X hits on this document

PDF document

A Prototype Optical Tracking System Investigation and Development - page 58 / 170

455 views

0 shares

0 downloads

0 comments

58 / 170

44

Black Spot Firmware

Using Figure 4.8 and Equation 4.6, preal

is

preal

= an + vn + pn.

(4.9)

In this case

smax

= |preal

ppredicted

|.

(4.10)

Substituting Equation 4.9 and Equation 4.7 into Equation 4.10 gives

smax

= |an| .

(4.11)

In Figure 4.8, it is the maximum marker acceleration amax

that defines smax

. Equating the

magnitude of this acceleration, a

0 max

, to that observed in Chapter3 gives

smax

= |amax| = a

0 max

.

(4.12)

Substituting a

0 max

for smax

in Equation 4.8 gives

ROI Size

2a

0 max

  • +

    D,

(4.13)

which states that, using this prediction scheme, the minimum ROI size required to track the typical scanner motion found in Chapter 3 is proportional to the magnitude of the max- imum acceleration. This was determined to be approximately 8 pixels / frame2 meaning that the maximum ROI size required for a diameter of 16 pixels (Equation 3.24) is 32 × 32.

This prediction algorithm is a limiting case of a discrete Kalman Filter [19, 66]. The system state is the position pn measured by the centroid calculation. The state prediction step is Equation 4.7. In this case, the centroid calculation is used directly to give the state estimate.

4.2.5

Future improvements

There are other algorithms that could improve the performance of the Black Spot module. They are proposed for future implementation.

Marker discriminator

The first algorithm is a marker discriminator algorithm. Ideally the module should be able to distinguish between a real marker and noise sources such as bright lights and reflections. lso, markers should be able to be ignored if they are too close or too far away to be useful. The design of this algorithm has not been investigated but the characteristics of marker images may be able to be used to determine whether a region of bright pixels is a marker

or an artefact.

possible, although computationally intensive, approach would be to use

Document info
Document views455
Page views455
Page last viewedSun Dec 11 12:29:31 UTC 2016
Pages170
Paragraphs6307
Words54996

Comments