A method for reconstructing three-dimensional structure of space targets using chain scoring of ISAR image sequences

Abstract

A method for reconstructing the three-dimensional structure of a spatial target using chain scoring of ISAR image sequences divides the radar-acquired echoes into multiple sub-apertures. RD imaging is then performed on each sub-aperture to obtain a set of ISAR image sequences. Images that meet both imaging quality and structural integrity requirements are selected, and strong scattering points are extracted from each image. The projection matrix of the three-dimensional target onto each two-dimensional ISAR image is then calculated. Each image is then aligned in two dimensions in range and azimuth. Scattering point matching is performed using a chain scoring matching method to obtain a score set and a matching combination set. Finally, a set of three-dimensional reconstructed point coordinates is obtained through target three-dimensional reconstruction. This method utilizes target position information obtained from radar measurements and, combined with the proposed scoring chain point matching method, not only can more reconstruction points be obtained, facilitating accurate structural restoration of the target, but also the target's accurate posture in the radar observation coordinate system can be obtained.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing technology, and in particular relates to a method for reconstructing the three-dimensional structure of a space target using chain scoring of an ISAR image sequence, which is used to reconstruct the three-dimensional structure and attitude of a space target such as a satellite in a radar observation coordinate system. Background Art

[0002] With the continuous development of aerospace technology, space orbit resources are becoming increasingly scarce, and the possibility of mutual impact between targets is increasing. Therefore, situational awareness of the space environment has become an urgent need.

[0003] Inverse Synthetic Aperture Radar (ISAR), as an active detection technology, plays a vital role in space situational awareness due to its ultra-long detection range, all-weather operation, and high-resolution detection capabilities. However, traditional ISAR imaging is limited to two-dimensional imaging of targets. Because ISAR 2D images are projections of a target's 3D structure onto a 2D plane, it is impossible to obtain precise 3D structural information from these images, making component-level detection and high-precision identification impossible.

[0004] To address this issue, several methods have been proposed within the industry to obtain the target's 3D structure using ISAR technology. Existing research literature can be categorized into two technical approaches: one is multi-base station ISAR detection, where multiple radars simultaneously observe the target and utilize interferometry to reconstruct the target's 3D structure. While effective, this approach requires multiple radars to observe simultaneously, resulting in high system complexity and cost. The other approach uses a single base station to obtain a multi-perspective ISAR 2D image sequence of the target over a long period of time, and then reconstruct the target's 3D structure from this ISAR 2D image sequence. This second approach has attracted widespread attention due to its low system requirements and the lack of need for new radar systems. Existing 3D reconstruction methods for 2D image sequences mostly use optical feature point matching methods to match scatter points between ISAR images, followed by SVD (substrate matrix decomposition) to achieve 3D reconstruction of the target. However, due to the angular scintillation phenomenon of microwave detection, optical feature point matching methods cannot effectively match the images, resulting in a small number of reconstructed scatter points and an inability to fully represent the target's actual 3D structure. In addition, since the SVD method does not utilize the target line of sight (LOS) information obtained by radar detection, the reconstruction result cannot reflect the target's posture information. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a method for reconstructing the three-dimensional structure of space targets using chain scoring of ISAR image sequences, so as to better reconstruct the three-dimensional structure of space targets and reflect the posture of the target in the radar observation coordinate system.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for reconstructing the three-dimensional structure of a space target using chain scoring of an ISAR image sequence comprises:

[0008] Step 1: Divide the echo collected by the radar into multiple sub-apertures, and then perform RD imaging on each sub-aperture to obtain a set of ISAR image sequences;

[0009] Step 2: Select the images that meet the requirements of imaging effect and result integrity from the obtained ISAR image sequence to obtain a set of image sequences I1, I2, ..., I k ,…,I K , I k Represents the k-th image, the number K is not less than 3;

[0010] Step 3, extract the strong scattering points in each image and obtain the coordinates of all strong scattering points in each image;

[0011] Step 4: Calculate the projection matrix of the three-dimensional target onto each two-dimensional ISAR image based on the radar's latitude β, the angle α between the radar meridian and the vernal equinox axis of the Earth's inertial coordinate system, and the target azimuth, elevation, and distance measured by the radar.

[0012] Step 5: perform range and azimuth two-dimensional alignment on each image in turn;

[0013] Step 6: Complete the scattering point matching through the chain scoring matching method to obtain the score set f base The matching combination set D is used to record the total number of successfully matched images in each set of matching points, as well as the corresponding image numbers and scattering point numbers;

[0014] Step 7: Obtain the three-dimensional reconstructed point coordinate set Sr through target three-dimensional reconstruction.

[0015] Compared to existing technologies, this invention utilizes target position information obtained from radar measurements and combines it with a proposed scoring chain point matching method. Compared to existing 3D ISAR reconstruction techniques, this method not only obtains more reconstruction points, facilitating accurate target structure restoration, but also accurately determines the target's pose in the radar observation coordinate system. Furthermore, by selecting an appropriate score threshold, the number of noise points in the reconstruction result can be controlled, achieving a balance between structural integrity and accuracy in the reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of the overall process of the present invention.

[0017] Figure 2 These are the three observation coordinate systems of the three-axis stabilized satellite.

[0018] Figure 3 The dominant image and extracted strong scattering points for the seven-scattering point model used in this example are shown in Figure 1. (a) Subaperture 1, (b) Subaperture 11, (c) Subaperture 21, (d) Subaperture 31, and (e) Subaperture 41.

[0019] Figure 4 The scatter point matching process of the seven scatter points in the implementation example using the chain scoring method. Among them: (a) first matching, (b) second matching, (c) third matching, and (d) fourth matching.

[0020] Figure 5 The following are the reconstruction results of seven scattering points in the implementation example: (a) 3D image, (b) reconstructed target XZ view, (c) reconstructed target YZ view, and (d) reconstructed target XY view.

[0021] Figure 6 This is a three-dimensional scattering point location map of the satellite model for the implementation example.

[0022] Figure 7 The dominant images and extracted strong scattering points selected for the satellite model of the implementation example include: (a) subaperture 1, (b) subaperture 11, (c) subaperture 21, (d) subaperture 31, and (e) subaperture 41.

[0023] Figure 8 The scatterer point matching process using the chain scoring method for the satellite model in the implementation example. (a) First matching, (b) Second matching, (c) Third matching, (d) Fourth matching.

[0024] Figure 9 The 3D reconstruction results of the satellite model for the implementation example include: (a) 3D image, (b) reconstructed target XZ view, (c) reconstructed target YZ view, and (d) reconstructed target XY view. DETAILED DESCRIPTION

[0025] The embodiments of the present invention are described in detail below with reference to the accompanying drawings and examples.

[0026] To improve the scatter point matching capability of ISAR image sequence reconstruction, this paper proposes a method for reconstructing the three-dimensional structure of space targets using chain scoring of ISAR image sequences. This method utilizes radar tracking data, such as observed LOS information, combined with a scoring chain image scatter point matching method to reconstruct the target's three-dimensional structure and pose. The resulting target structure has more reconstructed points. Furthermore, the target pose is obtained in the radar observation coordinate system.

[0027] refer to Figure 1 , which is a method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences designed by the present invention, describes that the present invention includes the following specific steps:

[0028] Step 1: Divide the echo collected by the radar into multiple sub-apertures, and then perform RD imaging on each sub-aperture to obtain a set of ISAR image sequences.

[0029] In an embodiment of the present invention, the sub-apertures are divided and the theoretical resolution of the target ISAR imaging is calculated based on the LOS data, so that each sub-aperture ISAR image meets a certain resolution while having imaging stability, that is, the rotation speed is relatively constant in a short period of time.

[0030] Step 2: Select some images with better imaging effects and more complete results from the obtained ISAR image sequence to obtain a set of image sequences I1, I2, ..., I k ,…,I K , I k Represents the k-th image, and the number K is not less than 3.

[0031] In this algorithm, the principles of image selection are as follows: as few noise points as possible; clear scattering points; and as complete an image as possible.

[0032] (1) Fewer spots

[0033] Since the algorithm uses the scattered point matching method, if the reference image set contains a large number of stray points, it may lead to incorrect matching in the subsequent scattered point matching. Therefore, it is necessary to select target ISAR images with fewer stray points as much as possible.

[0034] (2) Scattering points are clear

[0035] Clear scattering points mean good image resolution. The clearer the scattering points, the more conducive they are to subsequent scattering point matching. On the contrary, if the image scattering points are blurred, more erroneous point matching will occur.

[0036] (3) The image should be as complete as possible

[0037] Image completeness means that it contains as many target features as possible. In order to achieve more accurate three-dimensional reconstruction of the target, it is necessary to select as many ISAR images of the target from different perspectives as possible. At the same time, in order to ensure the integrity of the reconstruction of the target components, it is necessary to select as many relatively complete images as possible.

[0038] Step 3: Extract the strong scattering points in each image and obtain the coordinates of all strong scattering points in each image.

[0039] For the kth image, the method for extracting strong scattering points is as follows:

[0040] (1) According to the resolution of the image’s orientation and distance dimensions, a two-dimensional Sinker function convolution kernel Ccov is generated, and then the image is circularly convolved with the convolution kernel to obtain a two-dimensional matrix H with the same dimension as the image. Let the initial set of strong scattering points be Calculate the initial total energy of H as Where i and j are the row and column numbers of the matrix H respectively, and the number of scattering points extracted is initialized to q = 1.

[0041] (2) Extract the maximum energy position (x max ,y max ), update the strong scattering point set Then the maximum energy position (x max ,y max ) is set to 0 in the 9×9 neighborhood, and the remaining energy is calculated.

[0042] (3)(3), let q←q+1, repeat (2), and continuously extract new strong scattering points until the remaining energy E q Less than 5% of the initial total energy E0, the final strong scattering point set S is obtained k,q .

[0043] Step 4: Based on the radar's latitude β, the angle α between the radar meridian and the vernal equinox axis of the Earth's inertial coordinate system, and the target azimuth, elevation, and distance measured by the radar, the projection matrix of the 3D target onto each 2D ISAR image is calculated.

[0044] refer to Figure 2, which are the three observation coordinate systems for a three-axis stabilized satellite. O-XYZ represents the satellite orbital coordinate system, where O represents the satellite center, the OZ axis points from the satellite to the Earth's center, and the plane defined by the OZ and OX axes is called the orbital plane. OX lies on the orbital plane and points in the direction of the satellite's instantaneous motion. The OY axis is determined by the right-hand rule. O2-ENU represents the radar observation coordinate system, where the O2E and O2N axes are parallel to the radar's tangent plane to the Earth's surface, with O2E pointing east and O2N pointing north. O2U is determined by the right-hand rule. O1-X1Y1Z1 is the Earth-centered inertial coordinate system, where O1Z1 points to the North Pole, O1X1 lies in the equatorial plane and points toward the vernal equinox, and O1Y1 is determined by the right-hand rule.

[0045] First, according to the latitude β of the radar, the angle α between the radar meridian and the vernal equinox axis of the earth's inertial coordinate system, the transformation matrix T1 from the radar observation coordinate system to the earth's inertial coordinate system is calculated:

[0046]

[0047] Then, the target trajectory information, namely the azimuth angle, is extracted from the radar observation data. The elevation angle θ(t) and the distance r0(t) are used to obtain the rectangular coordinates of the satellite in the radar coordinate system:

[0048]

[0049] Then the coordinates of the satellite in the geocentric inertial coordinate system are obtained as follows:

[0050]

[0051] Where r is the position vector of the radar in the Earth-centered inertial coordinate system, and the calculation formula is:

[0052]

[0053] where R e is the radius of the Earth at the radar location. Then, the azimuth and elevation angles of the satellite in the Earth-centered inertial coordinate system are calculated, expressed as η(t) and γ(t), respectively, where η(0) and γ(0) represent the angles at time 0.

[0054] have

[0055]

[0056] The relative LOS in the radar coordinate system after removing the translation is calculated as:

[0057]

[0058] Among them, l ob(t) is the observed LOS obtained from the radar observation data,

[0059]

[0060] T2 is the rotation matrix caused by satellite attitude adjustment, which is:

[0061]

[0062] Next, calculate the equivalent azimuth corresponding to the equivalent LOS and pitch angle θ e (t)

[0063]

[0064] Then we get the projection matrix of the image 3D structure to the kth 2D ISAR image:

[0065] G k =[ρ a,k ,ρ r,k ] T (10)

[0066] Where, ρ a,k Represents the projection vector of the three-dimensional structure of the image to the Doppler domain in the two-dimensional ISAR image, ρ r,k The projection vector representing the distance dimension of the three-dimensional image structure to the two-dimensional ISAR image;

[0067]

[0068]

[0069] Where, t k represents the central moment of the kth image, θ e (t k ) are the equivalent azimuth and equivalent elevation angles at t k The value at the moment, λ is the radar wavelength, With w θ (t k ) are the first-order derivatives of the azimuth and elevation angles with respect to time at t k The value of the moment is

[0070]

[0071]

[0072] Step 5: Perform range and azimuth two-dimensional alignment on each image in turn as follows:

[0073] (1), let k = 1;

[0074] (2) Select the k-1th image as the reference image;

[0075] (3) Align the k-th image as follows:

[0076] (3.1), set the minimum value Smin, maximum value Smax, and interval value Slag of the distance and azimuth dimension offsets, and set the two-dimensional offset candidate set;

[0077] (3.2), calculate the projection position of the scattering point of the reference image in the kth image under each offset combination, and calculate the total energy of the kth image at the projection position based on the projection position;

[0078] (3.3), select the offset combination Sbest with the largest energy as the optimal value, perform offset correction on the scattering point position of the kth image, and use the scattering point position after offset correction as the center position of the new offset candidate set;

[0079] (3.4), create a new two-dimensional offset candidate set, whose minimum value of each dimension is Sbest-Slag, the maximum value is Sbest+Slag, and the interval value is Slag / 10;

[0080] (3.5), repeat (3.2) and (3.3) until the interval value is less than 1;

[0081] (4) Update k to k+1 and repeat (2) to (3);

[0082] (5) When k is greater than the total number of images K, the alignment is completed.

[0083] Step 6: Complete the scattering point matching through the chain scoring matching method to obtain the score set f base The matching combination set D is used to record the total number of successfully matched images in each set of matching points, as well as the corresponding image sequence numbers and scattering point sequence numbers. The method specifically includes:

[0084] (1) Initialization. Let k = 2 and initialize the reference scattering point set S base = S1, S1 is the coordinate set of strong scattering points in the first image. The reference scattering point set is the point set to be matched with the scattering point set of the kth image, which will continue to expand during the matching process. The initial score set is f base =[1 1…1] 1×M , where M is S base The number of scatter points in . Generate matching combination set

[0085] (2) Scattering point matching. base and the scatter point set S in the kth image kPerform angular alignment and correct S k Formation of S kc , then S base With S kc To match, let S base The fraction of scatter points that are successfully matched in the equation is increased by 1, i.e., f base (match base )←f base (match base )+1, where match base For S base The number of the successfully matched scattering points in f base (match base ) represents the fractional set f base In the example, the serial number is match base , and update the matching pairs Where D{match base} indicates that the sequence number in the matching combination set is match base Elements of match k is the serial number of the scattering point that is successfully matched in the kth image, and S k (match k ) and S base (match base ) one-to-one correspondence, where S k (match k ) and S base (match base ) represent the scattering point set S in the kth image. k With the benchmark scattering point set S base The scatter points that are successfully matched in .

[0086] (3) Matching point reconstruction: According to the following formula, the corresponding three-dimensional coordinates p of the N pairs of scattering points that are successfully matched are solved by the least squares method.

[0087]

[0088] Among them, G k,k-1 The projection matrix G of the kth and k-1th images k and G k-1 The combination of P k,k-1 is the position matrix composed of matched scattering points;

[0089]

[0090]

[0091] G k-1 , G kRepresents the projection matrix of the k-1th and kth images respectively, S base ,1~S base ,N and S k,1 ~S k,N Indicates the reference scattering point and S k The total number of all successfully matched scattering points in is N.

[0092] Then project the three-dimensional point coordinates to the k-th image to obtain the projection point set S of the reconstructed point in the k-th image match =G k p.

[0093] (4) Generate a new reference point set. base The point that matches successfully, let S base (match base )=S match , for S base There is no successful matching point S base (mis base ), replace it with the projection position of the kth image, that is, , mis base is the serial number of the point that failed to match the reference scattering point set, G k-1 The pseudo-inverse of S k The unmatched points in S are merged into base A new reference scattering point set S is formed base ←[S base S k (mis k )],mis base For S k The serial number of the point that failed to match. Then update the score set to f base ←[f base f add ], where f add =[1 1…1] 1×L , L is the number of unmatched scatter points in the k-th image. Update D←{DD add},in mis k (1) mis k (2) mis k (L) represents the scattered point set S in the kth image k The sequence numbers of the points that failed to match, a total of L.

[0094] (5) Let k←k+1 and repeat (2) to (4) until k is greater than the total number of images K, and obtain the score set f base And the set of matching groups D.

[0095] Step 7: Target 3D reconstruction is performed to obtain the 3D reconstruction point coordinate set Sr.

[0096] According to the specified score threshold f t , filter score set f base The middle score is greater than f t The serial number set Q corresponding to the items of , let the number of Q be I; then:

[0097] (1) Let i = 1, the coordinate set of the 3D reconstructed points

[0098] (2) Extract matching combinations Where in1,…,inV are the serial numbers of each image that is successfully matched, sn1,…,snV are the serial numbers of the matching scattering points corresponding to each image, and V is the score of this item, that is, the number of matching images.

[0099]

[0100]

[0101] D{Q(i)} is the i-th matching group in the matching combination set that meets the score threshold, Q(i) represents the sequence number of the i-th matching group that meets the score threshold in the matching combination set D, and i represents the sequence number in the sequence number set Q of all matching combinations that meet the threshold; S in1 (sct1),S in2 (sct2),...,S inV (sct V ) represent the scattering point sets sct1, sct1, ..., sct1 in the in1, in2, ..., inV images respectively. V Scattering points; G in1 ,G in2 ,...,G inV is the projection matrix corresponding to the in1, in2, ..., inV images, G V is the combined matrix of these projection matrices.

[0102] Similarly, the three-dimensional point coordinates corresponding to the i-th matching scattering point combination are obtained by the following formula:

[0103]

[0104] Update Sr←Sr∪p i .

[0105] (3) Let i←i+1 and repeat (2) until i>I.

[0106] The effects of the present invention are further illustrated by the following processing of simulation data:

[0107] In the experiment, two types of targets were used for simulation experiments. The first one was a seven-point scattering point model. The coordinate positions of the seven scattering points were: (0,0,0), (0 -5 0), (0 5 0), (-5 0 0), (5 0 0), (0 0 -5), (0 0 5).

[0108] The second one is the satellite model, and its three-dimensional scattering point position distribution map is shown in the attached figure. Figure 3 As shown in FIG, the model has an X dimension of approximately 16 meters, a Y dimension of approximately 12 meters, and a Z dimension of approximately 11 meters, and contains a total of 634 scattering points.

[0109] The radar simulation parameters used in the experiment are shown in Table 1.

[0110] Table 1 Implementation example parameters

[0111] parameter Value Radar center frequency / GHz 9 Bandwidth / GHz 1 PRF / Hz 300 Relative rotation angle / ° 134.7 Sampling rate / MHz 20 Azimuth resolution / m 0.15

[0112] The experimental steps are:

[0113] 1) Generate target echo and various tracking data based on STK simulation trajectory;

[0114] 2) Divide the sub-aperture and use RD imaging to obtain ISAR image sequences;

[0115] 3) Select some dominant images from all ISAR image sequences;

[0116] 4) Extract strong scattering points in all images and obtain the coordinates of the scattering points.

[0117] 5) Calculate the projection matrix corresponding to each image;

[0118] 6) Perform two-dimensional alignment of the range and azimuth of each image;

[0119] 7) Perform scattered point matching on all images using chain scoring method;

[0120] 8) Reconstruct the target 3D point position.

[0121] For the seven-point target, simulation is performed according to the parameters shown in Table 1. First, the sub-aperture is divided according to the requirement of 0.15m azimuth resolution, and then two-dimensional high-resolution imaging is performed respectively, generating a total of 41 sub-aperture images. The 1st, 11th, 21st, 31st, and 41st two-dimensional high-resolution ISAR images are selected, and their scattering points are extracted as follows: Figure 3 Then the scattering point matching of the chain scoring method is performed, and the matching results are shown in (a), (b), (c), (d), and (e). Figure 4As shown in (a), (b), (c), and (d), after four matchings, 7 groups of scattering points are matched, and the final reconstruction results are as follows: Figure 5 Table 2 shows a comparison of the reconstructed point coordinates and the actual point coordinates shown in (a), (b), (c), and (d). P1-P7 are the numbers of the seven scattering points, and ME is the maximum error in the X, Y, and Z coordinates. As can be seen, the three-axis reconstruction errors of all seven points are within 0.22 meters. Compared to the maximum coordinate value of 5 meters, the relative reconstruction error is less than 4.4%.

[0122] Table 2 Comparison of the coordinate values of the reconstructed points and the actual points

[0123]

[0124] Similarly, the satellite target model is Figure 6 As shown in , the same parameters are used to generate the echo of the satellite target and perform RD imaging. Figure 7 As shown in (a), (b), (c), (d), and (e), the 1st, 11th, 21st, 31st, and 41st images are selected and matched by chain scoring scatter points. The matching results are shown in Figure 8 As shown in (a), (b), (c), and (d), the reconstructed three-dimensional structure is finally obtained as shown in Figure 9 As shown in (a), (b), (c), and (d), from Figure 9 It can be seen that the reconstructed target posture, size, and shape are basically consistent with the original model, which verifies the effectiveness of this method.

[0125] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences, characterized in that: include: Step 1: Divide the echo collected by the radar into multiple sub-apertures, and then perform RD imaging on each sub-aperture to obtain a set of ISAR image sequences; Step 2: Select the images that meet the requirements of imaging effect and result integrity from the obtained ISAR image sequence to obtain a set of image sequences I1, I2, ..., I k ,…,I K , I k Represents the k-th image, the number K is not less than 3; Step 3, extract the strong scattering points in each image and obtain the coordinates of all strong scattering points in each image; Step 4: Calculate the projection matrix of the three-dimensional target onto each two-dimensional ISAR image based on the radar's latitude β, the angle α between the radar meridian and the vernal equinox axis of the Earth's inertial coordinate system, and the target azimuth, elevation, and distance measured by the radar. Step 5: perform range and azimuth two-dimensional alignment on each image in turn; Step 6: Complete the scattering point matching through the chain scoring matching method to obtain the score set f base The matching combination set D is used to record the total number of successfully matched images in each set of matching points, as well as the corresponding image numbers and scattering point numbers; Step 7: Obtain the three-dimensional reconstructed point coordinate set Sr through target three-dimensional reconstruction.

2. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: In step 1, the theoretical resolution of the target ISAR imaging is calculated based on the LOS data, so that each sub-aperture ISAR image meets the resolution requirement and has imaging stability, that is, the rotation speed is relatively constant in a short period of time.

3. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: In step 2, the principles of image selection are: as few noise points as possible, clear scattered points, and as complete an image as possible.

4. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: In step 3, for the k-th image, the method for extracting strong scattering points is as follows: (1) According to the resolution of the image’s orientation and distance dimensions, a two-dimensional Sinker function convolution kernel Ccov is generated, and then the image is circularly convolved with the convolution kernel to obtain a two-dimensional matrix H with the same dimension as the image. Let the initial set of strong scattering points be Calculate the initial total energy of H as Where i, j are the row and column numbers of the matrix H, respectively, and the number of scattering points extracted is initialized to q = 1; (2), extract the maximum energy position (x max ,y max ), update the strong scattering point set Then the maximum energy position (x max ,y max ) is set to 0 in the 9×9 neighborhood, and the remaining energy is calculated. (3), let q←q+1, repeat (2), and continuously extract new strong scattering points until the remaining energy E q Less than 5% of the initial total energy E0, the final strong scattering point set S is obtained k,q .

5. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: In step 4, the transformation matrix T1 from the radar observation coordinate system to the earth's inertial coordinate system is first calculated based on the latitude β of the radar and the angle α between the radar meridian and the vernal equinox axis of the earth's inertial coordinate system: Then, the target trajectory information, namely the azimuth angle, is extracted from the radar observation data. The elevation angle θ(t) and the distance r0(t) are used to obtain the rectangular coordinates of the satellite in the radar coordinate system: Then the coordinates of the satellite in the geocentric inertial coordinate system are obtained as follows: P earth (t)=T1·P radar (t)+r=[X1(t),Y1(t),Z1(t)] T (3) Where r is the position vector of the radar in the Earth-centered inertial coordinate system, and the calculation formula is: Among them, R e is the radius of the Earth at the radar location; The azimuth and elevation angles of the satellite in the geocentric inertial coordinate system are obtained, which are expressed as η(t) and γ(t) respectively. η(0) and γ(0) represent the angles at time 0, and we have: The relative LOS in the radar coordinate system after removing the translation is calculated as: Among them, l ob (t) is the observed LOS obtained from the radar observation data, T2(t) is the rotation matrix caused by satellite attitude adjustment, Next, calculate the equivalent LOS: e The corresponding equivalent azimuth and pitch angle θ e (t) Then we get the projection matrix of the image 3D structure to the kth 2D ISAR image: G k =[ρ a,k ,r r,k ] T (10) Where, ρ a,k Represents the projection vector of the three-dimensional structure of the image to the Doppler domain in the two-dimensional ISAR image, ρ r,k The projection vector representing the distance dimension of the three-dimensional image structure to the two-dimensional ISAR image; Where, t k represents the central moment of the kth image, θ e (t k ) are the equivalent azimuth and equivalent elevation angles at t k The value at the moment, λ is the radar wavelength, With w θ (t k ) are the first-order derivatives of the azimuth and elevation angles with respect to time at t k The value of the moment; 6. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: The step 5 comprises: (1), let k = 1; (2) Select the k-1th image as the reference image; (3) Align the k-th image as follows: (3.1), set the minimum value Smin, maximum value Smax, and interval value Slag of the distance and azimuth dimension offsets, and set the two-dimensional offset candidate set; (3.2), calculate the projection position of the scattering point of the reference image in the kth image under each offset combination, and calculate the total energy of the kth image at the projection position based on the projection position; (3.3), select the offset combination Sbest with the largest energy as the optimal value, perform offset correction on the scattering point position of the kth image, and use the scattering point position after offset correction as the center position of the new offset candidate set; (3.4), create a new two-dimensional offset candidate set, whose minimum value of each dimension is Sbest-Slag, the maximum value is Sbest+Slag, and the interval value is Slag / 10; (3.5), repeat (3.2) and (3.3) until the interval value is less than 1; (4) Update k to k+1 and repeat (2) to (3); (5) When k is greater than the total number of images K, the alignment is completed.

7. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: The step 6 comprises: (1), Initialization Let k = 2, initialize the reference scattering point set S base = S1, where S1 is the coordinate set of strong scattering points in the first image; the reference scattering point set is the point set to be matched with the scattering point set of the kth image, which will continue to expand as the matching process progresses; the initialization score set is f base =[1 1…1] 1×M , where M is S base The number of scatter points in the scattering point, generating a matching combination set (2) Scattering point matching The reference scattering point set S base and the scatter point set S in the kth image k Perform angular alignment and correct S k Formation of S kc , then S base With S kc To match, let S base The fraction of scatter points that are successfully matched in the equation is increased by 1, i.e., f base (match base )←f base (match base )+1, update matching pairs Among them, match base For S base The number of the successfully matched scattering points in f base (match base ) represents the fractional set f base In the example, the serial number is match base Elements of match k is the serial number of the scattering point that is successfully matched in the kth image, D{match base } indicates that the sequence number in the matching combination set is match base Elements; S k (match k ) and S base (match base ) one-to-one correspondence, where S k (match k ) and S base (match base ) represent the scattering point set S in the kth image. k With the benchmark scattering point set S base The scattering points that are successfully matched in ; (3) Matching point reconstruction According to the following formula, the corresponding three-dimensional coordinates p of the N pairs of scattering points that are successfully matched are solved by the least squares method. The formula is as follows: Among them, G k,k-1 The projection matrix G of the kth and k-1th images k and G k-1 The combination of P k,k-1 is the position matrix composed of matched scattering points; G k-1 , G k Represents the projection matrix of the k-1th and kth images respectively, S base,1 ~S base,N and S k,1 ~S k,N Indicates the reference scattering point and S k The total number of all successfully matched scattering points in is N; Then project the three-dimensional point coordinates to the k-th image to obtain the projection point set S of the reconstructed point in the k-th image match =G k p; (4) Generation of new reference point set For S base The point that matches successfully, let S base (match base )=S match , for S base There is no successful matching point S base (mis base ), replace it with the projection position of the kth image, that is, mis base is the serial number of the point that failed to match the reference scattering point set, G k-1 The pseudo-inverse of S k The unmatched points in S are merged into base A new reference scattering point set S is formed base ←[S base S k (mis k )],mis base For S k The serial number of the point that failed to match; then update the score set to f base ←[f base f add ], where f add =[1 1…1] 1×L , L is the number of unmatched scattering points in the k-th image, update D←{DD add },in mis k (1) mis k (2) mis k (L) represents the scattering point set S in the kth image k The serial numbers of the points that failed to match, a total of L; (5) Let k←k+1 and repeat (2) to (4) until k is greater than the total number of images K, and obtain the score set f base And the set of matching groups D.

8. The method for reconstructing the three-dimensional structure of a space target using chain scoring of ISAR image sequences according to claim 1, characterized in that: In step 7, according to the specified score threshold f t , filter score set f base The middle score is greater than f t The serial number set Q corresponding to the items of , let the number of Q be I; then: (1) Let i = 1, the coordinate set of the 3D reconstructed points (2) Extract matching combinations Among them, in1,…,inV are the serial numbers of each image that is successfully matched, sn1,…,snV are the serial numbers of the matching scattering points corresponding to each image, and V is the score of this item, that is, the number of matching images. D{Q(i)} is the i-th matching group in the matching combination set that meets the score threshold, Q(i) represents the sequence number of the i-th matching group that meets the score threshold in the matching combination set D, and i represents the sequence number in the sequence number set Q of all matching combinations that meet the threshold; S in1 (sct1),S in2 (sct2),...,S inV (sct V ) represent the scattering point sets sct1, sct1, ..., sct1 in the in1, in2, ..., inV images respectively. V scattering points; G in1 ,G in2 ,...,G inV is the projection matrix corresponding to the in1, in2, ..., inV images, G V is the combined matrix of these projection matrices; The three-dimensional coordinates of the i-th matching scattering point combination are obtained by the following formula: Update Sr←Sr∪p i； (3) Let i←i+1 and repeat (2) until i>I.

Images