Optical remote sensing satellite target stereo tracking method

Abstract

The application discloses a kind of optical remote sensing satellite target stereoscopic tracking methods, according to stereovision principle, through satellite area array camera imaging model, high-precision stereoscopic fast positioning of target is realized, and according to target three-dimensional position information, through cubic spline interpolation fitting target space flight trajectory, the first derivative of interpolation curve is solved to obtain rough speed estimation, and the curve after derivation is filtered, and high-precision target speed estimation is obtained.

Description

Technical Field

[0001] This invention relates to a method for three-dimensional tracking of optical remote sensing satellite targets, belonging to the field of information processing technology. Background Technology

[0002] Optical remote sensing satellite array cameras can simultaneously detect multiple space targets and obtain multiple frames of continuous flight trajectories of the targets. By using two satellites to view the target simultaneously, the satellite can fuse the two-dimensional image plane trajectory information of the target in real time, obtain the three-dimensional coordinates and three-dimensional velocity of the target, and thus achieve three-dimensional tracking of the target by the satellite.

[0003] Currently, the extended Kalman filter method is widely used to achieve stereo tracking of dual-satellite targets. By establishing a target motion state model and a dual-satellite measurement model and applying the extended Kalman filter algorithm, the target's flight position and velocity can be predicted.

[0004] Due to dynamic differences, the motion state of a target is difficult to estimate based on observation alone. Typically, it is necessary to establish a suitable target motion state model based on the target's flight stage or type. The process of establishing an accurate motion model is complex and cannot quickly construct accurate motion state models for new types of flying targets. Furthermore, the extended Kalman filter method requires a certain convergence time to provide a high-accuracy estimate of target position and velocity, making it difficult to achieve high-precision three-dimensional tracking of the target throughout its flight.

[0005] To address the above issues, a tracking method is needed that is universally adaptable to multiple types and stages of targets, while also providing faster target localization and velocity estimation, thereby improving the efficiency of dual-satellite stereo tracking. Summary of the Invention

[0006] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a method for stereo tracking of optical remote sensing satellite targets. The method is based on the principle of stereo vision, and realizes high-precision stereo rapid positioning of the target through the imaging model of the satellite array camera. Then, based on the target's three-dimensional position information, the target's spatial flight trajectory is fitted by cubic spline interpolation. The first derivative of the interpolation curve is obtained to obtain a rough velocity estimate. Finally, the median of the differentiated curve is filtered to obtain a high-precision target velocity estimate.

[0007] The technical solution of this invention is:

[0008] This invention discloses a method for stereo tracking of optical remote sensing satellite targets, comprising the following steps:

[0009] S1. Obtain the centroid trajectory points of the target on satellite A and satellite B, as well as the pixel coordinate information of the trajectory points;

[0010] S2. Establish a mapping model M1 from the pixel coordinates of the target to the camera coordinate system;

[0011] S3. Establish the target coordinates from satellite A camera coordinate system P. l To coordinates P in the satellite B camera coordinate system r The transformation relationship model M2;

[0012] S4. Calculate the three-dimensional coordinates of the target in the coordinate system of satellite A camera based on the mapping model M1, the transformation relationship model M2 and the pixel coordinate information;

[0013] S5. Based on the three-dimensional coordinates, calculate the nth positioning coordinate value of the target in the world coordinate system;

[0014] S6. If n is greater than or equal to the design quantity value m, proceed to step S7; otherwise, return to step S1.

[0015] S7. Based on the n positioning coordinate values, obtain a rough estimate curve of the three-axis velocity;

[0016] S8. Perform median filtering on the coarse triaxial velocity estimation curve to obtain a high-precision triaxial velocity estimation curve for the target;

[0017] S9. Repeat steps S2 to S8 until the target disappears, at which point the 3D tracking ends.

[0018] Furthermore, in the above tracking method, a mapping model M1 is established from the target pixel coordinates to the camera coordinate system, specifically as follows:

[0019] z c P im =KP c

[0020] Among them, P im It is the two-dimensional coordinate of the target pixel, P c It is the three-dimensional coordinate of the target in the satellite camera coordinate system, z c It is P c The component in the Z-axis direction, K is the camera intrinsic parameter matrix.

[0021] Furthermore, in the above tracking method, the expression for the camera intrinsic parameter matrix K is:

[0022]

[0023] Among them, f x f y These are the ratios of the camera's principal distance to the scaling factor, c. x c y It is the offset between the origin of the image coordinate system and the origin of the imaging plane coordinate system.

[0024] Furthermore, in the above tracking method, the transformation relationship model M2 is specifically as follows:

[0025] P r =R 12 P l -T b

[0026]

[0027] T b =R2(S B -S A )

[0028] Among them, R 12 Let T be the transition matrix from the coordinate system of camera A to the coordinate system of camera B. b P represents the coordinates of the vector pointing from the origin of the coordinate system of camera A to the origin of the coordinate system of camera B in the coordinate system of camera A. l Let P be the coordinates of the target in the satellite A camera coordinate system. r Let S be the target's coordinates in the satellite B camera coordinate system; A S B R1 and R2 are the position coordinates of the satellite in the inertial coordinate system at the current moment; R1 and R2 are the rotation matrices from the inertial coordinate system to the home system of satellite A and satellite B at the current moment.

[0029] Furthermore, in the above tracking method, the rotation matrix is ​​specifically:

[0030]

[0031]

[0032] Where R1 and R2 are the rotation matrices from the inertial coordinate system to the home systems of satellite A and satellite B at the current moment, respectively, and q 1A q 2A q 3A q 4A The quaternion data measured in real time by the sensor of satellite A, q 1B q 2B q 3B q 4B Quaternion data measured in real time by the sensors of Satellite B.

[0033] Furthermore, in the above tracking method, the three-dimensional coordinates of the target in the coordinate system of satellite A camera are calculated based on the mapping model M1, the transformation relationship model M2, and the pixel coordinate information, specifically as follows:

[0034]

[0035]

[0036] ul '=K -1 u l ,v l '=K -1 v l ,u' r =K -1 u r ,v' r =K -1 v r ,T b =(t1,t2,t3) T

[0037]

[0038] Among them, P l =(x l ,y l ,z l P represents the three-dimensional coordinates of the target in the coordinate system of satellite A camera. r =(x r ,y r ,z r K represents the three-dimensional coordinates of the target in the coordinate system of satellite B camera. -1 The inverse of the camera intrinsic parameter matrix, (u l v l ), (u r v r R represents the pixel coordinates of the target on camera A and camera B, respectively. 12 Let T be the transition matrix from the coordinate system of camera A to the coordinate system of camera B. b Let be the coordinates of the vector pointing from the origin of the coordinate system of camera A to the origin of the coordinate system of camera B in the coordinate system of camera A.

[0039] Furthermore, in the above tracking method, the three-dimensional coordinates P of the target in the inertial coordinate system are calculated. I Specifically:

[0040] P I =R I -1 (P c -T I )

[0041] P I＝ R1 -1 (P l+ S A ) = R2 -1 (P r +S B )

[0042] Among them, Pc Let R be the coordinates of the target in the camera coordinate system. I T is the rotation matrix that transforms the inertial coordinate system to the satellite camera coordinate system. I It is a translation from the origin of the inertial coordinate system to the origin of the camera coordinate system in the camera coordinate system.

[0043] Furthermore, in the above tracking method, a rough estimate curve of the three-axis velocity is obtained based on the n positioning coordinate values, specifically as follows:

[0044] The n positioning coordinate values ​​are placed in the corresponding observation time period [T] a ,T b Divide the time into n-1 time intervals t∈[t i-1 ,t i There are n nodes in total, i = 2, ..., n; each node t i Corresponding to t i The coordinates of the target in any dimension of the three-dimensional inertial coordinate system observed at any given time;

[0045] Perform cubic spline interpolation on each of the aforementioned time intervals to obtain the target trajectory during the observation period [T]. a ,T b The three-dimensional smooth motion trajectory curve f(t) within the space. x f(t) y f(t) z );

[0046] Differentiating the smooth motion trajectory curves in three-dimensional space respectively, we obtain a rough estimate curve for the three-axis velocity f'(t). x ), f'(t) y ), f'(t) z ).

[0047] Furthermore, in the above tracking method, the cubic spline interpolation is specifically performed as follows:

[0048] In each observation segment interval t∈[t i-1 ,t i On, f i (t)=a i +b i (tt i )+c i (tt i ) 2 +d i (tt i ) 3 The following conditions are met:

[0049]

[0050] Among them, according to the non-torsional boundary condition, f i (t) satisfies:

[0051]

[0052] Among them, f i (t) represents the trajectory curve function in any one dimension of three-dimensional space, where n is the observation time period [T]. a ,T b The number of target location coordinates within [T], i.e., a total of n nodes within the observation period, divided into n-1 observation intervals, where i is the observation period [T]. a ,T b Excluding the first and last observation points, a i b i c i d i These are the coefficients of the cubic spline interpolation function.

[0053] Furthermore, in the above tracking method, m is greater than 2.

[0054] The advantages of this invention over the prior art are as follows:

[0055] (1) Based on the principle of stereo vision, this invention completes the stereo positioning of the target through the camera imaging mapping model and the binary star pixel coordinate transformation relationship model. The modeling process is simple, and the model parameters are determined and adapted to multiple types and stages of flight targets.

[0056] (2) This invention achieves real-time target positioning through a single observation, without waiting for convergence, with low computational load and high positioning accuracy; based on cubic spline interpolation, a smooth motion curve of the target can be obtained, and then high-precision velocity estimation can be obtained through differentiation and median filtering. Compared with the Kalman filter algorithm, the computation speed is faster, improving the efficiency of stereo tracking and meeting the higher inter-satellite handover requirements in the future. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention;

[0058] Figure 2 This is a schematic diagram of the remote sensing satellite stereo tracking system equipped with an area array camera payload according to the present invention. Detailed Implementation

[0059] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] This invention discloses a method for stereo tracking of optical remote sensing satellite targets, comprising the following steps:

[0061] S1. Obtain the centroid trajectory points of the target on satellite A and satellite B, as well as the pixel coordinate information of the trajectory points;

[0062] S2. Establish a mapping model M1 from the pixel coordinates of the target to the camera coordinate system;

[0063] S3. Establish the target coordinates from satellite A camera coordinate system P. l To coordinates P in the satellite B camera coordinate system r The transformation relationship model M2;

[0064] S4. Calculate the three-dimensional coordinates of the target in the coordinate system of satellite A camera based on the mapping model M1, the transformation relationship model M2, and the pixel coordinate information;

[0065] S5. Based on the three-dimensional coordinates, calculate the nth positioning coordinate value of the target in the world coordinate system;

[0066] S6. If n is greater than or equal to the design quantity value m, proceed to step S7; otherwise, return to step S1.

[0067] S7. Based on the n positioning coordinate values, obtain a rough estimate curve of the three-axis velocity;

[0068] S8. Perform median filtering on the coarse three-axis velocity estimation curve to obtain a high-precision three-axis velocity estimation curve for the target;

[0069] S9. Repeat steps S2 to S8 until the target disappears, at which point the 3D tracking ends.

[0070] Preferably, a mapping model M1 from the pixel coordinates of the target to the camera coordinate system is established, specifically as follows:

[0071] z c P im =KP c

[0072] Among them, P im It is the two-dimensional coordinate of the target pixel, P c It is the three-dimensional coordinate of the target in the satellite camera coordinate system, z c It is P c The component in the Z-axis direction, K is the camera intrinsic parameter matrix.

[0073] Preferably, the expression for the camera intrinsic parameter matrix K is:

[0074]

[0075] Among them, f x f y These are the ratios of the camera's principal distance to the scaling factor, c. x c y It is the offset between the origin of the image coordinate system and the origin of the imaging plane coordinate system.

[0076] Preferably, the transformation relation model M2 is as follows:

[0077] P r =R 12 P l -T b

[0078] R 12 =R2·R1 -1

[0079] T b =R2(S B -S A )

[0080] Among them, R 12 Let T be the transition matrix from the coordinate system of camera A to the coordinate system of camera B. b P represents the coordinates of the vector pointing from the origin of the coordinate system of camera A to the origin of the coordinate system of camera B in the coordinate system of camera A. l Let P be the coordinates of the target in the satellite A camera coordinate system. r Let S be the target's coordinates in the satellite B camera coordinate system; A S B R1 and R2 are the position coordinates of the satellite in the inertial coordinate system at the current moment; R1 and R2 are the rotation matrices from the inertial coordinate system to the home system of satellite A and satellite B at the current moment.

[0081] Preferably, the rotation matrix is ​​as follows:

[0082]

[0083]

[0084] Where R1 and R2 are the rotation matrices from the inertial coordinate system to the home systems of satellite A and satellite B at the current moment, respectively, and q 1A q 2A q 3A q 4A The quaternion data measured in real time by the sensor of satellite A, q 1B q 2B q 3B q 4B Quaternion data measured in real time by the sensors of Satellite B.

[0085] Preferably, based on the mapping model M1, the transformation relationship model M2, and the pixel coordinate information, the three-dimensional coordinates of the target in the coordinate system of satellite A camera are calculated, specifically as follows:

[0086]

[0087]

[0088] u l '=K -1 u l ,v l '=K -1 v l ,u' r =K -1 u r ,v' r =K -1 v r ,T b =(t1,t2,t3) T

[0089]

[0090] Among them, P l =(x l ,y l ,z l P represents the three-dimensional coordinates of the target in the coordinate system of satellite A camera. r =(xr,yr,z) r K represents the three-dimensional coordinates of the target in the coordinate system of satellite B camera. -1 The inverse of the camera intrinsic parameter matrix, (u l v l ), (u r v r R represents the pixel coordinates of the target on camera A and camera B, respectively. 12 Let T be the transition matrix from the coordinate system of camera A to the coordinate system of camera B. b Let be the coordinates of the vector pointing from the origin of the coordinate system of camera A to the origin of the coordinate system of camera B in the coordinate system of camera A.

[0091] Preferably, the three-dimensional coordinates P of the target in the inertial coordinate system are calculated. I Specifically:

[0092] P I =R I -1 (P c -T I )

[0093] P I =R1 -1 (P l+ S A ) = R2 -1 (P r +S B )

[0094] Among them, P cLet R be the coordinates of the target in the camera coordinate system. I T is the rotation matrix that transforms the inertial coordinate system to the satellite camera coordinate system. I It is a translation from the origin of the inertial coordinate system to the origin of the camera coordinate system in the camera coordinate system.

[0095] Preferably, a rough estimation curve for the three-axis velocity is obtained based on n positioning coordinate values, specifically as follows:

[0096] The n positioning coordinate values ​​are placed in the corresponding observation time period [T] a ,T b Divide the time into n-1 time intervals t∈[t i-1 ,t i There are n nodes in total, i = 2, ..., n; each node t i Corresponding to t i The coordinates of the target in any dimension of the three-dimensional inertial coordinate system observed at any given time;

[0097] Perform cubic spline interpolation on each time interval to obtain the target trajectory during the observation period [T]. a ,T b The three-dimensional smooth motion trajectory curve f(t) within the space. x f(t) y f(t) z );

[0098] Differentiate the smooth motion trajectory curve in three-dimensional space to obtain a rough estimate of the three-axis velocity curve f'(t). x ), f'(t) y ), f'(t) z ).

[0099] Preferably, cubic spline interpolation is used, specifically as follows:

[0100] In each observation segment interval t∈[t i-1 ,t i On, f i (t)=a i +b i (tt i )+c i (tt i ) 2 +d i (tt i ) 3 The following conditions are met:

[0101]

[0102] Among them, according to the non-torsional boundary condition, f i (t) satisfies:

[0103]

[0104] Among them, f i (t) represents the trajectory curve function in any one dimension of three-dimensional space, where n is the observation time period [T]. a ,T b The number of target location coordinates within [T], i.e., a total of n nodes within the observation period, divided into n-1 observation intervals, where i is the observation period [T]. a ,T b Excluding the first and last observation points, a i b i c i d i These are the coefficients of the cubic spline interpolation function.

[0105] Preferably, m is greater than 2.

[0106] Example

[0107] like Figure 1 As shown, this embodiment provides a method for stereo tracking of optical remote sensing satellite targets, which includes the following steps:

[0108] (1) Satellite A and Satellite B both confirmed that they detected the same target P.

[0109] (2) The target processing systems of satellites A and B respectively receive the centroid trajectory point of target P in the pixel plane and the pixel coordinate information of the centroid trajectory point (u) at the current moment, output by the detection system. l ,v l ), (u r ,v r ), where (u l ,v l Let (u) be the coordinates of target P in the pixel plane of satellite A. r ,v r Let P be the coordinates of target P in the pixel plane of satellite B.

[0110] (3) Satellite A acquires the attitude quaternion matrix Q at the current moment. A The three-dimensional coordinates S of the satellite in space are obtained from the six roots of the current orbit. A Satellite B acquires the attitude quaternion matrix Q at the current moment. B The three-dimensional coordinates S of the satellite in space are obtained from the six roots of the current orbit. B .

[0111] (4) Establish a mapping model M1 from the target's pixel plane coordinates to the camera coordinate system, and establish the target's coordinates P in the dual-star camera coordinate system. l With P r The transformation relationship M2.

[0112] 4.1 Target coordinates from pixel coordinates (u, v) to camera coordinates (x) c ,y c ,z c The mapping model M1 is:

[0113]

[0114] In the formula, f is the camera's focal distance. For geostationary satellites, the value of f is approximately equal to the camera's focal length. λ1 and λ2 represent the distances between adjacent pixels in the horizontal and vertical directions on the sensor, respectively. c x c y It is the offset between the origin of the image coordinate system and the origin of the imaging plane coordinate system.

[0115] matrix Let K be the camera intrinsic parameter matrix.

[0116] 4.2 Let the target's coordinates in the satellite A camera coordinate system be P. l In the coordinate system of satellite B camera, the coordinates are P r The observation diagram of the A and B binary satellites is as follows: Figure 2 As shown, then P l With P r The transformation relationship M2 is:

[0117] P r =R 12 P l -T b

[0118] In the formula, Where R1 and R2 are the rotation matrices from the inertial coordinate system to the home system of satellite A and satellite B at the current moment, respectively, and Q is the quaternion currently observed by the dual star sensors. A (q 1A q 2A q 3A q 4A ), Q B (q 1B q 2B q 3B q 4B The calculation method is as follows:

[0119]

[0120]

[0121] T b The coordinates of the vector pointing from the origin of the coordinate system of satellite A to the origin of the coordinate system of satellite B are given by the formula: Tb =R2(S B -S A ), where S A S B These are the current position coordinates of the satellite in the inertial coordinate system, obtained from the six orbital roots of the satellite.

[0122] (5) Based on the mapping model M1, transformation relationship M2 and camera extrinsic matrix established in step (4), the three-dimensional coordinates of the target in the inertial coordinate system are solved.

[0123] 5.1 Based on the mapping model M1 and the transformation relation M2, establish the following system of equations:

[0124]

[0125] In the formula, P l =(x l ,y l ,z l ), P r =(x r ,y r ,z r K is the camera intrinsic parameter matrix described in 4.1.

[0126] 5.2 Let u l '=K -1 u l ,v l '=K -1 v l ,u' r =K -1 u r ,v' r =K -1 v r ,T b =(t1,t2,t3) T , The derivation yields the following system of matrix equations:

[0127]

[0128] The equation is solved to obtain (x) l ,y l ,z l ).

[0129] 5.3 Based on the camera extrinsic parameter matrix, the coordinates of target P in the inertial coordinate system are obtained as follows: where T I It is the vector from the current coordinates of satellite A in the inertial frame to the origin of the inertial coordinate system.

[0130] (6) Record the number of located trajectories n. When the number is greater than or equal to the design value m, proceed to step (7); otherwise, return to step (2).

[0131] (7) The observation time periods [T] corresponding to the n (n≥m) trajectory positions of the target are... a ,T b Divide into n-1 intervals t∈[t i-1 ,t i There are n nodes in total, i = 2, ..., n, and each node t i Corresponding to t i The target's coordinates in any dimension of the three-dimensional inertial coordinate system are observed at any given time. Through cubic spline interpolation with non-torsional boundary conditions, the smooth three-dimensional spatial trajectory curve function f(t) of the target within the interval t is obtained. x f(t) y f(t) z ).

[0132] 7.1 Taking the X-axis as an example, in each segmented time interval, f(t) = f i (t) are all cubic polynomials and satisfy the interpolation condition, that is:

[0133] f i (t)=a i +b i (tt i )+c i (tt i ) 2 +d i (tt i ) 3

[0134] Since the curve is smooth and its second derivative is continuous, its first derivative must also be continuous. The above constraints are expressed as follows:

[0135]

[0136] 7.2 To increase the number of equations, boundary conditions are added, and non-knotted boundaries are used, requiring that at interpolation nodes t2 and t... n-1 The third derivative of the spline interpolation function is continuous, and the constraint condition is expressed as:

[0137]

[0138] 7.3 Let f i (t i )=y i ,h i =t i+1 -t i M i =f” i(t i Based on the constraints in step 7.1, n-2 equations are finally derived:

[0139]

[0140] In the above formula, i = 2, 3, ..., n-1.

[0141] 7.4 According to the constraints in step 7.2, we know that d1 = d2, d n-2 =d n-1 Then by Finally, it is deduced that:

[0142]

[0143] 7.5 The n terms about M derived in steps 7.3 and 7.4 i Solve the system of equations to find the interval t∈[t]. i-1 ,t i The coefficients a of the cubic spline interpolation function on (i = 2, 3, ..., n) i b i c i d i Thus, the three-dimensional position estimation curve f(t) of the target trajectory is obtained. x f(t) y f(t) z ).

[0144] (8) Differentiate the target triaxial space curves obtained in step (7) to obtain the rough triaxial velocity estimation curve f'(t). x ), f'(t) y ), f'(t) z Due to the line-of-sight determination error of the satellite camera, the target coordinates obtained in step (2) are inaccurate compared to the actual coordinates. This results in the target trajectory curve obtained by interpolation in step (7) being less smooth than the actual trajectory curve. The lack of smoothness in the trajectory curve leads to oscillations in the velocity curve after differentiation, causing a large velocity error. Therefore, median filtering is then applied to the curves to obtain high-precision three-axis velocity estimation curves v(x), v(y), and v(z).

[0145] Satellite A and Satellite B detection systems determine whether the target has disappeared. If yes, return to step (2); otherwise, the 3D target tracking ends.

[0146] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

[0147] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A method for stereo tracking of optical remote sensing satellite targets, characterized in that, Includes the following steps: S1. Obtain the centroid trajectory points of the target on satellite A and satellite B, as well as the pixel coordinate information of the trajectory points; S2. Establish a mapping model M1 from the pixel coordinates of the target to the camera coordinate system; S3. Establish the target coordinates using the satellite A camera coordinate system. P l coordinates to satellite B camera coordinate system P r The transformation relationship model M2; S4. Calculate the three-dimensional coordinates of the target in the coordinate system of satellite A camera based on the mapping model M1, the transformation relationship model M2 and the pixel coordinate information; S5. Based on the three-dimensional coordinates, calculate the nth positioning coordinate value of the target in the world coordinate system; S6, if n Greater than or equal to the design quantity value m If yes, proceed to step S7; otherwise, return to step S1. S7, according to the above n Based on the positioning coordinates, a rough estimate curve of the three-axis velocity is obtained; S8. Perform median filtering on the coarse triaxial velocity estimation curve to obtain a high-precision triaxial velocity estimation curve for the target; S9. Repeat steps S2 to S8 until the target disappears, at which point the 3D tracking ends.

2. The optical remote sensing satellite target stereo tracking method according to claim 1, characterized in that: Establish a mapping model M1 from the target pixel coordinates to the camera coordinate system, specifically as follows: in, P im These are the two-dimensional coordinates of the target pixel. P c These are the three-dimensional coordinates of the target in the satellite camera coordinate system. z c yes P c The component in the Z-axis direction, K This is the camera intrinsic parameter matrix.

3. The optical remote sensing satellite target stereo tracking method according to claim 2, characterized in that: Camera intrinsic parameter matrix K The expression is: in, f x , f y These are the ratios of camera principal distance to scaling factor, respectively. c x , c y It is the offset between the origin of the image coordinate system and the origin of the imaging plane coordinate system.

4. The optical remote sensing satellite target stereo tracking method according to claim 1, characterized in that: The transformation relationship model M2 is specifically as follows: in, R 12 This is the transition matrix from the coordinate system of camera A to the coordinate system of camera B. T b The coordinates of the vector pointing from the origin of the coordinate system of satellite A camera to the origin of the coordinate system of satellite B camera in the coordinate system of satellite A camera. P l Let the target's coordinates be in the coordinate system of satellite A camera. P r The target's coordinates in the satellite B camera coordinate system; S A , S B These are the current position coordinates of the satellite in the inertial coordinate system; R 1. R 2 are the rotation matrices from the inertial coordinate system to the home system of satellite A and satellite B at the current moment.

5. The optical remote sensing satellite target stereo tracking method according to claim 4, characterized in that: The rotation matrix is ​​specifically: in, R 1. R 2 are the rotation matrices from the inertial coordinate system to the home system of satellite A and satellite B at the current moment, respectively. q 1A , q 2A 、 q 3A 、q 4A Quaternion data measured in real time by the sensors of satellite A. q 1B , q 2B 、q 3B 、q 4B Quaternion data measured in real time by the sensors of Satellite B.

6. The optical remote sensing satellite target stereo tracking method according to claim 1, characterized in that: Based on the mapping model M1, the transformation relationship model M2, and the pixel coordinate information, the three-dimensional coordinates of the target in the coordinate system of satellite A camera are calculated as follows: in, P l = ( x l ,y l ,z l () represents the three-dimensional coordinates of the target in the coordinate system of satellite A camera. P r =( x r ,y r ,z r () represents the three-dimensional coordinates of the target in the coordinate system of satellite B camera. K -1 The inverse of the camera intrinsic parameter matrix, ( u l , v l ), ( u r , v r ( ) represent the pixel coordinates of the target on camera A and the pixel coordinates of the target on camera B, respectively. R 12 This is the transition matrix from the coordinate system of camera A to the coordinate system of camera B. T b Let be the coordinates of the vector pointing from the origin of the coordinate system of camera A to the origin of the coordinate system of camera B in the coordinate system of camera A.

7. The optical remote sensing satellite target stereo tracking method according to claim 1, characterized in that: Calculate the three-dimensional coordinates of the target in the inertial coordinate system. P I Specifically: P I= R 1 -1 （ P l+ S A ）= R 2 -1 ( P r + S B ) in, P c Let the target's coordinates be in the camera coordinate system. R I It is the rotation matrix that transforms the inertial coordinate system to the satellite camera coordinate system. It is a translation from the origin of the inertial coordinate system to the origin of the camera coordinate system in the camera coordinate system; S A , S B These are the current position coordinates of the satellite in the inertial coordinate system; R 1. R 2 are the rotation matrices from the inertial coordinate system to the home system of satellite A and satellite B at the current moment.

8. The optical remote sensing satellite target stereo tracking method according to claim 1, characterized in that: According to the above n Based on the positioning coordinates, a rough estimate curve of the three-axis velocity is obtained, specifically: The n Each location coordinate value during the corresponding observation period Divided into time intervals common n Each node; t i Corresponding to t i The coordinates of the target in any dimension of the three-dimensional inertial coordinate system observed at any given time; Cubic spline interpolation is performed on each of the aforementioned time intervals to obtain the target trajectory during the observation period. Smooth motion trajectory curve in three-dimensional space f(t x f(t) y f(t) z ) ; Differentiate the smooth motion trajectory curves in three-dimensional space to obtain rough estimates of the three-axis velocities. f'(t x ) 、f’ (t y ) 、f'(t z ) .

9. The optical remote sensing satellite target stereo tracking method according to claim 8, characterized in that: The specific method for the cubic spline interpolation is as follows: In each observation segment interval superior, The following conditions are met: Among them, according to the non-torsional boundary condition, f i (t) satisfy: in, f i (t) A trajectory curve function representing any one dimension of three-dimensional space. n Observation period Number of target location coordinates within. i Observation period Excluding the first and last observation points, any observation time point. a i , b i , c i , d i These are the coefficients of the cubic spline interpolation function.

10. The optical remote sensing satellite target stereo tracking method according to claim 1, characterized in that: m Greater than 2.

Images