Reversable algorithm for to/from latitude and longitude and 3D point

3
votes

I know there are other questions similar to this one, but most of them deal with only converting one to the other. But I am searching for a algorithms that convert to and from each other. Simply using one of each has not produced the desired results.

For my purposes a unit sphere is more then acceptable. Any radius value would be 1.

Here are my current methods for performing this, in simple psudocode.

From latitude and longitude to a point on a unit sphere.

x = cos( longitude ) * sin( latitude )
y = sin( longitude ) * sin( latitude )
z = cos( latitude )

From 3D coordinates on a unit sphere to latitude and longitude.

latitude = acos( z )
longitude = atan2( x, y )

However these are not reversible and my trigonometry is not what it should be.

1
algorithmgeometrylatitude-longitudecoordinate-transformation
Many computer languages have an atan2(y, x) function, which may be what you need. - Andrew Morton
Yes, but that doesn't solve the inherent reversibility problem. - Chase
What is the reversibility problem you are seeing? Atan2 returns an angle in the correct quadrant, whereas atan won't necessarily do that. - Andrew Morton
What do you mean that these are not reversible? You just reversed the conversion from xyz to lat/long yourself in the question. The only detail is the fact that atan won't return the angle in the correct quadrant, which is what atan2 is for. - SirGuy
Edited the algorithm, but the problem remains. Try it with a latitude of 0 and a longitude of say 1 (which is about 57 degrees). The first produces 0,0,1, and feeding 0,0,1 into the second produces 0,0. Not reversible. - Chase

1 Answers

2
votes

Converting from lat/long to xyz is always possible, but going from xyz to lat/long fails when sin(lat) == 0. There is no solution to this in lat/long space so just stay away from it.
Other than that your formula just has a small error where atan2 takes y then x instead of x then y.