CODEWITHSUNDEEP
pythonloopsnumpy
oyildirim
oyildirim

Reputation: 61

Numpy finding 3d coordinate distance from a specified point

I am working on finding the 3D distance from a specified point in python. Right now, I am using for loops. But it's a little slow to compute.

Here is python code:

    for i_y in xrange(0,500,100):
        y = round(i_y/100.00,2)

        for i_x in xrange(0, 800, 1):
            x = round(i_x/100.00,2)

            for i_z in xrange(500, 0, -1):
                z = round(i_z/100.00,2)

                    for key in specifiedPoints.keys():
                    a = specifiedPoints[key]

                        subx1 = x-a.p1.x
                        suby1 = y-a.p1.y
                        subz1 = z-a.p1.z
                        subx2 = x-a.p2.x
                        suby2 = y-a.p2.y
                        subz2 = z-a.p2.z
                        subx3 = x-a.p3.x
                        suby3 = y-a.p3.y
                        subz3 = z-a.p3.z

                        distver1 = math.sqrt(subx1*subx1+suby1*suby1+subz1*subz1)
                        distver2 = math.sqrt(subx2*subx2+suby2*suby2+subz2*subz2)        
                        distver3 = math.sqrt(subx3*subx3+suby3*suby3+subz3*subz3)

                        if distver1 <= 1 or distver2<=1 or distver3<=1:
                            print "close point:", x, y, z

I worked a lot but I couldnt find a clear tutorial that shows an equal loop in numpy.

How can I make this in numpy to accelerate loop?

Thank you

Upvotes: 0

Views: 2079

Answers (2)

JoshAdel
JoshAdel

Reputation: 68682

I would look at scipy.spatial.distance and especially the cdist and pdist methods:

http://docs.scipy.org/doc/scipy/reference/spatial.distance.html

Upvotes: 2

Schuh
Schuh

Reputation: 1095

Your arange function could return the values you compute for x,y and z directly, what would save a lot of computations. The round function is not needed at all. You run the loop 5*800*500 = 2.000.000 times and every time you divide by 100 and round. Better do it like this:

    for y in np.arange(0,5,1):
        for x in np.arange(0,8,0.01):
            for z in np.arange(5,0,-0.01):

Collect the points in one array like in the following code. 

    point = np.array([x,y,z])
    a1 = np.array([a.p1.x,a.p1.y,a.p1.z])
    a2 = np.array([a.p2.x,a.p2.y,a.p2.z])
    a3 = np.array([a.p3.x,a.p3.y,a.p3.z])

    if np.linalg.norm(point-a1) <=1:
        print point
        continue
    if np.linalg.norm(point-a2) <=1:
        print point
        continue
    if np.linalg.norm(point-a3) <=1:
        print point
        continue

It is better to store the points directly as numpy arrays in your object specifiedPoints[key] and not to collect them again and again in every loop. This would get you this code:

    point = np.array([x,y,z])

    if np.linalg.norm(point-a.p1) <=1:
        print point
        continue
    if np.linalg.norm(point-a.p2) <=1:
        print point
        continue
    if np.linalg.norm(point-a.p3) <=1:
        print point
        continue

Upvotes: 1

Related Questions