Re: [boost] [gsoc17]The implement of Programming competency test From Ruoyun Jing

Subject: Re: [boost] [gsoc17]The implement of Programming competency test From Ruoyun Jing
From: Ruoyun Jing (jingry0321_at_[hidden])
Date: 2017-03-24 16:46:57

Hi Adam and Vissarion,

During these days I discussed the ideas of project 1 and begin to write my
proposal. Here are what I learn about these project.

1.For the approach of compute the 3D cartesian distances first and if these
numbers are not "far enough" then fall back to expensive geographic
distance calculation otherwise return the result by comparing those numbers
obtained by less expensive cartesian compuation.

The main problems is declare the "far enough". To figure out this problem,
I would implement the method, test the result between function
compare_distance and this method, then adjust parameter to limit deviation
in an acceptable range(maybe it would be 1e-6 correapongs, the accurate
value should be discussed by specific condition)

2.For the approach of perform a local spheroid approximation and return the
2D distance.

The main problem is to test the acceptable range of using spheroid
approximation. The solution is same as answer above.

3.For the approach of getting 2 intersections of plane (which would be
ellipse) with the help given points and compute the respective distances
using those ellipses.

In my understand of these approach, it means we need to get cross-section
through the center of ellipsoid and the geodesic segments. But I am not
sure that the geodesic segments of two points whether can get a
cross-section through the center of ellipsoid, and whether the
cross-section is ellipse that we can calculate the sectorial area easy to
compare. After check out the geodesic knowledge from differential geometry
for a long time. I got some prove, but I think it isn't very reliable.

4.I have an reliable approach is that we could use the algorithms in "map
projection" just like Lambert Conformal Conic, Gauss–Krüger projection,
Mercator projection...etc,

Those projections make the point on 3D to 2D with reliable functions,
although most of these projections use longitude and latitude, we could
transform our cartisian coordinate into Spherical coordinate then transform
them into longitude and latitude then use the projection to approximately
calculate.

All of above is what I discussed during these days, Could you please
provide me some advise about them so I can strenthen my proposal?And if
there has any problems about my poor English description please tell me, I
will try my best to explain more clearly:)

Thanks all for your reading!

Looking forward from you.:)

Ruoyun
-- 
Northwest University of China
Software Engineering
jingry0321_at_[hidden] <jingry0321_at_[hidden]>