An interactive multi-model tracking method based on ellipse and line constraints

Abstract

The present invention belongs to the technical field of target constraint tracking and discloses an interactive multi-model tracking method based on elliptical constraints and linear constraints. The method includes generating target motion data and generating measurement data through sensors; analyzing the constraints satisfied by the motion of elliptical trajectory targets and linear motion targets; establishing a tracking model; applying an interactive multi-model algorithm based on elliptical constraints and linear constraints; and obtaining observation results. Based on the characteristics of elliptical trajectory motion and linear trajectory motion, the present invention designs an unscented Kalman filter model for elliptical motion and an unscented Kalman filter model for linear motion. By introducing elliptical constraints and linear constraints into the interactive multi-model algorithm filtering, the tracking accuracy of targets with linear and elliptical trajectory motion characteristics is improved. In addition, the present invention establishes a powerful framework for introducing constraints into the interactive multi-model algorithm, making the model selection for different target motion states more targeted.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of target constraint tracking, and in particular to an interactive multi-model tracking method based on ellipse constraints and straight line constraints. Background Art

[0002] The interactive multi-model algorithm is an adaptive algorithm with Markov chain transition probabilities. This algorithm assumes that target motion follows a Markov chain with finite known transition probabilities between different models. It then considers the interaction of multiple models to estimate the target's state. In real-world target motion, targets often adhere to specific constraints. For example, vehicle trajectories on roads follow the shape of the road, satellites adhere to satellite orbital constraints, and specific jammers have runway or horseshoe-shaped flight paths. Incorporating these constraints into the interactive multi-model algorithm can improve target tracking accuracy.

[0003] Interactive multi-model algorithms typically involve multiple filters processed in parallel, each corresponding to a different state space that describes different target motion patterns. Incorporating prior constraint information into the filtering process can improve target tracking accuracy. Existing interactive multi-model algorithms have extensively studied various types of target maneuvers, but have failed to incorporate prior constraint information into these algorithms.

[0004] Currently, existing constraint information is primarily linear, introducing constraints as pseudo-measurements into tracking systems to improve target accuracy. However, linear constraints are overly idealistic and rarely used in practical applications. Nonlinear constraints are difficult to accurately represent mathematically, and can also lead to filter divergence due to strong nonlinearity.

[0005] In existing interactive multi-model algorithms, different filters process the same target dimension. Actual target motion may involve multiple models, each with different dimensions. A more comprehensive and effective solution for filtering different target state dimensions and weighted fusion to obtain a final estimate has yet to be proposed. Summary of the Invention

[0006] The purpose of the present invention is to provide an interactive multi-model tracking method based on elliptical constraints and straight line constraints. In view of the current lack of trajectory constraints to introduce interactive multi-models, the present invention is designed for scenarios where the target may perform elliptical and straight line motions, and introduces elliptical trajectory constraints and straight line trajectory constraints as pseudo-measurements into the filtering process of the interactive multi-model to improve the targeted selection of models for different motion states and the accuracy of target tracking.

[0007] To achieve the above object, the present invention provides an interactive multi-model tracking method based on ellipse constraints and line constraints, comprising the following steps:

[0008] Step 1: Generate motion data of elliptical trajectory targets and straight trajectory targets, and generate measurement data through sensors;

[0009] Step 2: Analyze the constraints satisfied by the elliptical trajectory target motion and obtain the elliptical trajectory constraint information;

[0010] Step 3: Introduce the elliptical trajectory constraint information into the tracking system through the augmented measurement equation and establish a tracking model, which includes the following steps:

[0011] Step 3.1: Use the constant speed turning motion model to describe the target moving along the elliptical trajectory. The target state vector expression under the constant speed turning motion model is: ;

[0012] The target state equation is , is the state transfer equation, is the noise distribution equation;

[0013] Step 3.2: Use the sensor to observe. The sensor can observe the distance and azimuth of the target, and the sensor measurement data is ;

[0014] Step 3.3: By expanding the state dimension, the elliptical trajectory constraint information is introduced into the sensor measurement data expression in step 3.2. The target state vector is expanded to , the target measurement equation is expanded to ;

[0015] Step 3.4: For targets subject to elliptical trajectory constraints, the noise mean corresponding to the corresponding constraints is zero. Assume that the target states after dimension expansion are uncorrelated. Diagonally augment the traditional noise covariance matrix. The augmented noise covariance matrix is , R k is the covariance matrix of observation distance noise and observation angle noise, O is a zero matrix, n is the number of constraints, is an n*n zero matrix;

[0016] Step 4: Analyze the constraints satisfied by the linear trajectory target motion and obtain the linear trajectory constraint information;

[0017] Step 5: Introduce the linear trajectory constraint information into the tracking system through the augmented measurement equation and establish a tracking model, which includes the following steps:

[0018] Step 5.1: Use the uniform motion model to describe the target moving along a straight line. The target state vector expression under the uniform motion model is: ;

[0019] Step 5.2: Expand the dimension of the target state vector expression. The target state vector is expanded to , the target measurement equation is expanded to ;

[0020] Step 5.3: Assume that the target states after dimension expansion are uncorrelated, and the noise covariance matrix is ​​diagonally augmented in the traditional noise covariance matrix. The augmented noise covariance matrix is , R k is the covariance matrix of observation distance noise and observation angle noise, O is a zero matrix, n is the number of constraints, is an n*n zero matrix;

[0021] Step 6: Apply an interactive multi-model algorithm based on elliptical and linear constraints; the interactive multi-model algorithm includes interaction, filtering, model probability update, and state estimation fusion;

[0022] Step 7: Obtain observation results.

[0023] Furthermore, in step 2, the constraints satisfied by the elliptical trajectory target motion are analyzed to obtain elliptical trajectory constraint information, including the following steps:

[0024] Step 2.1: Suppose a target moves in an elliptical trajectory in a two-dimensional plane, and the coordinates of the center of the ellipse are The semi-major axis and semi-minor axis of the ellipse are a and b respectively. The angle between the line connecting any point on the ellipse and the circle and the x-axis is , the standard equation of the ellipse is , the parametric equation of the ellipse is ;

[0025] Step 2.2: Derivative the standard equation and parametric equation of the ellipse respectively to obtain the position constraint expression of the ellipse trajectory: , the velocity constraint expression of the elliptical trajectory is , Indicates the angular velocity of the target moving along the elliptical trajectory;

[0026] Step 2.3: Arrange the expressions in step 2.1 and step 2.2 to obtain the elliptical trajectory constraint information. The three expressions contained in the elliptical trajectory constraint information are: .

[0027] Furthermore, in step 4, the constraints satisfied by the linear trajectory target motion are analyzed to obtain linear trajectory constraint information, which includes the following steps:

[0028] Step 4.1: A target moving along a straight line in a two-dimensional plane can be represented by a slope and an intercept. The general equation of a straight line is ;

[0029] Step 4.2: Derivative the general equation of the line, we can get the velocity constraint equation for any point: ;

[0030] Step 4.3: Arrange the equations in step 4.1 and step 4.2 to obtain the straight line trajectory constraint information. The two formulas included in the straight line trajectory constraint information are: .

[0031] Furthermore, it is characterized in that is the state estimate of filter j at time k-1, is the state covariance transfer matrix, is the probability of model j at time k-1;

[0032] The state estimates of the N filters after interaction are ;

[0033] The state covariance transfer matrix of N filters after interaction is .

[0034] Furthermore, the elliptical trajectory tracking model and the straight line estimation tracking model are independently subjected to unscented Kalman filtering processing, and the unscented Kalman filtering processing includes the following steps:

[0035] Step 6.1: Calculate 2nx+1 sampling points first and the corresponding weights , sampling point The expression is , the corresponding weight The expression is ;

[0036] Step 6.2: Use the state equation to obtain the one-step prediction and weight of the sampling point, and then get the state prediction estimate and state prediction covariance. The expression of the state prediction estimate is: , the expression of state prediction covariance is ;

[0037] Step 6.3: The update process is:

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] Where n represents the state vector Dimension, j is calculated as ;

[0047] Representation matrix The j-th row or j-th column of , and the calculation formula related to weight is ,in , and These are all empirical parameters of the sigma point in the unscented Kalman filter.

[0048] Furthermore, the model probability update includes the following steps:

[0049] Step 6.4, where the filter residual of model j is and covariance The expression is ;

[0050] Step 6.5: According to step 6.4, the updated probability of model j is .

[0051] Furthermore, the state estimation fusion step is to weight the state estimates of all models according to their latest model probabilities to form a comprehensive state estimate.

[0052] Beneficial effects: The present invention provides an interactive multi-model tracking method based on elliptical constraints and straight line constraints. According to the characteristics of elliptical trajectory motion and straight line trajectory motion, an unscented Kalman filter model of elliptical motion and an unscented Kalman filter model of straight line motion are designed. By introducing elliptical constraints and straight line constraints into the interactive multi-model algorithm filtering, the tracking accuracy of targets with straight line and elliptical trajectory motion characteristics is improved; in addition, the present invention establishes a powerful framework for introducing constraints into the interactive multi-model algorithm, and the model selection for different target motion states is more targeted. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of an interactive multi-model tracking method based on ellipse constraints and straight line constraints according to an embodiment of the present invention;

[0054] Figure 2 The target motion trajectory and tracking trajectory diagram involved in the embodiment of the present invention;

[0055] Figure 3 is a graph showing a root mean square error (RMS) curve of target tracking position involved in an embodiment of the present invention;

[0056] Figure 4 is a target tracking speed root mean square error curve diagram involved in an embodiment of the present invention;

[0057] Figure 5 is a model probability graph of an interactive multi-model with constraints according to an embodiment of the present invention;

[0058] Figure 6 It is a conventional interactive multi-model model probability graph involved in the embodiment of the present invention. DETAILED DESCRIPTION

[0059] The preferred structure and implementation method of the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0060] like Figures 1 to 6 As shown, an embodiment of the present invention discloses an interactive multi-model tracking method based on ellipse constraints and straight line constraints. Figure 1 This is a flowchart of an interactive multi-model tracking method based on ellipse constraints and line constraints according to an embodiment of the present invention.

[0061] In this method, the different conditions for elliptical and linear constraints, as well as the different number of constraint equations, lead to inconsistent dimensionality in the state vectors of the two models after dimensional expansion. To address this, the present invention employs dimensionality reduction before fusion. Specifically, only the common state components in each model are retained for fusion, while the remaining components are estimated using the single-model estimation results as the final estimate. The single-model estimation results are then used to expand the initial value of the mixture before filtering. Example 1

[0062] An interactive multi-model tracking method based on ellipse constraints and line constraints includes the following steps:

[0063] Step 1: Generate motion data of elliptical trajectory targets and straight trajectory targets, and generate measurement data through sensors;

[0064] Step 2: Analyze the constraints satisfied by the elliptical trajectory target motion and obtain the elliptical trajectory constraint information;

[0065] Step 2.1: Suppose a target moves in an elliptical trajectory in a two-dimensional plane, and the coordinates of the center of the ellipse are The semi-major axis and semi-minor axis of the ellipse are a and b respectively. The angle between the line connecting any point on the ellipse and the circle and the x-axis is , the standard equation of the ellipse is (Formula 1), the parametric equation of the ellipse is (Formula 2);

[0066] Step 2.2: Derivative the standard equation and parametric equation of the ellipse respectively to obtain the position constraint expression of the ellipse trajectory: (Equation 3), the velocity constraint expression of the elliptical trajectory is (Formula 4), Indicates the angular velocity of the target moving along the elliptical trajectory;

[0067] Step 2.3: Arrange the expressions in step 2.1 and step 2.2 to obtain the elliptical trajectory constraint information. The three expressions contained in the elliptical trajectory constraint information are: (Equation 5), the ellipse trajectory constraint includes the constraint relationship of position, velocity, angular velocity, ellipse center, and major and minor axes of the ellipse.

[0068] Step 3: Introduce the elliptical trajectory constraint information into the tracking system through the augmented measurement equation to establish a tracking model;

[0069] Step 3.1: Use the constant speed turning motion model (CT model) to describe the target motion along the elliptical trajectory. The target state vector expression under the constant speed turning motion model is: (Equation 6), the state of traditional two-dimensional target tracking is position and velocity. The target moves in an elliptical trajectory, and the state vector of the target increases the angular velocity;

[0070] The target state equation is (Formula 7), is the state transfer equation, is the noise distribution equation.

[0071] Step 3.2: Use sensors to observe. The present invention uses radar. The sensor can observe the distance and azimuth of the target, and the sensor measurement data is obtained as (Equation 8);

[0072] Step 3.3: The pseudo-measurement method introduces the elliptical trajectory constraint information into the sensor measurement data expression in step 3.2 by state dimension expansion. The target state vector is expanded to (Equation 9), the target measurement equation is expanded to (Equation 10);

[0073] Step 3.4: For targets subject to elliptical trajectory constraints, the noise mean corresponding to the corresponding constraints is zero. Assume that the target states after dimension expansion are uncorrelated, and the corresponding noise covariance matrix is ​​diagonally augmented in the traditional noise covariance matrix. The augmented noise covariance matrix is (Equation 11), the augmented noise covariance matrix is ​​a 5*5 diagonal matrix, R k is a 2*2 diagonal matrix, is the covariance matrix of the observation distance noise and the observation angle noise, O is a zero matrix, n is the number of constraints, is an n*n zero matrix.

[0074] Step 4: Analyze the constraints satisfied by the linear trajectory target motion and obtain the linear trajectory constraint information;

[0075] Step 4.1: A target moving along a straight line in a two-dimensional plane can be represented by a slope and an intercept. To describe a line perpendicular to the x-axis, the general equation of a line can be: (Equation 12);

[0076] Step 4.2: Derivative the general equation of the line, we can get the velocity constraint equation for any point: (Equation 13);

[0077] Step 4.3: Arrange the equations in step 4.1 and step 4.2 to obtain the straight line trajectory constraint information. The two formulas included in the straight line trajectory constraint information are: (Equation 14), the straight line constraint equation constrains the position and velocity relationship of any point on the straight line by introducing three additional parameters.

[0078] Step 5: Introduce the linear trajectory constraint information into the tracking system through the augmented measurement equation to establish a tracking model;

[0079] Step 5.1: Use the uniform motion model (CV model) to describe the target moving along a straight line. The target state vector expression under the uniform motion model is: (Equation 15);

[0080] Step 5.2: Expand the dimension of the target state vector expression. The target state vector after expansion is (Equation 16), the target measurement equation is expanded to (Equation 17);

[0081] Step 5.3: Assume that the target states after dimension expansion are uncorrelated, and the noise covariance matrix is ​​diagonally augmented in the traditional noise covariance matrix. The augmented noise covariance matrix is , R k is the covariance matrix of observation distance noise and observation angle noise, O is a zero matrix, n is the number of constraints, is an n*n zero matrix.

[0082] Step 6: Apply the interactive multi-model algorithm based on ellipse constraints and straight line constraints;

[0083] The interactive multi-model algorithm based on elliptical constraints and straight line constraints in step 6 includes four parts: interaction, filtering, model probability update, and state estimation fusion.

[0084] Part I Interaction

[0085] Interaction occurs between models to appropriately distribute the state estimates of each model at the previous moment to each model at the current moment. This process involves probabilistic mixing between models, creating a mixed state estimate and covariance for each model. Specifically, a mixing probability is calculated based on the model transition probability of each model and the model probability of the previous step. This probability is then used to weight the average state and error covariance of each model to generate a new initial condition.

[0086] make is the state estimate of filter j at time k-1, is the state covariance transfer matrix, is the probability of model j at time k-1;

[0087] The state estimates of the N filters after interaction are (Equation 18);

[0088] The state covariance transfer matrix of N filters after interaction is (Equation 19).

[0089] The second part of filtering

[0090] Each model is filtered independently. The filtering used in this method is unscented Kalman filtering, which includes the following steps:

[0091] Step 6.1: First calculate 2nx+1 sampling points and the corresponding weights , sampling point The expression is (Equation 20), The corresponding weight The expression is (Equation 21);

[0092] Step 6.2: Use the state equation to obtain the one-step prediction and weight of the sampling point, and then get the state prediction estimate and state prediction covariance. The expression of the state prediction estimate is: (Equation 22), the expression of state prediction covariance is (Equation 23);

[0093] Step 6.3: The update process is:

[0094] (Equation 24);

[0095] (Equation 25);

[0096] (Equation 26);

[0097] (Equation 27);

[0098] (Equation 28);

[0099] (Equation 29);

[0100] (Equation 30);

[0101] (Equation 31);

[0102] Where n represents the state vector Dimension, j is calculated as ;

[0103] Representation matrix The j-th row or j-th column of , and the calculation formula related to weight is ,in , and These are all empirical parameters of the sigma point in the unscented Kalman filter.

[0104] Part III Model Probability Update

[0105] Model probability updating involves updating the probability of each model after obtaining the filtering results. This step is performed based on the likelihood of each model generating the observed data (calculated using the likelihood function) and the previous model probability. Using the Bayesian rule, the probability of each model is updated based on the support of the current observation for each model.

[0106] Step 6.4: If the filter residual of model j is , and the corresponding covariance is , we can get the model probability of each filter output Updated to (Equation 32), where the filter residual of model j is and covariance The expression is (Equation 33);

[0107] Step 6.5: According to step 6.4, the updated probability of model j is (Equation 34).

[0108] Part IV State Estimation

[0109] State estimation fusion takes the weighted average of all model state estimates according to their latest model probabilities to form a comprehensive state estimate. This comprehensive state maximally integrates information from all possible models, providing an optimal estimate that comprehensively considers all possible scenarios.

[0110] Step 7: Obtain observation results. Example 2

[0111] The interactive multi-model tracking method based on ellipse constraint and line constraint in this embodiment is basically the same as that in embodiment 1, except that:

[0112] Assume that the motion trajectory of a target conforms to the elliptical trajectory and the straight line trajectory, such as Figure 1 As shown in , the target's trajectory is generated first. The target's position and velocity information is first generated in one dimension, and then the one-dimensional information is mapped to a two-dimensional trajectory. The one-dimensional motion noise variance of the target is 0.5 meters. The first part of the trajectory is an elliptical trajectory, and the second part is a straight trajectory. The target's trajectory shape is as follows Figure 2 As shown in Figure 2, the major semi-axis of the elliptical trajectory is 1000 meters, and the minor semi-axis is 700 meters. Assume that the sensor used is a radar, located at the origin, with a range observation noise variance of 5 meters and an azimuth observation noise variance of 0.05 degrees. The radar scan interval is 0.1 seconds, the simulation duration is 300 seconds, and 40 Monte Carlo simulations are performed.

[0113] In order to evaluate the accuracy of the interactive multi-model algorithm with or without constraints, the root mean square error is used to measure the deviation between the true value and the observed value. The calculation formula of the root mean square error (RMSE) is as follows:

[0114] (Equation 35), where n is the observation step length, is the target true value, Predicted value for the sensor.

[0115] like Figure 3 , Figure 4 As shown in the figure, the root mean square error of the constrained interactive multi-model algorithm is smaller than that of the conventional interactive multi-model algorithm, which proves that the constrained interactive multi-model algorithm has higher accuracy.

[0116] In order to evaluate the impact of constraints on model selection in interactive multiple models, Figure 5 、 Figure 6 The probabilities of the two models during the observation process are plotted. When the target follows an elliptical trajectory, the probability of the model for constant-speed circular motion is higher in the constrained interactive multi-model, while the probability of the model for linear motion is higher. There is no significant difference between the two model probabilities in the conventional interactive multi-model. This demonstrates that the proposed interactive multi-model tracking method with both elliptical and linear constraints can effectively improve tracking accuracy and model selection for targets in both elliptical and linear motion.

[0117] The present invention provides an interactive multi-model tracking method based on elliptical constraints and linear constraints. According to the characteristics of elliptical trajectory motion and linear trajectory motion, an unscented Kalman filter model for elliptical motion and an unscented Kalman filter model for linear motion are designed. By introducing the elliptical constraints and the linear constraints into the interactive multi-model algorithm filtering, the tracking accuracy of targets with linear and elliptical trajectory motion characteristics is improved. In addition, the present invention establishes a powerful framework for introducing constraints into the interactive multi-model algorithm, and the model selection for different target motion states is more targeted.

[0118] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent replacements for some of the technical features therein. However, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An interactive multi-model tracking method based on ellipse constraints and line constraints, characterized in that: The following steps are involved: Step 1: Generate motion data of elliptical trajectory targets and straight trajectory targets, and generate measurement data through sensors; Step 2: Analyze the constraints satisfied by the elliptical trajectory target motion and obtain the elliptical trajectory constraint information; Step 3: Introduce the elliptical trajectory constraint information into the tracking system through the augmented measurement equation and establish a tracking model, which includes the following steps: Step 3.1: Use the constant speed turning motion model to describe the target moving along the elliptical trajectory. The target state vector expression under the constant speed turning motion model is: ; The target state equation is , is the state transfer equation, is the noise distribution equation; Step 3.2: Use the sensor to observe. The sensor can observe the distance and azimuth of the target, and the sensor measurement data is ; Step 3.3: By expanding the state dimension, the elliptical trajectory constraint information is introduced into the sensor measurement data expression in step 3.

2. The target state vector is expanded to , the target measurement equation is expanded to ； Step 3.4: For targets subject to elliptical trajectory constraints, the noise mean corresponding to the corresponding constraints is zero. Assume that the target states after dimension expansion are uncorrelated. Diagonally augment the traditional noise covariance matrix. The augmented noise covariance matrix is , R k is the covariance matrix of observation distance noise and observation angle noise, O is a zero matrix, n is the number of constraints, is an n*n zero matrix; Step 4: Analyze the constraints satisfied by the linear trajectory target motion and obtain the linear trajectory constraint information; Step 5: Introduce the linear trajectory constraint information into the tracking system through the augmented measurement equation and establish a tracking model, which includes the following steps: Step 5.1: Use the uniform motion model to describe the target moving along a straight line. The target state vector expression under the uniform motion model is: ; Step 5.2: Expand the dimension of the target state vector expression. The target state vector is expanded to , the target measurement equation is expanded to ; Step 5.3: Assume that the target states after dimension expansion are uncorrelated, and the noise covariance matrix is ​​diagonally augmented in the traditional noise covariance matrix. The augmented noise covariance matrix is , R k is the covariance matrix of observation distance noise and observation angle noise, O is a zero matrix, n is the number of constraints, is an n*n zero matrix; Step 6: Apply an interactive multi-model algorithm based on elliptical and linear constraints; the interactive multi-model algorithm includes interaction, filtering, model probability update, and state estimation fusion; Step 7: Obtain observation results.

2. The interactive multi-model tracking method based on ellipse constraint and straight line constraint according to claim 1, characterized in that: In step 2, the constraints satisfied by the elliptical trajectory target motion are analyzed to obtain the elliptical trajectory constraint information, which includes the following steps: Step 2.1: Suppose a target moves in an elliptical trajectory in a two-dimensional plane, and the coordinates of the center of the ellipse are The semi-major axis and semi-minor axis of the ellipse are a and b respectively. The angle between the line connecting any point on the ellipse and the circle and the x-axis is , the standard equation of the ellipse is , the parametric equation of the ellipse is ; Step 2.2: Derivative the standard equation and parametric equation of the ellipse respectively to obtain the position constraint expression of the ellipse trajectory: , the velocity constraint expression of the elliptical trajectory is , Indicates the angular velocity of the target moving along the elliptical trajectory; Step 2.3: Arrange the expressions in step 2.1 and step 2.2 to obtain the elliptical trajectory constraint information. The three expressions contained in the elliptical trajectory constraint information are: .

3. The interactive multi-model tracking method based on ellipse constraint and straight line constraint according to claim 1, characterized in that: In step 4, the constraints satisfied by the linear trajectory target motion are analyzed to obtain the linear trajectory constraint information, which includes the following steps: Step 4.1: A target moving along a straight line in a two-dimensional plane can be represented by a slope and an intercept. The general equation of a straight line is ; Step 4.2: Derivative the general equation of the line, we can get the velocity constraint equation for any point: ; Step 4.3: Arrange the equations in step 4.1 and step 4.2 to obtain the straight line trajectory constraint information. The two formulas included in the straight line trajectory constraint information are: .

4. The interactive multi-model tracking method based on ellipse constraint and line constraint according to claim 1, characterized in that: make is the state estimate of filter j at time k-1, is the state covariance transfer matrix, is the probability of model j at time k-1; The state estimates of the N filters after interaction are ; The state covariance transfer matrix of N filters after interaction is 。 5. The interactive multi-model tracking method based on ellipse constraint and straight line constraint according to claim 1, characterized in that: The elliptical trajectory tracking model and the straight line estimation tracking model are independently processed by unscented Kalman filtering. The unscented Kalman filtering process includes the following steps: Step 6.1: Calculate 2nx+1 sampling points first and the corresponding weights , sampling point The expression is , the corresponding weight The expression is ; Step 6.2: Use the state equation to obtain the one-step prediction and weight of the sampling point, and then get the state prediction estimate and state prediction covariance. The expression of the state prediction estimate is: , the expression of state prediction covariance is ； Step 6.3: The update process is: ； ； ； ； ； ； ； ； Where n represents the state vector Dimension, j is calculated as ; Representation matrix The jth row or jth column of , and the calculation formula related to weight is ,in , and These are all empirical parameters of the sigma point in the unscented Kalman filter.

6. The interactive multi-model tracking method based on ellipse constraint and straight line constraint according to claim 1, characterized in that: Model probability update includes the following steps: Step 6.4, where the filter residual of model j is and covariance The expression is ； Step 6.5: According to step 6.4, the updated probability of model j is .

7. The interactive multi-model tracking method based on ellipse constraint and straight line constraint according to claim 1, characterized in that: The step of state estimation fusion is to weight the state estimates of all models according to their latest model probabilities to form a comprehensive state estimate.

Images