A space-based artificial satellite target trajectory extraction method based on geometric features and time sequence correlation cooperation complementation

Abstract

A kind of space-based artificial satellite target trajectory extraction method based on geometric features and time sequence association collaborative complementation.Belongs to the technical field of space-based target identification, specifically relates to the technical field of space-based artificial satellite target trajectory extraction.Can accurately extract weak artificial satellite target from staring image sequence, and reconstruct its motion trajectory, while effectively suppressing false alarm.The method comprises the following steps: space-based image preprocessing;Repair and enhancement of broken star trajectory, used to improve the integrity of star trajectory and reduce false alarm, realize the connection and reconstruction of broken star trajectory;Suspected artificial satellite extraction based on contour screening, used to improve the integrity and reliability of candidate target, provide input data for subsequent classification and trajectory analysis;Multi-frame trajectory association based on intersection over union, used to solve the problem of trajectory breakage or matching error caused by target displacement and noise interference factors;Trajectory optimization and target attribute extraction: including trajectory coordinate interpolation completion and target attribute extraction.

Description

Technical Field

[0001] This invention belongs to the field of space-based target recognition technology, specifically relating to the field of space-based artificial satellite target trajectory extraction technology. Background Technology

[0002] When observing specific artificial satellite targets using the "staring imaging" mode on space-based optical platforms (such as satellites and space telescopes), the resulting image sequences contain significant stellar trajectories and various types of noise in addition to the artificial satellite targets. Existing technologies face the following core challenges: Severe interference from stellar trajectories and fragmented segments increase false alarms: In staring imaging mode, background stars appear as line segments in the image. Due to weak signals, the trajectories of low-magnitude stars are often broken into discrete segments after preprocessing. Their shapes and sizes are highly similar to those of artificial satellite point targets, making them difficult to distinguish and resulting in a significant increase in the number of false alarms.

[0003] The point features of the target are affected by platform disturbances: Under ideal conditions, artificial satellites produce point-like images, but due to platform instability or shaking, the images may be blurred or deformed, which will destroy the consistency of point features and increase the difficulty of effective extraction.

[0004] Inter-frame target association is susceptible to motion inconsistencies: target satellites move slowly and have small displacements in the sequence, while star trajectories move significantly across frames. The large difference in their motion characteristics makes it easy for matching errors or trajectory breaks to occur when associating multiple frames, making it difficult to construct continuous and correct trajectories.

[0005] Target loss can lead to trajectory discontinuity: Due to noise or detection sensitivity, satellites may miss some frames, causing trajectory interruptions and affecting the accuracy of subsequent applications such as orbit determination and light variation analysis. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an efficient and robust method for extracting space-based targets. This method can accurately extract faint artificial satellite targets from staring image sequences and reconstruct their complete and smooth motion trajectories. At the same time, it effectively suppresses false alarms, providing a reliable data foundation for the identification, tracking, and characteristic analysis of artificial satellites.

[0007] The method includes the following steps: S1. Space-based image preprocessing: including image acquisition and staring imaging, noise suppression and image enhancement, as well as image binarization and foreground separation; S2. Repair and enhancement of broken stellar trajectories, used to improve the integrity of stellar trajectories and reduce false alarms, to achieve connectivity and reconstruction of broken stellar trajectories: including line segment detection and extraction, directional clustering and trajectory grouping, line segment fusion and trajectory reconstruction, as well as foreground enhancement and mask generation; S3. Suspected satellite extraction based on contour screening is used to improve the integrity and reliability of candidate targets and provide input data for subsequent classification and trajectory analysis: including contour extraction, geometric feature calculation, effective contour screening, contour merging and post-merging verification. S4. Multi-frame trajectory association based on cross-union ratio (CUNR) is used to solve the problem of trajectory breakage or matching errors caused by target displacement and noise interference: including target bounding box construction and expansion, CUNR calculation and initial screening, Hungarian algorithm optimization matching, and trajectory verification and screening. S5. Trajectory Optimization and Target Attribute Extraction: This includes trajectory coordinate interpolation and completion, and target attribute extraction.

[0008] Furthermore, image acquisition and staring imaging specifically involve: utilizing the staring imaging mode of the space-based platform, continuously acquiring multiple frames of images by keeping the sensor pointed at a fixed target; Noise suppression and image enhancement specifically involve: removing isolated noise points in the image by inter-frame differential denoising and applying mean filtering to the differential image, while preserving the edge structure of artificial satellite and star trajectories; Image binarization and foreground separation specifically involves performing binarization segmentation on the filtered image and converting the difference image into a binary image by setting a low threshold.

[0009] Furthermore, the line segment detection and extraction specifically involves applying the Hough transform to the binary foreground image to detect all potential line segment structures in the image. The Hough transform can identify regions with linear features and output a set of line segments, with each line segment represented by the coordinates of its endpoints. Directional clustering and trajectory grouping are specifically as follows: calculate the direction angle of each extracted line segment, and use a clustering algorithm to group these line segments according to direction. Since the trajectories of stars have a consistent direction of motion in the image, the dominant direction group can be obtained after clustering, thereby classifying the line segments that conform to the trend of star motion into the same category. The line segment fusion and trajectory reconstruction are as follows: For line segments within the same direction group, spatial relationship analysis is performed, distance thresholds and angle tolerances are set, and line segments that meet the connection conditions are fused to generate continuous and smooth trajectory line segments. Foreground enhancement and mask generation specifically involve performing a logical "OR" operation between the fused and reconstructed star trajectory and the original binary foreground image to obtain the enhanced foreground mask.

[0010] Furthermore, the contour extraction specifically involves: performing contour detection on the enhanced foreground mask, and using a boundary tracking algorithm to extract the outer contours of all connected regions in the image to form an initial contour set; The geometric feature calculation is as follows: for each contour, calculate its minimum bounding rectangle, obtain its long side length and short side length, and then calculate the ratio of the long side to the short side (Ratio). The effective contour selection process is as follows: set an upper limit threshold for the ratio of the long side to the short side, and set a range for the length of the long side; judge each contour, and if its ratio is less than the threshold and the length of the long side is within the set range, it is considered an effective contour and retained as a candidate target; contours that do not meet the conditions are removed. Contour merging specifically involves using the Euclidean distance between the center points of the contours as a measure of spatial proximity. and Its center point and Euclidean distance The calculation is as follows: ; in, express coordinates express The coordinates; Will With preset merging distance threshold Compare, if the following conditions are met:

[0011] Then determine the contour and Fragmented stellar trajectories originating from the same target but with low magnitude are merged; the merging operation employs a contour fusion algorithm. The post-merger verification specifically involves recalculating the minimum bounding rectangle for the merged new outline and re-judging the ratio of the long side to the short side and the range of the long side to ensure that it still meets the geometric requirements of an artificial satellite.

[0012] Furthermore, the construction and expansion of the target bounding box specifically involves: for each frame containing an extracted satellite target, assuming the target's pixel region in the image is... Its minimum bounding horizontal rectangle Defined as: ; in, Indicates the region The minimum x-coordinate of all points; Indicates the region The minimum y-coordinate of all points; Indicates the region The maximum x-coordinate of all points; Indicates the region The maximum y-coordinate of all points; The basic bounding box is expanded outwards, resulting in the expanded bounding box. The calculation is as follows: ; Right now: ,in, These are external parameters, determined based on image resolution and target motion characteristics.

[0013] Furthermore, the intersection-union ratio calculation and initial screening are as follows: For all target bounding boxes in two consecutive frames, calculate their intersection-union ratio (IoU) pairwise, which is the ratio of the area of ​​the intersection to the area of ​​the union of the two boxes. Set an IoU threshold and filter out box pairs with an IoU higher than the threshold as candidate matching pairs. The specific steps are as follows: S61. Calculation of intersection region: Let the bounding box in frame t be... Coordinates are The bounding box in frame t+1 Coordinates are ; The coordinates of the intersection region are calculated as follows: ; The width and height of the intersection region are: ; The area of ​​intersection is: ; S62. Union region calculation: The areas of the two bounding boxes are: ; The area of ​​the union is: ; S63. Calculation of intersection-union ratio: The intersection-union ratio of the bounding boxes between frame t and frame t+1 is calculated as follows: ; S64. Candidate Matching Pair Filtering: Set IoU threshold Select box pairs that meet the following conditions as candidate matching pairs: ; in, The value is determined based on the target's motion characteristics and image characteristics, and is preferably 0.3; All candidate matching pairs that meet the conditions constitute the initial association set. : .

[0014] Furthermore, the Hungarian algorithm optimizes the matching specifically by using the obtained initial set of associations. Based on this, a cost matrix between targets in two consecutive frames is constructed. The matrix elements reflect the matching cost. The Hungarian algorithm is applied to solve the minimum total cost matching of this cost matrix to achieve a one-to-one optimal association, ensuring that each target is matched with at most one target in the next frame, thereby forming a preliminary trajectory segment. The trajectory verification and screening process is as follows: For each candidate trajectory generated by matching, the number of valid frames in the entire image sequence is counted, and the proportion of the number of frames is calculated. If the proportion of the number of frames in a certain trajectory exceeds a preset threshold, it is determined to be a valid artificial satellite trajectory and is retained; otherwise, it is regarded as intermittent noise or false alarm and is removed.

[0015] Furthermore, the trajectory coordinate interpolation and completion specifically involves: for a valid artificial satellite trajectory, detecting any missing coordinate frames that may appear in its time series; based on the coordinate points of known frames in the trajectory, using an interpolation algorithm to reconstruct the estimated position of the artificial satellite in the missing frames, ensuring the continuity of the trajectory in the spatiotemporal dimension, and forming the target motion path; The target attribute extraction process specifically involves extracting the morphological and grayscale attributes of each target in the trajectory, including morphological features, grayscale features, background statistical features, and optical features.

[0016] Compared with the prior art, the present invention has the following advantages: 1. Significantly effective false alarm suppression and high detection accuracy: By introducing a "stellar trajectory repair" mechanism, the trajectories of low-magnitude stars, which are broken due to weak signals, are effectively connected and restored to continuous long line segments, fundamentally avoiding the problem of misidentifying discrete trajectory fragments as artificial satellites. Combined with geometric feature screening based on the ratio of long to short sides, the number of false alarms is greatly reduced, significantly improving the accuracy and reliability of target recognition.

[0017] 2. Complete and smooth target trajectory with excellent data quality: A strategy combining "contour merging" and "trajectory interpolation" is adopted to compensate for missing target information from both spatial and temporal dimensions. The former aggregates adjacent suspected targets within a single frame, while the latter fills in missing positions in the sequence, jointly ensuring the continuity and smoothness of the output trajectory and providing high-quality input for subsequent trajectory determination and characteristic analysis.

[0018] 3. Strong robustness and anti-interference capability of multi-frame association: The association method that integrates "intersection over union ratio screening" and "Hungarian algorithm matching" effectively addresses the challenge of large differences in the motion characteristics of stars and targets, achieves accurate tracking of targets with small displacements, avoids mismatch and trajectory breakage, and improves the stability of the system in complex imaging environments.

[0019] 4. High degree of automation and wide applicability: This method automates the entire process from image preprocessing to trajectory output, significantly reducing manual intervention and demonstrating good engineering practicality. It is applicable to space-based staring image sequences with different resolutions and noise levels, and has broad platform compatibility.

[0020] 5. Rich output information dimensions, supporting in-depth analysis: In addition to the target trajectory, it simultaneously extracts multiple features such as morphology, grayscale, background statistics and apparent stars, providing comprehensive data support for advanced applications such as satellite identification, status monitoring and light variation analysis. Attached Figure Description

[0021] Figure 1 This is an overall flowchart of the method described in the embodiments of the present invention; Figure 2 This is a diagram illustrating the imaging effect of space-based images in an embodiment of the present invention. Figure 3 This is a diagram showing the filtered effect in an embodiment of the present invention; Figure 4 This is a diagram showing the binarization result in an embodiment of the present invention; Figure 5 This is a diagram illustrating the line segment fusion and trajectory reconstruction results in an embodiment of the present invention. Figure 6 This is a comparison image of the foreground enhancement effect in an embodiment of the present invention; Figure 7 This is a diagram illustrating the effect of the geometric feature filtering and spatial merging strategy in an embodiment of the present invention. Figure 8 This is a diagram showing the results of multi-frame identification and association of artificial satellites in an embodiment of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0023] Existing space-based target extraction technologies mainly include single-frame-based processing methods (such as threshold segmentation and edge detection) and multi-frame-based methods (such as optical flow and background modeling). Their limitations are as follows: Single-frame method: sensitive to noise and background changes, threshold segmentation is prone to high false detection, and edge detection performs poorly in low-contrast images.

[0024] Limitations of multi-frame methods: such as the computational complexity of optical flow methods, sensitivity to motion blur, and difficulty in handling mixed scenes of small target displacements and large star movements.

[0025] Lack of inter-frame correlation mechanism: Most methods lack effective modeling of target motion consistency and temporal correlation, resulting in trajectory breakage or matching errors, which are difficult to meet the needs of practical applications.

[0026] This embodiment provides a space-based target extraction method based on feature selection and inter-frame correlation. Through image preprocessing, star trajectory restoration, geometric feature selection and multi-frame matching, it achieves robust extraction and trajectory construction of artificial satellite targets.

[0027] A schematic diagram of the overall method is shown below. Figure 1 As shown, the method will be described in detail.

[0028] Step 1: Space-based image preprocessing The space-based images processed in this embodiment are acquired using a staring imaging method. This staring imaging includes, but is not limited to, two forms: "platform-wide staring," which controls the overall attitude of the space-based platform to stably point the sensor's line of sight towards a specific artificial satellite target; or "payload-specific staring," which adjusts the internal components of the optical payload to track the target. In a preferred embodiment of the invention, the platform-wide staring method is used. In this mode, the target artificial satellite exhibits only a tiny shift of a few pixels in a continuous image sequence, while background stars, affected by the platform's orbital motion, appear as distinct long line segments in the image. This imaging characteristic is beneficial for distinguishing moving targets from stars, providing favorable conditions for subsequent extraction of the artificial satellite from the image. The specific preprocessing procedure is as follows: 1. Image Acquisition and Staring Imaging Staring imaging, utilizing a space-based platform (such as a satellite or space telescope), involves continuously acquiring multiple frames of images by keeping the sensor pointed at a fixed target. Staring imaging effectively reduces motion blur of the target, ensuring its relatively stable position within the image sequence. During this process, the satellite, due to its relatively small motion relative to the platform, experiences only a few pixels of displacement between consecutive frames; while background stars, affected by the platform's orbital motion and the Earth's rotation, appear as distinct line segments across frames, with displacements often exceeding tens of pixels. The imaging effect of space-based images is as follows: Figure 2 As shown, the figure contains high-magnitude star trails, low-magnitude star trails, artificial satellite targets, and a large amount of background noise.

[0029] 2. Noise Suppression and Image Enhancement Inter-frame differential denoising: This operation performs a forward difference operation on consecutive image frames, calculating the gray-level differences of pixels between adjacent frames to obtain a differential image sequence. This operation can effectively eliminate static backgrounds and uniformly distributed system noise, while enhancing regions in the image with inter-frame gray-level variations, such as satellite and star trails.

[0030] Median filtering: Median filtering is applied to the differencing image, preferably using a 3×3 filter window, to remove isolated noise points in the image while preserving the edge structure of satellite and star trails. Median filtering has good impulse noise suppression capabilities, smoothing the image while avoiding edge blurring, thus helping to improve the image signal-to-noise ratio. The filtered effect is shown in Figure 3.

[0031] 3. Image binarization and foreground separation The filtered image is binarized and segmented by setting a low threshold to convert the difference image into a binary image. Regions with higher grayscale values ​​are identified as foreground, primarily including satellite imagery, complete or fragmented stellar trails, and residual noise; regions with lower grayscale values ​​are identified as background. This step lays the foundation for subsequent target recognition and classification. The binarization result is shown below. Figure 4 As shown.

[0032] By employing the aforementioned staring imaging and combined denoising strategies, the complexity of image data can be effectively reduced and noise interference suppressed during the preprocessing stage, providing high-quality image input for the subsequent accurate identification and extraction of artificial satellites.

[0033] Step 2: Repairing and enhancing fractured stellar trajectories In the preprocessed binary foreground image, low-magnitude stars often exhibit broken and discrete trajectories due to their weak signals, making them easily misidentified as independent targets and interfering with subsequent satellite identification. To improve the integrity of stellar trajectories and reduce false alarms, this step proposes a trajectory repair method based on line segment detection and directional clustering to connect and reconstruct broken stellar trajectories. The specific process is as follows: 1. Line segment detection and extraction The Hough transform is applied to a binary foreground image to detect all potential line segment structures in the image. The Hough transform can effectively identify regions with linear features, including broken stellar trajectories and artificial satellite patches, and outputs a set of line segments, each represented by the coordinates of its endpoints.

[0034] 2. Orientation Clustering and Trajectory Grouping The orientation angles of each extracted line segment are calculated, and a clustering algorithm is used to group these line segments according to their orientation. Since the trajectories of stars have a consistent direction of motion in the image, clustering yields a dominant orientation group, thus classifying line segments that conform to the trend of star motion into the same category and eliminating interfering structures with obvious directional deviations.

[0035] 3. Line segment fusion and trajectory reconstruction Spatial relationship analysis is performed on line segments within the same directional group. Distance thresholds and angle tolerances are set, such as line segment spacing less than 100 pixels and directional angle difference less than 5 degrees. Line segments meeting the connection conditions are then fused to generate continuous and smooth trajectory segments. This process effectively bridges broken low-magnitude star trajectories, restoring their complete shape while maintaining directional consistency. The line segment fusion and trajectory reconstruction results are shown below. Figure 5 As shown.

[0036] 4. Foreground Enhancement and Mask Generation A logical OR operation is performed between the fused and reconstructed star trails and the original binary foreground image to obtain an enhanced foreground mask. This mask not only preserves the original satellite target but also includes the repaired continuous star trails, significantly improving the structural integrity and noise suppression of the foreground target, laying the foundation for subsequent differentiation between stars and satellites. Foreground enhancement effect comparison: Figure 6 As shown.

[0037] Step 3: Extraction of suspected artificial satellites based on contour filtering Based on the enhanced foreground mask obtained in the previous step, this step uses contour analysis technology to accurately extract suspected artificial satellite targets. To address issues such as target fragmentation and dispersion that may occur due to imaging conditions or noise interference, this method combines geometric feature screening and spatial merging strategies to effectively improve the integrity and reliability of candidate targets, providing high-quality input for subsequent classification and trajectory analysis. The specific implementation process is as follows: 1. Contour Extraction Contour detection is performed on the enhanced foreground mask, and a boundary tracking algorithm is used to extract the outer contours of all connected regions in the image, forming an initial contour set. Each contour is described by a series of continuous boundary points, covering artificial satellite candidate targets, star trajectory segments, and other noisy regions.

[0038] 2. Geometric Feature Calculation For each contour, its minimum bounding rectangle is calculated, and its long side length (L_long) and short side length (L_short) are obtained. Then, the ratio of the long side to the short side (Ratio = L_long / L_short) is calculated. This ratio can effectively reflect the shape characteristics of the contour: the closer the Ratio is to 1, the closer the contour is to a square, which is consistent with the characteristics of artificial satellites appearing as dot-like or nearly circular patches in images; the larger the Ratio is, the more the contour tends to be elongated, such as the trajectory of a star.

[0039] 3. Valid contour filtering Set an upper limit threshold for the ratio of the longer side to the shorter side (e.g., Ratio < 8) and a reasonable range for the length of the longer side (e.g., L_long ∈ [5 pixels, 50 pixels]). For each contour, if its Ratio is less than the threshold and the length of the longer side is within the set range, it is considered a valid contour and retained as a candidate target; contours that do not meet the conditions (e.g., excessively long star trails or excessively low noise) are discarded. This step initially filters out candidate targets whose shapes conform to the characteristics of artificial satellite imaging.

[0040] 4. Contour merging To address interference from satellite-like fragments caused by broken stellar trajectories of low-magnitude stars, contours with spatially adjacent locations within the effective contour set are merged. The Euclidean distance between the center points of each contour is calculated; if the distance is less than a preset merging threshold, they are determined to belong to the same target, and a contour fusion algorithm is used to merge them into a single continuous contour. This eliminates false detections of broken stellar fragments and improves target integrity. The spatial proximity between contours is used to determine whether they belong to the same target. Preferably, the Euclidean distance between the center points of the contours is used as the measure of spatial proximity. For contours... and Its center point and Euclidean distance The calculation is as follows: ; in, express coordinates express The coordinates; Will With preset merging distance threshold Compare, if the following conditions are met: ; Then determine the contour and Broken stellar trajectories originating from the same target but with lower magnitudes should be merged.

[0041] The merging operation employs a contour fusion algorithm. Preferably, all boundary points of the two contours are merged into a new point set. ,in and Outline and The set of boundary points. Then, for the point set... Calculate its convex hull to form a new, continuous, and complete contour. : ; The merged profile can more accurately reflect the stellar trajectory target, thus effectively eliminating false detections caused by broken trajectory fragments.

[0042] The effects of geometric feature selection and spatial merging strategies are as follows: Figure 7 As shown.

[0043] 5. Verification after merging The minimum bounding rectangle of the merged new contour is recalculated, and the ratio of the long side to the short side and the range of the long side are judged again to ensure that it still meets the geometric feature requirements of artificial satellites. The verified contours are finally included in the set of suspected artificial satellites as candidate targets for subsequent identification and tracking.

[0044] Step 4: Multi-frame trajectory association based on intersection-over-union ratio After obtaining suspected satellite targets in each frame of the image, this step proposes a multi-frame matching strategy combining the intersection-over-union metric and the Hungarian algorithm to construct and associate their trajectories in a continuous image sequence. This strategy effectively solves the problem of trajectory breaks or matching errors caused by target displacement, noise interference, and other factors, ultimately outputting stable and continuous satellite trajectories. The specific implementation process is as follows: 1. Target bounding box construction and expansion For each frame containing a suspected satellite target, its smallest bounding horizontal rectangle is first calculated as the basic bounding box. To accommodate possible target displacement between frames, this rectangle is expanded outwards; the number of pixels expanded can be set according to the image resolution and target motion characteristics (e.g., 100 pixels). The expanded bounding box better covers the possible areas where the target may appear in adjacent frames, providing sufficient spatial tolerance for subsequent matching. This bounding box expansion technique enhances the robustness of target matching between adjacent frames. Specifically, the bounding box expansion formula is as follows: (a) Basic bounding box definition Let the pixel region of the target in the image be... Its minimum bounding horizontal rectangle Defined as: ; in, Indicates the region The minimum x-coordinate of all points; Indicates the region The minimum y-coordinate of all points; Indicates the region The maximum x-coordinate of all points; Indicates the region The maximum y-coordinate of all points.

[0045] (b) Bounding box expansion formula To accommodate the target's motion displacement between frames, the basic bounding box is expanded outwards, resulting in an expanded bounding box. The calculation is as follows: ; Right now: ; in, These are external parameters, determined based on image resolution and target motion characteristics.

[0046] 2. Crossover and Union Ratio Calculation and Initial Screening For all target bounding boxes in two consecutive frames, calculate their intersection-to-union ratio (IoU), which is the ratio of the area of ​​their intersection to the area of ​​their union. Set an IoU threshold (e.g., 0.3) and filter out box pairs with an IoU higher than this threshold as candidate matching pairs. This step initially establishes the correspondence between targets in the frames and excludes obviously irrelevant targets. Specifically, the IoU calculation formula is as follows: (a) Calculation of intersection region Let the bounding box in frame t be... Coordinates are The bounding box in frame t+1 Coordinates are .

[0047] The coordinates of the intersection region are calculated as follows: ; The width and height of the intersection region are: ; The area of ​​intersection is: ; (b) Union region calculation The areas of the two bounding boxes are: ; The area of ​​the union is: ; (c) Calculation of intersection-union ratio The intersection-union ratio of the bounding boxes between frame t and frame t+1 is calculated as follows: ; (d) Candidate matching pair screening Set IoU threshold Select box pairs that meet the following conditions as candidate matching pairs: ; in, The value is determined based on the target's motion characteristics and image characteristics, and is preferably set to 0.3.

[0048] All candidate matching pairs that meet the conditions constitute the initial association set. : ; 3. Hungarian Algorithm for Optimized Matching With the obtained initial set of associations Based on this, a cost matrix is ​​constructed between targets in two consecutive frames, with matrix elements reflecting the matching cost (commonly represented by 1-IoU). The Hungarian algorithm is applied to solve for the minimum total cost matching of this cost matrix, achieving a one-to-one optimal association and ensuring that each target matches at most one target in the next frame, thus forming a preliminary trajectory segment.

[0049] 4. Trajectory Verification and Filtering For each candidate trajectory generated by matching, the number of valid frames appearing in the entire image sequence is counted, and the proportion of its appearance to the total number of frames is calculated. If the proportion of a trajectory's appearance exceeds a preset threshold (e.g., 50%), it is determined to be a valid satellite trajectory and retained; otherwise, it is considered intermittent noise or a false alarm and is discarded. This step ensures that the output trajectory has sufficient continuity and stability, improving the reliability of the detection results. The results of multi-frame satellite recognition and association are as follows: Figure 8 As shown.

[0050] Step 5: Trajectory Optimization and Target Attribute Extraction After completing the multi-frame trajectory association, this step optimizes and completes the confirmed valid satellite trajectories, extracting their key attribute features to improve trajectory quality and provide complete data support for subsequent target identification and analysis. The specific implementation method is as follows: 1. Trajectory coordinate interpolation completion For valid satellite trajectories, detect any missing frames in the time series (e.g., temporary loss of the target due to noise interference or detection omissions). Based on the coordinates of known frames in the trajectory, use interpolation algorithms (such as linear interpolation or spline interpolation) to reconstruct the estimated position of the satellite in the missing frames, ensuring the continuity of the trajectory in the spatiotemporal dimensions and forming a complete and smooth target motion path.

[0051] 2. Target attribute extraction For each target in the trajectory, its morphological and grayscale attributes are extracted for subsequent classification or behavior analysis. Extractable attributes include, but are not limited to: Morphological features: including the length, width, and pixel area of ​​the target's bounding rectangle; Gray-scale features include the sum of gray levels, mean gray level, and maximum gray level of pixels within the target area; Background statistical characteristics: including the mean and variance of the local background surrounding the target; Optical characteristics: Apparent star magnitude based on the conversion between target grayscale and background brightness.

[0052] Extracting these attributes can provide a multi-dimensional data foundation for the identification, motion state analysis, and light variation behavior research of artificial satellites.

Claims

1. A method for extracting the trajectory of a space-based artificial satellite target based on the complementary synergy of geometric features and temporal correlation, characterized in that, The method includes the following steps: S1. Space-based image preprocessing: including image acquisition and staring imaging, noise suppression and image enhancement, as well as image binarization and foreground separation; S2. Repair and enhancement of broken stellar trajectories, used to improve the integrity of stellar trajectories and reduce false alarms, to achieve connectivity and reconstruction of broken stellar trajectories: including line segment detection and extraction, directional clustering and trajectory grouping, line segment fusion and trajectory reconstruction, as well as foreground enhancement and mask generation; The line segment detection and extraction process involves applying the Hough transform to the binary foreground image to detect all potential line segment structures in the image. The Hough transform can identify regions with linear features and output a set of line segments, with each line segment represented by the coordinates of its endpoints. Directional clustering and trajectory grouping are specifically as follows: calculate the direction angle of each extracted line segment, and use a clustering algorithm to group these line segments according to direction. Since the trajectories of stars have a consistent direction of motion in the image, the dominant direction group can be obtained after clustering, thereby classifying the line segments that conform to the trend of star motion into the same category. The line segment fusion and trajectory reconstruction are as follows: For line segments within the same direction group, spatial relationship analysis is performed, distance thresholds and angle tolerances are set, and line segments that meet the connection conditions are fused to generate continuous and smooth trajectory line segments. Foreground enhancement and mask generation specifically involve performing a logical "OR" operation between the fused and reconstructed star trajectory and the original binary foreground image to obtain the enhanced foreground mask. S3. Suspected satellite extraction based on contour screening is used to improve the integrity and reliability of candidate targets and provide input data for subsequent classification and trajectory analysis: including contour extraction, geometric feature calculation, effective contour screening, contour merging and post-merging verification. Contour extraction specifically involves: performing contour detection on the enhanced foreground mask, and using a boundary tracking algorithm to extract the outer contours of all connected regions in the image to form an initial contour set; The geometric feature calculation is as follows: for each contour, calculate its minimum bounding rectangle, obtain its long side length and short side length, and then calculate the ratio of the long side to the short side (Ratio). The effective contour selection process is as follows: set an upper limit threshold for the ratio of the long side to the short side, and set a range for the length of the long side; judge each contour, and if its ratio is less than the threshold and the length of the long side is within the set range, it is considered an effective contour and retained as a candidate target; contours that do not meet the conditions are removed. Contour merging specifically involves using the Euclidean distance between the center points of the contours as a measure of spatial proximity. The Euclidean distance between the center points of each contour is calculated. If the distance is less than a preset merging threshold, the contours are determined to belong to the same target. The contour fusion algorithm is then used to merge them into a single continuous contour, thereby eliminating false detections of broken star fragments and improving target integrity. The verification after merging specifically involves: recalculating the minimum bounding rectangle for the merged new outline, and re-judging the ratio of the long side to the short side and the range of the long side to ensure that it still meets the geometric requirements of an artificial satellite; S4. Multi-frame trajectory association based on cross-union ratio (CUNR) is used to solve the problem of trajectory breakage or matching errors caused by target displacement and noise interference: including target bounding box construction and expansion, CUNR calculation and initial screening, Hungarian algorithm optimization matching, and trajectory verification and screening. S5. Trajectory Optimization and Target Attribute Extraction: This includes trajectory coordinate interpolation and completion, and target attribute extraction.

2. The method for extracting the trajectory of a space-based artificial satellite according to claim 1, characterized in that, Image acquisition and staring imaging specifically involves: using the staring imaging mode of a space-based platform, continuously acquiring multiple frames of images by keeping the sensor pointed at a fixed target; Noise suppression and image enhancement specifically involve: removing isolated noise points in the image by inter-frame differential denoising and applying mean filtering to the differential image, while preserving the edge structure of artificial satellite and star trajectories; Image binarization and foreground separation specifically involves performing binarization segmentation on the filtered image and converting the difference image into a binary image by setting a low threshold.

3. The method for extracting the trajectory of a space-based artificial satellite according to claim 2, characterized in that, When using the Euclidean distance between the center points of a contour as a measure of spatial proximity, for a contour... and Its center point and Euclidean distance The calculation is as follows: ； in, express coordinates express The coordinates; Will With preset merging distance threshold Compare, if the following conditions are met: Then determine the contour and Fragmented stellar trajectories originating from the same target but with low magnitude are merged; the merging operation employs a contour fusion algorithm.

4. The method for extracting the trajectory of a space-based artificial satellite according to claim 3, characterized in that, The construction and expansion of the target bounding box are as follows: For the artificial satellite target extracted in each frame, let the pixel region of the target in the image be... Its minimum bounding horizontal rectangle Defined as: ； in, Indicates the region The minimum x-coordinate of all points; Indicates the region The minimum y-coordinate of all points; Indicates the region The maximum x-coordinate of all points; Indicates the region The maximum y-coordinate of all points; The basic bounding box is expanded outwards, resulting in the expanded bounding box. The calculation is as follows: ； Right now: ,in, These are external parameters, determined based on image resolution and target motion characteristics.

5. The method for extracting the trajectory of a space-based artificial satellite according to claim 4, characterized in that, The calculation and initial screening of intersection-union ratio are as follows: For all target bounding boxes in two consecutive frames, calculate their intersection-union ratio (IoU) pairwise, which is the ratio of the area of ​​the intersection to the area of ​​the union of the two boxes. Set an IoU threshold and filter out box pairs with an IoU higher than the threshold as candidate matching pairs. The specific steps are as follows: S61. Calculation of intersection region: Let the bounding box in frame t be... Coordinates are The bounding box in frame t+1 Coordinates are ; The coordinates of the intersection region are calculated as follows: ； The width and height of the intersection region are: ； The area of ​​intersection is: ； S62. Union region calculation: The areas of the two bounding boxes are: ； The area of ​​the union is: ； S63. Calculation of intersection-union ratio: The intersection-union ratio of the bounding boxes between frame t and frame t+1 is calculated as follows: ； S64. Candidate Matching Pair Filtering: Set IoU threshold Select box pairs that meet the following conditions as candidate matching pairs: ； in, The value is determined based on the target's motion characteristics and image characteristics, and is set to 0.

3. All candidate matching pairs that meet the conditions constitute the initial association set. : 。 6. The method for extracting the trajectory of a space-based artificial satellite according to claim 5, characterized in that, The Hungarian algorithm optimizes matching by using the obtained initial set of associations. Based on this, a cost matrix between targets in two consecutive frames is constructed. The matrix elements reflect the matching cost. The Hungarian algorithm is applied to solve the minimum total cost matching of this cost matrix to achieve a one-to-one optimal association, ensuring that each target is matched with at most one target in the next frame, thereby forming a preliminary trajectory segment. The trajectory verification and screening process is as follows: For each candidate trajectory generated by matching, the number of valid frames in the entire image sequence is counted, and the proportion of the total number of frames is calculated. If the proportion of the number of frames in a certain trajectory exceeds a preset threshold, it is determined to be a valid artificial satellite trajectory and is retained. Otherwise, it will be considered intermittent noise or a false alarm and will be rejected.

7. The method for extracting the trajectory of a space-based artificial satellite according to claim 6, characterized in that, The trajectory coordinate interpolation and completion process is as follows: For a valid artificial satellite trajectory, detect any missing frames in the time series. Based on the coordinate points of known frames in the trajectory, use an interpolation algorithm to reconstruct the estimated position of the artificial satellite in the missing frames, ensuring the continuity of the trajectory in the spatiotemporal dimension and forming the target motion path. The target attribute extraction process specifically involves extracting the morphological and grayscale attributes of each target in the trajectory, including morphological features, grayscale features, background statistical features, and optical features.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-7.

9. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-7.

Images