A three-axis stabilized space target attitude estimation method

Abstract

The present invention discloses a method for estimating the attitude of a three-axis stabilized space target, comprising the following steps: extracting the target contour of each frame from a sequence of ISAR images; matching the extracted target contours with pre-generated template library contours, and outputting a list of optimal matching attitudes for each frame; calculating the imaging projection matrix of each frame based on the number of target orbital elements and the position of a ground observation station, performing attitude conversion on the attitude list of each frame, and obtaining a transformed attitude list of each frame in a satellite-based orbital coordinate system; performing attitude defuzzification on the attitude sequences of each frame, and outputting a unique attitude as a coarse attitude estimation result; using the coarse attitude estimation result as an initial value, performing an optimization search in the neighborhood of the result, using the three-dimensional attitude of the target model as an optimization variable, optimizing the average Hausdorff distance between its theoretical projection contour and the actual imaging contour to a minimum, and outputting an accurate target attitude. The present invention can achieve high-precision and high-efficiency attitude estimation of a three-axis stabilized space target.

Description

Technical Field

[0001] The present invention belongs to the field of target recognition based on ISAR images, and in particular relates to a three-axis stabilized space target attitude estimation method. Background Art

[0002] With the rapid development of space technology, effective space target state estimation and collision avoidance strategies are particularly important for ensuring space safety. The on-orbit attitude of a space target, as a key parameter, is crucial for determining its motion state. Estimating the attitude of a space target allows for diagnosis and analysis of the operating conditions of cooperative targets and identification of potential abnormal behavior of non-cooperative targets. Furthermore, estimating the target's attitude helps predict and correct its positional changes, preventing possible collisions with space debris. Accurate attitude estimation is also crucial for servicing missions and spacecraft docking. Compared to optical sensors, ISAR (inverse synthetic aperture radar) offers all-day, all-weather capabilities, circumventing daytime and weather restrictions. Therefore, ISAR offers significant advantages in space target attitude estimation.

[0003] Most existing ISAR-based three-axis stabilized space target attitude estimation methods start with the target model, extract features such as the satellite body and solar panels, and then perform feature recognition. They have the advantages of intuitive geometric features and clear attitude interpretation methods, and are the main methods for space target attitude estimation. However, the feature extraction of target components is difficult, resulting in the need to improve the efficiency and accuracy of the attitude estimation algorithm. Summary of the Invention

[0004] The purpose of the present invention is to provide a three-axis stabilized space target attitude estimation method, computer equipment, computer-readable storage medium and computer program product, which can solve the problems of low estimation efficiency and accuracy of existing methods.

[0005] In order to achieve the above object, one aspect of the present invention provides a three-axis stabilized space target attitude estimation method, comprising:

[0006] Step S1: extracting the target contour of each frame from the sequential ISAR image based on the Clean algorithm;

[0007] Step S2: Match the extracted target contours with the pre-generated template library contours respectively, use Hausdorff distance as the matching operator, and output a list of the first several postures with the smallest distance in each frame;

[0008] Step S3: Calculate the imaging projection matrix of each frame according to the number of target orbit elements and the position of the ground observation station, perform attitude conversion on the attitude list of each frame, and obtain the attitude list of each frame after conversion in the satellite-based orbit coordinate system;

[0009] Step S4: For the obtained pose list after conversion of each frame, a candidate pose is taken from each frame, all possible pose sequences are traversed, and the pose sequence with the smallest variance is found as the rough pose estimation result;

[0010] Step S5: Using the rough pose estimation result as the initial value, perform optimization search on the neighborhood of the result, use the 3D pose of the target model as the optimization variable, optimize the average Hausdorff distance between its theoretical projection profile and actual imaging profile to the minimum, and output the precise target pose.

[0011] Preferably, the step S1 includes:

[0012] Step S11, extracting all maximum scattering points of each frame image based on the Clean algorithm;

[0013] Step S12: using a connectivity analysis method to remove noise from the point set obtained in step S11;

[0014] Step S13: For the point set after noise removal in step S12, use Alpha-Shape and wavelet filtering to obtain the target contour.

[0015] Preferably, the template library in step S2 is generated as follows:

[0016] Assume that the positions of each point of the target at zero posture are , then the target point cloud position under any posture is:

[0017]

[0018] in, , , Represents yaw, pitch, and roll angles, respectively. For around Axis rotation angle The rotation matrix of For around Axis rotation angle The rotation matrix of For around Axis rotation angle The rotation matrix of

[0019] Traversal , , , establish the target position under all postures, and project the target three-dimensional model under each posture into a two-dimensional plane to form a two-dimensional closed contour, thereby generating a template library.

[0020] Preferably, step S3 includes:

[0021] Step S31: Construct the projection matrix:

[0022]

[0023] in, and are the distance and azimuth projection vectors respectively, and are the distance and azimuth unit vectors respectively, and are range and azimuth resolution, respectively;

[0024] Step S32: According to the number of target orbital elements and the position of the ground observation station, the radar line of sight at the middle moment of a single frame of imaging is calculated as the distance dimension unit vector, the range resolution is obtained by the radar parameters, and the azimuth dimension unit vector and azimuth resolution are obtained by calculating the LOS angle of a single frame, thereby obtaining the projection matrix. ;

[0025] Step S33: perform image registration on the posture list of each frame to achieve index posture from the template To actual posture conversion.

[0026] Preferably, in step S33, the following posture conversion formula is used to realize the index posture of the template: To actual posture Conversion:

[0027]

[0028] in, The quaternion required to rotate from the zero-pose target to the actual pose The vector expansion of :

[0029] ,

[0030] in, Represents the quaternion of the azimuth axis rotation, the quaternion of the distance axis rotation, and the index attitude The corresponding rotation matrix The converted quaternion.

[0031] Preferably, step S4 includes:

[0032] Step S41: Take the top matching postures of each frame from the converted posture list obtained in step S3;

[0033] Step S42: Using the minimum Euclidean distance method of posture angles, select a posture from the first frame, find the posture closest to it in the second frame, and continue this method until the last frame to form a posture sequence, and calculate the sequence variance;

[0034] Step S43, traverse the several postures taken in the first frame, and repeat step S42 to obtain several sequences and variances;

[0035] Step S44: Take the sequence with the smallest variance in the sequence of step S43, calculate the mean value and obtain a rough pose estimate of the target.

[0036] Preferably, in step S5, a PID-based search algorithm is used, and the optimization function is:

[0037]

[0038] in, For the The actual ISAR image of the frame, For the The projection matrix of the frame, is the position of the zero-attitude target, is the attitude rotation matrix, is the contour extraction function, is a function for calculating the Hausdorff distance between two contours, , , ,in, is the rough pose estimation result.

[0039] Another aspect of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0040] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0041] Yet another aspect of the present invention provides a computer program product, comprising a computer program, which implements the steps of the above method when executed by a processor.

[0042] The three-axis stabilized space target attitude estimation method, computer device, computer-readable storage medium and computer program product according to the above aspects of the present invention can achieve high-precision and high-efficiency attitude estimation of the three-axis stabilized space target. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the present invention, the following briefly introduces the drawings used in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive efforts.

[0044] Figure 1 is a flow chart of a method for estimating a three-axis stabilized space target attitude according to an embodiment of the present invention;

[0045] Figure 2 1 is a schematic diagram of a method for estimating a three-axis stabilized space target posture according to an embodiment of the present invention;

[0046] Figure 3 This is an example diagram of performing posture deblurring by combining posture sequences of each frame according to an embodiment of the present invention;

[0047] Figure 4 It is a structural diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0049] An embodiment of the present invention provides a method for estimating the attitude of a three-axis stabilized space target. Figure 1 As shown, the method of the embodiment of the present invention includes steps S1 to S5. Figure 2 In this embodiment, the radar system simulation parameters are: radar bandwidth 700MHz, wavelength 0.05m, pulse repetition frequency 50Hz, pulse width 4 The radar transmits a linear frequency modulation signal and generates five frames of images at equal intervals in a single observation cycle. The number of slow-time echo accumulation is 256, and the signal-to-noise ratio of a single echo is 10dB.

[0050] Step S1: extracting the space target contour from the sequence ISAR image.

[0051] In this step, the ISAR image target contour extraction method based on the Clean algorithm with improved iteration number is used to extract the target contour of each frame image.

[0052] Furthermore, the step S1 includes steps S11 to S13.

[0053] Step S11 : Record the maximum scattering point in each frame image, subtract the point spread function (PSF) from the image, and record the positions of all maximum scattering points in the image through an iterative function.

[0054] In this embodiment, the number of iterations uses a proxy optimization algorithm to select the optimal threshold to extract as many scattering points of the target body as possible without introducing too many noise points. The cost function is:

[0055]

[0056] in: is the threshold to be optimized, that is, the number of extraction points of the Clean algorithm, is the point on the extracted target, is the point outside the extracted target, that is, the noise point, is the weight of the noise point, which is set to 0.1 in this embodiment.

[0057] Step S12: Use a connectivity analysis method to remove noise points from the point set in step S11.

[0058] In this embodiment, isolated points in the image are first removed, and each point whose number of point sets in its neighborhood is less than 3 is determined to be an isolated point; then the largest connected set is retained to remove noise aggregation points, and the point sets with adjacent distances less than 2 are divided into a connected set, and the connected set containing the largest number of points is retained.

[0059] Step S13: For the point set outputted in step S12, use Alpha-Shape and wavelet filtering to obtain the target contour.

[0060] In this embodiment, the circle radius of the Alpha-Shape algorithm is set to 7, and the wavelet basis function is set to db wavelet.

[0061] Step S2: Match the target contour of each frame image with the template library and output the list of the best matching postures for each frame.

[0062] Specifically, the target contour extracted in step S1 is matched with the pre-generated template library contour, and the Hausdorff distance is used as the matching operator to output a list of the first several postures with the smallest distance in each frame.

[0063] Furthermore, the template library generation method of step S2 is as follows: assuming that the positions of each point of the target at zero posture are , then the target point cloud position under any posture is:

[0064]

[0065] in , , Represents yaw, pitch, and roll angles respectively, For around Axis rotation angle The rotation matrix of For around Axis rotation angle The rotation matrix of For around Axis rotation angle The rotation matrix of . , , , establish the target position in all postures, and project the target 3D model in each posture onto a 2D plane to form a 2D closed contour, which is the generated template library. In this embodiment, the reference projection plane is selected as the reference distance axis. and reference orientation axis , the distance and azimuth resolutions are set to .

[0066] Step S3: Calculate the imaging projection matrix of each frame according to the number of target orbit elements and the position of the ground observation station, and perform attitude conversion on the attitude list of each frame to obtain the converted attitude list of each frame in the satellite-based orbit coordinate system.

[0067] In this step, based on the number of target orbital elements and the position of the ground observation station, the radar line of sight (LOS) at the middle moment of a single-frame imaging is calculated as the distance unit vector; the range resolution is obtained through radar parameters; and the azimuth unit vector and azimuth resolution are obtained by calculating the LOS angle of a single frame.

[0068] For the posture list obtained in step S2, use quaternions to rotate the distance and orientation vectors of the template imaging surface and completely align them with the two-dimensional vectors of the actual imaging surface. At this time, the target in the index posture that follows the rotation of the template imaging surface will also eventually align with the actual posture, and the converted posture list will be output.

[0069] Furthermore, the step S3 includes steps S31 to S33:

[0070] Step S31: Construct the projection matrix:

[0071]

[0072] in and is the distance and azimuth dimension projection vector, and is the distance and azimuth unit vector, and is the range and azimuth resolution, Can be calculated from radar parameters, is the radar signal bandwidth, The speed of light.

[0073] Step S32: Obtain the LOS sequence direction in the target orbit coordinate system of each frame through coordinate conversion. The range dimension direction can be regarded as the LOS direction at the imaging center moment; the azimuth dimension direction is perpendicular to the LOS rotation axis and the range dimension direction; the azimuth dimension resolution is , where the imaging accumulation angle The projection matrix can be obtained by accumulating the LOS angle over time. .

[0074] Step S33: perform image registration on the pose list obtained in step S2 for each frame to achieve index pose from template To actual posture conversion.

[0075] Furthermore, the specific calculation method of step S33 is:

[0076] Constructing quaternions Distance dimension to template After quaternion rotation, we get , which is consistent with the actual distance dimension Alignment, this rotation also applies to the orientation dimension of the template and the target in the index pose , at this time, the rotation is aligned to one axis; the azimuth axis needs to be aligned again, and the quaternion is constructed using the same method After two rotations, the two image surfaces are aligned, and the template index pose target is aligned with the actual pose target. Next, the actual pose value needs to be solved.

[0077] Let the matching posture index be , the corresponding rotation matrix is , we can get

[0078]

[0079] The target in the index pose is

[0080]

[0081] in is the zero-pose target; the rotation matrix Convert to quaternion , according to the multiplication of quaternion rotation, the rotation from zero attitude target to actual attitude requires quaternion

[0082]

[0083] in Represents the azimuth axis rotation quaternion, the distance axis rotation quaternion, and the rotation matrix determined by the matching posture The converted quaternion is is the vector expansion of the quaternion. Apply the quaternion to the zero-attitude target to obtain the target after attitude transformation, that is,

[0084]

[0085] According to the relationship between quaternions and Euler angles, we have

[0086]

[0087] The above formula is the posture conversion formula, which realizes the index posture from the template To actual posture conversion to solve the posture mismatch problem.

[0088] Step S4: perform posture defuzzification on the posture sequences of each frame and output a unique posture.

[0089] For the pose list of each frame obtained in step S3, a candidate pose is taken from each frame, and all possible pose sequences are traversed to find the pose sequence with the smallest variance, which is the coarse pose estimation result.

[0090] An example of step S4 in this embodiment is as follows Figure 3 In this embodiment, the number of candidate poses selected for each frame is 5. It should be noted that under low signal-to-noise ratio, the number of candidate poses for each frame needs to be increased to ensure that the true pose appears in the candidates, sacrificing computational efficiency in exchange for accuracy.

[0091] In this embodiment, step S4 includes steps S41 to S44:

[0092] Step S41: Take the top matching postures of each frame from the converted posture list obtained in step S3. Figure 3 As shown, the top five candidate poses of the first frame are poses 0 to 4.

[0093] Step S42: Using the minimum Euclidean distance method for posture angles, select a posture from the first frame, find the posture closest to it in the second frame, and so on until the last frame to form a posture sequence, and calculate the sequence variance.

[0094] Step S43: traverse the several postures taken in the first frame and repeat step S42 to obtain several sequences and variances.

[0095] Step S44: Take the sequence with the smallest variance in the sequence of step S43, calculate the mean value and obtain a rough pose estimate of the target.

[0096] Step S5: Use the optimization algorithm to perform a refined estimate of the posture output in the previous step and output the accurate target posture.

[0097] Specifically, the rough pose estimate obtained in step S4 is used as the initial value. An optimization search is performed in the neighborhood of this estimate, directly using the target model's 3D pose as the optimization variable. The average Hausdorff distance between its theoretical profile and the actual five-frame imaging profile is optimized to a minimum, and the refined pose result is output. That is, without using the template library, the spatial target model is rotated directly using the pose angle as the optimization variable, and then projected to generate the theoretical profile. Minimizing the difference between the theoretical projected profile and the actual profile results in the optimal pose angle.

[0098] Furthermore, step S5 adopts a PID-based search algorithm (PSA), which has good optimization capabilities in constrained optimization problems and has advantages in computational cost and complexity. The optimization function is

[0099]

[0100] in, For the The actual ISAR image of the frame, The projection matrix calculated for this frame; is the position of the zero-attitude target, is the attitude rotation matrix, is the contour extraction function, is the function for calculating the Hausdorff distance between two contours. The feasible domain of this constrained optimization problem is , , ,in, is the rough pose estimation result.

[0101] In this embodiment, the initial population size of PSA is set to 50, and the number of iterations is set to 100.

[0102] In this embodiment, the posture is set to The target pose is estimated, and the output result of the method in the embodiment of the present invention is , the Euler angle rotation error is , the calculation time is 113s; the error of the traditional shape context matching method is , the calculation time is 207s; the method of the embodiment of the present invention is superior to the traditional method in both accuracy and efficiency.

[0103] In summary, the three-axis stabilized space target pose estimation method of the embodiment of the present invention is a method based on ISAR image target contour matching. It innovatively proposes an image plane registration algorithm, derives the image plane registration expression in detail, aligns the image planes of different frames with the template image plane to obtain the actual pose of the template matching of different frames. Next, the actual poses of multiple frames are combined to solve the multi-solution problem, and the PSA algorithm is used to efficiently obtain the target's precise pose. The method of the embodiment of the present invention can be used for three-axis stabilized targets of known models to achieve high-precision pose estimation with low computer resource consumption, and has high engineering application value.

[0104] The embodiment of the present invention further provides a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 4 As shown. The computer device includes a processor, a memory and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store operating parameter data of each framework. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps of the method of the embodiment of the present invention are implemented.

[0105] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0106] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method of the embodiment of the present invention are implemented.

[0107] An embodiment of the present invention further provides a computer program product, including a computer program, which implements the steps of the method of the embodiment of the present invention when executed by a processor.

[0108] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A three-axis stabilized space target attitude estimation method, characterized in that: include: Step S1: extracting the target contour of each frame from the sequential ISAR image based on the Clean algorithm; Step S2: Match the extracted target contours with the pre-generated template library contours respectively, use Hausdorff distance as the matching operator, and output a list of the first several postures with the smallest distance in each frame; Step S3: Calculate the imaging projection matrix of each frame according to the number of target orbit elements and the position of the ground observation station, perform attitude conversion on the attitude list of each frame, and obtain the attitude list of each frame after conversion in the satellite-based orbit coordinate system; Step S4: For the obtained pose list after conversion of each frame, a candidate pose is taken from each frame, all possible pose sequences are traversed, and the pose sequence with the smallest variance is found as the rough pose estimation result; Step S5: Using the rough pose estimation result as the initial value, perform optimization search on the neighborhood of the result, use the 3D pose of the target model as the optimization variable, optimize the average Hausdorff distance between its theoretical projection profile and actual imaging profile to the minimum, and output the precise target pose.

2. The method according to claim 1, wherein The step S1 comprises: Step S11, extracting all maximum scattering points of each frame image based on the Clean algorithm; Step S12: using a connectivity analysis method to remove noise from the point set obtained in step S11; Step S13: For the point set after noise removal in step S12, use Alpha-Shape and wavelet filtering to obtain the target contour.

3. The method according to claim 1 or 2, wherein: The template library in step S2 is generated as follows: Assume that the positions of each point of the target at zero posture are , then the target point cloud position under any posture is: in, , , Represents yaw, pitch, and roll angles respectively, For around Axis rotation angle The rotation matrix of For around Axis rotation angle The rotation matrix of For around Axis rotation angle The rotation matrix of Traversal , , , establish the target position under all postures, and project the target three-dimensional model under each posture into a two-dimensional plane to form a two-dimensional closed contour, thereby generating a template library.

4. The method according to claim 1 or 2, wherein: The step S3 comprises: Step S31: Construct the projection matrix: in, and are the distance and azimuth projection vectors respectively, and are the distance and azimuth unit vectors respectively, and are range and azimuth resolution, respectively; Step S32: According to the number of target orbital elements and the position of the ground observation station, the radar line of sight at the middle moment of a single frame of imaging is calculated as the distance dimension unit vector, the range resolution is obtained by the radar parameters, and the azimuth dimension unit vector and azimuth resolution are obtained by calculating the LOS angle of a single frame, thereby obtaining the projection matrix. ; Step S33: perform image registration on the posture list of each frame to achieve index posture from the template To actual posture conversion.

5. The method according to claim 4, wherein In step S33, the following posture conversion formula is used to realize the index posture from the template To actual posture Conversion: in, The quaternion required to rotate from the zero-pose target to the actual pose The vector expansion of : ， in, Represents the quaternion of the azimuth axis rotation, the quaternion of the distance axis rotation, and the index attitude The corresponding rotation matrix The converted quaternion.

6. The method according to claim 1 or 2, wherein: The step S4 comprises: Step S41: Take the top matching postures of each frame from the converted posture list obtained in step S3; Step S42: Using the minimum Euclidean distance method of posture angles, select a posture from the first frame, find the posture closest to it in the second frame, and continue this method until the last frame to form a posture sequence, and calculate the sequence variance; Step S43, traverse the several postures taken in the first frame, and repeat step S42 to obtain several sequences and variances; Step S44: Take the sequence with the smallest variance in the sequence of step S43, calculate the mean value and obtain a rough pose estimate of the target.

7. The method according to claim 1 or 2, wherein: In step S5, a PID-based search algorithm is used, and the optimization function is: in, For the The actual ISAR image of the frame, For the The projection matrix of the frame, is the position of the zero-attitude target, is the attitude rotation matrix, is the contour extraction function, is a function for calculating the Hausdorff distance between two contours, , , ,in, is the rough pose estimation result.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images