Suggestions for impementing an image processing and motor control algorithm.

hello,

we are five final year electronics and communication engineering students
of college of engineering chengannur, india. We are planning to make a
system based on DSP for our final year project.

It must be able to find a moving object in the frame of the camera and
point a laser beam on it as it moves within a specified angle. Like
automatic tracking of targets. And when the laser is replaced by a high
power laser (or a gun), it will be able to destroy the targets. I need
some help regarding how to go about doing it and how to implement this
finally (on a processor, or just using LabVIEW or using MATLAB or things
like that). Any help is highly appreciated.

Thanks
Githin F Alapatt

crazydiamond skrev:
> It must be able to find a moving object in the frame of the camera and
> point a laser beam on it as it moves within a specified angle. Like
> automatic tracking of targets. And when the laser is replaced by a high
> power laser (or a gun), it will be able to destroy the targets. I need
> some help regarding how to go about doing it and how to implement this
> finally (on a processor, or just using LabVIEW or using MATLAB or things
> like that). Any help is highly appreciated.

These are two different projects: Object tracking in an image, and
weapons guidance. You say "*like* tracking of targets". Do you
mention this as an example, or is this actually what you try to do?

If you are trying to implement an automatic target tracking system
for guns, be aware that you need a stereoscopic vision system.
The reason is that you need an estimate for target range as well
as elevation and azimuth angles. You can't get range estimates
from a single camera.

Rune

Rune Allnor wrote:

> ... You can't get range estimates from a single camera.

Not unless you know the size of the target and the magnification of the 
camera.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Rune Allnor wrote:
> crazydiamond skrev:
> 
>>It must be able to find a moving object in the frame of the camera and
>>point a laser beam on it as it moves within a specified angle. Like
>>automatic tracking of targets. And when the laser is replaced by a high
>>power laser (or a gun), it will be able to destroy the targets. I need
>>some help regarding how to go about doing it and how to implement this
>>finally (on a processor, or just using LabVIEW or using MATLAB or things
>>like that). Any help is highly appreciated.
> 
> 
> These are two different projects: Object tracking in an image, and
> weapons guidance. You say "*like* tracking of targets". Do you
> mention this as an example, or is this actually what you try to do?
> 
> If you are trying to implement an automatic target tracking system
> for guns, be aware that you need a stereoscopic vision system.

Not necessarily so. I'll admit you want "stereoscopic data". But 
"stereoscopic vision" is not the only way, just the conventional way.

I'll give one hint. It would serve as a good teaching tool om the idea 
of engineering trade offs. Depending on what cost/effort parameter needs 
to be minimized, it might be demonstratively "better" than conventional 
solution.

I'll borrow a page from Jerry's (and others) playbook and say "think 
about it" ;)

> The reason is that you need an estimate for target range as well
> as elevation and azimuth angles. You can't get range estimates
> from a single camera.
> 
> Rune
> 

Richard Owlett skrev:

> I'll give one hint. It would serve as a good teaching tool om the idea
> of engineering trade offs. Depending on what cost/effort parameter needs
> to be minimized, it might be demonstratively "better" than conventional
> solution.

You are thinking about focus lengths? To what extent it works,
it depends on the target being very close to the camera. Don't
know a lot of optics, but the term "Fresnel zone" popped up
from some deep dark corner of my mind...

Rune

Rune Allnor wrote:
> Richard Owlett skrev:
> 
>> I'll give one hint. It would serve as a good teaching tool om the idea
>> of engineering trade offs. Depending on what cost/effort parameter needs
>> to be minimized, it might be demonstratively "better" than conventional
>> solution.
> 
> You are thinking about focus lengths? To what extent it works,
> it depends on the target being very close to the camera. Don't
> know a lot of optics, but the term "Fresnel zone" popped up
> from some deep dark corner of my mind...

I judge distance pretty well even with one eye. The relative sizes of 
things is a strong clue (and subject to manipulation when deception is 
wanted). Parallax cause by the motion of my eye is a fine tool also.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins skrev:
> Rune Allnor wrote:
> > Richard Owlett skrev:
> >
> >> I'll give one hint. It would serve as a good teaching tool om the idea
> >> of engineering trade offs. Depending on what cost/effort parameter needs
> >> to be minimized, it might be demonstratively "better" than conventional
> >> solution.
> >
> > You are thinking about focus lengths? To what extent it works,
> > it depends on the target being very close to the camera. Don't
> > know a lot of optics, but the term "Fresnel zone" popped up
> > from some deep dark corner of my mind...
>
> I judge distance pretty well even with one eye. The relative sizes of
> things is a strong clue (and subject to manipulation when deception is
> wanted).

It works as there are objects nearby that serve for scale comparisions.
For objects at sea or in the air, you can become fooled very easy.

> Parallax cause by the motion of my eye is a fine tool also.

"One-eyed-man's stereoscopy"...

Rune

Rune Allnor wrote:
> Jerry Avins skrev:
>> Rune Allnor wrote:
>>> Richard Owlett skrev:
>>>
>>>> I'll give one hint. It would serve as a good teaching tool om the idea
>>>> of engineering trade offs. Depending on what cost/effort parameter needs
>>>> to be minimized, it might be demonstratively "better" than conventional
>>>> solution.
>>> You are thinking about focus lengths? To what extent it works,
>>> it depends on the target being very close to the camera. Don't
>>> know a lot of optics, but the term "Fresnel zone" popped up
>>> from some deep dark corner of my mind...
>> I judge distance pretty well even with one eye. The relative sizes of
>> things is a strong clue (and subject to manipulation when deception is
>> wanted).
> 
> It works as there are objects nearby that serve for scale comparisions.
> For objects at sea or in the air, you can become fooled very easy.
> 
>> Parallax cause by the motion of my eye is a fine tool also.
> 
> "One-eyed-man's stereoscopy"...

The first interstellar distances were measures in parsecs. One parsec is 
the distance that produces one arc-second of parallax of nearby stars 
against the background of very distant stars when the earth has gone 
halfway around its orbit. Beyond ten meters, we might as well all be one 
eyed.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Rune Allnor wrote:
> Richard Owlett skrev:
> 
> 
>>I'll give one hint. It would serve as a good teaching tool om the idea
>>of engineering trade offs. Depending on what cost/effort parameter needs
>>to be minimized, it might be demonstratively "better" than conventional
>>solution.
> 
> 
> You are thinking about focus lengths? 

Nope ;)

But I have to admit I've realized there would be a problem implementing 
my solution.

My key sentence was "Not necessarily so. I'll admit you want 
"stereoscopic data". But "stereoscopic vision" is not the only way, just 
the conventional way."

I'm on my way to work, my answer when I get home.




To what extent it works,
> it depends on the target being very close to the camera. Don't
> know a lot of optics, but the term "Fresnel zone" popped up
> from some deep dark corner of my mind...
> 
> Rune
> 

Richard Owlett wrote:
> Rune Allnor wrote:
> 
>> Richard Owlett skrev:
>>
>>
>>> I'll give one hint. It would serve as a good teaching tool om the idea
>>> of engineering trade offs. Depending on what cost/effort parameter needs
>>> to be minimized, it might be demonstratively "better" than conventional
>>> solution.
>>
>>
>>
>> You are thinking about focus lengths? 
> 
> 
> Nope ;)
> 
> But I have to admit I've realized there would be a problem implementing 
> my solution.

My error was thinking i made an error :)

> 
> My key sentence was "Not necessarily so. I'll admit you want 
> "stereoscopic data". But "stereoscopic vision" is not the only way, just 
> the conventional way."
> 
> I'm on my way to work, my answer when I get home.

Have 2 lasers whose beams are *parallel*
apparent separation in image inversely proportional to distance.


How about 1 emitter and one almost off the shelf camera which would 
yield 6 meter resolution out to ~50 miles.

Hint: at least 2 members of group deal/dealt with this methodology on a 
daily basis.

To borrow from a current commercial -- think outside the bun