find Centre of ball/3D model of picture

i want to find the centre of ball (almost solved thanks every one for helping)

http://imageshack.us/photo/my-images/685/97503698.png/

Now i want to create the 3D model of this picture which includes (Pitch ,wickets,and grass) how I can do this kindly help

THANKS

[Walter Roberson]

http://dictionary.reference.com/browse/urgent

"urgent

adjective

1. compelling or requiring immediate action or attention; imperative; pressing: an urgent matter."

So basically you are saying that this is sufficiently important that I should cancel my appointment that I have been waiting a month for, and pay the $150 cancellation penalty, because it requires IMMEDIATE action? Is there an international crisis in automated ball detection in cricket that YOU personally were chosen to solve on a day's notice? Will there be riots in some countries if you are unable to solve it in time?

[awais rehman]

LOL if u have some thing important then just dont answer here and stop wasting ur time as well as mine thanks

[Walter Roberson]

A matter is either urgent or it isn't.

I have removed the "urgent" indicator from the title.

Answer 1

[awais rehman]

automatically.......when i convert this to binary there are many other white regions i dun know how to get only ball .....as am new to this

[Anton Semechko]

what kind of a priori information are you willing to provide? just that you are looking for a small round object? or maybe in addition you are you willing to specify the regions where the object is most likely to be found and its colour?

[awais rehman]

i just want to calculate the center of ball and plot it ....if u can help with some new stuff it will b really good

Thanks

[Anton Semechko]

So do you still insist that you want to detect the ball automatically? Or are you willing to click on the image a couple of times to help locate the ball?

[awais rehman]

automaticallly

Answer 2

[Image Analyst]

Look for dark regions, not white. I'd take one of the color channels and then threshold for dark stuff. Then do some size filtering and shape filtering. See my BlobsDemo for an example of how to do segmentation. http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

Answer 3

[Pramod Bhat]

k. first crop out the unwanted portion of the image - that is background. Now only the ball remains in the image. now consider the middle row and middle column of the matrix. find the indices of the first and last non-zero element of both the mentioned column and row. find the length of them. the intersecting point is the center of the ball.

[awais rehman]

i am new to this ....previously i easily find the ball with black/white background but now i am stuck at this one as i am not able to find the center/exact location of the ball. can u explain in simple words?

[Image Analyst]

Of course finding the wanted and unwanted portions of the image is the whole crux of the problem. Then, you don't want to crop as that just messes up your location of the ball - you'd have to add the cropping location to the centroid location of the ball after cropping to get the coordinates in the original image. All of that is unnecessary. I really think you should have looked at my demo which gives you the concepts you'll need. You just need to threshold for dark things instead of bright things like I did.

[awais rehman]

Image analyst I download your Demo programs really helpful and with the help of that i manage to detect ball but along with 3 other centroids:S

[Matt Tearle]

Are you doing this on a sequence (ie tracking frames of a video), or do you know anything else about the setup that could help you (like the camera position & zoom is fixed)? For example, in a video sequence, you expect the ball to appear at the bottom center of the frame as it leaves the bowler's hand. For most of it's flight, you expect it to be against a background of the pitch. And, in a video sequence, you expect the position to change in small increments from frame to frame. This could help you eliminate the false findings.

Also, you might be able to use the fact that the ball color is probably unique.

[Image Analyst]

Yes, there are other dark regions there so you will have to do something to determine which is the ball, like compare video frames before and after and look for the regions that didn't change position much, or check for the shape (perimeter^2/(4*pi*area)), or something like that.

Answer 4

[awais rehman]

http://imageshack.us/photo/my-images/819/19028011.jpg/

output of my code which shows three marker on picture one at ball second on bowler arm third on batsman leg but i need only one on balll....plz help

Answer 5

[awais rehman]

cmon no 1 can do this?

[Image Analyst]

I probably can. At least for certain frames.

[awais rehman]

how?

[awais rehman]

can u do it 4 me?m stucked

[Image Analyst]

Sorry I'm too busy with my real job to take on unpaid consulting jobs. To make a robust system is going to take more than a couple of minutes. Most of the code I paste here in Answers is just pulled from and adapted from existing demos I have done, so I don't really spend that much time on them even though it may look like a lot of code.