X hits on this document

PDF document

A Prototype Optical Tracking System Investigation and Development - page 39 / 170

318 views

0 shares

0 downloads

0 comments

39 / 170

3.2 Black Spot module analysis

25

ROI Width

Marker Height

ROI Height

Marker Width

Figure 3.3 A Region of Interest (ROI) is a rectangular window that surrounds a marker.

that the background remains constant, foreground objects can be emphasised using this ap- proach. The method cannot be used in this system as the Black Spot modules themselves are moving. Therefore, there cannot be any stationary reference image.

nother method of segmentation is template matching [44–46].

template image that

is to be detected is compared with an image frame. The template image must be shifted across the entire captured image and the similarity between the template object and the image data in the template position determined. If the similarity is high enough then the algorithm determines that the object is present at this shifted position. Similarity can be defined in a number of ways including the cross-correlation between the image and the background it covers, as well as the sum of the squared differences between the pixels

from the template and image [44–46].

Template matching relies on the image template remaining constant. However, as the cam- era moves in relation to the markers, the images of the markers will change in size. This can be accounted by using a number of templates of markers at different distances but this requires more memory. Template matching is also computationally intensive (due to the template being shifted many times) and has not been considered here.

The

approach

taken

to

segment

the

markers

is

computationally

cheap

and

can

run

in

real-

time on an embedded system. Under work in, the markers are assumed to

the limited conditions that the system is designed to be the brightest objects in the image. The algorithm

looks

for

bright

pixels

in

each

image.

If

a

bright

pixel

is

located,

then

a

ROI

is

constructed.

The

pixels

above

a

certain

threshold

are

assumed

to

be

bright

image

pixels.

Therefore,

segmentation is a combination of using a ROI and this threshold.

It is assumed that the

marker will be contained within this ROI. This is typically the case but, depending on the position of the bright pixel, the marker may be only partially contained by the ROI. lso, the bright pixel may correspond to a non-marker object, such as a reflection or image noise.

Despite these potential issues this method Further explanation is given in Chapter 4.

works

well

in

the

office

setting

used

for

testing.

Document info
Document views318
Page views318
Page last viewedSat Dec 03 10:08:23 UTC 2016
Pages170
Paragraphs6307
Words54996

Comments