Space spin non-cooperative target capturing method and system

Abstract

The application provides a space spin non-cooperative target capturing method and system, and the method comprises the following steps: pre-processing a current image, adjusting the gray mean value of the image to a preset range; collecting point cloud data of a target star in real time; performing point cloud matching on a point cloud data frame; state estimation, obtaining the final position and attitude of the target object; and adjusting the end speed and joint angle through visual servo control. The application solves the technical problems of relying on the magnetic properties of the target material, being easily interfered by factors such as a magnetic field, strong radiation and high temperature difference, and low control accuracy in the initial training period.

Description

Technical Field

[0001] This invention relates to the field of robotics, specifically to a method and system for capturing non-cooperative targets using spatial spin. Background Technology

[0002] With the continued increase in human space activities, the issue of space debris has become a global concern. Space orbits are widely used for communication, meteorological, and reconnaissance satellites. To address this challenge, on-orbit servicing technologies have developed significantly in recent years, including satellite repair, refueling, and life extension. These technological advancements necessitate effective methods for recovering and disposing of defunct satellites to ensure the sustainability of the orbital environment. Recovery technologies can not only prevent defunct satellites from becoming space debris but also recover valuable resources, reducing the need to launch new satellites and thus lowering costs.

[0003] The rapid development of the commercial space sector is also a significant factor driving advancements in satellite recovery technology. With the rise of private companies like SpaceX and Blue Origin, commercial space activities are becoming increasingly frequent. Effective satellite recovery technology not only protects existing commercial assets but also provides these companies with new service areas, such as in-orbit satellite maintenance and upgrades. This technology not only extends the lifespan of satellites but also improves their economic efficiency, further propelling the development of the commercial space market.

[0004] However, satellite recovery technology faces several technical challenges. First, high-precision orbit measurement and control require accurate calculation and prediction of the satellite's and its debris's orbits to achieve safe approach and capture. Second, highly reliable capture and recovery technology must operate in high vacuum and microgravity environments, while also dealing with the complexities of satellite rotation and irregular shapes. Furthermore, the complex space operation environment places higher demands on the system's autonomy and intelligence.

[0005] In summary, the research and application of satellite recovery technology not only helps solve the space debris problem and ensure orbital safety, but also promotes the overall progress of aerospace technology. The successful implementation of this technology will lay a solid foundation for more complex future space missions, including deep space exploration, on-orbit assembly, and maintenance of large space structures. With continuous technological advancements and deepening international cooperation, satellite recovery technology will undoubtedly play an increasingly important role in future aerospace endeavors.

[0006] The existing invention patent application document CN114674179A, entitled "A Capture System and Method for Space Targets," describes a system that employs multiple capture devices. Through the cooperation of a transmission and launch device, these devices can be sequentially launched to the target orbit, achieving efficient capture of failed space targets. The design of multiple capture devices improves the system's capture efficiency, enabling multiple captures in a single mission. However, the magnetic capture mechanism in this existing system has limited effectiveness against non-metallic or non-magnetic targets. This capture method relies on the magnetic properties of the target material, making it difficult to effectively capture targets made of non-magnetic materials. Furthermore, magnetic capture may be affected by other magnetic fields, reducing the stability and reliability of the capture process.

[0007] The existing invention patent application document CN108508749A, entitled "An Anti-interference Iterative Learning Control Method for a Space Robotic Arm System for Capturing Non-cooperative Targets," describes an existing technology that uses iterative learning control to enable a space robotic arm to capture non-cooperative targets in complex space environments and continuously optimize its control strategy to improve operational accuracy and anti-interference capabilities. However, in extreme space environments, such as those with strong radiation or extreme temperature differences, the system's stability and performance may still be affected. Furthermore, the initial training period and data quality also significantly impact the system's control accuracy.

[0008] In summary, existing technologies suffer from technical problems such as reliance on the magnetic properties of the target material, susceptibility to interference from factors such as magnetic fields, strong radiation, and temperature differences, and low control precision in the early stages of training. Summary of the Invention

[0009] The technical problem to be solved by this invention is: how to solve the technical problems of existing technologies that rely on the magnetic properties of the target material, are easily affected by factors such as magnetic fields, strong radiation, and high temperature differences, and have low control accuracy in the early stage of training.

[0010] This invention solves the above-mentioned technical problems by employing the following technical solution: A space spin non-cooperative target capture method includes: S1. Obtain and preprocess the current image, adjust the average grayscale value of the current image to a preset range, and obtain the preprocessed image; S2. Using a TOF camera, point cloud data of the target star is acquired in real time based on the pre-processed image; S3. Perform point cloud matching operation on the point cloud data frame of the point cloud data to obtain the relative pose; S4. Based on the relative attitude, combined with the initial attitude and position of the template point cloud, perform state estimation and calculate the final position and attitude of the target object; S5. Through visual servo control, based on the principle of relative motion and the final position and posture of the target, the absolute motion velocity of the target is obtained; the differential kinematic equations of the space robot are constructed, and the motion velocity of the end effector of the robotic arm is planned.

[0011] This invention provides a spin-based non-cooperative target capture technology based on a space-driven, agile robotic arm. By employing a highly flexible and agile cable-driven system, the robotic arm can accurately lock onto and capture these spin-based targets, ensuring the success rate and safety of the operation. Furthermore, the technology's stability and performance in extreme space environments have been verified through thermal vacuum experiments and microgravity tests.

[0012] In a more specific technical solution, S1 includes: S11. Calculate the mean gray level mP of the current image; S12. The system sets a predetermined grayscale value threshold G; S13. Determine whether the absolute value of the difference between the image grayscale mean mP and the set grayscale threshold G satisfies: |mP-G|>3, and obtain the threshold range determination result; S14. Adjust the exposure time based on the threshold range determination result.

[0013] In a more specific technical solution, S3 includes: S31. In the point cloud matching operation, determine whether the current point cloud data frame is the first frame; if so, execute S32 for the point cloud data of the first frame: S32. For the point cloud data of the first frame, use the Random Sample Consensus (RANSAC) algorithm for coarse registration; S33. Perform precise registration using the generalized iterative nearest point (GICP) algorithm; S34. Using the NICP algorithm, perform non-rigid registration on the point cloud data of subsequent frames.

[0014] In a more specific technical solution, in S33, the transformation matrices of the reference point cloud and the target point cloud are initialized, and the covariance matrix of each point is calculated to characterize the local geometric properties of the point. In each iteration, for each point in the reference point cloud data, find the nearest corresponding point pair in the target point cloud, combine the covariance matrix to optimize the transformation matrix, minimize the error objective function, and iteratively update the position of the reference point cloud until the transformation matrix converges or the maximum number of iterations is reached, and output the optimized transformation matrix and aligned reference point cloud.

[0015] In a more specific technical solution, S33 utilizes the following logic to perform fine registration based on the GICP algorithm:

[0016] In the formula,T Represents the transformation matrix. p i Points representing the reference point cloud, q i Represents the points in the target point cloud.

[0017] In a more specific technical solution, S34 initializes the transformation model and calculates the initial position of each point; In the target point cloud, the nearest corresponding point is found to align each point in the reference point cloud, forming an initial matching point pair; Based on the initial matching point pairs, update the deformation model and introduce smoothing constraints; Construct an error function, and iteratively optimize by minimizing the error function to update the position and deformation parameters of the points; Through iterative operations, the initial matching point pairs and deformation models are updated until the deformation models converge or the maximum number of iterations is reached; the optimized deformation model and aligned point cloud are obtained, and non-rigid registration is completed.

[0018] In a more specific technical solution, within S34, the NICP algorithm uses the following logic to process the point cloud data of subsequent frames:

[0019] In the formula, This represents a point from the previous frame. p i Represents the point in the current frame. q j Let λ represent the target point, and λ is a balancing parameter used to weigh two errors.

[0020] This invention enables the system to maintain high-precision target tracking and capture capabilities in noisy environments through multi-step point cloud processing and attitude estimation. In particular, the use of RANSAC, GICP, and NICP algorithms for coarse and fine registration of point clouds improves the accuracy and robustness of attitude estimation.

[0021] In a more specific technical solution, S5 uses the following logic to plan the desired end-effector velocity:

[0022]

[0023] In the above formula, E v h , E ω h The absolute velocity of the target in E The representation in Δ OThis refers to the attitude pointing error; Based on the principle of relative motion, the absolute velocity of the target is obtained using the following logic:

[0024]

[0025] In the formula, E v e h The linear velocity of the target star's handle relative to the robotic arm's end effector, expressed in the end effector coordinate system:

[0026]

[0027] Determine the target's velocity using the following logic:

[0028] According to the differential kinematic equations of the space robot, we have:

[0029] in, E J g The Jacobian matrix of the robotic arm's end effector. θ Given the current angular velocity of the robotic arm, derived from the above formula:

[0030] Plan the end effector speed of the robotic arm according to the following logic:

[0031] In a more specific technical solution, the attitude pointing error Δ is calculated according to the following logic. O for:

[0032] In the formula, [ n e , o e , a e ]and[ n h , o h , a h They are respectively E and H The attitude transformation matrix, EAH This is the attitude matrix of the handle relative to the end-effector coordinate system, measured by the hand-eye camera. It is a weight matrix.

[0033] The robotic arm of this invention employs advanced sensor technology and precision control algorithms at its end effector, enabling it to sense and adjust gripping force and angle in real time, thereby improving gripping stability and accuracy. Simultaneously, the use of a rope-driven system gives the robotic arm greater flexibility and responsiveness, allowing it to adapt to complex and dynamic spatial environments.

[0034] In more specific technical solutions, the space spin non-cooperative target capture system includes: The grayscale preprocessing module is used to acquire and preprocess the current image, adjusting the average grayscale value of the current image to a preset range; The point cloud acquisition module is used to acquire point cloud data of the target star in real time using a TOF camera based on preprocessed images. The point cloud acquisition module is connected to the grayscale preprocessing module. The point cloud matching module is used to perform point cloud matching operations on point cloud data frames to obtain relative poses. The point cloud matching module is connected to the point cloud acquisition module. The state estimation module is used to estimate the state based on the relative attitude, combined with the initial attitude and position of the template point cloud, and calculate the final position and attitude of the target object. The state estimation module is connected to the point cloud matching module. The visual servo module is used to obtain the absolute motion velocity of the target based on its final position and orientation according to the principle of relative motion through visual servo control; it constructs the differential kinematic equations of the space robot and plans the motion velocity of the robotic arm's end effector. The visual servo module is connected to the state estimation module.

[0035] The present invention has the following advantages over the prior art: This invention provides a spin-based non-cooperative target capture technology based on a space-driven, agile robotic arm. By employing a highly flexible and agile cable-driven system, the robotic arm can accurately lock onto and capture these spin-based targets, ensuring the success rate and safety of the operation. Furthermore, the technology's stability and performance in extreme space environments have been verified through thermal vacuum experiments and microgravity tests.

[0036] This invention enables the system to maintain high-precision target tracking and capture capabilities in noisy environments through multi-step point cloud processing and attitude estimation. In particular, the use of RANSAC, GICP, and NICP algorithms for coarse and fine registration of point clouds improves the accuracy and robustness of attitude estimation.

[0037] The invented robotic arm employs advanced sensor technology and precision control algorithms at its end effector, enabling it to sense and adjust gripping force and angle in real time, thereby improving gripping stability and accuracy. Simultaneously, the use of a cable-driven system gives the robotic arm greater flexibility and responsiveness, allowing it to adapt to complex and dynamic spatial environments.

[0038] This invention solves the technical problems of existing technologies, such as dependence on the magnetic properties of the target material, susceptibility to interference from factors such as magnetic fields, strong radiation, and high temperature differences, and low control accuracy in the early stages of training. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the basic steps of the space spin non-cooperative target capture method in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the basic data stream processing of the space spin non-cooperative target capture method in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram illustrating the specific steps of preprocessing the current image in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram illustrating the specific steps of point cloud registration in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the point cloud registration data stream in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram illustrating the specific steps of visual servoing in Embodiment 1 of the present invention; Figure 7 This is a schematic diagram of visual servo data stream processing in Embodiment 1 of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, 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.

[0041] Example 1 like Figure 1 and Figure 2 As shown, the space spin non-cooperative target capture method provided by the present invention includes the following basic steps: S1. Preprocess the current image to adjust the average grayscale value to a preset range; like Figure 3 As shown, in this embodiment, step S1 of preprocessing the current image further includes the following specific steps: S11. Calculate the mean gray level mP of the current image; S12. The system sets a predetermined grayscale value threshold G; S13. Determine whether the absolute value of the difference between the grayscale mean mP and the set grayscale threshold G satisfies: |mP-G|>3; In this embodiment, the system compares the calculated grayscale mean value mP with the set grayscale threshold G. If the absolute value of the difference is greater than 3, it indicates that the brightness of the current image is not within an acceptable range and the exposure time needs to be adjusted; otherwise, the image brightness is considered to be within an acceptable range and no adjustment is required.

[0042] S14. If the difference between the gray mean mP and the threshold G is greater than 3, then further determine whether the difference between mP and G is positive or negative. S15. If mP - G < -3, it means the current image is too dark, so increase the exposure time. In this embodiment, the system increases the exposure time by N microseconds; If mP - G > 3, it means the current image is too bright and the exposure time needs to be reduced. In this case, the system will reduce the exposure time by N microseconds, which can be, for example, 50 μs.

[0043] S16. If not, reduce the exposure time; In this embodiment, the reduction in exposure time can be, for example, 50 μs. After the above adjustment, the system will continue image acquisition. If the adjusted average grayscale value of the image is still not within the predetermined range, the system will repeat the above steps until the average grayscale value mP of the image approaches the set grayscale threshold G, ensuring that the quality of the acquired image meets the requirements.

[0044] S2. Real-time acquisition of point cloud data of the target star; In this embodiment, a TOF camera is used to acquire point cloud data of the target star in real time. To improve the quality of the point cloud data, a pass-through filter is first performed to retain the point cloud data of the region of interest; then a voxel filter is performed to reduce the amount of point cloud data by dividing the point cloud into voxels of fixed size while preserving the overall shape; finally, a statistical filter is performed to remove isolated points and noise points, resulting in smooth and clear point cloud data.

[0045] S3. Perform point cloud matching on the point cloud data frame; like Figure 4 and Figure 5 As shown, in this embodiment, step S3 of point cloud registration further includes the following specific steps: S31. In the point cloud matching process, first determine whether the current point cloud data is the first frame; if it is the first frame of point cloud data, then the following processing steps are adopted: S32. For the first frame of point cloud data, use the RANSAC (Random Sample Consensus) algorithm for coarse registration; In this embodiment, the RANSAC algorithm can quickly find the approximate relative pose by randomly selecting a subset and iteratively optimizing. This process can filter out most noise points and determine the preliminary point cloud alignment relationship.

[0046] S33. Use the GICP algorithm for precise registration; In this embodiment, after coarse registration, the GICP (Generalized Iterative Closest Point) algorithm is used for precise registration. The GICP algorithm can handle a wide range of pose errors, exhibiting high registration accuracy and fast convergence speed. GICP optimizes the covariance matrix at each point, considering the local geometric features of the point cloud data, thus further refining the alignment of the point cloud and obtaining accurate relative pose. First, the transformation matrices of the reference and target point clouds are initialized, and the covariance matrix of each point is calculated to characterize its local geometric properties. Then, in each iteration, each point in the reference point cloud finds its nearest corresponding point pair in the target point cloud. By combining the covariance matrix information, the transformation matrix is ​​optimized to minimize the error objective function. This process is repeated continuously, updating the position of the reference point cloud until the transformation matrix converges or the maximum number of iterations is reached. Finally, the algorithm outputs the optimized transformation matrix and the precisely aligned reference point cloud. The fine registration formula of the GICP algorithm is as follows:

[0047] in, T Represents the transformation matrix. p i Points representing the reference point cloud, q i Represents the points in the target point cloud.

[0048] S34. Use the NICP algorithm to process the point cloud data of subsequent frames; In this embodiment, if the current point cloud data is not the first frame, the point cloud data of the previous frame is used as a reference, and point cloud registration is performed using the NICP (Non-rigid ICP) algorithm. The NICP algorithm can better handle local deformation and small pose changes, maintain the continuity and accuracy of point cloud data, and is suitable for pose estimation in dynamic environments. The algorithm flow first initializes the transformation model and calculates the initial position of each point; then, it finds the nearest corresponding point in the target point cloud for each point in the reference point cloud, forming an initial matching point pair; next, it updates the deformation model based on the matching point pair and introduces a smoothing constraint to ensure the deformation consistency between adjacent points; it constructs an error function, and iteratively optimizes by minimizing the error function, updating the position and deformation parameters of the points; iteratively updates the matching point pair and deformation model until convergence or the maximum number of iterations is reached; finally, it outputs the optimized deformation model and the aligned point cloud, completing the non-rigid registration.

[0049] The NICP algorithm is optimized when processing point cloud data in subsequent frames using the following formula:

[0050] in, This represents a point from the previous frame. p i Represents the point in the current frame. q j Let λ represent the target point, and λ is a balancing parameter used to weigh two errors.

[0051] S4. State estimation: Obtain the final position and orientation of the target object; In this embodiment, after obtaining the relative pose through point cloud matching, the final position and pose of the target object are calculated by combining the initial pose and position of the template point cloud. This process includes fusing data from multiple point cloud frames to eliminate noise and errors, resulting in a stable estimate of the target position and pose. Using this information, a state estimation model can be constructed to predict the motion state of the target object, providing reliable data support for the subsequent visual servo control of the robotic arm and ensuring that the system can approach and capture the target object in real time and accurately.

[0052] S5. Adjust the end effector speed and joint angle through visual servo control.

[0053] In this embodiment, the desired end-effector velocity is planned using the following formula:

[0054]

[0055] In the above formula, E v h , E ω h The absolute velocity of the target in E The representation in Δ O The orientation error (i.e., the orientation of the handle coordinate system relative to the robot arm end effector coordinate system) is calculated using the following formula:

[0056] in,[ n e , o e , a e ]and[ n h , o h , a h They are respectively E and H The attitude transformation matrix (relative to the inertial frame), and EAH This is the attitude matrix of the handle relative to the end-effector coordinate system, measured by the hand-eye camera. As a weight matrix, since the target being grasped is a spinning, non-cooperative target with a very large range of attitude changes, while the robotic arm's motion space is limited, therefore, The value can be adjusted according to the strategy shown in the figure below, combined with different working conditions. Proportional coefficient It is determined based on the limitations of the linear velocity and angular velocity at the end of the robotic arm.

[0057] According to the principle of relative motion, the absolute velocity of the target can be calculated using the following formula:

[0058]

[0059] in, E v e h The linear velocity of the target star's handle relative to the end effector of the robotic arm, expressed in the end effector coordinate system, is as follows:

[0060]

[0061] Target's speed:

[0062] According to the differential kinematic equations of the space robot, we have:

[0063] in, E J g The Jacobian matrix of the robotic arm's end effector. θ Given the current angular velocity of the robotic arm, we have:

[0064] Finally, the end effector speed of the robotic arm can be planned using the following formula:

[0065] In practical applications, the movement speed of the robotic arm's end effector is generally not allowed to be too high. Therefore, when selecting K, it is best to specify the upper limit of the speed.

[0066] like Figure 6 and Figure 7 As shown, in this embodiment, step S5 of visual servoing further includes the following specific steps: S51. Input the current joint angle, target measurement confidence, and pose; S52. Determine if the vision servoing time has expired; S53. If so, then visual servo capture is deemed to have failed. S54. If not, determine whether the camera measurement data is reliable; S55. If not, the current joint angle and angular velocity are the output joint angle and angular velocity; S56. If so, has the target reached the capture area? S57. If so, the capture flag is set to 1; S58. If not, the capture flag is 0; S59. Correct the position difference to within the threshold; S510, visual servo end-effector speed planning; S511. Plan the terminal velocity based on the terminal velocity threshold; S512, Calculate the Jacobian matrix between the end-effector pose and joint angles of the robotic arm; S513. Calculate the pose of the target position relative to the robot arm's base coordinate system; S514. Calculate the pose difference (K*dp) between the target position and the end effector of the robotic arm. S515, speed-level Jacobian pseudo-inverse solution for manipulator joint angle increment; S516. Correct the joint angle based on the joint angular velocity threshold; S517. Correct the joint angle based on the joint angle threshold; S518. Determine if the capture flag is 1; S519. If yes, the robotic arm reaches the capture position; if no, it jumps to the start.

[0067] In summary, this invention provides a spin-based non-cooperative target capture technology based on a space-driven, agile robotic arm. By employing a highly flexible and agile cable-driven system, the robotic arm can accurately lock onto and capture these spin-based targets, ensuring the success rate and safety of the operation. Furthermore, the technology's stability and performance in extreme space environments have been verified through thermal vacuum experiments and microgravity tests.

[0068] This invention enables the system to maintain high-precision target tracking and capture capabilities in noisy environments through multi-step point cloud processing and attitude estimation. In particular, the use of RANSAC, GICP, and NICP algorithms for coarse and fine registration of point clouds improves the accuracy and robustness of attitude estimation.

[0069] The invented robotic arm employs advanced sensor technology and precision control algorithms at its end effector, enabling it to sense and adjust gripping force and angle in real time, thereby improving gripping stability and accuracy. Simultaneously, the use of a cable-driven system gives the robotic arm greater flexibility and responsiveness, allowing it to adapt to complex and dynamic spatial environments.

[0070] This invention solves the technical problems of existing technologies, such as dependence on the magnetic properties of the target material, susceptibility to interference from factors such as magnetic fields, strong radiation, and high temperature differences, and low control accuracy in the early stages of training.

[0071] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for capturing non-cooperative targets using space spin, characterized in that: The method includes: S1. Acquire and preprocess the current image, adjust the average grayscale value of the current image to a preset range, and obtain a preprocessed image; S2. Using a TOF camera, collect point cloud data of the target star in real time based on the preprocessed image; S3. Perform point cloud matching operation on the point cloud data frames of the point cloud data to obtain the relative pose; S3 includes: S31. In the point cloud matching operation, determine whether the current point cloud data frame is the first frame; if so, execute S32 for the point cloud data of the first frame: S32. For the point cloud data of the first frame, use the Random Sample Consensus (RANSAC) algorithm for coarse registration; S33. Perform precise registration using the generalized iterative nearest point (GICP) algorithm; S34. Using the NICP algorithm, perform non-rigid registration on the point cloud data of subsequent frames; S4. Based on the relative attitude, combined with the initial attitude and position of the template point cloud, perform state estimation and calculate the final position and attitude of the target object; S5. Using visual servo control, based on the principle of relative motion and according to the final position and posture of the target object, the absolute motion speed of the target is obtained; the differential kinematic equations of the space robot are constructed, and the motion speed of the robotic arm end effector is planned.

2. The method for capturing non-cooperative targets using space spin according to claim 1, characterized in that, S1 includes: S11. Calculate the average gray level mP of the current image; S12. The system sets a predetermined grayscale value threshold G; S13. Determine whether the absolute value of the difference between the average gray level mP of the image and the set gray level threshold G satisfies: |mP-G|>3, and obtain the threshold range determination result; S14. Adjust the exposure time based on the threshold range determination result.

3. The method for capturing non-cooperative targets using space spin as described in claim 1, characterized in that, In step S33, the transformation matrices of the reference point cloud and the target point cloud are initialized, and the covariance matrix of each point is calculated to characterize the local geometric properties of the points in the point cloud data. In each iteration, referencing each point in the point cloud data, find the nearest corresponding point pair in the target point cloud, combine the covariance matrix, optimize the transformation matrix, minimize the error objective function, iteratively update the position of the reference point cloud until the transformation matrix converges or the maximum number of iterations is reached, and output the optimized transformation matrix and aligned reference point cloud.

4. The method for capturing a non-cooperative target by a space spin according to claim 1, characterized in that, In step S33, the precise registration is performed using the GICP algorithm based on the following logic: In the formula, T Represents the transformation matrix. p i Points representing the reference point cloud, q i Represents the points in the target point cloud.

5. The spatial spin non-cooperative target capture method according to claim 3, characterized in that, Initialize the transformation model and calculate the initial position of each point; In the target point cloud, find the nearest corresponding point for each point in the alignment reference point cloud to form an initial matching point pair; Based on the initial matching point pairs, update the transformation model and introduce smoothing constraints; Construct an error function, and iteratively optimize by minimizing the error function to update the position and deformation parameters of the point; The initial matching point pairs and the transformation model are updated through iterative operations until the transformation model converges or the maximum number of iterations is reached. The optimized transformation model and aligned point cloud are obtained, and non-rigid registration is completed.

6. The method for capturing a non-cooperative target by a space spin according to claim 1, characterized in that, In step S34, the NICP algorithm processes the point cloud data of subsequent frames using the following logic: In the formula, This represents a point from the previous frame. Represents the point in the current frame. Let λ represent the target point, and λ is a balancing parameter used to weigh two errors.

7. The method for capturing a non-cooperative target by a space spin according to claim 1, characterized in that, In S5, the desired end-effector velocity is planned using the following logic: In the above formula, Δ O This refers to the attitude pointing error; Based on the principle of relative motion, the absolute velocity of the target is obtained using the following logic. : In the formula, E v e h The linear velocity of the target star's handle relative to the robotic arm's end effector, expressed in the end effector coordinate system: Determine the target's velocity using the following logic: According to the differential kinematic equations of the space robot, we have: in, E J g Let Jacobian matrix be the end effector matrix of the robotic arm. θ Given the current angular velocity of the robotic arm, derived from the above formula: The end effector speed of the robotic arm is planned according to the following logic: 。 8. The method for capturing a non-cooperative target by a space spin according to claim 7, characterized in that, The attitude pointing error Δ is calculated according to the following logic. O for: In the formula, [ n e , o e , a e ]and[ n h , o h , a h They are respectively E and H The attitude transformation matrix, This is the attitude matrix of the handle relative to the end effector coordinate system, measured by a hand-eye camera. It is a weight matrix.

9. A space spin non-cooperative target capture system, used to execute the space spin non-cooperative target capture method according to any one of claims 1 to 8, characterized in that, The system includes: The grayscale preprocessing module is used to acquire and preprocess the current image, adjusting the average grayscale value of the current image to a preset range; The point cloud acquisition module is used to acquire point cloud data of the target star in real time based on the preprocessed image using a TOF camera. The point cloud acquisition module is connected to the grayscale preprocessing module. The point cloud matching module is used to perform point cloud matching operations on the point cloud data frames of the point cloud data to obtain the relative pose. The point cloud matching module is connected to the point cloud acquisition module. The state estimation module is used to estimate the state based on the relative attitude, combined with the initial attitude and position of the template point cloud, and calculate the final position and attitude of the target object. The state estimation module is connected to the point cloud matching module. The visual servo module is used to obtain the absolute motion velocity of the target by processing the final position and posture of the target object according to the principle of relative motion through visual servo control; to construct the differential kinematic equations of the space robot and plan the motion velocity of the end effector of the robotic arm; the visual servo module is connected to the state estimation module.

Images