using inpolygon as a matrix operation

Dear All,

I'm trying to do the following. I have 2 very large vecotrs and two matrices. The two vecotrs are X and Y coordinates of random points. In the matrices are the x and y data points for circles aroud certain random points (from the first two vectors). To clarify a bit more: each row of a matrix is the X or Y coordinates for a circle, with i number of circles below eachother. So I'm trying to see which random points are in each of the circles.

How i do that now is with the following code:

for i=1:length(matrix(:,1))
    % xcoord and ycoord are the coordinate vectors of random points, x and y are the
    % coordinates of the circles
    in = inpolygon(xcoord,ycoord,x(i,:),y(i,:)); 
    % only take the indices where inpolygon is 1 and save those as cells
    nz(i,:)={find(in)}; 
end

But this is slow progress. Is there a way to do this using matrix operations to speed it up?

With kind regards,

Tom

Answer 1

[Matt J]

I'm not sure why you are using a polygon approximation to the circles. You could just calculate the distance of all (x,y) from the centers of the circles and test whether that distance is less than the circle radius.

pdist2([X(:),Y(:)],[xc(:),yc(:)])<=radii(:).' ;

[Tom Wagemans]

Hi Matt,

wouldn't pdist2 work on a vector basis? As X and Y are both matrices (248345 by 100), and will be needing a forloop to do it several times? It's an interesting idea though.

kind regards,

Tom

[John D'Errico]

You would essentially need loops to use inpolygon anyway. But you really don't want to use inpolygon, if these points are just circles!

One circle intersects with another circle if the Euclidean distance between the centers of the circles is less than the sum of the two radii.

So I'm not even sure if your complete approach here seems wrong. Don't work with circles ay all by creating a set of points around the circumference, but treat them properly, as circles.

[Matt J]

wouldn't pdist2 work on a vector basis? As X and Y are both matrices (248345 by 100)

I don't understand the question. You told us in your initial post that X and Y are both vectors not matrices. How many circles do you have? And how many query points?

pdist2 will result in an MxN matrix where M is the number of query points and N is the number of circles. If M*N is prohibitively large, you may indeed need to break the calculation into smaller vectorized chunks, say Mx10.

Answer 2

[darova]

Use colon (:) operator

in = inpolygon(xcoord,ycoord,x(:),y(:));

[Tom Wagemans]

Hi darova,

I've used the colon operator as you can see in my OP. Unfortunately this doesn't solve the need for a forloop. As X Y are matrices of 248345 by 100, in need to do inpolygon 248345 times. Each loop accesses the X Y matrix as a 1 by 100 vector. So you can imagine this taking a while.

kind regards,

Tom

[darova]

SOrry, didn't unerstand the question. Matt's idea should work. Find centers, radii and use pdist2()

[Sourav Mukherjee]

I am facing the same problem. Did you come up with solution? In my case, I am looking for the points within a square and not circle.

[darova]

Just do something like

ind1 = x0-a/2 < x & x < x0+a/2;
ind2 = y0-b/2 < y & y < y0+b/2;
ind = ind1 & ind2;

a,b - sides of your square

x0,y0 specific point (blue points)

x,y - your data (red points)