CN103115627B - Multi-track combined on-track geometrical checking method of remote sensing satellite linear array sensor - Google Patents

Abstract

The invention discloses a multi-track combined on-track geometrical checking method of a remote sensing satellite linear array sensor. The multi-track combined on-track geometrical checking method has the advantages that a large quantity of image data obtained by using a remote sensing satellite linear array sensor is utilized to fully play the role of existing geographic information, a method of multi-track data combined block adjustment is adopted for carrying out on-track geometrical checking on the remote sensing satellite linear array sensor, the checking cost is greatly reduced, the checking efficiency is increased, and the checking accuracy is ensured.

Description

Geometry calibration method in-orbit combined by the many rails of remote sensing satellite line array sensor

Technical field

The invention belongs to Surveying Science and Technology field, relate to the many rails of a kind of remote sensing satellite line array sensor and combine geometry calibration method in-orbit, be mainly used in the geometry calibration in-orbit of Optical remote satellite line array sensor.

Background technology

Three-line imagery surveying & mapping as far back as nineteen ninety Germany MOMS system time, just achieve good result.In addition, the SPOT5 satellite of France is also the linear array push-broom type design adopted, front-and rear-view camera HRS1 and HRS2 primary optical axis with under depending on angle be respectively ± 20 °, what the base-height ratio of formation was maximum reaches 0.8.Equally, SPOT5 also utilizes linear array image mapping to achieve successfully, and the stereo mapping satellite of identical type mainly also has the IRS-P5 of India and the ALOS of Japan, and after these passing of satelline calibrations, the precision of direct geometry location is all within 100m.And utilize the appearance rail data that resource No. three stars directly pass down to carry out direct geo-location, its positioning precision is 1.3 kilometers, plane, elevation about 300m, therefore in order to improve its geometric positioning accuracy, elements of interior orientation self calibration and band control point block adjustment are absolutely necessary process.

Sensor calibration is also called camera calibration, mainly comprise the error of fixed angles of three line scanner three between camera and satellite body to correct, camera focus Correction of Errors, lens distortion Correction of Errors, camera internal CCD (charge coupled cell) array rotation, translation, convergent-divergent, the bending distortion inaccuracy that waits corrects, before satellite launch, these parameters all can carry out calibration in laboratory, but satellite is after injection, the elements of interior orientation of camera can be variant with the calibration result in laboratory before, it is irrational for only utilizing the elements of interior orientation result of laboratory calibration to carry out process, therefore the work of calibration in-orbit of sensor must be carried out.For Space-borne, all have very strong correlation between sensor inner orientation and elements of exterior orientation and between all kinds of parameter of elements of interior orientation, in block adjustment, solution asks these parameters to be impossible simultaneously.The common ground of tradition calibration method is all have employed Ground Nuclear Magnetic Resonance control point, and calibration multiclass intrinsic parameter.But Ground Nuclear Magnetic Resonance control point often needs field operation measurement to obtain or measures on large scale topographical map to obtain, cost is larger, and the control point obtained is relative to long strip image, distribution is too little, in addition in calibration process, between elements of interior orientation and elements of exterior orientation, certain correlation is all there is between all kinds of parameter of elements of interior orientation, cannot strictly by these parameters separated in adjustment, suitable adjustment strategy must be taked, just can avoid the interdependence effects between these parameters.

Summary of the invention

The object of this invention is to provide the many rails of a kind of remote sensing satellite line array sensor and combine geometry calibration method in-orbit, overcome above-mentioned the deficiencies in the prior art.

Technical scheme of the present invention is that geometry calibration method in-orbit combined by the many rails of a kind of remote sensing satellite line array sensor, comprises the following steps:

Step 1, imports the tie point of empty three couplings of many tracks, the control point coordinate data of Auto-matching and the track of satellite and attitude data;

Step 2, builds the error equation of sensor established angle calibration model, and picture point carries out methodization and changing process one by one, generates the normal equation after changing;

Step 3, the normal equation after utilizing least square principle solution to ask changing of step 2 gained, obtains the correction of each unknown number;

Step 4, determines whether to need to carry out next step iterative processing according to the correction of step 3 gained unknown number, if do not meet exit criteria, then forwards step 2 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 5 to;

Step 5, exports established angle calibration parameter;

Step 6, uses the established angle parameter after calibration, builds the error equation of CCD skew and convergent-divergent calibration model, and picture point carries out methodization and changing process one by one, generates the normal equation after changing;

Step 7, the normal equation after utilizing least square principle solution to ask changing of step 6 gained, obtains the correction of each unknown number;

Step 8, determines whether to need to carry out next step iterative processing according to the correction of step 7 gained unknown number, if do not meet exit criteria, then forwards step 6 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 9 to;

Step 9, exports CCD skew and convergent-divergent calibration parameter;

Step 10, if sensor ccd array is spliced by sub-linear CCD array, then forwards step 11 to, otherwise exits;

Step 11, the established angle after use calibration and CCD offset, zooming parameter, build the error equation of the skew of sub-line array CCD and convergent-divergent calibration model, and picture point carries out methodization and changing process one by one, the normal equation after changing of generation;

Step 12, the normal equation after utilizing least square principle solution to ask changing of step 11 gained, obtains the correction of each unknown number;

Step 13, determines whether to need to carry out next step iterative processing according to the correction of step 12 gained unknown number, if do not meet exit criteria, then forwards step 11 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 14 to;

Step 14, exports the skew of sub-line array CCD and convergent-divergent calibration parameter.

And sensor established angle calibration model is as follows described in step 2,

$\left[\begin{array}{c}\mathrm{Xs}-Y\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\λ}{R}_{\mathrm{obj}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]=\mathrm{\λR}(p+\mathrm{\Δp},l+\mathrm{\Δl},k+\mathrm{\Δk})\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]$

Wherein, Xs, Ys, Zs represent the object coordinates of photo centre, and X, Y, Z represent topocentric object coordinates, and x, y represent the image space coordinate that ground point is corresponding, and f represents the focal length of camera, R _obj2satrepresent the spin matrix of satellite body coordinate system relative to object coordinates system, R _sat2camrepresent the spin matrix of camera to satellite body, λ is proportionality coefficient; L, p, k represents sensor that laboratory measurement obtains three established angles relative to satellite body coordinate system, Δ l, Δ p, Δ k represents the error of fixed angles of sensor relative to satellite body coordinate system, R (p+ Δ p, l+ Δ l, k+ Δ k) represents p+ Δ p, the spin matrix that l+ Δ l, k+ Δ k tri-anglecs of rotation are formed.

And, described in step 6 CCD skew and convergent-divergent calibration model as follows,

$\left[\begin{array}{c}\mathrm{Xs}-Y\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\λ}{R}_{\mathrm{obj}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x+\mathrm{\Δx}\\ y+\mathrm{\Δy}\\ -f\end{array}\right]$

Δx＝x ₀

Δy＝y ₀+y ₁c

Wherein, Δ x, Δ y represent image space coordinate correction amount respectively, x ₀for CCD is along the deviant of heading, y ₀, y ₁for the linear drift parameter perpendicular to heading of CCD, c is CCD row number.

And, sub-line array CCD skew described in step 11 and convergent-divergent calibration model as follows,

Being located at camera focal plane is be spliced into a large ccd array by n sub-linear CCD array, then every a line image is all divided into n part according to ccd array position, and the CCD number of each part is respectively N ₁, N ₂... N _n-1, N _n, then total CCD number is N ₁, N ₂... N _n-1, N _nsum;

$\begin{array}{c}\mathrm{\&Delta;x}=\left\{\begin{array}{cc}{x}_0& 0<c<N_1\\ x_1& N_1<c<N_1+N_2\\ x_2& N_1+N_2<c<N_1+N_2+N_3\\ & ...\\ x_{n-1}& N_1+...+N_{n-1}<c<N_1+...+N_n\end{array}\right.\\ \mathrm{\&Delta;y}=\left\{\begin{array}{cc}{y}_0+y_{1}c& 0<c<N_1\\ y_2+y_{3}c& N_1<c<N_1+N_2\\ y_4+y_{5}c& N_1+N_2<c<N_1+N_2+N_3\\ & ...\\ y_{2n-2}+y_{2n-1}c& N_1+...+N_{n-1}<c<N_1+...+N_n\end{array}\right.\end{array}$

Wherein, x ₀, x ₁, x ₂... x _n-1for CCD is along the deviant of heading, y ₀, y ₂, y ₄... y _2n-2for the offset parameter perpendicular to heading of CCD, y ₁, y ₃, y ₅... y _2n-1for the linear drift parameter perpendicular to heading of CCD, c is that image is perpendicular to heading row coordinate.

Advantage of the present invention is that the strategy of many rails data aggregate process effectively inhibits elements of exterior orientation error on the impact of geometry calibration, and what adopt is the control point of Auto-matching, without the need to manually carrying out field operation measuring point, do not need calibration field data, significantly reduce calibration cost, whole calibration process is full-automatic, and calibration efficiency significantly improves, and precision can reach the calibration field calibration way of even beyond tradition.

Accompanying drawing explanation

Fig. 1 is the flow chart of the embodiment of the present invention.

Detailed description of the invention

Technical solution of the present invention is described in detail below in conjunction with drawings and Examples.

See Fig. 1, the present invention is that geometry calibration method in-orbit combined by the many rails of a kind of remote sensing satellite line array sensor.The concrete methods of realizing of embodiment comprises the following steps:

Step 1, imports the tie point of empty three couplings of many tracks, the control point coordinate data of Auto-matching and the track of satellite and attitude data.

During concrete enforcement, read adjustment project file and associated data files.Embodiment concrete methods of realizing is: adopt engineering management way, organize each data file, generate a project file, after importing project file, reading project file can obtain related data information and (comprise control point file path, the data file path such as tie point, control point, the track of each remote sensing satellite and attitude data file path, and relevant systematic parameter).Thus desired data can be read further.

Can in existing geographic information database Auto-matching control point data, existing geographic information database comprises image database, Law of DEM Data storehouse etc., the databases such as such as GoogleMap, ETM, SRTM, ASTER.

The present invention adopts the data of many tracks to carry out geometry calibration in-orbit simultaneously, between many rails data can overlap also can not be overlapping.

Step 2, builds the error equation of sensor established angle calibration model, and picture point carries out methodization and changing process one by one, generates the normal equation after changing.Build the error equation of sensor established angle calibration model, conventionally can realize according to calibration model.

Angle of eccentricity between sensor and satellite as unknown parameter, solves by the present invention in the process of calibration simultaneously.During concrete enforcement, pointwise can build established angle calibration error equation and methodization and changing.Embodiment concrete methods of realizing is: the relevant knowledge utilizing matrix operation, picture point calculates local approach equation corresponding to picture point and this local approach equation position in the normal equation after final changing one by one, then these local approach equations are filled in the rear normal equation of final changing, i.e. updating method equation.Namely the normal equation after complete changing is obtained after all picture points are disposed.

For ease of implementing reference, provide the sensor that the present invention relates to as follows relative to the established angle calibration model of satellite body:

Introduce sensor relative to after satellite established angle calibration parameter, traditional collinearity equation becomes following form:

$\left[\begin{array}{c}\mathrm{Xs}-Y\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\&lambda;}{R}_{\mathrm{obj}2\mathrm{orb}}{R}_{\mathrm{orb}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]---\left(1\right)$

Wherein, Xs, Ys, Zs represent the object coordinates of photo centre, and X, Y, Z represent topocentric object coordinates, and x, y represent the image space coordinate that ground point is corresponding, and f represents the focal length of camera, and λ is proportionality coefficient, R _obj2orbrepresent the spin matrix of satellite orbit coordinate system relative to object coordinates system, R _orb2satrepresent the spin matrix between satellite body to satellite orbit coordinate system, R _sat2camrepresent the spin matrix of camera to satellite body.

Because the anglec of rotation of satellite body coordinate system relative to object coordinates system is obtained by front-end processing, namely directly use satellite body coordinate system directly relative to the anglec of rotation of object coordinates system in adjustment, camera is made up of to the spin matrix of satellite body three established angles of camera:

Wherein, R _obj2satrepresent the spin matrix of satellite body coordinate system relative to object coordinates system, ω, κ represent three anglecs of rotation of satellite body coordinate system relative to object coordinates system; represent the spin matrix that ω, κ are formed, l, p, k represent sensor that laboratory measurement obtains three established angles relative to satellite body coordinate system.Δ l, Δ p, Δ k represent the error of fixed angles of sensor relative to satellite body coordinate system, the unknown number namely in self calibration parameter model of the present invention.R (p+ Δ p, l+ Δ l, k+ Δ k) represents p+ Δ p, l+ Δ l, the spin matrix that k+ Δ k tri-anglecs of rotation are formed.

Formula (2) is substituted into formula (1) and obtains sensor established angle calibration model:

$\left[\begin{array}{c}\mathrm{Xs}-Y\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\&lambda;}{R}_{\mathrm{obj}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]=\mathrm{\&lambda;R}(p+\mathrm{\&Delta;p},l+\mathrm{\&Delta;l},k+\mathrm{\&Delta;k})\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]---\left(3\right)$

Step 3, the normal equation after utilizing least square principle solution to ask changing of step 2 gained, obtains the correction of each unknown number.

Solving method equation be embodied as prior art, according to least square principle, normal equation inverse matrix and constant term vector dot can obtain unknown number correction vector, shown in (7).

▽X＝(A ^TPA) ^-1A ^TPL (4)

Wherein ▽ X represents unknown number vector, and A represents error equation coefficient matrix, and L represents error equation constant term vector, and P observation represents weight matrix.

Step 4, determines whether to need to carry out next iteration process according to the correction of step 3 gained unknown number, if do not meet exit criteria, then forwards step 2 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 5 to.

Embodiment concrete methods of realizing is: arrange a unknown number correction threshold value and iterations threshold value, exit criteria is: the correction maximum of unknown number exceedes this unknown number correction threshold value or iterations exceedes iterations threshold value.During concrete enforcement, those skilled in the art can preset threshold value as the case may be.

Step 5, exports established angle calibration parameter.

Embodiment concrete methods of realizing is: established angle calibration parameter exported in camera parameter file, so that next step calibration uses.

Step 6, uses the established angle parameter after calibration, builds the error equation of CCD skew and convergent-divergent calibration model, and picture point carries out methodization and changing process one by one, generates the normal equation after changing.Build the error equation of CCD skew and convergent-divergent calibration model, conventionally can realize according to calibration model.

CCD position offset parameter and scaling parameter as unknown number, solve by the present invention in the process of calibration simultaneously.During concrete enforcement, pointwise can build whole CCD calibration error equation and methodization and changing.Embodiment concrete methods of realizing is: the relevant knowledge utilizing matrix operation, picture point calculates local approach equation corresponding to picture point and this local approach equation position in the normal equation after final changing one by one, then these local approach equations are filled in the normal equation after final changing, i.e. updating method equation.Namely the normal equation after complete changing is obtained after all picture points are disposed.

For ease of implement reference, provide the present invention relates to CCD skew and scaling error calibration model as follows: introduce CCD skew and scaled error calibration parameter after, convert traditional collinearity condition equation obtain CCD skew and convergent-divergent calibration model:

$\left[\begin{array}{c}\mathrm{Xs}-Y\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\&lambda;}{R}_{\mathrm{obj}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x+\mathrm{\&Delta;x}\\ y+\mathrm{\&Delta;y}\\ -f\end{array}\right]---\left(5\right)$

Δx＝x ₀

(6)

Δy＝y ₀+y ₁c

Wherein, Δ x, Δ y represent image space coordinate correction amount respectively, x ₀for CCD is along the deviant of heading, y ₀, y ₁for the linear drift parameter perpendicular to heading of CCD, be respectively constant term and once item, c is CCD row number.

Step 7, the normal equation after utilizing least square principle solution to ask changing of step 6 gained, obtains the correction of each unknown number.

Embodiment concrete methods of realizing is: according to least square principle, and normal equation inverse matrix and constant term vector dot can obtain the correction vector of unknown number, shown in (4).

Step 8, determines whether to need to carry out next iteration process according to the correction of step 7 gained unknown number, if do not meet exit criteria, then forwards step 6 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 9 to.

Embodiment concrete methods of realizing is: arrange a unknown number correction threshold value and iterations threshold value, exit criteria is: the correction maximum of unknown number exceedes unknown number correction threshold value or iterations exceedes iterations threshold value.

Step 9, exports CCD skew and convergent-divergent calibration parameter.

Embodiment concrete methods of realizing is: CCD skew and convergent-divergent calibration parameter are exported in camera parameter file, so that next step calibration uses.

Step 10, if sensor ccd array is spliced by sub-linear CCD array, then forward step 11 to, otherwise calibration terminates.

Step 11, the established angle after use calibration and CCD offset, zooming parameter, build the error equation of the skew of sub-line array CCD and convergent-divergent calibration model, and picture point carries out methodization and changing process one by one, the normal equation after changing of generation.Build the error equation of the skew of sub-line array CCD and convergent-divergent calibration model, conventionally can realize according to calibration model.

The present invention using the position of sub-line array CCD skew and zooming parameter as unknown number, solve in the process of calibration simultaneously.During concrete enforcement, pointwise can build sub-line array CCD calibration error equation and methodization and changing.Embodiment concrete methods of realizing is: the relevant knowledge utilizing matrix operation, picture point calculates local approach equation corresponding to picture point and this local approach equation position in the normal equation after final changing one by one, then these local approach equations are filled in the rear normal equation of final changing, i.e. updating method equation.Namely the normal equation after complete changing is obtained after all picture points are disposed.

For ease of implementing reference, provide the sub-line array CCD calibration model that the present invention relates to as follows:

After introducing the skew of sub-linear CCD array and scaling error calibration parameter, obtain sub-linear CCD array skew and scaling error calibration model such as formula (7), if be spliced into a large ccd array by n sub-linear CCD array in camera focal plane, then every a line image is all divided into n part according to ccd array position, and the CCD number of each part is respectively N ₁, N ₂... N _n-1, N _n, then total CCD number is N ₁, N ₂... N _n-1, N _nsum.

$\begin{array}{c}\mathrm{\&Delta;x}=\left\{\begin{array}{cc}{x}_0& 0<c<N_1\\ x_1& N_1<c<N_1+N_2\\ x_2& N_1+N_2<c<N_1+N_2+N_3\\ & ...\\ x_{n-1}& N_1+...+N_{n-1}<c<N_1+...+N_n\end{array}\right.\\ \mathrm{\&Delta;y}=\left\{\begin{array}{cc}{y}_0+y_{1}c& 0<c<N_1\\ y_2+y_{3}c& N_1<c<N_1+N_2\\ y_4+y_{5}c& N_1+N_2<c<N_1+N_2+N_3\\ & ...\\ y_{2n-2}+y_{2n-1}c& N_1+...+N_{n-1}<c<N_1+...+N_n\end{array}\right.\end{array}---\left(7\right)$

Wherein, x ₀, x ₁, x ₂... x _n-1for CCD is along the deviant of heading, y ₀, y ₂, y ₄... y _2n-2for the offset parameter perpendicular to heading of CCD, y ₁, y ₃, y ₅... y _2n-1for the linear drift parameter perpendicular to heading of CCD, c is that image is perpendicular to heading row coordinate.

For three slice, thin piece line array CCDs, sub-line array CCD calibration model is as follows:

$\begin{array}{c}\mathrm{\&Delta;x}=\left\{\begin{array}{cc}{x}_0& 0<c<N_1\\ x_1& N_1<c<N_1+N_2\\ x_2& N_1+N_2<c<N_1+N_2+N_3\end{array}\right.\\ \mathrm{\&Delta;y}=\left\{\begin{array}{cc}{y}_0+y_{1}c& 0<c<N_1\\ y_2+y_{3}c& N_1<c<N_1+N_2\\ y_4+y_{5}c& N_1+N_2<c<N_1+N_2+N_3\end{array}\right.\end{array}---\left(8\right)$

Wherein, x ₀, x ₁, x ₂for CCD is along the deviant of heading, y ₀, y ₂, y ₄, be the offset parameter perpendicular to heading of CCD, y ₁, y ₃, y ₅for the linear drift parameter perpendicular to heading of CCD, c is total CCD row number.N ₁, N ₂, N ₃be respectively the 1st, the CCD of the 2nd and the 3rd slice, thin piece linear CCD array visits first number.

Step 12, the normal equation after utilizing least square principle solution to ask changing of step 11 gained, obtains the correction of each unknown number.

Embodiment concrete methods of realizing is: according to least square principle, and normal equation inverse matrix and constant term vector dot can obtain the correction vector of unknown number, shown in (4).

Step 13, determines whether to need to carry out next iteration process according to the correction of step 12 gained unknown number, if do not meet exit criteria, then forwards step 11 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 14 to.

Embodiment concrete methods of realizing is: arrange a unknown number correction threshold value and iterations threshold value, exit criteria is: the correction maximum of unknown number exceedes unknown number correction threshold value or iterations exceedes iterations threshold value.

Step 14, export the skew of sub-line array CCD and convergent-divergent calibration parameter, calibration terminates.

Embodiment concrete methods of realizing is: sub-line array CCD skew and convergent-divergent calibration parameter are exported in camera parameter file.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.

Claims (4)

1. a geometry calibration method in-orbit combined by the many rails of remote sensing satellite line array sensor, comprises the following steps:

Step 1, imports the tie point of empty three couplings of many tracks, the control point coordinate data of Auto-matching and the track of satellite and attitude data;

Step 2, builds the error equation of sensor established angle calibration model, and picture point carries out methodization and changing process one by one, generates the normal equation after changing;

Step 3, the normal equation after utilizing least square principle solution to ask changing of step 2 gained, obtains the correction of each unknown number;

Step 4, determines whether to need to carry out next step iterative processing according to the correction of step 3 gained unknown number, if do not meet exit criteria, then forwards step 2 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 5 to;

Step 5, exports established angle calibration parameter;

Step 6, uses the established angle parameter after calibration, builds the error equation of CCD skew and convergent-divergent calibration model, and picture point carries out methodization and changing process one by one, generates the normal equation after changing;

Step 7, the normal equation after utilizing least square principle solution to ask changing of step 6 gained, obtains the correction of each unknown number;

Step 8, determines whether to need to carry out next step iterative processing according to the correction of step 7 gained unknown number, if do not meet exit criteria, then forwards step 6 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 9 to;

Step 9, exports CCD skew and convergent-divergent calibration parameter;

Step 10, if sensor ccd array is spliced by sub-linear CCD array, then forwards step 11 to, otherwise exits;

Step 11, the established angle after use calibration and CCD offset, zooming parameter, build the error equation of the skew of sub-line array CCD and convergent-divergent calibration model, and picture point carries out methodization and changing process one by one, the normal equation after changing of generation;

Step 12, the normal equation after utilizing least square principle solution to ask changing of step 11 gained, obtains the correction of each unknown number;

Step 13, determines whether to need to carry out next step iterative processing according to the correction of step 12 gained unknown number, if do not meet exit criteria, then forwards step 11 to, regenerate error equation and normal equation, until meet exit criteria namely forward step 14 to;

Step 14, exports the skew of sub-line array CCD and convergent-divergent calibration parameter.

2. geometry calibration method in-orbit combined by the many rails of remote sensing satellite line array sensor according to claim 1, it is characterized in that: described in step 2, sensor established angle calibration model is as follows,

$\left[\begin{array}{c}\mathrm{Xs}-X\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\&lambda;}{R}_{\mathrm{obj}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]\mathrm{\&lambda;R}(p+\mathrm{\&Delta;p},l+\mathrm{\&Delta;l},k+\mathrm{\&Delta;k})\left[\begin{array}{c}x\\ y\\ -f\end{array}\right]$

Wherein, Xs, Ys, Zs represent the object coordinates of photo centre, and X, Y, Z represent topocentric object coordinates, and x, y represent the image space coordinate that ground point is corresponding, and f represents the focal length of camera, R _obj2satrepresent the spin matrix of satellite body coordinate system relative to object coordinates system, R _sat2camrepresent the spin matrix of camera to satellite body, λ is proportionality coefficient; L, p, k represents sensor that laboratory measurement obtains three established angles relative to satellite body coordinate system, Δ l, Δ p, Δ k represents the error of fixed angles of sensor relative to satellite body coordinate system, R (p+ Δ p, l+ Δ l, k+ Δ k) represents p+ Δ p, the spin matrix that l+ Δ l, k+ Δ k tri-anglecs of rotation are formed.

3. geometry calibration method in-orbit combined by the many rails of remote sensing satellite line array sensor according to claim 2, it is characterized in that: described in step 6 CCD skew and convergent-divergent calibration model as follows,

$\left[\begin{array}{c}\mathrm{Xs}-X\\ \mathrm{Ys}-Y\\ \mathrm{Zs}-Z\end{array}\right]=\mathrm{\&lambda;}{R}_{\mathrm{obj}2\mathrm{sat}}{R}_{\mathrm{sat}2\mathrm{cam}}\left[\begin{array}{c}x+\mathrm{\&Delta;x}\\ y+\mathrm{\&Delta;y}\\ -f\end{array}\right]$

Δx＝x ₀

Δy＝y ₀+y ₁c

Wherein, Δ x, Δ y represent image space coordinate correction amount respectively, x ₀for CCD is along the deviant of heading, y ₀, y ₁for the linear drift parameter perpendicular to heading of CCD, c is CCD row number.

4. geometry calibration method in-orbit combined by the many rails of remote sensing satellite line array sensor according to claim 3, it is characterized in that: sub-line array CCD skew described in step 11 and convergent-divergent calibration model as follows,

Being located at camera focal plane is be spliced into a large ccd array by n sub-linear CCD array, then every a line image is all divided into n part according to ccd array position, and the CCD number of each part is respectively N ₁, N ₂... N _n-1, N _n, then total CCD number is N ₁, N ₂... N _n-1, N _nsum;

$\mathrm{\&Delta;x}=\left\{\begin{array}{cc}{x}_0& 0<c<N_1\\ x_1& N_1<c<N_1+N_2\\ x_2& N_1+N_2<c<N_1+N_2+N_3\\ & ...\\ x_{n-1}& N_1+...+N_{n-1}<c<N_1+...+N_n\end{array}\right.$

$\mathrm{\&Delta;y}=\left\{\begin{array}{cc}{y}_0+y_{1}c& 0<c<N_1\\ y_2+y_{3}c& N_1<c<N_1+N_2\\ y_4+y_{5}c& N_1+N_2<c<N_1+N_2+N_3\\ & ...\\ y_{2n-2}+y_{2n-1}c& N_1+...+N_{n-1}<c<N_1+...+N_n\end{array}\right.$

Wherein, x ₀, x ₁, x ₂... x _n-1for CCD is along the deviant of heading, y ₀, y ₂, y ₄... y _2n-2for the offset parameter perpendicular to heading of CCD, y ₁, y ₃, y ₅... y _2n-1for the linear drift parameter perpendicular to heading of CCD, c is that image is perpendicular to heading row coordinate.