is obtained by processing the sequence images frame by frame. 目标s (t) is a one-dimensional curve, the one-dimensional Fourier transform is used to obtain the Fourier transform result value T 目标s (v), and the amplitude and phase values are separated to obtain A 目标s (v) and P 目标s (v), similarly, the Fourier transform of E 背景s (t) is performed to obtain the amplitude A 背景s (v) after Fourier transform, and the final light scattering intensity curve of the target is obtained, and the results are shown in Figure 4 , and the specific formula is:

[0063] E 目标 (t) = ABS(iFFT((A 目标s (v)-A 背景s (v)).*P 目标r (v))));

[0064] Finally, the formula M 目标 (t) = M 恒 -2.5*log(E 目标 (t) / E 恒 ) is used to obtain the magnitude of the target.

[0065] In summary, the above is only a preferred embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for extracting photometric data of a space object, characterized in that: In the two-dimensional image space dimension, the method for eliminating the influence of the background in the frequency domain on the image and the influence of the background sequence on the light scattering intensity sequence in the frequency domain is adopted, and the method specifically includes five steps of image preparation, image region selection, two-dimensional image frequency domain background noise elimination, one-dimensional light intensity frequency domain background noise elimination, and star magnitude data solving; The two-dimensional image frequency domain background noise elimination further includes: Firstly, two-dimensional Fourier transforms are performed on the target region, the background reference region 1 and the background reference region 2 of each frame respectively, two-dimensional amplitude distribution results and phase value distribution results in the Fourier transform results of the target region and the background reference region 1 are obtained, and two-dimensional amplitude distribution results in the Fourier transform results of the background reference region 2 are also obtained; Then, the two-dimensional amplitude distribution results of the target region are subtracted from the two-dimensional amplitude distribution results of the reference region 2 to obtain the two-dimensional distribution results of the residual intensity amplitude of the target region, and the two-dimensional amplitude results of the reference region 1 are subtracted from the two-dimensional amplitude distribution results of the reference region 2 to obtain the two-dimensional distribution results of the residual intensity amplitude of the reference region 1; Finally, the two-dimensional distribution results of the residual intensity amplitude of the target region are multiplied by the phase value distribution results of the target region, and inverse Fourier transform and amplitude value calculation are sequentially performed to obtain the two-dimensional distribution results of the denoised intensity value of the target region, and the two-dimensional distribution results of the residual intensity amplitude of the reference region 1 are multiplied by the phase value distribution results of the background reference region 1, and inverse Fourier transform and amplitude value calculation are sequentially performed to obtain the two-dimensional distribution results of the denoised intensity value of the background reference region 1; The one-dimensional light intensity frequency domain background noise elimination further includes: Firstly, the number of actual target covering pixels in the two-dimensional distribution results of the denoised intensity value of the target region is counted, and the sum of the denoised intensity values corresponding to the pixels is calculated, that is, the light scattering intensity value of the target is obtained, and the same number of pixels as the number of actual target covering pixels is randomly extracted from the two-dimensional distribution results of the denoised intensity value of the background reference region 1, and the sum of the denoised intensity values corresponding to the pixels is calculated, that is, the light scattering intensity value of the reference region 1 is obtained; Then, the preliminary curve of the target light scattering intensity value with respect to time is obtained according to the target light scattering intensity value of each optical image and the time corresponding to each optical image, and the preliminary curve of the reference region 1 light scattering intensity value with respect to time is obtained according to the reference region 1 light scattering intensity value of each optical image and the time corresponding to each optical image; Finally, the Fourier transform is performed on the preliminary light scattering intensity value curve of the target, and the amplitude and phase values are separated. The Fourier transform is performed on the preliminary light scattering intensity value curve of the reference region 1, and the amplitude and phase values are separated. The Fourier amplitude of the preliminary light scattering intensity value curve of the target is subtracted from the Fourier amplitude of the preliminary light scattering intensity value curve of the reference region 1. The result is multiplied by the phase value of the preliminary light scattering intensity value curve of the target. Then, the inverse Fourier transform is performed, and the amplitude is calculated to obtain the final light scattering intensity curve E of the target. 目标 (t).

2. The method of claim 1, wherein: The image region selection further includes: optical image data is extracted frame by frame, three regions are selected in each frame of image, which are the target region, the background reference region 1 and the background reference region 2, the areas of the three regions are consistent, and the target region contains the spatial target image to be extracted.

3. The method of claim 1, wherein: The star magnitude data solving further includes: Using the target final light scatter intensity curve E 目标 (t) to get the equal brightness curve, i.e. spatial target photometric data, in comparison with the standard photometric.

4. The method of claim 1 or 3, wherein: Space object luminosity curve M 目标 (t) is represented as follows: M 目标 (t) = M 恒 - 2.5 * log(E 目标 (t) / E 恒 ) where M 恒 represents the luminosity value of the reference star, E 恒 represents the brightness value of the reference star.

Images