fastest way to detect a red ball?

asked 2016-03-22 11:15:43 -0600

215
230 ●3 ●10 ●22

I at moment trying to implement a simple an robust way of detecting an red ball.

I've converted into HSV and filtered using inRange(..), such that the red ball is clear and some other noise.

But what approach would be appropriate to detect the ball?

findContour and Hough seems a bit overkill computation wise, but what other alternatives do i have. I have to detect the position of the ball at all time, so the speed of the algorithm has to be appropriate for tracking.

what do you suggest?

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2016-03-22 12:19:26 -0600

kbarni

4134 ●1 ●13 ●49

I think you are looking for the image moments.

Moments mom=moments(segm_image);
int area,xc,yc;
area=mom.m00; //area of the ball in pixels
xc=mom.m10/mom.m00; //x and y center of the ball
yc=mom.m01/mom.m00;

edit flag offensive delete link

Comments

I Wont know the distance for sure as the size of Ball of the image is not known.. I am using stereo vision

215 ( 2016-03-22 12:50:05 -0600 )edit

The disadvantage of moments is, that it is executed on the whole image. So if there are just a few pixels left far away from the ball position, the center of that ball will be shifted a lot. The next question is if moments is that much faster than findContours.

Why do you think, that findContours is too slow? You can also resize your image to speed up findContours at the cost of accuracy.

matman ( 2016-03-23 16:16:55 -0600 )edit

add a comment

Links

Question Tools

1 follower

Stats

Asked: 2016-03-22 11:15:43 -0600

Seen: 707 times

Last updated: Mar 22 '16

fastest way to detect a red ball? edit