Space target detection and self-calibration photometry method and system for space-borne star field imaging

Abstract

This invention discloses a method and system for space target detection and self-calibration photometry in spaceborne star field imaging. The method includes the following steps: acquiring the raw telemetry data stream from the satellite in spaceborne star field imaging mode, and reconstructing a two-dimensional raw image sequence; performing self-calibration radiometric correction, background noise modeling, multi-frame registration, source parameter extraction, and space target discrimination on the raw image sequence to obtain the corresponding stellar background, space targets, and single-frame artifacts; constructing a standard star candidate set and fitting a mapping model from standard star pixel coordinates to celestial coordinates; converting the pixel coordinates of the space target to the corresponding celestial coordinates; performing photometric zero-point fitting and space target photometry operations; and outputting structured space target observation results. This invention relates to the field of space target observation technology. By introducing multi-frame self-calibration radiometric correction, inter-frame geometric registration and multi-frame consistency discrimination, and robust photometric zero-point fitting, reliable observation of spaceborne space targets is achieved.

Description

Technical Field

[0001] This invention relates to the field of space target observation technology, and more specifically to a space target detection and self-calibration photometry method and system for spaceborne star field imaging. Background Technology

[0002] With increasingly frequent space activities, the number of low-Earth orbit and high-Earth orbit space targets (including on-orbit satellites, rocket final stages, debris, etc.) has increased dramatically, placing higher demands on space situational awareness (SSA) capabilities. Optical observation, due to its passive nature, high resolution, and ability to acquire target luminosity and shape characteristics, has become an important component of ground-based and space-based space target monitoring systems. In recent years, microsatellite platforms carrying staring optical payloads have been increasingly used for space target survey observation missions due to their low cost, flexible deployment, and ability to achieve global coverage. These missions typically operate in "starfield imaging" mode—that is, staring at a specific area of ​​the sky for an extended period, continuously acquiring two-dimensional image sequences containing stellar backgrounds and potential space targets.

[0003] However, space target observation in a spaceborne star field imaging environment faces a series of unique challenges, making it difficult to directly apply traditional ground-based astronomical surveys or dedicated space target monitoring methods. The problems include:

[0004] Calibration data missing or mismatched:

[0005] Ground-based telescopes can acquire dedicated dark frames and flat fields for radiometric correction before and after observations. However, spaceborne platforms are often limited by storage, downlink bandwidth, and mission planning, making it impossible to schedule calibration frames before and after each scientific observation. Even if historical calibration data is available, the detector response may drift with temperature, radiation dose, and time, leading to inaccurate radiometric correction and affecting the detection and photometric accuracy of small targets.

[0006] High false alarm rate and artifact interference:

[0007] Spaceborne CCD / CMOS detectors are susceptible to transient events such as cosmic rays, hot pixels, and single-event upsets, which can produce bright anomalies in single-frame images. Traditional single-frame threshold detection methods struggle to distinguish these artifacts from real space targets, resulting in a persistently high false alarm rate and severely impacting the efficiency of subsequent orbit processing.

[0008] Poor platform stability:

[0009] The attitude control precision of microsatellite platforms is limited, resulting in micro-vibrations and drifts, which lead to geometric distortions or displacements between multiple image frames. Without precise registration, cross-frame target association and trajectory discrimination will fail, making it impossible to effectively utilize temporal information to improve detection robustness.

[0010] High requirements for automation and autonomy:

[0011] Limited onboard computing resources and communication delays with the ground necessitate a high degree of autonomy in the observation and processing link. The entire process, from raw data to structured target products, must be completed without ground intervention, including star identification, astronomical positioning, photometric calibration, and quality assessment.

[0012] Therefore, there is an urgent need for an end-to-end, adaptive, and highly cohesive method for space target observation and processing, which can achieve reliable detection, accurate celestial positioning, robust absolute photometry, and controllable output quality throughout the entire process, even in the absence of dedicated calibration data and under conditions of platform jitter and transient interference. Summary of the Invention

[0013] In view of the above problems, this invention is proposed to provide a space target detection and self-calibration photometry method and system for spaceborne star field imaging that overcomes or at least partially solves the above problems. By introducing multi-frame self-calibration radiometric correction, unified background noise modeling, inter-frame geometric registration and multi-frame consistency discrimination, and robust photometric zero-point fitting, a closed-loop, autonomous, and reproducible processing link is constructed, which significantly improves the reliability, accuracy, and engineering practicality of spaceborne space target observation.

[0014] To achieve the above objectives, the present invention adopts the following technical solution:

[0015] In a first aspect, embodiments of the present invention provide a space target detection and self-calibration photometry method for spaceborne star field imaging, comprising the following steps:

[0016] S1. Acquire the raw telemetry data stream of the satellite in the onboard satellite field imaging mode, perform frame-by-frame analysis, and reconstruct the two-dimensional raw image sequence;

[0017] S2. Perform self-calibrated radiometric correction on the original image sequence and output the corrected image sequence; perform background noise modeling based on the corrected image sequence to obtain the background model and noise estimation;

[0018] S3. Based on the corrected image sequence, background model and noise estimation, perform multi-frame registration, source parameter extraction and spatial target discrimination to obtain the corresponding stellar background, spatial target and single-frame artifact;

[0019] S4. Construct a standard star candidate set based on the stellar background, and fit a mapping model from standard star pixel coordinates to celestial coordinates; convert the pixel coordinates of the space target into corresponding celestial coordinates based on the mapping model;

[0020] S5. Based on the space target, noise estimation, and standard star candidate set, perform photometric zero-point fitting and space target photometry; output structured space target observation results; the structured space target observation results include the pixel coordinates, celestial coordinates, magnitude, and magnitude uncertainty of the space target.

[0021] Furthermore, in step S2, self-calibrated radiometric correction is performed on the original image sequence, including:

[0022] Based on the metadata corresponding to each frame in the original image sequence, a set of multi-frame observation images is selected from the original image sequence according to the principle of metadata similarity; wherein, the metadata includes frame timestamp, exposure time, gain, readout mode, platform attitude information and temperature parameters;

[0023] The background of each frame in the multi-frame observation image is estimated and normalized; robust statistics are performed on the pixels of all frames in the normalized multi-frame observation image in the time dimension to obtain robust statistical results.

[0024] The robust statistical results are separated in the spatial dimension to obtain a stable bias component as the synthesized dark field and a stable response non-uniform component as the synthesized flat field.

[0025] Radiometric correction is performed on each frame of the original image sequence using synthetic dark field and synthetic flat field.

[0026] Furthermore, in step S2, background noise modeling is performed based on the corrected image sequence to obtain a background model and noise estimation; specifically including:

[0027] Each frame of the corrected image sequence is divided into several background estimation units;

[0028] Within each background estimation unit, robust background estimation is performed on the pixel distribution to obtain the background value of that unit;

[0029] Interpolate and filter the background values ​​of all units to generate a background model;

[0030] Background subtraction is performed on each frame of the corrected image to obtain the corresponding residual image;

[0031] Robust variance or standard deviation estimation is performed on all residual images to obtain noise estimates.

[0032] Furthermore, step S3 specifically includes:

[0033] Based on the background model and noise estimation, background subtraction and threshold detection are performed on each frame of the corrected image sequence, source parameters are extracted and their image plane positions are calculated.

[0034] Using stars that appear in most frames as control points, the inter-frame geometric transformation model is estimated, and the source coordinates of each frame are unified to the same reference coordinate system.

[0035] After unifying the reference coordinate system, cross-frame correlation and statistical discrimination are performed on the source parameters of multiple frames. Based on the frequency of occurrence, persistence and trajectory consistency, stellar background, space targets and single-frame artifacts are distinguished.

[0036] Furthermore, the threshold detection employs a dynamic threshold based on background and noise estimation.

[0037] The dynamic threshold is expressed by the formula:

[0038]

[0039] in, Represents the background model. This represents noise estimation, and K represents the threshold coefficient.

[0040] Furthermore, in step S4, a standard star candidate set is constructed based on the stellar background, including:

[0041] Based on the stellar background and the corresponding platform attitude information, the estimated field of view of the current image on the celestial sphere is determined; candidate stars that meet the magnitude threshold within the estimated field of view are extracted from the reference star catalog.

[0042] Based on the stellar positions of the candidate stars, geometric pattern matching is performed with the stars in the star catalog, and mismatched star pairs are eliminated using the RANSAC algorithm to obtain high-confidence matching star pairs, forming a standard star candidate set for photometric calibration.

[0043] Furthermore, step S5 specifically includes:

[0044] Photometric measurements were performed on the standard stars in the candidate set of standard stars, and the photometric zero point was fitted.

[0045] A photometric operation is performed on the space target, its magnitude is calculated in combination with the photometric zero point, and error propagation is performed based on the noise estimate and the zero point uncertainty to obtain the magnitude uncertainty;

[0046] The output includes structured observation results of space targets, including pixel coordinates, celestial coordinates, magnitude, and magnitude uncertainty.

[0047] Secondly, embodiments of the present invention provide a space target detection and self-calibration optical measurement system for spaceborne star field imaging, comprising the following modules:

[0048] Acquisition module: used to acquire the raw telemetry data stream of the satellite in the onboard satellite field imaging mode, perform frame parsing and reconstruct the two-dimensional raw image sequence;

[0049] Correction module: used to perform self-calibrated radiometric correction on the original image sequence and output the corrected image sequence; based on the corrected image sequence, background noise modeling is performed to obtain the background model and noise estimation;

[0050] The discrimination module is used to perform multi-frame registration, source parameter extraction, and spatial target discrimination based on the corrected image sequence, background model, and noise estimation, so as to obtain the corresponding stellar background, spatial target, and single-frame artifacts.

[0051] Coordinate transformation module: used to construct a standard star candidate set based on the stellar background, and fit a mapping model from standard star pixel coordinates to celestial coordinates; based on the mapping model, convert the pixel coordinates of the space target into the corresponding celestial coordinates;

[0052] Fitting and Output Module: Used to perform photometric zero-point fitting and photometric operations on the space target based on the space target, noise estimation and standard star candidate set; output structured space target observation results; the structured space target observation results include the pixel coordinates, celestial coordinates, magnitude and magnitude uncertainty of the space target.

[0053] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a space target detection and self-calibration photometry method and system for spaceborne star field imaging, which has the following beneficial effects:

[0054] This invention firstly generates and synthesizes dark and flat fields from multi-frame observation images, eliminating the reliance on dedicated calibration data and demonstrating strong on-orbit adaptability, effectively addressing operational mismatch issues such as temperature drift and exposure variations. Secondly, it introduces inter-frame geometric registration and multi-frame consistency discrimination mechanisms, establishing detection under a unified reference coordinate system, reliably distinguishing real space targets, stellar backgrounds, and single-frame artifacts (such as cosmic rays and hot pixels), significantly reducing the false alarm rate. Thirdly, the photometric process employs a robust zero-point fitting strategy, automatically eliminating saturated or contaminated standard stars, and combining noise models with positioning accuracy for error propagation, outputting absolute magnitudes with uncertainty. Finally, the closed-loop design of the entire process enables automated processing from raw telemetry to structured products, with each stage sharing background, registration, and calibration parameters, avoiding manual parameter tuning and improving result consistency and reproducibility.

[0055] This invention breaks through the path dependence of traditional spaceborne space target observation methods on calibration data, human experience, and single-frame processing. It constructs a new observation and processing paradigm that is self-calibrated, multi-frame collaborative, quality controllable, and end-to-end autonomous. It has outstanding advantages in improving the detection capability of weak targets, suppressing false alarms, ensuring photometric accuracy, and enhancing engineering practicality. It can be widely applied to mission scenarios such as microsatellite constellations, space-based space surveillance systems, and deep space exploration. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0057] Figure 1 This is a flowchart of the space target detection and self-calibration photometry method for spaceborne star field imaging provided in this embodiment of the invention;

[0058] Figure 2 This is a schematic diagram of the space target detection and self-calibration photometry method for spaceborne star field imaging provided in this embodiment of the invention;

[0059] Figure 3 This is a block diagram of the space target detection and self-calibration photometry system for spaceborne star field imaging provided in an embodiment of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1

[0062] This invention discloses a space target detection and self-calibration photometry method for spaceborne star field imaging, referring to... Figure 1 As shown, it includes the following steps:

[0063] S1. Acquire the raw telemetry data stream of the satellite in the onboard satellite field imaging mode, perform frame-by-frame analysis, and reconstruct the two-dimensional raw image sequence;

[0064] S2. Perform self-calibrated radiometric correction on the original image sequence and output the corrected image sequence; perform background noise modeling based on the corrected image sequence to obtain the background model and noise estimation;

[0065] S3. Based on the corrected image sequence, background model and noise estimation, perform multi-frame registration, source parameter extraction and spatial target discrimination to obtain the corresponding stellar background, spatial target and single-frame artifact;

[0066] S4. Construct a standard star candidate set based on the stellar background, and fit a mapping model from standard star pixel coordinates to celestial coordinates; convert the pixel coordinates of the space target into corresponding celestial coordinates based on the mapping model;

[0067] S5. Based on the space target, noise estimation, and standard star candidate set, perform photometric zero-point fitting and space target photometry; output structured space target observation results; the structured space target observation results include the pixel coordinates, celestial coordinates, magnitude, and magnitude uncertainty of the space target.

[0068] This embodiment applies to a satellite equipped with an optical imaging payload operating in a 600km sun-synchronous orbit to observe space targets in low Earth orbit, such as satellites and space debris. This embodiment performs staring observations of the target area, identifies and tracks space targets within that area, and records their trajectories and brightness changes.

[0069] In this embodiment, the satellite continuously captures 30 frames of satellite field images during a single transit, with each frame having an exposure time of 2 seconds and an interval of 0.5 seconds between frames. The downlink telemetry data is received by the ground station and then referenced... Figure 2 As shown, the implementation steps are as follows:

[0070] Step 01: Acquire downlink telemetry data and perform data parsing.

[0071] This embodiment takes downlink telemetry data streams such as CADU / bin and auxiliary data as input. In this embodiment, the auxiliary data is metadata, which includes at least timecode, exposure time, gain / readout mode, platform attitude quaternion or equivalent attitude information, and temperature and other parameters.

[0072] This embodiment first segments and unpacks the original telemetry data stream according to the frame synchronization word and frame header / tail structure, and performs consistency checks on the length, sequence number, and checksum (e.g., CRC). Then, it extracts the image payload, performs bit-width expansion, end-order processing, and row / column rearrangement according to pixel packing rules, thereby reconstructing a two-dimensional grayscale image sequence, i.e., the two-dimensional original image sequence. ; k This refers to the image frame number.

[0073] Simultaneously, this embodiment aligns each image frame with metadata such as timestamp, exposure, gain, and pose quaternions to establish an "image-metadata" mapping relationship. This outputs the original image sequence. and its corresponding metadata This embodiment Meta ( k It should include at least the frame timestamp, exposure time, gain and readout mode, image size, ROI parameters, and pose quaternion.

[0074] Step 02: Perform image preprocessing on each frame.

[0075] The image preprocessing in this embodiment includes image DN value correction and background noise modeling; wherein, background noise modeling is used for deep space background suppression.

[0076] This step takes the original observed image sequence as input, in the presence of a calibration library dark field. Draw The calibration library is used preferentially for correction. For the k-th frame image... Perform darkening and flattening to obtain the corrected image. It can be expressed by the formula:

[0077]

[0078] in, To calibrate the dark field of the library, To calibrate the Kuping field.

[0079] When the dark / flat field calibration library data is missing, the operating conditions do not match, or the matching reliability is insufficient, this embodiment enters the self-calibration mode.

[0080] Based on the principle of similar exposure / gain / temperature, the two-dimensional original image sequence is processed. Grouping and filtering out multi-frame subsets with highly similar imaging conditions K represents the total number of image frames. Specifically, it includes:

[0081] Similar exposure times: Only select image frames whose exposure time difference does not exceed a preset threshold (such as ±10% or absolute difference ≤0.2 seconds).

[0082] Frames with identical or linearly comparable gain settings are preferred; if there are slight differences (such as due to automatic gain control), the ratio should be close to 1 and linear scaling compensation should be performed during subsequent normalization.

[0083] Detector temperatures are similar: using the focal surface temperature in the metadata, image frames with a temperature difference less than the tolerance are selected.

[0084] In this embodiment, a multi-frame observation image set was selected. First, the background of each frame in the set was estimated and the necessary normalization was performed. Then, robust statistics were performed on the pixels in the time dimension, and outlier removal (such as sigma-clipping) was combined to suppress stellar point sources and transient highlight components.

[0085] In this embodiment, for each original image frame, physical quantities are first normalized based on its metadata to eliminate differences in operating conditions; for example, the units of all frames are unified to "response value per unit time and unit gain". Secondly, for each pixel, its normalized value sequence across multiple frames is collected, and robust sigma-clipping statistics are performed (e.g., iterative 3σ removal of outliers followed by median selection) to obtain the stable total signal for that pixel. This signal contains two physical components:

[0086] Stable bias component: Consists of readout bias + dark current, and is independent of illumination;

[0087] The background of stable response modulation: pixel response gain (i.e., flat field) and sky background brightness (slowly changing).

[0088] Since the background brightness of the sky is low-frequency smooth in space, and the stable bias component is fixed pattern noise (which may contain high-frequency structures such as column noise), this embodiment extracts the bias by spatial high-pass or low-frequency suppression to obtain the spatial distribution that reflects the inherent bias and dark current of the detector, i.e., the stable bias component.

[0089] After obtaining the bias and background estimates, this embodiment reconstructs the response gain; the bias is subtracted from the stable total signal to obtain the background response; the background response is globally or in blocks normalized (making its mean 1) to obtain the non-uniform component of the stable response, which characterizes the relative sensitivity difference of each pixel.

[0090] Based on this, this embodiment extracts the fixed bias component that "stable with each frame" to form the synthetic dark field / fixed pattern correction term. The stable component of "pixel response inhomogeneity" is extracted to form the synthetic flat / response correction term. (and normalize it so that its mean is 1).

[0091] Then, the self-calibration results are used to perform correction on each frame, expressed by the formula:

[0092]

[0093] in, This represents the corrected image sequence.

[0094] This embodiment obtains Simultaneously, background noise modeling is performed on the corrected image to generate a background model. With noise level This information is shared for subsequent testing and photometry.

[0095] This embodiment uses a block-based background estimation method to divide the entire frame into several background estimation units. Within each unit, robust background estimation of the pixel distribution is performed. This embodiment combines a sigma-clipping median / SExtractor-like background estimator to obtain block-based background values. Then, through interpolation and filtering, a two-dimensional background model is reconstructed, expressed by the formula:

[0096]

[0097] Output the residual image after background subtraction:

[0098]

[0099] in, Let be the background estimate of the ij-th block in the k-th frame, where ij is the block index; Interp represents the set of all block background estimates for the k-th frame; () is an interpolation operator used to interpolate the block background estimate to the original value. Figure 1 Smooth resolution; () is a smoothing operator used to smooth the interpolated background field to suppress block boundary effects; This represents the background model value for the k-th frame. This is the radiometrically corrected image of the k-th frame; This is the residual image after background subtraction in the k-th frame. Simultaneously, this embodiment performs robust variance / standard deviation estimation on the background pixel set (or each block unit) to obtain the noise level. It can form a whole-frame scalar according to application needs. Used for threshold calibration or block noise mapping It is used for modeling spatially non-uniform noise.

[0100] Further combining image metadata, such as gain g With readout noise r By unifying the noise into the electron number domain, a noise model that can be used to address the propagation of photometric errors can be constructed; for example, for a photometric region with an aperture area of ​​A, the net signal electron number is denoted as... The standard deviation of the background noise electron count is denoted as . The total noise is then expressed as:

[0101]

[0102] Based on this, quality indicators such as signal-to-noise ratio are obtained, providing a unified standard for subsequent target detection, photometry, and uncertainty assessment.

[0103] To ensure reproducible results and cross-module compatibility, this embodiment outputs and records intermediate results in a structured manner; the corrected image sequence is output. Self-calibration results , Background Model and noise estimation results , and .

[0104] Step 03: Space target detection.

[0105] This embodiment uses For each frame of the image as input, background subtraction is performed first, followed by threshold detection under noise constraints to obtain a candidate pixel set. The threshold can be expressed in the form of "background + noise factor", as shown by the formula:

[0106]

[0107] Where K is the threshold coefficient, which can also be determined in this embodiment using a local adaptive method.

[0108] This embodiment then analyzes the candidate region metadata imaging pattern and performs target detection to obtain the image plane position XY of the target / star point. This embodiment first performs connected component analysis and morphological screening, and calculates measurements such as the source's centroid / sub-pixel position, shape parameters, flux, and signal-to-noise ratio to form a frame-by-frame source table. To mitigate the impact of attitude jitter and drift on cross-frame consistency, this embodiment further utilizes star points that appear stably in most frames as control points to estimate inter-frame geometric transformations. The source coordinates of each frame are unified to the same reference coordinate system. Under the unified reference system, cross-frame correlation and statistical discrimination are performed on the source tables of multiple frames. By using features such as frequency of occurrence, persistence, brightness change and trajectory consistency, sources that exist stably in most frames are identified as stellar backgrounds, and sources that appear only in a few frames or change significantly or show a consistent motion trend are identified as spatial targets. At the same time, artifacts such as cosmic rays, hot pixels, and bad pixels are eliminated or downweighted by combining morphological and temporal consistency.

[0109] In this embodiment, the stellar background appears in the vast majority of frames, e.g., >90%, with its position remaining almost unchanged, i.e., residual <0.5 pixels. Space targets appear in multiple frames, e.g., ≥3 frames, exhibiting linear or near-linear trajectories with velocities conforming to orbital dynamics. Single-frame artifacts appear only in 1–2 frames, lacking trajectory continuity or exhibiting irregular positional jumps.

[0110] This outputs a frame-by-frame record or trajectory fragment of the spatial target:

[0111]

[0112] Where C represents the set of frame-by-frame records or trajectory segments of the space target; k is the image frame number; x(k) and y(k) represent the image plane coordinates of the space target in the k-th frame, i.e., column coordinates and row coordinates, respectively; flux(k) represents the photometric flux value of the space target in the k-th frame, which in this embodiment is the net flux obtained by aperture measurement or PSF fitting, determined according to the photometric method used; SNR(k) represents the signal-to-noise ratio of the space target in the k-th frame; flag(k) represents the quality flag of the target record in the k-th frame, used to indicate whether the measurement in this frame is valid and its anomaly type, such as: normal, suspected artifact, saturation, neighboring star contamination, edge truncation, or matching uncertainty, etc. The registration model is also output simultaneously. Set of stellar background sources.

[0113] Step 04: Star identification and star catalog matching.

[0114] This embodiment takes a set of stellar background sources, attitude quaternions, and other field-of-view pointing priors, as well as a reference star catalog (including magnitude attributes), as inputs. Based on the attitude priors, this embodiment estimates the field-of-view center and range, and prunes a subset of candidate sky regions from the reference star catalog to reduce the matching search space. Subsequently, a robust star catalog matching strategy (e.g., based on inter-star angular distance, triangle invariants, or hash features) is used to establish a correspondence between the "image star point set" and the "star catalog subset," and erroneous matches are eliminated through RANSAC or equivalent robust estimation to obtain a standard star candidate set M.

[0115] Step 05: Astronomical positioning.

[0116] This embodiment uses a standard star candidate set M and the pixel coordinates of the space target in the k-th frame image. As input, a mapping model from pixel coordinates to celestial coordinates is fitted by minimizing the reprojection error of the standard star; subsequently, the candidate pixel positions of the space target are... Mapped to celestial coordinates That is, the target location.

[0117] First, in the k-th frame image, successfully matched standard star samples are selected from set M, and the pixel coordinates of each standard star and its celestial coordinates (right ascension, declination) in the star catalog are obtained. Then, using these standard star samples as constraints, the mapping relationship from pixel coordinates to celestial coordinates is established and solved. By adjusting the parameters of the mapping relationship, the overall deviation between the celestial coordinates calculated from the pixel coordinates of each standard star and the celestial coordinates given in the star catalog is minimized, thus obtaining the astronomical positioning / tablet solution result for the k-th frame. Finally, the candidate pixel positions (x(k), y(k)) of the space target are substituted into the mapping relationship to obtain the celestial coordinates (RA(k), Dec(k)) of the space target in the k-th frame, and these celestial coordinates are output as the position of the space target.

[0118] Step 06: Magnitude Calculation.

[0119] This embodiment uses image correction. Background / Noise Estimation The standard star candidate set M and the candidate locations of space targets (pixels AND or RA / Dec) are used as inputs. Aperture photometry or PSF photometry is performed on the successfully matched standard stars to obtain the instrument flux. During the fitting process, standard stars that are saturated, have nearby contamination, or have abnormal residuals are automatically removed or reduced in weight, thereby obtaining the photometric zero point and its uncertainty.

[0120] The fitted photometric zero-point relationship is expressed by the formula:

[0121]

[0122] ZP represents the zero point of the instrument's photometric readings. For standard star magnitudes, This represents the instrument flux of the j-th standard star obtained in the current frame through aperture photometry or PSF fitting.

[0123] Subsequently, the instrument flux of the target was obtained by performing photometry at each frame position of the space target. And combined with the zero-point calculation of the target magnitude, the formula is expressed as:

[0124]

[0125] in, The instrument flux of the space target in the kth frame is obtained by aperture measurement or PSF fitting, which is the net flux after background subtraction. Let be the magnitude of the space target in the kth frame.

[0126] This embodiment propagates errors based on noise estimation and zero-point uncertainty, outputs magnitude uncertainty, and finally outputs a structured spatial target directory or trajectory file, which includes at least pixel coordinates, RA / Dec, magnitude and magnitude uncertainty.

[0127] This embodiment proposes a space target detection and self-calibration photometry method for spaceborne star field imaging. Unlike existing technologies that commonly employ "loading specialized dark-field / flat-field calibration data for correction and distributing decoding, calibration, detection, star catalog matching, and photometry across multiple tools in a cascaded manner," this embodiment differs primarily in four aspects:

[0128] First, a self-calibration mechanism based on multi-frame observation data is introduced in the calibration stage to generate usable dark / flat field correction terms and background noise estimates when dark / flat field references are lacking or mismatched. Second, in the target detection stage, inter-frame registration and multi-frame statistical discrimination are mandatory steps to distinguish between stable stellar backgrounds and transient / moving targets, thereby reducing the sensitivity of single-frame thresholds and manual parameter tuning. Third, in the photometry stage, the standard star set obtained from star catalog matching, zero-point fitting, target photometry, and quality diagnosis are managed in a unified manner, outputting uncertainty and quality indicators such as contamination / saturation along with the output magnitude. Fourth, in terms of engineering organization, images, metadata, background noise models, registration models, astronomical positioning models, and photometric calibration parameters are managed in a closed loop within the same data structure, realizing an integrated link from telemetry input to catalog output.

[0129] Through the aforementioned differentiated technical means, this embodiment can reduce the dependence of radiometric correction on specialized calibration data when dark flat field reference is missing or operating conditions change, improve the temporal consistency of space target candidate extraction by using multi-frame statistical discrimination, and simultaneously provide uncertainty and quality indicators in photometric output, thereby improving the traceability and engineering usability of the results.

[0130] Regarding target detection performance, this embodiment can adapt to different observation modes and target morphology differences. In scanning or target-line (trailing) scenarios, the direction, length, and other geometric quantities of the trail can be output through linear structure detection and fitting. In spatial staring scenarios where the target is point-like, combining motion consistency across multiple frames for confirmation and correlation can significantly reduce false alarms and discontinuities caused by the sensitivity of single-frame threshold detection to random noise and transient artifacts (such as isolated bright spots), making it easier for point-like targets to form stable and continuous observation records, thus providing reliable input for subsequent positioning and photometry. The comparison of its performance in different working modes is shown in Table 1 below:

[0131] Table 1 Comparison of different working modes

[0132]

[0133] In the astronomical positioning process, this embodiment first completes star identification and star catalog matching in the image, providing reliable input for subsequent positioning solutions. By extracting and filtering star points in the image, and combining the projection results of the reference star catalog within a given field of view, the correspondence between stars in the image and stars in the star catalog is established. The matching results have good consistency in geometric structure, forming a stable set of matching pairs to support subsequent plate-based solutions; the relevant matching relationships can be visualized and verified through the source point and target point, as well as the matching results based on quadrilateral invariants.

[0134] After obtaining a stable set of standard star candidates, astronomical positioning is further performed. A mapping relationship between pixel coordinates and celestial coordinates (RA / Dec) is established using a plate-based model, and the astronomical direction of the image center is calculated from this. By comparing the astronomical direction obtained from the plate-based model with the original image direction, the consistency and correction effect of the positioning results can be evaluated. The relevant results are presented in tabular form to reflect the performance differences of the positioning solution under different data source conditions. The above process realizes a complete closed loop from star identification, star catalog matching to astronomical positioning, enabling space targets to directly output their corresponding celestial locations and providing verifiable positioning data for engineering applications. The astronomical positioning right ascension comparison table of this embodiment is shown in Table 2 below:

[0135] Table 2 Comparison of Right Ascension for Astronomical Positioning

[0136]

[0137] Regarding photometric calibration, this embodiment uses stars successfully matched to the star catalog as a standard star set for photometric zero-point fitting and converts the target flux into magnitude output. In this embodiment, the photometric results of the stars are compared with the true values ​​of the external star catalog, yielding an average magnitude error of approximately 0.405 magnitudes for 11 stars; simultaneously, an estimated target magnitude of approximately 6.075 magnitudes is provided. These results demonstrate that this embodiment can establish a quantifiable photometric link of "standard star zero-point calibration—target magnitude output" and possesses a certain level of magnitude consistency.

[0138] Regarding engineering feasibility, this embodiment supports large-format image processing and has conducted computation time evaluation; the image size of the embodiment is 4096×4118, and processing tests were carried out on the Arm architecture platform, indicating that this embodiment has the engineering foundation for deployment and optimization on resource-constrained platforms.

[0139] Example 2

[0140] This invention discloses a space target detection and self-calibration optical measurement system for spaceborne star field imaging, referring to... Figure 3 As shown, it includes the following modules:

[0141] Acquisition module: used to acquire the raw telemetry data stream of the satellite in the onboard satellite field imaging mode, perform frame parsing and reconstruct the two-dimensional raw image sequence;

[0142] Correction module: used to perform self-calibrated radiometric correction on the original image sequence and output the corrected image sequence; based on the corrected image sequence, background noise modeling is performed to obtain the background model and noise estimation;

[0143] The discrimination module is used to perform multi-frame registration, source parameter extraction, and spatial target discrimination based on the corrected image sequence, background model, and noise estimation, so as to obtain the corresponding stellar background, spatial target, and single-frame artifacts.

[0144] Coordinate transformation module: used to construct a standard star candidate set based on the stellar background, and fit a mapping model from standard star pixel coordinates to celestial coordinates; based on the mapping model, convert the pixel coordinates of the space target into the corresponding celestial coordinates;

[0145] Fitting and Output Module: Used to perform photometric zero-point fitting and photometric operations on the space target based on the space target, noise estimation and standard star candidate set; output structured space target observation results; the structured space target observation results include the pixel coordinates, celestial coordinates, magnitude and magnitude uncertainty of the space target.

[0146] This embodiment achieves an integrated processing chain from raw telemetry to a "position + magnitude" catalog through acquisition, correction, discrimination, coordinate transformation, and fitting and output modules. Starting with raw telemetry data from CADU / bin, the following steps are completed sequentially within the same processing framework: telemetry decoding and image reconstruction, image preprocessing, spatial target detection and multi-frame association, star identification, star catalog matching and plate decomposition, target photometry and photometric calibration, and finally, a target catalog containing pixel coordinates, RA / Dec, magnitude, and quality information. Each stage shares a unified data structure and intermediate parameters (such as attitude information, background noise estimation, registration / plate decomposition model, zero-point parameters, etc.), avoiding parameter inconsistencies and untraceable results caused by multiple scripts and software splicing, facilitating engineering deployment and batch processing.

[0147] This embodiment employs a target detection and artifact suppression mechanism combining "inter-frame registration + multi-frame consistency discrimination." First, stable stars in the image are used to estimate inter-frame geometric transformations, unifying the source tables of each frame to the same reference coordinate system. Based on this, statistical analysis is performed on the frequency of occurrence, duration, brightness variations, and trajectory consistency of the sources. Sources that "stable in most frames" are considered stellar backgrounds, sources that "appear isolated in a single frame without a reasonable motion model" are considered artifacts or subject to weighting, and sources that "satisfy continuous motion characteristics" are considered spatial target candidates. Especially in staring scenarios with point targets and complex backgrounds, this mechanism can significantly reduce the false alarm rate of artifacts such as cosmic rays and hot pixels in a single frame, improving the stability of target detection and trajectory continuity.

[0148] This embodiment also provides robust photometry and quality indicator output based on a standard star set. A standard star set is obtained through star identification and catalog matching. Aperture or PSF photometry is then performed on this set, and the luminosity zero point is fitted using the catalog magnitude. During the fitting process, standard stars with saturation, neighbor contamination, or residual anomalies are automatically identified and removed or weighted. When performing photometry on space targets at each frame position, robust background estimation and anomalous pixel removal are employed, and neighboring stars and strong background structures within the aperture are detected. The final output includes not only the target magnitude but also zero-point uncertainty, target photometry uncertainty, and quality indicators such as contamination / saturation / anomalies. This allows for subsequent filtering and weighting based on quality information during orbit determination, target classification, and light variation analysis, improving the engineering usability of the results.

[0149] This embodiment utilizes self-calibrated radiometric correction and background noise estimation based on multi-frame observations. When specialized calibration data, such as dark fields and flat fields, are missing or significantly mismatched with current operating conditions (temperature, gain, exposure, etc.), the statistical characteristics of multi-frame scientific observation data are used to construct synthetic dark and flat fields, or equivalent fixed bias and response inhomogeneity correction terms are extracted. Multi-frame fusion suppresses stellar and transient bright structures while preserving stable system components, providing a unified background level and noise estimate. Thus, even with missing or mismatched calibration, radiometric correction can still be achieved, providing a consistent noise scale for detection and photometry, reducing reliance on specialized calibration observations, and improving adaptability to changes in on-orbit operating conditions.

[0150] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0151] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A space target detection and self-calibration photometry method for spaceborne star field imaging, characterized in that, Includes the following steps: S1. Acquire the raw telemetry data stream of the satellite in the onboard satellite field imaging mode, perform frame-by-frame analysis, and reconstruct the two-dimensional raw image sequence; S2. Perform self-calibrated radiometric correction on the original image sequence and output the corrected image sequence; perform background noise modeling based on the corrected image sequence to obtain the background model and noise estimation; The self-calibrated radiometric correction of the original image sequence includes: Based on the metadata corresponding to each frame in the original image sequence, a set of multi-frame observation images is selected from the original image sequence according to the principle of metadata similarity; wherein, the metadata includes frame timestamp, exposure time, gain, readout mode, platform attitude information and temperature parameters; The background of each frame in the multi-frame observation image is estimated and normalized; robust statistics are performed on the pixels of all frames in the normalized multi-frame observation image in the time dimension to obtain robust statistical results. The robust statistical results are separated in the spatial dimension to obtain a stable bias component as the synthesized dark field and a stable response non-uniform component as the synthesized flat field. Radiometric correction is performed on each frame of the original image sequence using synthetic dark field and synthetic flat field. S3. Based on the corrected image sequence, background model and noise estimation, perform multi-frame registration, source parameter extraction and spatial target discrimination to obtain the corresponding stellar background, spatial target and single-frame artifact; S4. Construct a standard star candidate set based on the stellar background, and fit a mapping model from standard star pixel coordinates to celestial coordinates; convert the pixel coordinates of the space target into corresponding celestial coordinates based on the mapping model; S5. Based on the space target, noise estimation, and standard star candidate set, perform photometric zero-point fitting and space target photometry; output structured space target observation results; the structured space target observation results include the pixel coordinates, celestial coordinates, magnitude, and magnitude uncertainty of the space target.

2. The method as described in claim 1, characterized in that, In step S2, background noise modeling is performed based on the corrected image sequence to obtain a background model and noise estimation; specifically including: Each frame of the corrected image sequence is divided into several background estimation units; Within each background estimation unit, robust background estimation is performed on the pixel distribution to obtain the background value for that unit; Interpolate and filter the background values ​​of all units to generate a background model; Background subtraction is performed on each frame of the corrected image to obtain the corresponding residual image; Robust variance or standard deviation estimation is performed on all residual images to obtain noise estimates.

3. The method as described in claim 1, characterized in that, Step S3 specifically includes: Based on the background model and noise estimation, background subtraction and threshold detection are performed on each frame of the corrected image sequence, source parameters are extracted and their image plane positions are calculated. Using stars that appear in most frames as control points, the inter-frame geometric transformation model is estimated, and the source coordinates of each frame are unified to the same reference coordinate system. After unifying the reference coordinate system, cross-frame correlation and statistical discrimination are performed on the source parameters of multiple frames. Based on the frequency of occurrence, persistence and trajectory consistency, stellar background, space targets and single-frame artifacts are distinguished.

4. The method as described in claim 3, characterized in that, The threshold detection is performed using a dynamic threshold based on background and noise estimation; The dynamic threshold is expressed by the formula: in, Represents the background model. This represents noise estimation, and K represents the threshold coefficient.

5. The method as described in claim 3, characterized in that, In step S4, a standard star candidate set is constructed based on the stellar background, including: Based on the stellar background and the corresponding platform attitude information, the estimated field of view of the current image on the celestial sphere is determined; candidate stars that meet the magnitude threshold within the estimated field of view are extracted from the reference star catalog. Based on the stellar positions of the candidate stars, geometric pattern matching is performed with the stars in the star catalog, and mismatched star pairs are eliminated using the RANSAC algorithm to obtain high-confidence matching star pairs, forming a standard star candidate set for photometric calibration.

6. The method as described in claim 1, characterized in that, Step S5 specifically includes: Photometric measurements were performed on the standard stars in the candidate set of standard stars, and the photometric zero point was fitted. A photometric operation is performed on the space target, its magnitude is calculated in combination with the photometric zero point, and error propagation is performed based on the noise estimate and the zero point uncertainty to obtain the magnitude uncertainty; The output includes structured observation results of space targets, including pixel coordinates, celestial coordinates, magnitude, and magnitude uncertainty.

7. A space target detection and self-calibration photometry system for spaceborne star field imaging, characterized in that, Includes the following modules: Acquisition module: used to acquire the raw telemetry data stream of the satellite in the onboard satellite field imaging mode, perform frame parsing and reconstruct the two-dimensional raw image sequence; Correction module: used to perform self-calibrated radiometric correction on the original image sequence and output the corrected image sequence; based on the corrected image sequence, background noise modeling is performed to obtain the background model and noise estimation; The self-calibrated radiometric correction of the original image sequence includes: Based on the metadata corresponding to each frame in the original image sequence, a set of multi-frame observation images is selected from the original image sequence according to the principle of metadata similarity; wherein, the metadata includes frame timestamp, exposure time, gain, readout mode, platform attitude information and temperature parameters; The background of each frame in the multi-frame observation image is estimated and normalized; robust statistics are performed on the pixels of all frames in the normalized multi-frame observation image in the time dimension to obtain robust statistical results. The robust statistical results are separated in the spatial dimension to obtain a stable bias component as the synthesized dark field and a stable response non-uniform component as the synthesized flat field. Radiometric correction is performed on each frame of the original image sequence using synthetic dark field and synthetic flat field. The discrimination module is used to perform multi-frame registration, source parameter extraction, and spatial target discrimination based on the corrected image sequence, background model, and noise estimation, so as to obtain the corresponding stellar background, spatial target, and single-frame artifacts. Coordinate transformation module: used to construct a standard star candidate set based on the stellar background, and fit a mapping model from standard star pixel coordinates to celestial coordinates; based on the mapping model, convert the pixel coordinates of the space target into the corresponding celestial coordinates; Fitting and Output Module: Used to perform photometric zero-point fitting and photometric operations on the space target based on the space target, noise estimation and standard star candidate set; output structured space target observation results; the structured space target observation results include the pixel coordinates, celestial coordinates, magnitude and magnitude uncertainty of the space target.

Images