Finding center of a line

Hi i need to find the center of a line.A robot needs to foolow this line in real time but my problem is how to find the center of a line.I used bwmorph(image,'thin',inf); and i get the that my thick line thins to a line with 1 pixel size.

http://img651.imageshack.us/img651/8264/87688912.jpg

http://img687.imageshack.us/img687/3962/87281852.jpg

so now the only thing that i need is to calculate the distance between the center of my image (blue +) and the line so i could say to my robot move to left so that the center of the line is equal to the image center. .

Answer 1

[Kevin Moerman]

If your edge/line is the only segmented object (only the line consists of ones) in your image then if M is your binary image (containing ones for white) then: [I,J]=find(M); %equivalent to find(M==1) gives you the I and J indices or image coordinates of the whole line. If your line is straight then the mean of these coordinates is the same as the mean of the first and last coordinate and thus gives you the middle.

If you have multiple objects in the image with ones, not just your line. Then you could filter those out based on size/shape criteria and do the above. Alternatively could create an adjacency matrix whereby you check whether pixels are one and also have neighbours in the line direction that are one. For the end points of the line the latter will be false which allows you to find them. But this depends on the quality and thickness of your line depictions. So not sure if this would work. Also you could develop a type of mask (filter) that will highlight the ends of lines. But I would go for the simple mean of the object approach.

Kevin

[samir]

well i used find for finding all x and y and then i used mean x and mean y so i got the mean values of x and y where every pixel is 1.

Answer 2

[Image Analyst]

How about using the Pythagorean Theorem? Look up MATLAB's hypot() function.

[samir]

well the harder part is to find the center of the line when i have the centers i can find the distance between them

[Image Analyst]

You're right, if the "line" is not straight. You can't just find the centroid because the centroid may not line on or in the line. You could try bwboundaries() and play around with what that returns.

[samir]

i will try it with bwboundaries and see what i get

Answer 3

[Kevin Moerman]

If its a straight line then you could simply take the mean of your line coordinates (or even centre of gravity of the image object).

If the line is curved then sum up line segments to get total curve length (calculate point-to-point distances and use cumsum for instance), and then solve for (e.g. interpolate) what coordinate describes half of the curve length.

[samir]

ok thank you for your help.

I did some modifications i zoomed with the camera to the line so the line is now more or less straight.I used edge() function and i get the edges .So now i have to find the coordinates of first white and the last white pixel pixel in middle row

[x,y]=size(image);

middle_x=x/2;

With the funciton find() i get every white pixel.So my question is how to get only the twho white pixels where x=middle_x