CN102928836A - Ground target tracking method - Google Patents

Abstract

The invention discloses a ground target tracking method which is put forward on the basis of road information auxiliary, specifically defines parameters of road information, and puts forward three judgment conditions, namely a judgment condition of a target leaving road section, a judgment condition of a target at the position of a point and a judgment condition of the target at the position of a certain road section. An interactive multi-model state model is timely updated according to the judgment conditions, the road information, forecasting states of the target and estimation states of the target, namely model quantity, model structures and the parameters in the model are updated. Compared with the existing scheme, the ground target tracking method uses the road information and improves tracking performance of the ground target.

Description

A kind of Ground Target Tracking method

Technical field

The invention belongs to the Radar Technology field, be specifically related to a kind of Ground Target Tracking method.

Background technology

Target following is the main research field of modern radar, and terrain object is its main research object, and terrain object is difficult because a large amount of False Intersection Points marks and strong maneuverability cause it to follow the tracks of.The method for tracking target of traditional simple based on motion mathematic(al) parameter, interacting multiple model algorithm (Interacting Multiple Model for example, IMM) etc., be faced with stern challenge: the 1) mismatch of trace model causes the tracking performance of maneuvering target sharply to descend; 2) traditional IMM algorithm model is fixed, and requires in the motion model of maneuvering target is included in, and this can increase calculated amount undoubtedly, reduces the degree of accuracy of Target state estimator.But most of terrain object, normally along the track or river operation, road limits the position of target, and utilize the motion state that road information can target of prediction, such as a straight road, target is uniformly accelerated motion or uniform motion, can not be turning motion; The approach way of road is known, and namely the ensuing motion model of target is known.The speed of terrain object and state model etc. all are subjected to the constraint of road, so effectively utilize road information can improve the tracking performance of terrain object.

Summary of the invention

The objective of the invention is to have proposed a kind of Ground Target Tracking method in order to solve the problems referred to above of traditional method for tracking target existence.

Technical scheme of the present invention is: a kind of Ground Target Tracking method comprises the steps:

S1. the initialization setting of parameter is finished in initialization, specifically comprises step by step following:

S11. according to the road scene of reality, the road information parameter is set, described road information parameter comprises road slice parameter and node parameter, and described road slice parameter comprises: the motion model of road segment number, road fragment, road fragment start node; Described node parameter comprises: the variation of the motion state model of the motion model of the numbering in node serial number, adjacent highway section and number, Nodes, Nodes, adjacent road fragment and current road;

S12. make variable sign_road=0, sign_point=0; Variable sign_road represents whether target leaves current highway section, and sign_road=1 represents that target leaves current highway section, and sign_road=0 represents that target is on certain highway section; Whether variable sign_point represents target at the adjacent node place, and sign_point=1 represents target at the adjacent node place, and sign_point=0 represents that target is not or not the adjacent node place;

S2. read k metric data constantly, be designated as { z _i(k), i=1,2...n}; Read the flight path data, be designated as { t _j, j=1,2...p}; Wherein, n refers to the measurement number that k is constantly total, and variable p refers to the flight path number that k is constantly total;

S3. target of prediction state;

If sign_road=0 S4., then according to the dbjective state of predicting among the step S3, and target leaves the decision condition in highway section, judges whether target leaves current road; Do not leave current road if judge target, it is related then the measurement of reading among the step S2 and flight path to be made data, obtains the relating dot mark of each flight path; Leave current road if judge target, then make sign_road=1, and according to the road information described in the step S11, upgrade the predicted state model, namely select the model of the Nodes adjacent with target, again the target of prediction state, do the data association, obtain the relating dot mark of each flight path; If sign_road=1 then directly does the data association, obtain the relating dot mark of each flight path;

S5. according to the relating dot mark that obtains, make tracking filter, obtain the estimated state of target;

S6. judge the sign_road value, if sign_road=0 then continues execution in step S2; If sign_road=1 then according to the decision condition of target at Nodes, judges that whether target is at Nodes; If target at Nodes, then makes sign_point=1, otherwise makes sign_point=0;

If S7. sign_point=0 then according to the decision condition of target in certain highway section, judges whether target enters the adjacent road fragment; If judge that target does not enter the adjacent road fragment, then continue execution in step S2; Enter the adjacent road fragment if judge target, then make sign_road=0, and according to the road information described in the S11 step by step, upgrade the filter state model, namely select the model of adjacent road fragment, continue execution in step S2.

Further, it is as follows that the target described in the step S4 is left the decision condition in highway section:

The coordinate of supposing the node j that k is constantly adjacent with target is (x _j, y _j), if the predicted position state of target and node j coordinate satisfy formula (1) and formula (2), judge that then target leaves current road.

$x_j-\hat{x}(k-1)\≤v_{\max }T+2\sqrt{p_{\mathrm{xx}}(k-1)}---\left(1\right)$

$y_j-\hat{y}(k-1)\≤v_{\max }T+2\sqrt{p_{\mathrm{yy}}(k-1)}---\left(2\right)$

Wherein, p _Xx(k-1) and p _Yy(k-1) be location status error covariance value among the state error covariance matrix P (k-1),

Be respectively target and estimate v at k-1 location status constantly _MaxBe the maximal rate of target, T is the sampling period.

Further, the predicted state model described in the step S4 is the predicted state model in the interactive multi-model.

Further, the target described in the step S6 is as follows at the decision condition of Nodes:

Centered by target k predicted value constantly, set up the elliptical wave door, if adjacent node j(x _j, y _j) in this elliptical wave door, namely satisfy formula (3), then judge target at node j place,

${\left[\begin{array}{c}{x}_j-\hat{x}\left(k\right)\\ y_j-\hat{y}\left(k\right)\end{array}\right]}^{\′}{\left[\begin{array}{cc}{p}_{\mathrm{xx}}\left(k\right)& p_{\mathrm{xy}}\left(k\right)\\ p_{\mathrm{xy}}\left(k\right)& p_{\mathrm{yy}}\left(k\right)\end{array}\right]}^{-1}\left[\begin{array}{c}{x}_j-\hat{x}\left(k\right)\\ y_j-\hat{y}\left(k\right)\end{array}\right]\≤\mathrm{\α}---\left(3\right)$

Wherein, α is predefined threshold value,

Be respectively target and estimate at k location status constantly, $P_{\mathrm{pos}}\left(k\right)=\left[\begin{array}{cc}{p}_{\mathrm{xx}}\left(k\right)& p_{\mathrm{xy}}\left(k\right)\\ p_{\mathrm{xy}}\left(k\right)& p_{\mathrm{yy}}\left(k\right)\end{array}\right]$ The position submatrix of the state error covariance matrix P (k) of model probability maximum, the transposition computing of [] ' representing matrix, [] ^-1The inverse operation of representing matrix.

Further, the target described in the step S7 is as follows at the decision condition in certain highway section:

Target is at the decision condition in certain highway section, and whether the state estimation of namely judging target on certain highway section, if there is any point (x, y) to satisfy (4) formula on certain highway section, judges that then target is on this highway section;

${\left[\begin{array}{c}x-\hat{x}\left(k\right)\\ y-\hat{y}\left(k\right)\end{array}\right]}^{\′}{P_{\mathrm{pos}}}^{-1}\left(k\right)\left[\begin{array}{c}x-\hat{x}\left(k\right)\\ y-\hat{y}\end{array}\right]\≤\mathrm{\α}---\left(4\right)$

Wherein, P _Pos ^-1(k) be matrix P _Pos(k) contrary.

Further, the filter state model described in the step S7 is the filter state model in the interactive multi-model.

Beneficial effect of the present invention: method of the present invention proposes on the basis of assisting based on road information, specific definition the parameter of road information, three decision conditions have been proposed, it is the decision condition that target is left the highway section, target is at the decision condition of Nodes, target is at the decision condition in certain highway section, and the state model according to predicted state and the estimated state of decision condition, road information and target comes the real-time update interactive multi-model namely upgrades its pattern number, model structure and Model Parameter.Compare with existing scheme, method of the present invention has been utilized road information, has improved the tracking performance of terrain object.

Description of drawings

Fig. 1 is the mileage chart of the embodiment of the invention.

Fig. 2 is road and the target travel scene simulation synoptic diagram of the embodiment of the invention.

Fig. 3 is the schematic flow sheet of the inventive method.

Fig. 4 is that clutter density is 0.01km ²During/s, the position square error curve map of method of the present invention and existing IMM algorithm.

Fig. 5 is that clutter density is 0.1km ²During/s, the position square error curve map of method of the present invention and existing IMM algorithm.

Fig. 6 is that clutter density is 0.5km ²During/s, the position square error curve map of method of the present invention and existing IMM algorithm.

Embodiment

Below in conjunction with accompanying drawing, provide specific embodiments of the invention.Need to prove: the parameter among the embodiment does not affect generality of the present invention.

The speed of terrain object and state model etc. are subjected to the constraint of landform, road, can improve the tracking performance of terrain object so utilize these information, based on this, the present invention proposes a kind of tracking that is used for realizing terrain object based on the auxiliary change dimension interactive model method (RVS-IMM) of road information, the method hypothetically Area Objects operates in the centre of road always, below Ground Target Tracking method of the present invention is designated as: RVS-IMM.

In order to verify the performance of the inventive method, method of the present invention is followed the tracks of emulation under different clutter density conditions, and compare with traditional IMM algorithm.Traditional algorithm IMM adopts MODEL C V model, CT ⁺Model and CT ^-Model, angular velocity is respectively 0.01rod/s, 0.125rod/s ,-0.04rod/s.

The road scene of emulation as shown in Figure 1, the road information parameter of Fig. 1 is as shown in table 1, Fig. 3 has provided the schematic flow sheet of the inventive method.

Table 1

The direction of the parameter obey section of showing the way in the table 1 definition, highway section AB for example, obey=1 refers to the direction from A to B, obey=0 refers to the direction from B to A, model M ₁Represent uniform motion MODEL C V, M ₂Represent uniformly accelerated motion MODEL C A, M ₃Represent the motion model CT that at the uniform velocity turns left ⁺, M ₄Represent the motion model CT that at the uniform velocity turns right ^-Need to prove,

Implication be that the angle of adjacent road fragment and current road changes, be clockwise if current road runs to the adjacent road fragment, then $-\mathrm{\&pi;}<\&PartialD;<0;$ If counterclockwise, then $0<\&PartialD;<\mathrm{\&pi;}.$

Road and target travel scene simulation are seen Fig. 2.The coordinate of road circuit node is respectively among the figure: A (1000,2000), B (3000,2000), C (4000,2000), D (5500,2000), E (5500,2500), F (6000,2000) (unit is m).Suppose to have at road the target of three motions, three targets operate in respectively on highway section AB and the BC, and on highway section EF and the GF, and on highway section EF and the FH, their original state is respectively:

System noise and observation noise are the white Gaussian noise of zero-mean, and the measurement noise standard deviation of all directions is identical, and σ=10m, sweep spacing T=5s, the emulation step number is 40, detection probability PD=0.8111, thresholding probability P G=0.9997, the investigative range of radar is 7000 * 4000 (m ²), clutter is evenly distributed in the search coverage of radar at random, and clutter density λ is respectively 0.01/km ²S ^-1, 0.1/km ²S ^-1, 0.5/km ²S ^-1, Monte Carle simulation times is 200 times.

Simulation result shows is compared the IMM algorithm, and the method that the present invention proposes is Tracking Ground Targets more reliablely and stablely.It is 0.01km that Fig. 4 ~ 6 are respectively clutter density ²/ s, 0.1km ²/ s, 0.5km ²In/s the situation, two kinds of algorithms are to the directions X of the tracking of three targets and the position square error (RMSE) of Y-direction.From Fig. 4 ~ 6, can see, method RVS-IMM in the position root-mean-square error of X, Y-direction under the different clutter environments nearly all in scope [0,10m].Compare the X of two kinds of algorithms, the position root-mean-square error of Y-direction, the tracking accuracy of the inventive method all will be far superior to traditional IMM algorithm under three kinds of clutter environments, and tracking performance is all good and stable.

Those of ordinary skill in the art will appreciate that embodiment described here is in order to help reader understanding's principle of the present invention, should to be understood to that the protection domain of inventing is not limited to such special statement and embodiment.Everyly make various possible being equal to according to foregoing description and replace or change, all be considered to belong to the protection domain of claim of the present invention.

Claims (6)

1. a Ground Target Tracking method comprises the steps:

S1. the initialization setting of parameter is finished in initialization, specifically comprises step by step following:

S11. according to the road scene of reality, the road information parameter is set, described road information parameter comprises road slice parameter and node parameter, and described road slice parameter comprises: the motion model of road segment number, road fragment, road fragment start node; Described node parameter comprises: the variation of the motion state model of the motion model of the numbering in node serial number, adjacent highway section and number, Nodes, Nodes, adjacent road fragment and current road;

S12. make variable sign_road=0, sign_point=0; Variable sign_road represents whether target leaves current highway section, and sign_road=1 represents that target leaves current highway section, and sign_road=0 represents that target is on certain highway section; Whether variable sign_point represents target at the adjacent node place, and sign_point=1 represents target at the adjacent node place, and sign_point=0 represents that target is not or not the adjacent node place;

S2. read k metric data constantly, be designated as { z _i(k), i=1,2...n}; Read the flight path data, be designated as { t _j, j=1,2...p}; Wherein, n represents the measurement number that k is constantly total, and variable p represents the flight path number that k is constantly total;

S3. target of prediction state;

If sign_road=0 S4., then according to the dbjective state of predicting among the step S3, and target leaves the decision condition in highway section, judges whether target leaves current road; Do not leave current road if judge target, it is related then the measurement of reading among the step S2 and flight path to be made data, obtains the relating dot mark of each flight path; Leave current road if judge target, then make sign_road=1, and according to the road information described in the step S11, upgrade the predicted state model, namely select the model of the Nodes adjacent with target, again the target of prediction state, do the data association, obtain the relating dot mark of each flight path; If sign_road=1 then directly does the data association, obtain the relating dot mark of each flight path;

S5. according to the relating dot mark that obtains, make tracking filter, obtain the estimated state of target;

S6. judge the sign_road value, if sign_road=0 then continues execution in step S2; If sign_road=1 then according to the decision condition of target at Nodes, judges that whether target is at Nodes; If target at Nodes, then makes sign_point=1, otherwise makes sign_point=0;

If S7. sign_point=0 then according to the decision condition of target in certain highway section, judges whether target enters the adjacent road fragment; If judge that target does not enter the adjacent road fragment, then continue execution in step S2; Enter the adjacent road fragment if judge target, then make sign_road=0, and according to the road information described in the S11 step by step, upgrade the filter state model, namely select the model of adjacent road fragment, continue execution in step S2.

2. Ground Target Tracking method according to claim 1 is characterized in that, it is as follows that the target described in the step S4 is left the decision condition in highway section:

The coordinate of supposing the node j that k is constantly adjacent with target is (x _j, y _j), if the predicted position state of target and node j coordinate satisfy formula (1) and formula (2), judge that then target leaves current road.

$x_j-\hat{x}(k-1)\&le;v_{\max }T+2\sqrt{p_{\mathrm{xx}}(k-1)}---\left(1\right)$

$y_j-\hat{y}(k-1)\&le;v_{\max }T+2\sqrt{p_{\mathrm{yy}}(k-1)}---\left(2\right)$

Wherein, p _Xx(k-1) and p _Yy(k-1) be location status error covariance value among the state error covariance matrix P (k-1),

Be respectively target and estimate v at k-1 location status constantly _MaxBe the maximal rate of target, T is the sampling period.

3. Ground Target Tracking method according to claim 1 and 2 is characterized in that, the target described in the step S6 is as follows at the decision condition of Nodes:

Centered by target k predicted value constantly, set up the elliptical wave door, if adjacent node j(xi, _Yj) in this elliptical wave door, namely satisfy formula (3), judge that then target is at node j place.

${\left[\begin{array}{c}{x}_j-\hat{x}\left(k\right)\\ y_j-\hat{y}\left(k\right)\end{array}\right]}^{\&prime;}{\left[\begin{array}{cc}{p}_{\mathrm{xx}}\left(k\right)& p_{\mathrm{xy}}\left(k\right)\\ p_{\mathrm{xy}}\left(k\right)& p_{\mathrm{yy}}\left(k\right)\end{array}\right]}^{-1}\left[\begin{array}{c}{x}_j-\hat{x}\left(k\right)\\ y_j-\hat{y}\left(k\right)\end{array}\right]\&le;\mathrm{\&alpha;}---\left(3\right)$

Wherein, α is predefined threshold value,

Be respectively target and estimate at k location status constantly, $P_{\mathrm{pos}}\left(k\right)=\left[\begin{array}{cc}{p}_{\mathrm{xx}}\left(k\right)& p_{\mathrm{xy}}\left(k\right)\\ p_{\mathrm{xy}}\left(k\right)& p_{\mathrm{yy}}\left(k\right)\end{array}\right]$ The position submatrix of the state error covariance matrix P (k) of model probability maximum, the transposition computing of [] ' representing matrix, [] ^-1The inverse operation of representing matrix.

4. Ground Target Tracking method according to claim 3 is characterized in that, the target described in the step S7 is as follows at the decision condition in certain highway section:

Target is at the decision condition in certain highway section, and whether the state estimation of namely judging target on certain highway section, if there is any point (x, y) to satisfy (4) formula on certain highway section, judges that then target is on this highway section;

${\left[\begin{array}{c}x-\hat{x}\left(k\right)\\ y-\hat{y}\left(k\right)\end{array}\right]}^{\&prime;}{P_{\mathrm{pos}}}^{-1}\left(k\right)\left[\begin{array}{c}x-\hat{x}\left(k\right)\\ y-\hat{y}\end{array}\right]\&le;\mathrm{\&alpha;}---\left(4\right)$

Wherein, P _Pos ^-1(k) be matrix P _Pos(k) contrary.

5. according to claim 1 to the described Ground Target Tracking method of 4 arbitrary claims, it is characterized in that the predicted state model described in the step S4 is the predicted state model in the interactive multi-model.

6. according to claim 1 to the described Ground Target Tracking method of 4 arbitrary claims, it is characterized in that the filter state model described in the step S7 is the filter state model in the interactive multi-model.

Images