1. ## Lets solve this

Hi people..
About problem statement. Lets this time work on little bit on Image processing.
Okay…?
Ya Actually this is a problem I have to solve.. and thought why not be a little bit selfish and take you people’s help in solving this problem?

Situation
1. There are 4 IR LED placed on an object, emits constant IR light.
2. The LEDs are placed at 4 corners of a square having 4 inch sides.
3. There is a cam which is on top of the robo is facing the IR LEDs, receives the light emitted from IR LED.
4. The distance between IR LEDs and cam is 4-10 mtrs. (variable)

5. The movement of the object can be on both X & Y direction .. ie the object can move left right & up down..

6. The object moves slowly .. it is some thing like some one’s face’s movement while he is talking ( max can be 2 - 3 inches in both X &Y direction) does not go any where else.. ( or imagine you are cam and watching some other person's eyes.. eyes are LEDs..)

Problems
1. The motion detection module of the robo monitors the motion of the object with the help of cam.
2. Should provide this information to control module using some Wireless method.
3. But if there is any other movement the cam should not detect those.
4. The module should be able to determine the distance between the robo and object correctly.
5. If required the module can switch on /off the IR LEDs
6. Once robo starts monitoring, the distance between the object & cam remains same. But object can be placed any where between 4 to 10 mtrs in front of the cam and system should tell how far it is..

Optional

1. Try to make the cam & detection module as a single unit.
2. We will be appreciated if we could achieve to provide only the diff between two positions in any two conjugate frames, to the control unit than providing the entire streaming video data and expect the control unit do the processing.
3. Think of some good cam .. Web cam or CCTV cam ...

People I am also going to work with you people to solve this problem. I have some solutions to this problem .. but lets listen form you people first .. then I will put my ideas.

We will talk and will find out if there is any “Must required ” type requirements exists and we will modify the requirement accordingly.

I have also added a Wiki page to project the current status of the project.
http://www.roboticsindia.com/modules...otionDetection

Hope you find this problem interesting.

Thanks
Regards

2. 8O 8O Whew!! Good one Mishra bhai!

3. first thing is the isolation..

Let the intensity of the LED visible to the cam be higher than the ambient light and the surroundings..

Then we grab the frame and threshold the image so that we are left with just a black frame with white dots...

now we can monitor the movement of the LED only and other movements wont be visible.. thats the first phase

For calculating the distance.. we suppose that the square just moves in a straight line running through the cam bot away or closer to the bot and not in a direction perpendicular to that line..

In that case, we calculate the pixel distance between the front two LEDs and calibrate that..
when the square moves forward or backward, the pixel distance between the white dots representing the LEDs increases or decreases respectively,
Using the calibrated values, we can substitute values and hence find the distance between the robot and the square..

Will think about other cases later and put them on..

Rao

5. my IP algo skills are in infancy but can contribute to test it....

if we use threshold we can locate the position of edges of square (IR led), but if blink the led for certain duration initially for sensing the change in bit map, we can reduce the possibility of false location due ambient light and the surroundings holding same threshold value and know exact position of led on bitmap....

i had not done actual implementation of any IP problem but thought transforming the bitmap into matrix and just comparing the change of elements of required position we will be able to know is the object is moving or not....

6. People,
Let me share one information which i have found..

If we can some how filter the visible lights falling on cam .. we can only see the IR ray...

To achieve this I found that a photo film ( black portion) (Not sure if Xray could also help) .. in front of the web cam .. we are going to see the IRs only...

regards

7. yes... red filters will do...

It will cut off all the components other than the lower red ones..
hopefully,the lower red component in the visible light will be far less in intensity then the IR... Unlesss you have a strong source of red light..
I wonder if you can use those low cost red laser beams... good in intensity.. but will it affect the camera??
:roll:

as shashak said, Modulation is also a good option, but we might have to take care of the frame rate and the blinking frequency...

and how accurately do we have to measure the distance??

8. Hi allbits..
There are problems in using the red laser.

1. they travel stright .. so cam can not detect the beam until the cam comes just in front of the laser beam.

2. And if 4 Lasers are pointed in such a way that they fall on cam then cam will go blind..
3. And when we place the object in a different location, the lasers are required to be adjusted .. so that they again fall on cam..

I think we should think a bit more on Modulation..

About the distance? we should be able to measure the distance in MM.CM format.. for ex 08.34 cm..

okay..

mishra

9. Here is a link which details a technique i would have used

http://www.pages.drexel.edu/~twd25/w...er_ranger.html

For your example here are the modifications required ...

In place of Laser , we have you four IR emitters to look at,track and multiply with pre calibrated values for getting final values of x,y movement.
For interfearance problem we will use IR filters.

So what do you think now ?

10. Hey Vikas,
gr8 link... hope it works as per design..

But only one thing is, hope the code is not slow.. ie .. should not produce delays..

Thats why I was thinking if we could find out the x, y position using some h/w ( image processor ), the control system can do many heavy rendering things with that info..
i am yet to finalize the rendering requirements..

Sorry I have no idea about the DSP chips and how they work..

Sorry .. i am not going to thank any one.. project is still going on.. I will try these ideas and will publish the results ..

regards