A space-based, highly reliable, wide dynamic range space debris capture and tracking method

Abstract

This invention proposes a space-based, highly reliable, wide dynamic range space debris capture and tracking method, applicable to the initial capture and continuous tracking measurement of distant space debris. After acquiring multiple image frames, stellar image points are identified and discarded. The remaining image points are placed in a suspected target queue for motion pattern discrimination, retaining those conforming to the target's motion pattern. A new image frame is captured, the predicted position of the target in the image is calculated, and image points near the target are extracted using a windowing method. The image point with the highest confidence is selected as the target point for the current frame. Based on the inertial attitude of the newly captured image, the right ascension and declination of the target image point are calculated, replacing the earliest data in the suspected target queue, and the relationship between the target's right ascension and declination and time is refitted. This invention overcomes the optical resolution limits of instruments and the accuracy limitations of image point extraction, achieving highly reliable, wide dynamic range target capture and improving the real-time performance of target capture and tracking.

Description

Technical Field

[0001] This invention relates to a space-based, highly reliable, wide dynamic range space debris capture and tracking method, belonging to the field of space awareness and protection. Background Technology

[0002] With the increasing number of spacecraft in orbit, the amount of debris generated by spacecraft impacts and disintegration is rising rapidly. Debris monitoring and protection have become urgent needs for current space missions. Traditional debris target monitoring methods mostly rely on ground-based observation or space-based observation and ground processing, which suffer from time lags and cannot provide all-day, all-weather coverage. Space-based optical sensors offer better imaging conditions and have become a hot topic for autonomous space debris monitoring.

[0003] Currently, common methods for space-based optical debris capture and tracking include masking, stacking, inter-frame difference, and star identification. Masking methods have low accuracy, stacking methods have a narrow dynamic range, inter-frame difference methods have high computational complexity, while star matching methods are mature and have stable time and space complexity. Existing technologies suffer from insufficient centroid positioning accuracy and low update rates in some cases, and excessive computational complexity in others. Furthermore, they are easily limited by optical system and detector noise, resulting in a narrow dynamic range for target detection and a lack of stability in space debris detection. Summary of the Invention

[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a space-based high-reliability wide dynamic range space debris capture and tracking method. It distinguishes image points based on the different motion characteristics of stars and targets, and improves the signal-to-noise ratio and dynamic range through long-period calculation, thereby enhancing the ability of autonomous target capture and tracking.

[0005] The technical solution of this invention is: a space-based, highly reliable, wide dynamic range space debris capture and tracking method, comprising:

[0006] Step 1. The space-based optical sensor acquires multiple frames of superimposed images of the star and the target, identifies the image points of the star using the triangle matching method, and removes the image points of the star from the image;

[0007] Step 2. Based on the star catalog information, locate stars not identified by the triangle matching method and remove the corresponding image points from the image;

[0008] Step 3. After removing star image points, the remaining image points are placed into the suspected target queue, which is divided into a fast suspected target queue and a slow suspected target queue according to the target speed.

[0009] Step 4. Determine the motion patterns of suspected target image points, retain image points whose right ascension and declination conform to the motion patterns, and fit the parameters of the relationship between the target's right ascension and declination and time.

[0010] Step 5. Take a new image frame, input the image time into the fitting formula of right ascension and declination, and obtain the predicted position of the target in the image;

[0011] Step 6. Extract image points near the predicted position of the target in the newly captured image using a window method. If multiple image points are extracted, compare the right ascension and declination of the image points with the predicted position of the target in the image, and select the image point with the highest confidence as the target point of the current frame.

[0012] Step 7. Based on the inertial attitude of the newly captured image, calculate the right ascension and declination of the target image point, and replace the earliest data in the suspected target queue with the first-in-first-out data according to the first-in-first-out principle, and refit the parameters of the relationship between the target's right ascension and declination and time.

[0013] Step 8. Repeat steps 5-7 to predict and track the target position in the next frame image.

[0014] Preferably, the method for identifying the image points of stars in step 1 is as follows:

[0015] The angular distance between each pair of observed stars is calculated. The angular distances of the stars are found by searching the pre-stored navigation star database for values ​​that are close to the angular distances of the image points. The stars are then identified by a triangle matching method.

[0016] Preferably, in step 2, finding stars not identified by the triangle matching method and removing the corresponding image points from the image specifically involves:

[0017] First, the unit vector of image point i in the space-based optical sensor system is transformed to the inertial coordinate system. The vector of image point i in the inertial coordinate system is cross-multiplied with the vector of the navigation star in the star table in the inertial space to obtain the angular distance between image point i and the navigation star. If the angular distance between image point i and the navigation star is less than the star discrimination threshold, image point i is considered to be a star and is removed. Otherwise, the image point is retained.

[0018] Furthermore, the angular distance between the two unremoved image points in any two frames in the inertial frame is compared with the multi-frame marking threshold. If the angular distance between the two unremoved image points in any two frames in the inertial frame is less than the multi-frame marking threshold, the points are marked as stars and the remaining stars are removed.

[0019] Preferably, after removing stellar image points in step 3, the remaining image points are placed into a suspected target queue, which is divided into a fast suspected target queue and a slow suspected target queue according to the target speed.

[0020] The frame interval of the image containing the image point in the fast suspected target queue is equal to the exposure frame period;

[0021] The frame interval of the image containing the image point in the slow-moving suspected target queue is the shortest time required to distinguish the target with the smallest angular velocity from the star. The specific calculation method is as follows:

[0022] Let the frame interval of the image containing the image point in the slow suspected queue be Δt, and the lower limit of the target's angular velocity be ω. l The multi-frame labeling threshold is Θ t The detection noise is θ n Then the frame interval of the image containing the image point in the slow queue is

[0023] Preferably, in step 4, motion pattern discrimination is performed on suspected target image points, and image points that conform to the target motion pattern are retained, specifically:

[0024] (1) First, the inter-frame angular velocities of the target image points are calculated as ω1, ω2, ..., ω using the angular distance and time difference between corresponding image points in two adjacent frames. n-1 :

[0025] If ω1, ω2, ..., ω n-1 Meets the preset target angular velocity threshold range [ω l ,ω u If the target is detected, further motion determination is performed.

[0026] If ω1, ω2, ..., ω n-1 If the target angular velocity does not meet the set target angular velocity threshold range, then the entire sequence of target image points is discarded.

[0027] (2) Let the time of the target parameter queue relative to the first frame be t1, t2, ..., t n Right ascension and declination are α1, α2, ..., α n and δ1,δ2,…,δ n During the observation period, the relationship between right ascension and time is as follows:

[0028] α j =k α t j +b α ,

[0029] The relationship between declination and time is as follows:

[0030] δ j =k δ t j +b δ Here, 1≤j≤n;

[0031] By fitting the relationship between right ascension and time, the parameter k characterizing the relationship between right ascension and time is obtained. α b α and the coefficient of determination of right ascension fitting Right Ascension Fit Standard Deviation Similarly, the parameter k for fitting the relationship between declination and time... δ b δ and the coefficient of determination for declination fitting Declination fit standard deviation

[0032] like and If the sequence of suspected target images conforms to the target's motion pattern, it is considered a target; otherwise, the entire sequence of target images is discarded. Here, fitThresh represents the fitting determination coefficient threshold.

[0033] Preferably, in step 5, a new image is captured, and the image time is substituted into the fitting formula for right ascension and declination to obtain the predicted position of the target in the image, specifically:

[0034] After capturing a new image, input the image time into the following formula:

[0035] α p =k α t p +b α

[0036] δ p =k δ t p +b δ

[0037] The predicted right ascension position α of the target is obtained. p Declination position δ p And convert it into an inertial frame vector; where t p Indicates: the time of the newly captured image relative to the first frame of the target parameter queue;

[0038] The attitude obtained through star map identification is further used to convert the inertial frame vector into the body frame vector of the space-based optical sensor;

[0039] Next, the body vector of the space-based optical sensor is converted into planar coordinates through the camera calibration model, which is the predicted position of the target in the image plane.

[0040] Preferably, in step 6, image points are extracted near the predicted location of the target in the newly captured image using a windowing method. If multiple image points are extracted, the right ascension and declination of the image points are compared with the predicted values ​​of the target, and the image point with the highest confidence is selected as the target point of the current frame. Specifically:

[0041] Using the right ascension α of all extracted image points cur δ declination cur And the predicted value of right ascension α p The predicted value of declination δ pFind the residual r α ,r δ :

[0042] r α =|α p -α cur |

[0043] r δ =|δ p -δ cur |

[0044] If the residual satisfies and If a pixel conforms to the target pattern, it is considered to meet the target pattern. If multiple pixels conform, the pixel with the smallest residual, i.e., the pixel with the highest confidence, is selected. Here, `outlierRatio` is the residual discrimination threshold. The standard deviation of the right ascension fit. is the standard deviation of the declination fit.

[0045] Preferably, in step 7, based on the inertial attitude of the newly captured image, the right ascension and declination of the target image point are calculated, and the right ascension and declination are replaced with the earliest data in the suspected target queue according to the first-in-first-out principle. The parameters relating the target's right ascension and declination to time are then refitted, specifically as follows:

[0046] Let the latest frame time be t. n+1 Right ascension is α n+1 Declination is δ n+1 The queue used for fitting is then updated to: t2, t3, ..., t n+1 Right ascension and declination are α2, α3, ..., α n+1 and δ2,δ3,…,δ n+1 ,get:

[0047] (1) The parameters characterizing the relationship between right ascension and time, as well as the right ascension fitting determination coefficient and right ascension fitting standard deviation, are refitted and assigned to k. α b α ,

[0048] (2) The parameters characterizing the relationship between declination and time, as well as the declination fitting determination coefficient and the declination fitting standard deviation, are refitted and assigned to k. δ b δ ,

[0049] Based on this, the target location in subsequent captured images is predicted and extracted, thus enabling continuous tracking.

[0050] Compared with the prior art, the present invention has the following advantages:

[0051] (1) The present invention is based on a mature high-precision star identification method, which provides accurate prior information for space debris capture and tracking, thereby improving accuracy.

[0052] (2) The present invention uses a sub-pixel centering method, which has high centroid positioning accuracy and high update rate;

[0053] (3) This invention improves the signal-to-noise ratio of slow targets by independently calculating and increasing the frame interval of slow targets, breaking the limitations of optical system and detector noise, and significantly improving the dynamic range of target detection;

[0054] (4) The present invention performs target capture and tracking calculations based on image point coordinates rather than the original image, which reduces computational complexity and improves the real-time performance of target capture and tracking;

[0055] (5) This invention filters multiple image points within a window to avoid false tracking. That is, it uses least squares fitting and residual filtering to filter multiple image points within a window, ensuring that the correct target image points are output and improving the stability of space debris detection.

[0056] (6) The present invention uses a target queue management method to associate and match current data with historical data, thereby eliminating interference between multiple targets;

[0057] (7) This invention uses a mature optical sensor as a hardware platform. After reliable verification, it can output target pointing information while maintaining inertial attitude. Attached Figure Description

[0058] Figure 1 This is an overall flowchart of the method of the present invention;

[0059] Figure 2 This is a flowchart of the target acquisition method of the present invention;

[0060] Figure 3 This is a flowchart of the target tracking method of the present invention. Detailed Implementation

[0061] like Figure 1 The diagram shown is an overall flowchart of a space-based high-reliability wide dynamic range space debris capture and tracking method involved in the present invention.

[0062] Based on the principle that stars are stationary in inertial space while space debris moves, stars are filtered out from the image. Then, the target is captured and tracked by utilizing the continuity of the multi-frame vector angular distance of image points. First, image points are compared with a star catalog, and stars in the catalog are removed. Second, by comparing the positions of image points across multiple frames, points with inter-frame inertial frame vector angular distances less than a threshold are marked as stars, thus establishing a potential target queue. The vector distance between adjacent frames is calculated for potential targets; points that meet the vector distance requirement are considered target points. Different target points between frames are correlated to obtain multi-frame position information of the target, such as... Figure 2 and Figure 3 As shown. While performing the above calculations, the signal-to-noise ratio is improved through long-period computation to achieve slow space debris capture.

[0063] The technical solution of this invention is:

[0064] 1. Star map recognition

[0065] Space-based optical sensors acquire multiple frames of superimposed images of stars and targets through an optical system, and use triangle matching to identify stars as prior information for target recognition.

[0066] The angular distance between each pair of observed stars is calculated. The angular distances of the navigation stars are searched for values ​​that are close to the angular distances of the image points through the pre-stored navigation star database. The stars are then identified by the intersection of the three sets of angular distances, which is the triangle matching method.

[0067] 2. Stellar removal

[0068] Triangle matching only identifies some stars. The remaining image points include unidentified stars and target points. Based on prior information from the star catalog, it is necessary to further search for and remove stars that were not identified by triangle matching, and retain the target image points.

[0069] First, the unit vector of image point i in the camera coordinate system of the space-based optical sensor is transformed to the inertial coordinate system. Then, the vector of image point i in the inertial coordinate system is cross-multiplied with the vector of the navigation star in the star catalog in inertial space to obtain the angular distance between image point i and the navigation star. If the angular distance between image point i and the navigation star is less than the star discrimination threshold, image point i is considered a star and is discarded; otherwise, the image point is retained. Specifically:

[0070] If the inertial attitude matrix corresponding to the image at time t1 is known to be A:

[0071] The image coordinates of the i-th image point are: (u i ,v i The unit vector of the image point in the camera coordinate system is V. i Then the vector of image point i in the inertial coordinate system is:

[0072] V inertial =A -1 Vi

[0073] Let V be the vector of the navigation star in inertial space. guStar Image point and navigation star angular distance Θ i for:

[0074] Θ i = <V inertial ×V guStar >

[0075] If Θ i <Θ FixedStar If the value is Θ, then the image point is considered a star and is discarded. FixedStar The threshold for stellar discrimination.

[0076] Due to limitations in the star catalog, not all stars can be removed; therefore, a multi-frame comparison method is needed for star removal. Since the vector of a star remains unchanged in the inertial frame, comparing the inertial frame vectors across multiple frames allows for the removal of remaining stars, leaving only potential target image points. The specific method is as follows:

[0077]

[0078] 3. Establishment of a suspected target queue

[0079] After stars are removed, the remaining image points are placed in the suspected target queue. Due to the limitations of the sensor resolution, targets moving at slower speeds cannot show any difference in motion from stars over a short period of time, but their motion trend becomes more obvious over a longer period of time.

[0080] Based on different target velocities, fast and slow suspected target queues are established. The frame interval for the fast suspected target queue is equal to the exposure frame period; the frame interval for the slow suspected target queue is the shortest time required to distinguish the target with the smallest angular velocity from the star.

[0081] Let the frame interval of the slow suspected queue be Δt, and the lower limit of the target's angular velocity be ω. l The multi-frame labeling threshold is Θ t The detection noise is θ n The frame interval of the slow queue is:

[0082]

[0083] Each frame of an image contains suspected target points that form a suspected queue of one frame. Suspected target points from multiple frames of images form a suspected target queue. A "sequence" refers to a sequence of target image points. A suspected target queue may not contain a sequence of target image points, or it may contain multiple sequences of target image points.

[0084] 4. Target motion detection

[0085] Suspected target images need to be analyzed for their motion patterns, and images that conform to the target's motion patterns should be retained. The target's angular velocity in the inertial frame is a continuous value, and the target's right ascension and declination conform to certain motion patterns.

[0086] Let the sequential positions of the target in the inertial frame be {V1, V2, ..., V...} n Then, by using the angular distance and time difference between two adjacent frames, the inter-frame angular velocity of the target can be calculated as ω1, ω2, ..., ω n-1 If the target angular velocity sequence is ω1, ω2, ..., ω n-1 Meets the preset target angular velocity threshold range [ω l ,ω u If the target's angular velocity sequence does not meet the set threshold range, then the entire sequence of target images is discarded.

[0087] Let the time of the target parameter queue relative to the first frame be t1, t2, ..., t n Right ascension and declination are α1, α2, ..., α n and δ1,δ2,…,δ n During the observation period, the right ascension and declination of the target in the inertial frame can be approximated by a first-order Taylor formula, i.e.:

[0088] Let the relationship between right ascension and time be:

[0089] The relationship between right ascension and time is: α j =k α t j +b α ,

[0090] The relationship between declination and time is: δ j =k δ t j +b δ Here, 1≤j≤n;

[0091] Using right ascension and time fitting, k is obtained. α b α and the coefficient of determination of fit Fit standard deviation Coefficient of determination Fit standard deviation The calculation method belongs to the well-known content of least squares fitting.

[0092] Similarly, by fitting the declination, we obtain k. δ b δ ,

[0093] like and If the queue of suspected target images conforms to the motion law, it is confirmed as a target, where fitThresh represents the fitting determination coefficient threshold.

[0094] 5. Target location prediction

[0095] After capturing a new image, the image time is substituted into the fitting formula for right ascension and declination to obtain the predicted right ascension and declination α. p ,δ p .

[0096] α p =k α t p +b α

[0097] δ p =k δ t p +b δ

[0098] Where t p This indicates the time of the newly captured image relative to the first frame of the target parameter queue.

[0099] Convert right ascension and declination into inertial frame vector V. inertial The attitude obtained through star chart identification is used to convert the inertial frame vector into the space-based optical sensor's body frame vector V. body The body vector of the space-based optical sensor is transformed into planar coordinates (u) through the camera calibration model. p ,v p ), which is the predicted location of the target.

[0100]

[0101] 6. Target point selection

[0102] The newly captured image in step 5 (u p ,v p Image points are extracted near the target location using a windowing method. If multiple image points are extracted, the right ascension and declination of the image points are compared with the target prediction values, and the image point with the highest confidence is selected as the target point of the current frame.

[0103] Right ascension α of all extracted image points cur δ declination cur The predicted value α is obtained. p ,δ p residual r α ,r δ .

[0104] r α =|α p -α cur |

[0105] r δ =|δ p -δ cur |

[0106] If the residual satisfies and If a pixel conforms to the target pattern, it is considered to meet the target pattern. If multiple pixels conform, the pixel with the smaller residual is selected, where outlierRatio is the residual discrimination threshold. The standard deviation of the right ascension fit. is the standard deviation of the declination fit.

[0107] 7. Target queue update

[0108] Based on the inertial attitude corresponding to the newly captured image in step 5, the right ascension and declination of the target image point are calculated. The right ascension and declination are then replaced with the earliest data in the queue in step 3 according to the first-in-first-out principle. The relationship between the target's right ascension and declination and time is then refitted for target position prediction and tracking.

[0109] Let the latest frame time be t. n+1 Right ascension is α n+1 Declination is δ n+1 The queue used for fitting is then updated to: t2, t3, ..., t n+1 Right ascension and declination are α2, α3, ..., α n+1 and δ2,δ3,…,δ n+1 .

[0110] Obtain the refitted parameters and assign them to k. α b α , and k δ b δ ,

[0111] 8. Repeat steps 5-7 to predict and track the target position in the next frame image.

[0112] The above-mentioned fitting parameters for each frame are used to select target points and predict target positions in the next frame of newly captured images using the methods in steps 5, 6, and 7 above, thereby maintaining target tracking.

[0113] This invention solves the technical challenge of high-reliability, wide dynamic range, spatial non-cooperative multi-target acquisition, tracking, and measurement, enabling on-orbit applications.

[0114] This invention is based on the characteristic that stars are stationary in an inertial frame while space debris is moving. It extracts image points from images, identifies and removes stars, and captures and tracks space debris based on its motion characteristics.

[0115] This invention is applicable to the acquisition of long-range visible light targets on high-orbit space-based platforms, and can be used for initial target acquisition and continuous tracking and measurement. It provides prior information for threat warning and guides close-range reconnaissance.

[0116] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. A space-based high-reliability wide dynamic range space debris capture and tracking method, characterized in that... include: Step 1. The space-based optical sensor acquires multiple frames of superimposed images of the star and the target, identifies the image points of the star using the triangle matching method, and removes the image points of the star from the image; Step 2. Based on the star catalog information, locate stars not identified by the triangle matching method and remove the corresponding image points from the image; Step 3. After removing star image points, the remaining image points are placed into the suspected target queue, which is divided into a fast suspected target queue and a slow suspected target queue according to the target speed. Step 4. Determine the motion patterns of suspected target image points, retain image points whose right ascension and declination conform to the motion patterns, and fit the parameters of the relationship between the target's right ascension and declination and time. Step 5. Take a new image frame, input the image time into the fitting formula of right ascension and declination, and obtain the predicted position of the target in the image; Step 6. Extract image points near the predicted position of the target in the newly captured image using a window method. If multiple image points are extracted, compare the right ascension and declination of the image points with the predicted position of the target in the image, and select the image point with the highest confidence as the target point of the current frame. Step 7. Based on the inertial attitude of the newly captured image, calculate the right ascension and declination of the target image point, and replace the earliest data in the suspected target queue with the first-in-first-out data according to the first-in-first-out principle, and refit the parameters of the relationship between the target's right ascension and declination and time. Step 8. Repeat steps 5-7 to predict and track the target position in the next frame image.

2. The space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 1, characterized in that: The method for identifying the image points of stars in step 1 is as follows: The angular distance between each pair of observed stars is calculated. The angular distances of the stars are found by searching the pre-stored navigation star database for values ​​that are close to the angular distances of the image points. The stars are then identified by a triangle matching method.

3. The space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 2, characterized in that: Step 2 involves finding stars not identified by the triangle matching method and removing the corresponding image points from the image. Specifically: First, the unit vector of image point i in the space-based optical sensor system is transformed to the inertial coordinate system. The vector of image point i in the inertial coordinate system is cross-multiplied with the vector of the navigation star in the star table in the inertial space to obtain the angular distance between image point i and the navigation star. If the angular distance between image point i and the navigation star is less than the star discrimination threshold, image point i is considered to be a star and is removed. Otherwise, the image point is retained. Furthermore, the angular distance between the two unremoved image points in any two frames in the inertial frame is compared with the multi-frame marking threshold. If the angular distance between the two unremoved image points in any two frames in the inertial frame is less than the multi-frame marking threshold, the points are marked as stars and the remaining stars are removed.

4. The space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 3, characterized in that: In step 3, after removing stellar image points, the remaining image points are placed into a suspected target queue. Based on the target's velocity, these are divided into a fast suspected target queue and a slow suspected target queue. The frame interval of the image containing the image point in the fast suspected target queue is equal to the exposure frame period; The frame interval of the image containing the image point in the slow-moving suspected target queue is the shortest time required to distinguish the target with the smallest angular velocity from the star. The specific calculation method is as follows: Let the frame interval of the image containing the image point in the slow suspected queue be Δt, and the lower limit of the target's angular velocity be ω. l The multi-frame labeling threshold is Θ t The detection noise is θ n Then the frame interval of the image containing the image point in the slow queue is 5. The space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 4, characterized in that: In step 4, motion patterns are determined for suspected target image points, and image points that conform to the target's motion patterns are retained. Specifically: (1) First, the inter-frame angular velocities of the target image points are calculated as ω1, ω2, ..., ω using the angular distance and time difference between corresponding image points in two adjacent frames. n-1 : If ω1, ω2, ..., ω n-1 Meets the preset target angular velocity threshold range [ω l ,ω u If the target is detected, further motion determination is performed. If ω1, ω2, ..., ω n-1 If the target angular velocity does not meet the set target angular velocity threshold range, then the entire sequence of target image points is discarded. (2) Let the time of the target parameter queue relative to the first frame be t1, t2, ..., t n Right ascension and declination are α1, α2, ..., α n and δ1,δ2,…,δ n During the observation period, the relationship between right ascension and time is as follows: a j =k α t j +b α , The relationship between declination and time is as follows: δ j =k δ t j +b δ Here, 1≤j≤n; By fitting the relationship between right ascension and time, the parameter k characterizing the relationship between right ascension and time is obtained. α b α and the coefficient of determination of right ascension fitting Right Ascension Fit Standard Deviation Similarly, the parameter k for fitting the relationship between declination and time... δ b δ and the coefficient of determination for declination fitting Declination fit standard deviation like and If the sequence of suspected target images conforms to the target's motion pattern, it is considered a target; otherwise, the entire sequence of target images is discarded. Here, fitThresh represents the fitting determination coefficient threshold.

6. The space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 5, characterized in that: In step 5, a new image is captured, and the image time is substituted into the fitting formula for right ascension and declination to obtain the predicted position of the target in the image, specifically: After capturing a new image, input the image time into the following formula: a p =k α t p +b α δ p =k δ t p +b δ The predicted right ascension position α of the target is obtained. p Declination position δ p And convert it into an inertial frame vector; where t p Indicates: the time of the newly captured image relative to the first frame of the target parameter queue; The attitude obtained through star map identification is further used to convert the inertial frame vector into the body frame vector of the space-based optical sensor; Next, the body vector of the space-based optical sensor is converted into planar coordinates through the camera calibration model, which is the predicted position of the target in the image plane.

7. A space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 6, characterized in that: In step 6, image points are extracted near the predicted location of the target in the newly captured image using a windowing method. If multiple image points are extracted, the right ascension and declination of the image points are compared with the predicted values ​​of the target, and the image point with the highest confidence is selected as the target point of the current frame. Specifically: Using the right ascension α of all extracted image points cur δ declination cur And the predicted value of right ascension α p The predicted value of declination δ p Find the residual r α r δ : r α ＝|α p -α cur | r δ =|δ p -d cur | If the residual satisfies and If a pixel conforms to the target pattern, it is considered to meet the target pattern. If multiple pixels conform, the pixel with the smallest residual, i.e., the pixel with the highest confidence, is selected. Here, `outlierRatio` is the residual discrimination threshold. The standard deviation of the right ascension fit. denoted as the standard deviation of the declination fit.

8. The space-based high-reliability wide dynamic range space debris capture and tracking method according to claim 7, characterized in that: In step 7, based on the inertial attitude of the newly captured image, the right ascension and declination of the target image point are calculated. Then, the right ascension and declination are replaced with the earliest data in the suspected target queue according to a first-in, first-out principle. The parameters relating the target's right ascension and declination to time are then refitted. Specifically: Let the latest frame time be t. n+1 Right ascension is α n+1 Declination is δ n+1 The queue used for fitting is then updated to: t2, t3, ..., t n+1 Right ascension and declination are α2, α3, ..., α n+1 and δ2, δ3, ..., δ n+1 ,get: (1) The parameters characterizing the relationship between right ascension and time, as well as the right ascension fitting determination coefficient and right ascension fitting standard deviation, are refitted and assigned to k. α b α , (2) The parameters characterizing the relationship between declination and time, as well as the declination fitting determination coefficient and the declination fitting standard deviation, are refitted and assigned to k. δ b δ , Based on this, the target location in subsequent captured images is predicted and extracted, thus enabling continuous tracking.

Images