Main Content

initcvimm

IMM initialization with two constant velocity models

Since R2023b

Description

imm = initcvimm(detection) initializes a trackingIMM object with two constant velocity models based on the information provided in detection. The first model assumes a smooth trajectory and uses low process noise. The second model assumes a maneuvering trajectory and uses high process noise. For more details, see Algorithms and trackingIMM.

example

Examples

collapse all

Create a 3-D position detection in the rectangular frame. The measured coordinates are x = 1, y = 3, and z = 0. Set the covariance of the measurement noise as [1 0.4 0; 0.4 4 0; 0 0 1].

detection = objectDetection(0,[1;3;0],MeasurementNoise=[1 0.4 0; 0.4 4 0; 0 0 1]);

Use the initcvimm function to create a trackingIMM object based on the detection.

imm = initcvimm(detection);

As expected, the initialized filter has two tracking filters with low and high process noise, respectively.

imm.TrackingFilters
ans=2×1 cell array
    {1x1 trackingEKF}
    {1x1 trackingEKF}

imm.TrackingFilters{1}.ProcessNoise
ans = 3×3

    0.0400         0         0
         0    0.0400         0
         0         0    0.0400

imm.TrackingFilters{2}.ProcessNoise
ans = 3×3

   100     0     0
     0   100     0
     0     0   100

Verify the filter state has the same position components as those of the detection.

imm.State
ans = 6×1

     1
     0
     3
     0
     0
     0

Verify that the filter measurement noise is the same as that of the detection.

imm.MeasurementNoise
ans = 3×3

    1.0000    0.4000         0
    0.4000    4.0000         0
         0         0    1.0000

Create a 3-D position detection in the spherical frame. The measurement has an azimuth of 40 degrees, an elevation of 6 degrees, a range of 100 meters, and a rang-rate of 5 m/s. Also, specify the measurement noise for the measurement.

meas = [40;6;100;5];
measNoise = diag([2.5,2.5,0.5,1].^2);

Specify the measurement parameters and create the detection.

measParams = struct(Frame="Spherical",HasAzimuth=true, ...
    HasElevation=true,HasRange=true, ...
    HasVelocity=true);
detection = objectDetection(0,meas,MeasurementNoise=measNoise, ...
    MeasurementParameters=measParams);

Use the initcvimm function to create a trackingIMM object based on the detection.

imm = initcvimm(detection);

As expected, the initialized filter has two tracking filters with low and high process noise, respectively.

imm.TrackingFilters
ans=2×1 cell array
    {1x1 trackingEKF}
    {1x1 trackingEKF}

imm.TrackingFilters{1}.ProcessNoise
ans = 3×3

    0.0400         0         0
         0    0.0400         0
         0         0    0.0400

imm.TrackingFilters{2}.ProcessNoise
ans = 3×3

   100     0     0
     0   100     0
     0     0   100

Verify that the filter measurement noise is the same as that of the detection.

imm.MeasurementNoise
ans = 4×4

    6.2500         0         0         0
         0    6.2500         0         0
         0         0    0.2500         0
         0         0         0    1.0000

Input Arguments

collapse all

Detection report, specified as an objectDetection object.

Example: detection = objectDetection(0,[1;4.5;3],'MeasurementNoise', [1.0 0 0; 0 2.0 0; 0 0 1.5])

Output Arguments

collapse all

Interacting multiple model filter, returned as a trackingIMM object. The object has two constant velocity motion models.

Algorithms

  • The function initializes a trackingIMM object. The state of the filter is defined as [x; vx; y; vy; z; vz], where x, y, and z are the position coordinates and vx, vy, and vz are the corresponding velocities.

  • The TrackingFilters property of the initialized trackingIMM object contains these tracking filters.

    • The first filter is a trackingEKF object based on the constant velocity motion model. The ProcessNoise property of the filter is set to 0.04*eye(3) m/s2. The rest of the filter properties are initialized by using the initcvekf function. See the Algorithm section of the initcvekf function page for more details.

    • The second filter is the same as the first filter except its ProcessNoise property is set to 100*eye(3) m/s2.

  • You can use this function as the FilterInitializationFcn property of a tracker object such as the trackerGNN object.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Version History

Introduced in R2023b