CN105631063A - Intersection moment prediction method based on object features - Google Patents

Abstract

The invention belongs to the technical field of radio wave distance measurement, and especially relates to an intersection moment prediction method based on object features. The method comprises following steps of detecting the geometrical size feature information, gray feature information and geometrical area feature information of an observed platform by utilizing an observation platform; building a space intersection moment prediction method according to the geometrical size feature information of the observed platform; building a kalman filter or a proportion feature resolving method based on the geometrical size feature information, the gray feature information and the geometrical area feature information of the observed platform; and predicting the intersection moment of the observation platform and the observed platform. According to the method provided by the invention, through utilizing the features that the geometrical size feature information, the gray feature information and the geometrical area feature information of the observed platform obtained through passive observation are relatively stable, the intersection moment can be predicted accurately; compared with the conventional passive distance measurement method, the method provided by the invention does not need a transverse maneuvering process; and moreover, compared with the conventional passive distance measurement method, the method has higher convergence efficiency and convergence precision.

Description

A kind of moment forecasting procedure that crossing of based target feature

Technical field

The invention belongs to and utilize radio wave ranging technology field, be specifically related to the moment forecasting procedure that crosses of a kind of based target feature.

Background technology

In platform process close to each other, the accurate forecast in the moment that crosses there is important Practical significance in Guidance and control, platform anticollision, the docking operation that crosses etc., the cross forecast in moment of tradition mainly has two class methods, one class is to adopt initiative range measurement method that platform spacing and closing speed are measured, then the time information that crosses is calculated, there is advantage real-time, that precision is high, but need to increase complicated initiative range measurement system, it is difficult to adapt to the process that remote (dozens of kilometres to the distance of several hundred kilometers) target quickly crosses. One class is to adopt passive ranging method, the observation angle information utilizing target is adjusted the distance by the method filtered and velocity information is settled accounts, but need observation platform and be observed between platform and have the crossrange maneuvering process of certain distance, the many solutions problem existed in filtering could be eliminated, obtain comparatively accurate distance and velocity information, crossrange maneuvering needs additional power and measurement time, is aloft very restricted with space application.

Summary of the invention

For above-mentioned prior art, it is an object of the invention to provide the moment forecasting procedure that crosses of a kind of based target feature, the gray feature information of platform that is observed, geometries characteristic information and the metastable feature of geometric area characteristic information that passive measurement obtains can be utilized, when need not observation platform and be observed between platform crossrange maneuvering and when other range sensor, to crossing, the moment carries out accurate forecast.

In order to achieve the above object, the present invention is by the following technical solutions.

The moment forecasting procedure that crosses of a kind of based target feature of the present invention, the method comprises the following steps:

(1) observation platform detects and is observed the geometries characteristic information of platform, gray feature information and geometric area characteristic information, according to the geometries characteristic information being observed platform that observation platform imaging detection system provides, set up space and cross moment forecasting procedure;

When being observed platform the physical dimension of imaging being less than or equal to 3 pixel in observation platform detection system, perform step (2);

When being observed platform the physical dimension of imaging being more than 3 pixel in observation platform detection system, perform step (3);

(2) set up based on the moment forecasting procedure that crosses being observed platform gray feature, comprise the steps:

1st step: set up and be observed platform gray feature and the analytical relation in the moment that crosses, derives the parsing relation being observed platform gray feature with remaining the time that crosses;

2nd step: judge to be observed platform signal to noise ratio of gray feature in observation platform detection system, when being observed platform the signal to noise ratio of gray feature being less than or equal to 20 in observation platform detection system, performs the 3rd step in step (2); When being observed platform the signal to noise ratio of gray feature being more than 20 in observation platform detection system, perform the 4th step in step (2);

3rd step: according to the analytical relation being derived by the 1st step in step (2), selecting initial proportion coefficient, remaining the time that crosses, remain the time-derivative that crosses is quantity of state, being observed platform gray feature is observed quantity, sets up based on the Kalman filter being observed platform gray feature;

4th step: be observed platform gray feature and the parsing relation remaining the time that crosses according to what the 1st step in step (2) was set up, the selection difference observation moment is observed the gray feature of platform, sets up residue and crosses the time and the calculation method of the ratio characteristic being observed platform gray feature;

5th step: observation platform and the moment that crosses being observed platform are forecast, output residue crosses the time and crosses moment predicted value;

(3) set up based on the moment forecasting procedure that crosses being observed platform geometric area feature or geometries characteristic, comprise the steps:

1st step: set up the analytical relation in the moment that crosses being observed platform geometric area feature or geometries characteristic, derives and is observed platform geometric area feature or geometries characteristic and residue crosses the parsing relation of time;

2nd step: judge to be observed platform signal to noise ratio of geometric area feature or geometries characteristic in observation platform detection system, when being observed platform the signal to noise ratio of geometric area feature or geometries characteristic being less than or equal to 20 in observation platform detection system, perform the 3rd step in step (3); When being observed platform the signal to noise ratio of geometric area feature or geometries characteristic being more than 20 in observation platform detection system, perform the 4th step in step (3);

3rd step: according to the analytical relation being derived by the 1st step in step (3), selection is observed platform geometric area feature or geometries characteristic, residue crosses the time, residue crosses, and time-derivative is quantity of state, it is observed platform geometric area feature or geometries characteristic is observed quantity, set up the Kalman filter being observed platform geometric area feature or geometries characteristic;

4th step: according to the 1st step in step (3) set up be observed platform geometric area feature or geometries characteristic and residue crosses the parsing relation of time, the selection difference observation moment is observed geometric area feature or the geometries characteristic of platform, sets up residue and crosses the time and the calculation method of the ratio characteristic being observed platform geometric area feature or geometries characteristic;

5th step: observation platform and the moment that crosses being observed platform are forecast, output residue crosses the time and crosses moment predicted value.

In described step (2), the 1st step is observed platform gray feature and the analytical relation in the moment that crosses is:

$g_{\mathrm{Tar}}\left(T\right)=\frac{K\·I_{\mathrm{Tar}}}{R{\left(T\right)}^2}=\frac{K\·I_{\mathrm{Tar}}}{{(v\·T)}^2}=\frac{K\·I_{\mathrm{Tar}}}{v^2}\·\frac{1}{T^2}$

In formula: T represents that residue crosses the time, unit is s; g_Tar(T) gray feature of platform it is observed for the T moment; R is the distance being observed platform between observation platform optical imaging system, and unit is m; K is observation platform detection system conversion coefficient; I_TarFor being observed platform radiant intensity; V is observation platform and is observed between platform closing speed, and unit is m/s;For observation platform detection system conversion coefficient K, it is observed platform radiant intensity I_TarAnd observation platform and be observed between platform the ratio of closing speed v square, estimates according to observation platform result of detection.

In described step (3), the 1st step is observed the cross analytical relation of time of platform geometric area feature and residue and is:

$S_{\mathrm{sen}}\left(T\right)=S_{\mathrm{Obj}}{(\frac{f}{v}\·\frac{1}{S_{\mathrm{Pix}}})}^2\·\frac{1}{T^2}$

In formula: S_sen(T) being the geometric area feature being observed platform the T moment in observed image, unit is pixel; S_ObjBeing the geometric area feature being observed platform, unit is pixel; F is optical imaging system focal length, and unit is m; V is observation platform and is observed between platform closing speed, and unit is m/s; S_PixBeing observation platform detection system Pixel Dimensions, unit is m; T is that residue crosses the time, and unit is s;

In described step (3), the 1st step is observed the cross analytical relation of time of platform geometries characteristic and residue and is:

$L_{\mathrm{sen}}=L_{\mathrm{Obj}}\·\frac{f}{v}\·\frac{1}{S_{\mathrm{Pix}}}\·\frac{1}{T}$

In formula: L_sen(T) being be observed platform geometries characteristic in observed image the T moment, unit is pixel; L_ObjBeing the geometries characteristic being observed platform, unit is pixel; F is optical imaging system focal length, and unit is m; V is observation platform and is observed between platform closing speed, and unit is m/s; S_PixBeing observation platform detection system Pixel Dimensions, unit is m; T is that residue crosses the time, and unit is s.

In described step (2), in the 3rd step and step (3), in the 3rd step, Kalman filter relational expression is:

$X(k+1)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& \mathrm{\Δt}\\ 0& 0& 1\end{array}\right)X\left(k\right)=\mathrm{\ΦX}\left(k\right)$

In formula: X (k+1) is the quantity of state X value in the k+1 moment; X (k) is the quantity of state X value in the k moment; �� t is interval, and unit is s; �� is state-transition matrix.

In described step (2), in the 3rd step and step (3), in the 3rd step, the moment forecast iterative process that crosses of Kalman filter is:

A. predictive equation is calculated:

${\hat{X}}_{K+1/K}={\mathrm{\ΦX}}_k$

In formula:For state a step of forecasting value; �� is state-transition matrix; X_kFor the vector X value in the k moment;

B. computation and measurement method is generalized into the Jacobi matrix at future position:

$H_{K+1}={\frac{\∂G\left[X\right]}{\∂X}|}_{X={\hat{X}}_{k+1/K}}$

In formula: H_k+1Jacobi matrix for k+1; The right-hand member nonlinear function vector that G [��] is systematic observation equation group; X is state variable;For state a step of forecasting value;

C. prediction covariance and Kalman gain are calculated:

$P_{K+1/K}={\mathrm{\Φ}}_{K+1/K}{P}_{K/K}{\mathrm{\Φ}}_{K+1/K}^T$

$K_{K+1}=P_{K+1/K}{H}_{k+1}^{+T}{(H_{k+1}{P}_{K+1/K}{H}_{k+1}^T+R_{k+1})}^{-1}$

In formula: P_K+1/KFor a step of forecasting variance matrix; ��_K+1/KFor state-transition matrix; P_K/KFor filtering error variance matrix;For matrix ��_K+1/KTransposed matrix; K_K+1For filtering gain matrix; H_k+1Jacobi matrix for k+1;For matrix H_k+1Transposed matrix; R_K+1For systematic observation noise variance matrix;

D. filtering estimated value is obtained:

${\hat{X}}_{K+1/K+1}={\hat{X}}_{K+1/K}+K_{K+1}[G_{K+1}-G\left({\hat{X}}_{K+1/K}\right)]$

In formula:It it is the vector X value according to the k+1 moment estimated value to k+1 moment value;For state a step of forecasting value; K_K+1For filtering gain matrix; G_K+1Observation vector for the k+1 moment;For according to state a step of forecasting valueThe forecast vector calculated;

If E. adopting Modified covariance extended Kalman filter device, reusing future position data and recalculating

Jacobi matrix:

$H_{K+1}^{+}={\frac{\∂G\left[X\right]}{\∂X}|}_{X={\hat{X}}_{k+1/K+1}}$

In formula:The Jacobi matrix of the k+1 for recalculating by future position data; The right-hand member nonlinear function vector that G [��] is systematic observation equation group; X is state variable;It it is the estimation to k+1 moment value of the vector X value according to the k+1 moment;

F. prediction covariance and Kalman gain are recalculated:

$K_{K+1}^{+}=P_{K+1/K}{H}_{k+1}^{+T}{(H_{k+1}{P}_{K+1/K}{H}_{k+1}^T+R_{k+1})}^{-1}$

$P_{K+1/K+1}=(I-K_{K+1}^{+}{H}_{K+1}^{+})P_{K+1/K}{(I-K_{K+1}^{+}{H}_{K+1}^{+})}^T+K_{K+1}^{+}{R}_{K+1}{K}_{K+1}^{T+}$

In formula:For the filtering gain matrix recalculated by future position data;For matrixTransposed matrix; P_K+1/KFor a step of forecasting variance matrix; P_K+1/K+1For; H_k+1Jacobi matrix for k+1;The Jacobi matrix of the k+1 for recalculating by future position data;For sensing amountTransposed matrix; R_K+1For systematic observation noise variance matrix; I is unit vector matrix; T is that residue crosses the time, unit s;

G. can utilize after being predicted the outcome residue cross time T forecast estimate filter value;

Repeating step A to step F, to crossing, the moment forecasts, obtains residue and crosses the time and cross moment predicted value.

Setting up in process in Kalman filter, the residue time-derivative that crosses is set to fixed value-1, and residue crosses the value that the standard variance of time-derivative is set between 0.01��0.00001.

In described step (2), the cross calculation method relational expression of time and the ratio characteristic that is observed platform gray feature of the 4th step residue is:

$T=\frac{\mathrm{\Δt}}{1-{\left(\frac{g_{\mathrm{Tar}}\left(T\right)}{g_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}^{1/2}}$

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; g_Tar(T) it is the gray value T moment being observed platform, accordingly, g_Tar(T-�� t) is the gray value that T-�� t is observed platform.

In described step (3), the cross calculation method relational expression of time and the ratio characteristic that is observed platform geometric area feature of the 4th step residue is:

$T=\frac{\mathrm{\Δt}}{1-{\left(\frac{S_{\mathrm{Tar}}\left(T\right)}{S_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}^{1/2}}$

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; S_Tar(T) being be observed the two dimensional image geometric area feature that platform presents in observation platform detection system the T moment, unit is pixel; Accordingly, S_Tar(T-�� t) is the two dimensional image geometric area feature that T-�� t is observed that platform presents in observation platform detection system, and unit is pixel;

In described step (3), the cross resolving relational expression of time and the ratio characteristic that is observed platform geometries characteristic of the 4th step residue is:

$T=\frac{\mathrm{\Δt}}{1-\left(\frac{L_{\mathrm{Tar}}\left(T\right)}{L_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}$

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; L_Tar(T) being be observed the one dimensional image geometries characteristic that platform presents in observation platform detection system the T moment, unit is pixel; Accordingly, L_Tar(T-�� t) is the one dimensional image geometries characteristic that T-�� t is observed that platform presents in observation platform detection system, and unit is pixel.

Described Kalman filter is Modified covariance extended Kalman filter device.

The technical scheme that the embodiment of the present invention provides has the benefit that

The moment forecasting procedure that crosses of a kind of based target feature of the present invention, accurate forecast is carried out to close to the moment that crosses between platform, this method utilizes the platform gray feature information that is observed, geometries characteristic information and the metastable feature of geometric area characteristic information that passive measurement obtains, it is thus achieved that the accurate forecast to the moment that crosses; More traditional passive ranging method, it is not necessary to crossrange maneuvering process, more traditional passive ranging method has higher convergent probability and convergence precision simultaneously.

Accompanying drawing explanation

Fig. 1 is the moment forecasting procedure flow chart that crosses of based target feature of the present invention;

Fig. 2 is based on the platform grayscale emulational sequence that is observed being observed platform gray feature and estimates gray value curve chart with Kalman filter;

Fig. 3 is based on the Kalman filter being observed platform gray feature and crosses direct Predictor and filter value and the actual value curve chart of moment forecasting procedure;

Fig. 4 is based on the Kalman filter being observed platform gray feature and crosses the absolute error curve chart of moment forecasting procedure;

Fig. 5 is based on the Kalman filter being observed platform gray feature and crosses the relative error curve chart of moment forecasting procedure;

When Fig. 6 is interval �� t=2.5s, it is observed platform grayscale emulational sequence curve figure based on what ratio resolved that relation crosses moment forecasting procedure;

When Fig. 7 is interval �� t=2.5s, resolves relation based on the ratio being observed platform gray feature and cross predictive value and the actual value curve chart of moment forecasting procedure;

When Fig. 8 is interval �� t=2.5s, the absolute error curve chart of the moment forecasting procedure that crosses based on the ratio resolving relation being observed platform gray feature;

When Fig. 9 is interval �� t=2.5s, the relative error curve chart of the moment forecasting procedure that crosses based on the ratio resolving relation being observed platform gray feature;

What Figure 10 was based on that the Kalman filter being observed platform geometric properties crosses moment forecasting procedure is observed platform area characteristics simulation sequence curve figure;

Figure 11 is based on the Kalman filter being observed platform geometric properties and crosses direct Predictor and filter value and the actual value curve chart of moment forecasting procedure;

Figure 12 is based on the Kalman filter being observed platform geometric properties and crosses the absolute error curve chart of moment forecasting procedure;

When Figure 13 is interval �� t=2.5s, resolves relation based on the ratio being observed platform geometric properties and cross predictive value and the actual value curve chart of moment forecasting procedure;

When Figure 14 is interval �� t=2.5s, the absolute error curve chart of the moment forecasting procedure that crosses based on the ratio resolving relation being observed platform geometric properties.

Detailed description of the invention

Elaborate below in conjunction with the drawings and specific embodiments space of a kind of based target observational characteristic of the present invention crossed moment forecasting procedure.

The moment forecasting procedure as it is shown in figure 1, the space of a kind of based target observational characteristic of the present invention crosses, comprises the following steps:

Observation platform imaging detection system provides and is observed the gray feature information of platform, geometries characteristic information and geometric area characteristic information, according to the geometries characteristic information being observed platform, sets up the moment forecasting procedure that crosses;

One, when being observed platform the physical dimension of imaging being less than or equal to 3 pixel in observation platform detection system, set up based on the moment forecasting procedure that crosses being observed platform gray feature, comprise the steps:

(1) foundation is observed platform gray feature and the analytical relation in the moment that crosses, and derives the parsing relation being observed platform gray feature with remaining the time that crosses;

Imaging process according to observation platform passive measurement system, is observed platform, the radiation of background is converted to digital picture through optical imaging system, is observed platform, gray scale that the radiant intensity of background is presented as in target digital image. After observation platform system compensation, demarcating, being observed the relation between the radiation feature of platform and gray feature can represent with formula 1.

$\begin{array}{c}{g}_{(\mathrm{Bac}+\mathrm{Tar})}=K\·\frac{I_{\mathrm{Tar}}+I_{\mathrm{Bac}}}{R^2}+B=\frac{K\·I_{\mathrm{Tar}}}{R^2}+(\frac{K\·I_{\mathrm{Bac}}}{R^2}+B)\\ \⇒g_{\mathrm{Tar}}=g_{(\mathrm{Bac}+\mathrm{Tar})}-g_{\mathrm{Bac}}=\frac{K{\·I}_{\mathrm{Tar}}}{R^2}\end{array}$ Formula 1

In formula: g_(Bac+Tar)For being observed the gray scale that platform and background radiation are formed on observation platform optical imaging system; I_TarFor being observed the radiant intensity of platform, unit is W/Sr; I_BacFor the radiant intensity of background, unit is W/Sr; R is the distance being observed platform between observation platform optical imaging system, and unit is m; K is the observation platform optical imaging system conversion proportion coefficient to radiant intensity to gradation of image, and unit is m²Sr/W; B is observation platform optical imaging system background image gray scale; g_TarFor being observed the gray scale that platform is formed on observation platform optical imaging system; g_BacFor the gray scale that background radiation is formed on observation platform optical imaging system.

Representing that residue crosses the time with T, unit is s, remains the time T that crosses and is gradually reduced; V is observation platform and is observed between platform closing speed, and unit is m/s; Then formula 1 is rewritable is:

$g_{\mathrm{Tar}}\left(T\right)=\frac{K\·I_{\mathrm{Tar}}}{R{\left(T\right)}^2}=\frac{K\·I_{\mathrm{Tar}}}{{(v\·T)}^2}=\frac{K\·I_{\mathrm{Tar}}}{v^2}\·\frac{1}{T^2}$ Formula 2

In formula: T represents that residue crosses the time, unit is s; g_Tar(T) gray feature of platform it is observed for the T moment; R is the distance being observed platform between observation platform optical imaging system, and unit is m; K is observation platform detection system conversion coefficient; I_TarFor being observed platform radiant intensity; V is observation platform and is observed between platform closing speed, and unit is m/s;For observation platform detection system conversion coefficient K, it is observed platform radiant intensity I_TarAnd observation platform and be observed between platform the ratio of closing speed v square, estimates according to observation platform result of detection.

Due in formula (2)It is observation platform detection system conversion coefficient K, is observed platform radiant intensity I_TarAnd observation platform and be observed between platform the ratio of closing speed v square, according to close assumed conditions be observed the metastable feature of platform radiation feature, observation platform detection system conversion coefficient K, is observed platform radiant intensity I_TarAnd observation platform and to be observed between platform closing speed v be definite value; It is therefore contemplated that it is relevant to be observed the platform gray feature time T that only crosses with residue, and become inverse square relation.

From the foregoing, it will be observed that utilize passive measurement to be observed platform gray feature, it is possible to estimate the moment that crosses exactly, it is not necessary to the motor-driven grade of additional lateral other operation.

(2) judge to be observed platform signal to noise ratio of gray feature in observation platform detection system, when being observed platform the signal to noise ratio of gray feature being less than 20 in observation platform detection system, based on (1st) step observation platform gray feature in step one and the analytical relation remaining the time that crosses, select to be observed platform initial gray G_Gray0, cross time T, the residue time-derivative T' that crosses of residue be state variable, be observed the gray feature g of platform_Tar(T) for observed quantity, Kalman filter is set up.

Utilize formula 2, select state variable to select X=(G_Gray0TT') ', wherein G_Gray0Crossing the time for initial gray, T residue, T' is the first derivative of T. Ideally, state equation $A=\left(\begin{array}{ccc}0& 0& 0\\ 0& 0& 1\\ 0& 0& 0\end{array}\right),$ Observational equationThe foundation of concrete Kalman filter filtering is as follows:

X=(G_Gray0TT') '=(G₀TT')^TFormula 3

$\stackrel{\·}{X}\left(t\right)=\mathrm{AX}=\left(\begin{array}{ccc}0& 0& 0\\ 0& 0& 1\\ 0& 0& 0\end{array}\right)X$ Formula 4

$g\left(T\right)=G_{\mathrm{Gray}0}\·\frac{1}{T^2}+V\left(T\right)$ Formula 5

After discretization it is:

$X(k+1)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& \mathrm{\Δt}\\ 0& 0& 1\end{array}\right)X\left(k\right)=\mathrm{\ΦX}\left(k\right)$ Formula 6

In formula: G_Gray0For being observed platform initial gray value, unit is pixel; T residue crosses the time, and unit is s; T' is the first derivative of T; V (T) is background noise; X (k+1) is the quantity of state X value in the k+1 moment; X (k) is the quantity of state X value in the k moment; �� t is interval, and unit is s; �� is state-transition matrix.

Due to observational equationContaining nonlinear terms, therefore Kalman filter uses extended Kalman filter (EKF), in order to increase the non-linear estimation thought, it would however also be possible to employ Modified covariance extended Kalman filter device (MCEKF), it is possible to improve the common EKF wave filter situation to convergence.

Due to the existence of the nonlinear terms of EKF and MCEKF, when initial estimation parameter arranges unreasonable, it will causing that the convergence rate of wave filter is slow, therefore parameter is arranged has large effect to filter effect, is especially observed the initial gray value G of platform_Gray0, residue cross time T, be observed platform initial gray value G_Gray0Square estimated value E [G₀ ²], residue crosses the estimated value E [T of T square of initial value of time₀ ²] initial value set, affect Kalman filter convergence rate and restrain after precision, when being observed platform initial gray value G_Gray0Square estimated value E [G₀ ²], the estimated value E [T of T square of initial value of time₀ ²] parameter is when arranging bigger, fast convergence rate, but after convergence, concussion is big, and filtering accuracy is relatively low; When being observed platform initial gray value G_Gray0Square estimated value E [G₀ ²], the estimated value E [T of T square of initial value of time₀ ²] parameter arranges too small, convergence rate is affected, but convergence post filtering value stabilization, precision is high.

Based on the Design on Kalman Filter method of passive measurement process in the design of wave filter, the information such as distance closing speed does not occur, being observed platform radiant intensity (being observed platform geometric area feature or geometries characteristic), observation platform and be observed between platform, can be seen that from formula 6, the Kalman filter set up only occurs in that 3 quantity of states, the letter of such design of filter structure, variable is few, it is possible to the significantly convergent probability of boostfiltering device and convergence rate.

Set up in process in Kalman filter, in the state equation selected, remain the time T that crosses to gradually decrease, the time first derivative T' that can residue be crossed is set to fixed value-1, so reducing the number of unknown quantity in quantity of state space, improve the convergent probability of wave filter, its standard variance is set to the value between 0.01��0.00001 simultaneously, in acceptable filtering error, improve the convergence rate of wave filter.

(3) Kalman filter set up based on (2nd) step in step one, forecast observation platform and the moment that crosses being observed platform, obtains residue and crosses time T and cross moment predicted value;

The moment forecast iterative process that crosses based on Kalman filter is as follows:

A. predictive equation is calculated:

${\hat{X}}_{K+1/K}={\mathrm{\ΦX}}_k$ Formula 7

In formula:For state a step of forecasting value; �� is state-transition matrix; X_kFor the vector X value in the k moment;

B. computation and measurement method is generalized into Jacobi (Jacobi) matrix at future position:

$H_{K+1}={\frac{\∂G\left[X\right]}{\∂X}|}_{X={\hat{X}}_{k+1/K+1}}$ Formula 8

In formula: H_k+1Jacobi (Jacobi) matrix for k+1; The right-hand member nonlinear function vector that G [��] is systematic observation equation group; X is state variable;For state a step of forecasting value;

C. prediction covariance and Kalman (Kalman) gain are calculated:

$P_{K+1/K}={\mathrm{\Φ}}_{K+1/K}{P}_{K/K}{\mathrm{\Φ}}_{K+1/K}^T$ Formula 9

$K_{K+1}=P_{K+1/K}{H}_{k+1}^{+T}{(H_{k+1}{P}_{K+1/K}{H}_{k+1}^T+R_{k+1})}^{-1}$ Formula 10

In formula: P_K+1/KFor a step of forecasting variance matrix; ��_K+1/KFor state-transition matrix; P_K/KFor filtering error variance matrix;For matrix ��_K+1/KTransposed matrix; K_K+1For filtering gain matrix; H_k+1Jacobi (Jacobi) matrix for k+1;For matrix H_k+1Transposed matrix; R_K+1For systematic observation noise variance matrix;

D. filtering estimated value is obtained:

${\hat{X}}_{K+1/K+1}={\hat{X}}_{K+1/K}+K_{K+1}[G_{K+1}-G\left({\hat{X}}_{K+1/K}\right)]$ Formula 11

In formula:It it is the vector X value according to the k+1 moment estimated value to k+1 moment value;For state a step of forecasting value; K_K+1For filtering gain matrix; G_K+1Observation vector for the k+1 moment;For according to state a step of forecasting valueThe forecast vector calculated

If E. adopting Modified covariance extended Kalman filter device (MCEKF), reusing future position data and recalculating Jacobi (Jacobi) matrix:

$H_{K+1}^{+}={\frac{\∂G\left[X\right]}{\∂X}|}_{X={\hat{X}}_{k+1/K+1}}$ Formula 12

In formula:Jacobi (Jacobi) matrix of the k+1 for recalculating by future position data; The right-hand member nonlinear function vector that G [��] is systematic observation equation group; X is state variable;It it is the estimation to k+1 moment value of the vector X value according to the k+1 moment;

F. prediction covariance and Kalman (Kalman) gain are recalculated:

$K_{K+1}^{+}=P_{K+1/K}{H}_{k+1}^{+T}{(H_{k+1}{P}_{K+1/K}{H}_{k+1}^T+R_{k+1})}^{-1}$ Formula 13

$P_{K+1/K+1}=(I-K_{K+1}^{+}{H}_{K+1}^{+})P_{K+1/K}{(I-K_{K+1}^{+}{H}_{K+1}^{+})}^T+K_{K+1}^{+}{R}_{K+1}{K}_{K+1}^{T+}$ Formula 14

In formula:For the filtering gain matrix recalculated by future position data;For matrixTransposed matrix; P_K+1/KFor a step of forecasting variance matrix; P_K+1/K+1For; H_k+1Jacobi (Jacobi) matrix for k+1;Jacobi (Jacobi) matrix of the k+1 for recalculating by future position data;For sensing amountTransposed matrix; R_K+1For systematic observation noise variance matrix; I is unit vector matrix; T is that residue crosses the time, unit s;

G. can utilize after being predicted the outcome residue cross time T forecast estimate filter value;

Repeating step A to step F, to crossing, the moment forecasts, obtains residue and crosses the time and cross moment predicted value.

(4) when being observed platform the signal to noise ratio of gray feature being more than 20 in observation platform detection system, it is observed platform gray feature and the parsing relation remaining the time that crosses according to what (1st) step in step one was set up, the selection difference observation moment is observed the gray feature of platform, sets up residue and crosses the time and the calculation method of the ratio characteristic being observed platform gray feature;

Being observed platform gray feature sequence based on what obtain in observation platform detection system, the two observation moment selecting interval to be �� t are observed the gray value g of platform_Tar(T)��g_Tar(T-�� t), can obtain according to formula 3:

$T=\frac{\mathrm{\Δt}}{1-{\left(\frac{g_{\mathrm{Tar}}\left(T\right)}{g_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}^{1/2}}$ Formula 15

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; g_Tar(T) it is the gray value T moment being observed platform, accordingly, g_Tar(T-�� t) is the gray value that T-�� t is observed platform.

Be observed platform gray feature and the residue that utilize that formula 15 sets up cross the parsing relation of time, it is possible to directly calculation goes out residue and crosses the time.

Above-mentioned it is set as based on the Kalman filter forecasting process initial parameter being observed platform gray feature:

Assuming that under vacuum conditions, observation platform and be observed between platform initial distance 13000m, at the uniform velocity close between platform, closing speed is 1000m/s; Being observed platform is initially point target on the detection system, and the radiant intensity feature being observed platform remains stable for, and according to above-mentioned condition, observation platform and the moment that crosses being observed between platform is forecast; Wherein known observation platform maximum detectable range 10000m, observation platform and the typical closing speed being observed between platform are 700m/s��1200m/s. Assume optical imaging system Space Angle 0.05mrad, be observed platform size at 1m��10m.

Initial predicted time value is set as 16, owing to the time is fixing reduction, so being T'=-1; Gray scale initial estimate sets 100, then filter original state ${\hat{x}}_{\%}={\left(\begin{array}{ccc}100& 16& -1\end{array}\right)}^T,$ In formula,For filter error variance matrix setup values.

Systematic observation error statistics feature-set is as follows: the estimated value E [V of observation noise V initial value₀]=0, the estimated value E [V of observation noise V square of initial value_K ²]=50.

Initial value observational characteristic is set as follows according to observing capacity: the estimated value E [G of gray value G initial value₀]=0, the estimated value E [G of G square of initial value of gray value₀ ²]=5 �� 10⁴; Estimated value E [the T of time T initial value₀]=0, the estimated value E [T of T square of initial value of time₀ ²]=25; The first derivative estimated value E [T' of time T₀]=0, the initial value estimated value E [T of the first derivative square of time T₀'²]=0.0001. Work as T'₀It is fixed value-1, belongs to known quantity, work as T'₀Deviation-1, filter result will appear from mistake, in order to ensure the correct convergence of wave filter, arranges E [T₀'²]=0.0001, it is ensured that T'₀With very little deviation profile near-1.

Fig. 2 to Fig. 5 is the emulation forecast result schematic diagram of moment forecasting procedure of crossing based on the Kalman filter being observed platform gray feature. From Fig. 2 to Fig. 5 it can be seen that according to simulation analysis, based on the wave filter that said method is set up, prediction error is less than 10%, and wave filter converges to prediction error less than in the scope of 10% in 4��6s.

When Fig. 6 to Fig. 9 gives �� t=2.5s, the emulation forecast result schematic diagram of the moment forecasting procedure that crosses based on the ratio resolving relation being observed platform gray feature. From Fig. 6 to Fig. 9 it can be seen that when signal to noise ratio is higher, be observed the change of platform gray feature very fast time, prediction error is less than 10%.

Two, when being observed platform the physical dimension of imaging being more than 3 pixel in observation platform detection system, set up based on the moment forecasting procedure that crosses being observed platform geometric properties, comprise the steps:

(1) foundation is observed platform geometric properties and the analytical relation in the moment that crosses, and derives the parsing relation being observed platform geometric properties with remaining the time that crosses;

According to observation platform optical system imaging relation, being observed the imaging session of platform, be observed platform size of Pixel Dimensions in observation platform detection system, its expression formula is as follows:

$L_{\mathrm{sen}}=L_{\mathrm{Obj}}\·\frac{f}{R}\·\frac{1}{S_{\mathrm{Pix}}}$ Formula 16

In formula, L_senBeing the pixel size being observed platform in observation platform detection system, unit is pixel; The distance of R target range detection system, unit is m; F is the focal length of observation platform optical imagery detection system, and unit is m; L_ObjBeing the one-dimensional physical dimension being observed platform, unit is m; S_PixBeing the Pixel Dimensions of observation platform detection system, unit is pixel.

With the derivation of formula 2, representing that residue crosses the time with T, unit is s, remains the time T that crosses and is gradually reduced; V is observation platform and is observed between platform closing speed, and unit is m/s; Can be derived by from formula 17:

$L_{\mathrm{sen}}\left(T\right)=L_{\mathrm{Obj}}\·\frac{f}{R}\·\frac{1}{S_{\mathrm{Pix}}}=L_{\mathrm{Obj}}\·\frac{f}{v\·T}\·\frac{1}{S_{\mathrm{Pix}}}=L_{\mathrm{Obj}}\·\frac{f}{v}\·\frac{1}{S_{\mathrm{Pix}}}\·\frac{1}{T}$ Formula 17

Then can derive the geometric area feature being observed platform in observation platform detection system:

$S_{\mathrm{sen}}\left(T\right)=S_{\mathrm{Obj}}{(\frac{f}{v}\·S_{\mathrm{Pix}})}^2\·\frac{1}{T^2}=S_{\mathrm{Obj}}{(\frac{f}{v}\·\frac{1}{S_{\mathrm{Pix}}})}^2\·\frac{1}{T^2}$ Formula 18

Utilize formula 18 to carry out remaining the forecast of the time of crossing by the geometric area feature being observed platform, can also forecast that friendship is carved according to formula 17 with the one-dimensional geometries characteristic being observed platform equally; The geometric properties information prediction being observed platform is utilized to cross the moment, it is not necessary to additional any crossrange maneuvering process.

(2) judge to be observed platform signal to noise ratio of geometric area feature or geometries characteristic in observation platform detection system, when being observed platform the signal to noise ratio of geometric area feature or geometries characteristic being less than 20 in observation platform detection system, the analytical relation can derived according to (1st) step in step 2, selects to be observed the initial geometric surface product value S of platform_sen0Or initial physical dimension value L_sen0, cross time T, the residue time-derivative T' that crosses of residue be state variable, be observed the geometric area feature S of platform_senOr geometries characteristic L (T)_sen(T) for observed quantity, Kalman filter is set up.

Below to be observed the geometric area feature S of platform_sen(T) be example for observed quantity, instruction card Thalmann filter set up process, be observed platform geometries characteristic L_sen(T) process of Kalman filter is set up for observed quantity prefabricated similar.

Utilize formula 18, select state variable to select X=(S_sen0TT') ', wherein S_sen0Crossing the time for initial geometric surface product value, T residue, T' is the first derivative of T. Ideally, state equation $A=\left(\begin{array}{ccc}0& 0& 0\\ 0& 0& 1\\ 0& 0& 0\end{array}\right),$ Observational equationKalman filter residue is crossed time T is utilized to be predicted and estimation.

The foundation of concrete Kalman filter filtering is as follows:

X=(S_sen0TT') '=(S_sen0TT')^TFormula 19

$\stackrel{\·}{X}\left(t\right)=\mathrm{AX}=\left(\begin{array}{ccc}0& 0& 0\\ 0& 0& 1\\ 0& 0& 0\end{array}\right)X$ Formula 20

$S_{\mathrm{sen}}\left(T\right)=S_{\mathrm{Sen}0}\·\frac{1}{T^2}+V\left(T\right)$ Formula 21

After discretization it is:

$X(k+1)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& \mathrm{\Δt}\\ 0& 0& 1\end{array}\right)X\left(k\right)=\mathrm{\ΦX}\left(k\right)$ Formula 22

In formula: S_sen0For initial geometric surface product value, unit is pixel; T residue crosses the time, and unit is s; T' is the first derivative of T; V (T) is background noise; X (k+1) is the quantity of state X value in the k+1 moment; X (k) is the quantity of state X value in the k moment; �� t is interval, and unit is s; �� is state-transition matrix.

Due to observational equationContaining nonlinear terms, therefore Kalman filter uses extended Kalman filter (EKF), in order to increase the non-linear estimation thought, it would however also be possible to employ Modified covariance extended Kalman filter device (MCEKF), it is possible to improve the common EKF wave filter situation to convergence.

Based on the Design on Kalman Filter method of passive measurement process in the design of wave filter, the information such as distance closing speed does not occur, being observed platform radiant intensity (being observed platform geometric properties), observation platform and be observed between platform, can be seen that from formula 22, the Kalman filter set up only occurs in that 3 quantity of states, the letter of such design of filter structure, variable is few, it is possible to the significantly convergent probability of boostfiltering device and convergence rate.

Set up in process in Kalman filter, in the state equation selected, remain the time T that crosses to gradually decrease, the time first derivative T' that can residue be crossed is set to fixed value-1, so reducing the number of unknown quantity in quantity of state space, improve the convergent probability of wave filter, its standard variance is set to the value between 0.01��0.00001 simultaneously, in acceptable filtering error, improve the convergence rate of wave filter.

(3) based on the Kalman filter set up in step 2 (2nd) step, observation platform and the moment that crosses being observed platform are forecast, obtain residue and cross time T and cross moment predicted value; The iterative process of Kalman filter is consistent with (3rd) step iterative process A to F in step one.

(4) judge to be observed platform signal to noise ratio of geometric area feature or geometries characteristic in observation platform detection system, when being observed platform the signal to noise ratio of geometric area feature or geometries characteristic being more than 20 in observation platform detection system, it is observed platform geometric properties and the parsing relation remaining the time that crosses according to what (1st) step in step 2 was set up, the selection difference observation moment is observed the geometric properties of platform, sets up residue and crosses the time and the calculation method of the ratio characteristic being observed platform geometric properties;

It is observed platform geometric area feature or geometries characteristic sequence based on what observation platform detection system obtained, the two observation moment selecting interval to be �� t are observed the geometric properties of platform, it is possible to obtain remaining the time of crossing and the resolving relational expression of the ratio characteristic being observed platform geometric area feature is:

$T=\frac{\mathrm{\Δt}}{1-{\left(\frac{S_{\mathrm{Tar}}\left(T\right)}{S_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}^{1/2}}$ Formula 23

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; S_Tar(T) being be observed the two dimensional image geometric area feature that platform presents in observation platform detection system the T moment, unit is pixel; Accordingly, S_Tar(T-�� t) is the two dimensional image geometric area feature that T-�� t is observed that platform presents in observation platform detection system, and unit is pixel.

With formula 23, residue crosses the time and the resolving relational expression of the ratio characteristic being observed platform geometries characteristic is:

$T=\frac{\mathrm{\Δt}}{1-\left(\frac{L_{\mathrm{Tar}}\left(T\right)}{L_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}$ Formula 24

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; L_Tar(T) being be observed the one dimensional image geometries characteristic that platform presents in observation platform detection system the T moment, unit is pixel; Accordingly, L_Tar(T-�� t) is the one dimensional image geometries characteristic that T-�� t is observed that platform presents in observation platform detection system, and unit is pixel.

Utilize that formula 23 or formula 24 set up be observed platform geometric area feature or geometries characteristic and residue crosses the parsing relation of time, it is possible to directly calculation goes out residue and crosses the time.

Above-mentioned it is set as based on the Kalman filter forecasting process initial parameter being observed platform geometric area feature or geometries characteristic:

Assuming that under vacuum conditions, observation platform and be observed between platform initial distance 130m, at the uniform velocity close between platform, closing speed is 10m/s; Being observed platform is initially speckle target on the detector, and the radiant intensity feature of target remains stable for, and according to above-mentioned condition, the intersection moment between platform is forecast; Wherein known passive measurement platform maximum detectable range 200m, the typical closing speed between platform is 7m/s��15m/s. Assuming optical imaging system Space Angle 1.0mrad, target size is at 0.5m��3m.

Systematic observation error statistics feature-set is as follows: the estimated value E [V of observation noise V initial value₀]=0, the estimated value E [V of observation noise V square of initial value_K ²]=1.

Initial value observational characteristic is set as follows according to observing capacity: maximum detectable range 200m, the about 7��15m/s of closing speed, and therefore initial predicted time value is set as 18s, owing to the time is fixing reduction, so being T'=-1; Gray scale initial estimate sets 20, then filter original state ${\hat{X}}_{\%}={\left(\begin{array}{ccc}20& 18& -1\end{array}\right)}^T;$ Systematic observation error statistics feature E [V₀]=0, E [V_K ²]=10; Initial value observational characteristic is set as follows E [G according to observing capacity₀]=0, E [G₀ ²]=2 �� 10²; E [T₀]=0, E [T₀ ²]=25; E [T'₀]=0, works as T'₀Deviation-1, filter result will appear from mistake, in order to ensure the correct convergence of wave filter, arranges E [T₀'²]=0.0001, it is ensured that T'₀With very little deviation profile near-1.

Figure 10 to Figure 12 gives the emulation forecast result schematic diagram of the moment forecasting procedure that crosses based on the Kalman filter being observed platform area geometric properties, and Figure 11 cathetus is actual value, and the curve of cyclical fluctuations is Kalman filtering value. When Figure 13 and Figure 14 gives �� t=2.5s, the emulation forecast result schematic diagram of the moment forecasting procedure that crosses based on the ratio resolving relation being observed platform area geometric properties, Figure 13 cathetus is actual value, and the curve of cyclical fluctuations is Kalman filtering value.

The moment forecasting procedure that crosses from step one and step 2 can be seen that, utilize and be observed platform gray feature, geometric area feature or geometries characteristic and the residue relation of time moment that carries out crossing that crosses and forecast, completely not by the crossrange maneuvering of Additional Observations platform device or other supplementary observation process, just it is capable of the forecast remaining the time that crosses.

Claims (9)

1. the moment forecasting procedure that crosses of a based target feature, it is characterised in that: comprise the following steps:

(1) observation platform detects and is observed the geometries characteristic information of platform, gray feature information and geometric area characteristic information, according to the geometries characteristic information being observed platform that observation platform imaging detection system provides, set up space and cross moment forecasting procedure;

When being observed platform the physical dimension of imaging being less than or equal to 3 pixel in observation platform detection system, perform step (2);

When being observed platform the physical dimension of imaging being more than 3 pixel in observation platform detection system, perform step (3);

(2) set up based on the moment forecasting procedure that crosses being observed platform gray feature, comprise the steps:

1st step: set up and be observed platform gray feature and the analytical relation in the moment that crosses, derives the parsing relation being observed platform gray feature with remaining the time that crosses;

2nd step: judge to be observed platform signal to noise ratio of gray feature in observation platform detection system, when being observed platform the signal to noise ratio of gray feature being less than or equal to 20 in observation platform detection system, performs the 3rd step in step (2); When being observed platform the signal to noise ratio of gray feature being more than 20 in observation platform detection system, perform the 4th step in step (2);

3rd step: according to the analytical relation being derived by the 1st step in step (2), selecting initial proportion coefficient, remaining the time that crosses, remain the time-derivative that crosses is quantity of state, being observed platform gray feature is observed quantity, sets up based on the Kalman filter being observed platform gray feature;

4th step: be observed platform gray feature and the parsing relation remaining the time that crosses according to what the 1st step in step (2) was set up, the selection difference observation moment is observed the gray feature of platform, sets up residue and crosses the time and the calculation method of the ratio characteristic being observed platform gray feature;

5th step: observation platform and the moment that crosses being observed platform are forecast, output residue crosses the time and crosses moment predicted value;

(3) set up based on the moment forecasting procedure that crosses being observed platform geometric area feature or geometries characteristic, comprise the steps:

1st step: set up the analytical relation in the moment that crosses being observed platform geometric area feature or geometries characteristic, derives and is observed platform geometric area feature or geometries characteristic and residue crosses the parsing relation of time;

2nd step: judge to be observed platform signal to noise ratio of geometric area feature or geometries characteristic in observation platform detection system, when being observed platform the signal to noise ratio of geometric area feature or geometries characteristic being less than or equal to 20 in observation platform detection system, perform the 3rd step in step (3); When being observed platform the signal to noise ratio of geometric area feature or geometries characteristic being more than 20 in observation platform detection system, perform the 4th step in step (3);

3rd step: according to the analytical relation being derived by the 1st step in step (3), selection is observed platform geometric area feature or geometries characteristic, residue crosses the time, residue crosses, and time-derivative is quantity of state, it is observed platform geometric area feature or geometries characteristic is observed quantity, set up the Kalman filter being observed platform geometric area feature or geometries characteristic;

4th step: according to the 1st step in step (3) set up be observed platform geometric area feature or geometries characteristic and residue crosses the parsing relation of time, the selection difference observation moment is observed geometric area feature or the geometries characteristic of platform, sets up residue and crosses the time and the calculation method of the ratio characteristic being observed platform geometric area feature or geometries characteristic;

5th step: observation platform and the moment that crosses being observed platform are forecast, output residue crosses the time and crosses moment predicted value.

2. the moment forecasting procedure that crosses of based target feature according to claim 1, it is characterised in that: in described step (2), the 1st step is observed platform gray feature and the analytical relation in the moment that crosses is:

$g_{\mathrm{Tar}}\left(T\right)=\frac{K\·I_{\mathrm{Tar}}}{R{\left(T\right)}^2}=\frac{K\·I_{\mathrm{Tar}}}{{(v\·T)}^2}=\frac{K\·I_{\mathrm{Tar}}}{v^2}\·\frac{1}{T^2}$

In formula: T represents that residue crosses the time, unit is s; g_Tar(T) gray feature of platform it is observed for the T moment; R is the distance being observed platform between observation platform optical imaging system, and unit is m; K is observation platform detection system conversion coefficient; I_TarFor being observed platform radiant intensity; V is observation platform and is observed between platform closing speed, and unit is m/s;For observation platform detection system conversion coefficient K, it is observed platform radiant intensity I_TarAnd observation platform and be observed between platform the ratio of closing speed v square, estimates according to observation platform result of detection.

3. the moment forecasting procedure that crosses of based target feature according to claim 1, it is characterised in that: in described step (3), the 1st step is observed the cross analytical relation of time of platform geometric area feature and residue and is:

$S_{\mathrm{sen}}\left(T\right)=S_{\mathrm{Obj}}{(\frac{f}{v}\·\frac{1}{S_{\mathrm{Pix}}})}^2\·\frac{1}{T^2}$

In formula: S_sen(T) being the geometric area feature being observed platform the T moment in observed image, unit is pixel; S_ObjBeing the geometric area feature being observed platform, unit is pixel; F is optical imaging system focal length, and unit is m; V is observation platform and is observed between platform closing speed, and unit is m/s; S_PixBeing observation platform detection system Pixel Dimensions, unit is m; T is that residue crosses the time, and unit is s;

In described step (3), the 1st step is observed the cross analytical relation of time of platform geometries characteristic and residue and is:

$L_{\mathrm{sen}}\left(T\right)=L_{\mathrm{Obj}}\frac{f}{v}\·\frac{1}{S_{\mathrm{Pix}}}\·\frac{1}{T}$

In formula: L_sen(T) being be observed platform geometries characteristic in observed image the T moment, unit is pixel; L_ObjBeing the geometries characteristic being observed platform, unit is pixel; F is optical imaging system focal length, and unit is m; V is observation platform and is observed between platform closing speed, and unit is m/s; S_PixBeing observation platform detection system Pixel Dimensions, unit is m; T is that residue crosses the time, and unit is s.

4. the moment forecasting procedure that crosses of the based target feature according to Claims 2 or 3, it is characterised in that: in described step (2), in the 3rd step and step (3), in the 3rd step, Kalman filter relational expression is:

$X(k+1)=\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& \mathrm{\Δt}\\ 0& 0& 1\end{array}\right)X\left(k\right)=\mathrm{\ΦX}\left(k\right)$

In formula: X (k+1) is the quantity of state X value in the k+1 moment; X (k) is the quantity of state X value in the k moment; �� t is interval, and unit is s; �� is state-transition matrix.

5. the moment forecasting procedure that crosses of based target feature according to claim 4, it is characterised in that: the moment forecast iterative process that crosses of Kalman filter is:

A. predictive equation is calculated:

${\hat{X}}_{K+1/K}=\mathrm{\Φ}{X}_k$

In formula:For state a step of forecasting value; �� is state-transition matrix; X_kFor the vector X value in the k moment;

B. computation and measurement method is generalized into the Jacobi matrix at future position:

$H_{K+1}=\frac{\∂G\left[X\right]}{\∂X}{|}_{X={\hat{X}}_{k+1/K}}$

In formula: H_k+1Jacobi matrix for k+1; The right-hand member nonlinear function vector that G [��] is systematic observation equation group; X is state variable;For state a step of forecasting value;

C. prediction covariance and Kalman gain are calculated:

$P_{K+1/K}={\mathrm{\Φ}}_{K+1/K}{P}_{K/K}{\mathrm{\Φ}}_{K+1/K}^T$

$K_{K+1}=P_{K+1/K}{H}_{k+1}^T{(H_{k+1}{P}_{K+1/K}{H}_{k+1}^T+R_{k+1})}^{-1}$

In formula: P_K+1/KFor a step of forecasting variance matrix; ��_K+1/KFor state-transition matrix; P_K/KFor filtering error variance matrix;For matrix ��_K+1/KTransposed matrix; K_K+1For filtering gain matrix; H_k+1Jacobi matrix for k+1;For matrix H_k+1Transposed matrix; R_K+1For systematic observation noise variance matrix;

D. filtering estimated value is obtained:

${\hat{X}}_{K+1/K+1}={\hat{X}}_{K+1/K}+K_{K+1}[G_{K+1}-G\left({\hat{X}}_{K+1/K}\right)]$

In formula:It it is the vector X value according to the k+1 moment estimated value to k+1 moment value;For state a step of forecasting value; K_K+1For filtering gain matrix; G_K+1Observation vector for the k+1 moment;For according to state a step of forecasting valueThe forecast vector calculated;

If E. adopting Modified covariance extended Kalman filter device, reusing future position data and recalculating Jacobi matrix:

$H_{k+1}^{+}=\frac{\∂G\left[X\right]}{\∂X}{|}_{X={\hat{X}}_{k+1/K+1}}$

In formula:The Jacobi matrix of the k+1 for recalculating by future position data; The right-hand member nonlinear function vector that G [��] is systematic observation equation group; X is state variable;It it is the estimation to k+1 moment value of the vector X value according to the k+1 moment;

F. prediction covariance and Kalman gain are recalculated:

$K_{K+1}^{+}=P_{K+1/K}{H}_{k+1}^{+T}{(H_{k+1}{P}_{K+1/K}{H}_{k+1}^T+R_{k+1})}^{-1}$

$P_{K+1/K+1}=(I-K_{K+1}^{+}{H}_{K+1}^{+})P_{K+1/K}{(I-K_{k+1}^{+}{H}_{K+1}^{+})}^T+K_{K+1}^{+}{R}_{K+1}{K}_{K+1}^{T+}$

In formula:For the filtering gain matrix recalculated by future position data;For matrixTransposed matrix; P_K+1/KFor a step of forecasting variance matrix; P_K+1/K+1For; H_k+1Jacobi matrix for k+1;The Jacobi matrix of the k+1 for recalculating by future position data;For sensing amountTransposed matrix; R_K+1For systematic observation noise variance matrix; I is unit vector matrix; T is that residue crosses the time, unit s;

G. can utilize after being predicted the outcome residue cross time T forecast estimate filter value;

Repeating step A to step F, to crossing, the moment forecasts, obtains residue and crosses the time and cross moment predicted value.

6. the moment forecasting procedure that crosses of based target feature according to claim 4, it is characterized in that: set up in process in Kalman filter, remain the time-derivative that crosses and be set to fixed value-1, remain the value that the standard variance of time-derivative that crosses is set between 0.01��0.00001.

7. the moment forecasting procedure that crosses of based target feature according to claim 2, it is characterised in that: in described step (2), the cross calculation method relational expression of time and the ratio characteristic that is observed platform gray feature of the 4th step residue is:

$T=\frac{\mathrm{\Δt}}{1-{\left(\frac{g_{\mathrm{Tar}}\left(T\right)}{g_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}^{1/2}}$

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; g_Tar(T) it is the gray value T moment being observed platform, accordingly, g_Tar(T-�� t) is the gray value that T-�� t is observed platform.

8. the moment forecasting procedure that crosses of based target feature according to claim 3, it is characterised in that: in described step (3), the cross calculation method relational expression of time and the ratio characteristic that is observed platform geometric area feature of the 4th step residue is:

$T=\frac{\mathrm{\Δt}}{1-{\left(\frac{S_{\mathrm{Tar}}\left(T\right)}{S_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}^{1/2}}$

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; S_Tar(T) being be observed the two dimensional image geometric area feature that platform presents in observation platform detection system the T moment, unit is pixel; Accordingly, S_Tar(T-�� t) is the two dimensional image geometric area feature that T-�� t is observed that platform presents in observation platform detection system, and unit is pixel;

In described step (3), the cross resolving relational expression of time and the ratio characteristic that is observed platform geometries characteristic of the 4th step residue is:

$T=\frac{\mathrm{\Δt}}{1-\left(\frac{L_{\mathrm{Tar}}\left(T\right)}{L_{\mathrm{Tar}}(T-\mathrm{\Δt})}\right)}$

In formula: T is that residue crosses the time, and unit is s; �� t is interval, and unit is s; L_Tar(T) being be observed the one dimensional image geometries characteristic that platform presents in observation platform detection system the T moment, unit is pixel; Accordingly, L_Tar(T-�� t) is the one dimensional image geometries characteristic that T-�� t is observed that platform presents in observation platform detection system, and unit is pixel.

9. the moment forecasting procedure that crosses of based target feature according to claim 1, it is characterised in that: described Kalman filter is Modified covariance extended Kalman filter device.