Tip:
Highlight text to annotate it
X
(Fred) I am Fred DePiero, been at Cal Poly for ten years and I am a
professor of Electrical Engineering.
Certainly the first day, everyday and definitely early in quarter you really have
to get your finger on the pulse as what the students know and don't
know and the best way to do that is to be interactive with the student.
(Fred) On the second page of the syllabus there is a list of topics and
usually adjust that a little bit based on interest of the class.
Imaging processing is a becoming more and more common place
now a day, okay.
Any particular interests?
(Student) Are we going to do OCR this quarter?
(Fred) We could do OCR, I have done in past years, but that's an example of...
I put down in the machine vision category.
Interest in OCR, maybe?
(Students) Yes.
(Fred) Maybe, yes ok.
Any other topics, that... because these are things that we could emphasis
a little bit more in the course.
I will ask question again, you know, as move along in the course.
(Fred) And then I sort... I have taxonomy for the topics that use to present
them on the first as I walk through the syllabus, that, you know, tries to unify
the topics a little bit to, instead of making them appear like,
you know, separate things.
(Fred) All right, on the second page of the syllabus there are topics for the
course and those are... there is a taxonomy there that I am going
describe a little to you and give some examples and the taxonomy
splits the course first of all in half, with the first topics being labeled IP
for image processing and the second half of the topics labeled MV
for machine vision.
Now image processing, I generally refer to as the idea of some sort of
manipulation of an image so that eventually a person would use that image,
look at that image.
Okay, contrast that with the idea of machine vision where we have a
computer analyzing images in order to maybe make some kind of
measurement or something like that.
So the computer is interpreting the image, not a person.
Can anybody think of some examples of machine vision or computer
vision that are around us in use today?
Any examples you can think of?
(Student 1) Traffic light cameras.
(Fred) Traffic light cameras, yeah, to do what?
(Student 1) Make sure there is cars there or not.
[Laughter]
(Fred) Yeah, well I don't know exactly what they do, but the Civil Liberties
Union has a couple things to say about this.
Some for license plates, for example, there are automatic license plate
readers and if blow through a red light they will try and snap an image
of your vehicle and mail you a ticket.
I don't think that's what the cameras around town are for though.
(Student 2) Those detect cars and counts the cars going though and stuff.
(Fred) There you go, ok, for... adjust the traffic lights.
(Fred) I try and mention the... what I call transferable skills that could help
them over a wide range of engineering topics.
I also give them bit of warning about what we will be doing in the course,
because a lot of electrical engineers students they don't really like
software and this course is, you know, oriented toward software.
The students have programs and so I get a kind of mixed reaction to that,
but I try and I warn them of this.
I try and push them out of their comfort zone, because I think it is a good
life long learning experience, for them to be... you know, have to learn
new things and may be get comfortable with new kinds if things that are a
little bit less familiar.
(Fred) I would like to push you out of your comfort zone to some degree
in this class and I was hoping software would be a way to do that, because
a lot of people aren't that comfortable with it.
But I will try to do it as best as I can some how.
Also try and emphasis what I call transferable skills, mathematical skills,
other topics such as calibration, sensitivity analysis.
These are kinds of skill that you can use on various engineering project.
They are things you will run into in engineering.
As opposed to just image processing per say.
Probably not many folks are going go out and get a job
in image processes.
You know, it' s not a real common area.
DSP would be more common than image processing.
But never the less I think there is some... there is some value in this
course and a lot of it has to do transferable skills that you can
use in other areas.
(Fred) One thing I do is to present some multimedia demonstrations
and this is a little software tool that run on a laptop and project it on the
screen and it gives examples of different image processing techniques
and I will go through a variety of these examples and these are actual
projects that they will be working on during the course.
So I will go through some those examples and it provides a nice,
you know, concentrate example of the kinds of processing that will be doing.
(Fred) In the upper left you see an image of a comet, it is the Hal Bopp comet.
So I'm going to do that... I am going to do a contrast enhancement
operation to this.
Here we go and what you see there is a second tail, in the comet.
Okay, so this is an example of how image processing benefited astronomers,
because of astronomers had this and you can almost see the second tail
in there, but it's a lot easier to see here, you know.
Would this type of scene astronomers can deduce there are actually
two different major elements or major components to the comet,
because they are spewing off in two different directions.
We are going to find edges in this image.
Let me show you what I mean by that and then try and define it also.
Now this is a book that you see here over to the left and the book just
happens to have a black section, a darker section, near the binding
and this mid-grey portion on the cover a little further away from the binding.
So strictly speaking if your looking for the edges of objects this ain't an
edge of an object, ok.
But it shows up do to the edge detector.
Suppose we had this picture, all right, and this is white up here this
is black down here.
I am glad you asked this is good, okay.
But now I want to go from this picture to that picture on the left.
Okay, I am looking down on the top, I am seeing white, white, white,
white, black, black, black, black, black, black, black like that.
Okay, and now I want to generate that entire image looking down
from the top here.
I'll just take profile and I'll rotate about that point.
Are you okay with that?
So now if we know that the transform of this rectangular pulse is
this sync function.
Imagine what this what this would look like as grey level image.
Okay, brightest in the middle, periodic dark places as move farther
away from the middle, but now I am also going to spin that around
by the center here and we would see something like this.
Okay, there is the sync function and you can see... you don't see the
lighter variations when you use a linear type scale for the intensity.
(Fred) I like the idea of, you know, disseminating those demonstrations
over the web, so they can immediately start exploring the different
techniques, even if they have a limited understanding that,
you know, might just be based on their life experience, but they
can immediately start playing around with it a little bit and get them interested.