THE OPEN UNIVERSITY OF SRI LANKA

LEVEL 5
ACADEMIC YEAR 2010/11
ASSIGNMENT #01

MEX5271 MACHINE VISION

(1) Only the soft copy of the answer script must be uploaded to the OUSL Virtual
Classroom on or before 10 th October 2010, in PDF format. The answer script must be
less than seven pages.

(2) When you are submitting the assignment, first click the “Assignment’ module and
select the appropriate assignment in the given item list. Then answer the particular
assignment through the virtual class. If you do not follow the instructions, then your
submission can be seen by anyone in the MEX5271 class.

(3) Late submission will not be accepted.

Q1.

a) A photographer has taken the image below at night. 0 corresponds to black and 15 to
white. In this image we can recognize one bright house in the left (the pixels with
gray levels ≥ 8) and the moon on the right (the pixels with gray level ≤ 6 and ≥ 2). We
want to extract a good image of the moon.

0 0 1 0 0 0 2 3 0
0 8 0 0 0 5 4 6 3
9 14 9 0 0 3 4 4 2
15 12 10 0 0 0 2 2 0
11 13 13 0 1 0 0 0 0
I. Perform the histogram of this image.
II. Design a point processing letting the moon unchanged and transforming the
house and the background to black.
III. Give the histogram of the negative of the image transformed after II.

b) To add contrast to the image resulting from a). II, find the linear point processing that
maps the gray levels on the moon to cover the full range of light intensities from 0 to
15.

Q2.

Discuss four phases in Image Acquisition.

Q3.

a) An image has the gray level probability density function Pr(r) shown in Fig. 1(a). We
wish to transform the gray levels of the image so that they will have the specified Pz(z)
shown in Fig. 1(b). Assume continuous quantities, find the transformation (in terms of
r and z) that will accomplish this.

Pr(r) Pz(z)

2 2

1 r 1 z

Fig. 1(a) Fig. 1(b)

b) An engineer is asked to inspect a certain class of images generated by a machine vision

system. He has decided to use digital image enhancement and has examined a set of
representative images. He has found the following problems:
i. bright and isolated dots are of no interest;
ii. lack of sharpness;
iii. insufficient contrast in some images;
iv. shifts in the average gray- level value from a pre-determined value

He wishes to correct these problems and then colour in constant red all gray levels
between a given band between I1 and I2 , while keeping normal tonality in the
remaining gray levels. Propose a sequence of processing steps that the engineer can
follow to achieve the desired goal.
Q4.

(a) Design, with the aid of structured lighting, a computer vision based system to measure
the heights of the two protrusions of the machine component shown in Fig. 2.

(b) Consider an automated machine vision based system for the sorting of the gender of
juvenile guppy. The objective is to detect the presence and the size of a small black
dot at a certain part of the body. The current set up of the system provides an image
capturing window measuring 2cm by 2cm square. The CCD camera has a solid state
sensor array 0.8 by 0.8 cm square, which can have a resolution of 512 by 512, 1024
by 1024, or 2048 by 2048 pixels. For the lenses, you are only allowed to use the
available sizes, which are 25 mm, 35 mm or 50 mm. If the required tolerance of the
measurement is 20 µm, and the object distance should be between 100 to 125 mm,
design the system by choosing the appropriate lens and sensor array resolution. You
are to note that the camera (with any of the stated array size) costs much more than
the lenses. In addition, the higher the resolution of the sensor array, the more
expensive is the camera. Discuss briefly your design.