# Object Tracking using kalman filter

Started by July 31, 2011
```Hey guys im trying to track an object using kalman filter. I wanted to ask
that how can we interface a stepper motor with the kalman filter? (the
motor is connected with the camera to control its movement).

Also how can we calculate measurement and process error matrices in kalman
filter?
thnx.

```
```On Sun, 31 Jul 2011 21:34:08 -0500, AE1 wrote:

> Hey guys im trying to track an object using kalman filter. I wanted to
> ask that how can we interface a stepper motor with the kalman filter?
> (the motor is connected with the camera to control its movement).
>
> Also how can we calculate measurement and process error matrices in
> kalman filter?
> thnx.

Oi.  Oi oi oi.

Oi #1:

A Kalman filter is a theoretical construct.  A stepper motor is a
physical construct.  How do you interface a theoretical construct with a
physical one?

Presumably the output of your Kalman filter is a pointing direction.
Step the stepper in such a way that your camera points closer to the
direction that the Kalman filter is estimating.

(Forward looking hint: in what reference frame is your Kalman
functioning, and what happens to this frame when you move the camera?
What happens to the object's apparent motion when you move the camera?)

Oi #2:

A Kalman filter is a filter that has been designed to a certain set of
optimality criterion, using as inputs known values of process and
measurement noise.  If you don't know what your process and measurement
noise are, then you haven't designed a Kalman filter.

You _estimate_ your process and measurement noise the same way that you
arrive at your object's motion model: you make informed guesses.  You can
then verify all these guesses by checking your measurement error and
seeing if it is white -- if it isn't, then you've done something wrong.

Since the process noise is, in this case, not necessarily known, or
white, or Gaussian, or anything else that is assumed by the Kalman filter
design process, you may be better off thinking in terms of an H-infinity
filter.

--
www.wescottdesign.com
```
```On 8/1/11 2:01 PM, Tim Wescott wrote:
> On Sun, 31 Jul 2011 21:34:08 -0500, AE1 wrote:
>
>> Hey guys im trying to track an object using kalman filter. I wanted to
>> ask that how can we interface a stepper motor with the kalman filter?
>> (the motor is connected with the camera to control its movement).
>>
>> Also how can we calculate measurement and process error matrices in
>> kalman filter?
>> thnx.
>
> Oi.  Oi oi oi.
>
> Oi #1:
>
> A Kalman filter is a theoretical construct.  A stepper motor is a
> physical construct.  How do you interface a theoretical construct with a
> physical one?

well you talk to the sales engineer where you buy either the Kalman
filter or the stepper motor.  it's probably best to find a place that
sells 'em both.

--

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

```
```On Aug 1, 2:34&#4294967295;pm, "AE1" <ahsanejaz1@n_o_s_p_a_m.gmail.com> wrote:
> Hey guys im trying to track an object using kalman filter. I wanted to ask
> that how can we interface a stepper motor with the kalman filter? (the
> motor is connected with the camera to control its movement).
>
> Also how can we calculate measurement and process error matrices in kalman
> filter?
> thnx.

Make sure you replace the Kalman filters filter bag regularly.
```
```On Tue, 02 Aug 2011 13:15:46 -0700, HardySpicer wrote:

> On Aug 1, 2:34&nbsp;pm, "AE1" <ahsanejaz1@n_o_s_p_a_m.gmail.com> wrote:
>> Hey guys im trying to track an object using kalman filter. I wanted to
>> ask that how can we interface a stepper motor with the kalman filter?
>> (the motor is connected with the camera to control its movement).
>>
>> Also how can we calculate measurement and process error matrices in
>> kalman filter?
>> thnx.
>
> Make sure you replace the Kalman filters filter bag regularly.

Lest they get clogged, and the accumulated errors overflow into your data?

--
www.wescottdesign.com
```
```Where I used to work we developed a system that used a Kalman filter
to track and count people using a thermal sensor array, it's been very
successful. It was hooked up to a servo motor system with an ordinary
camera, and worked very well with people at a distance. Close up,
though, it always slewed to the groin area. :)
```