最接近点对的算法java,用于在有限集中找到最接近点的高效算法

I have a list L of ~30k locations (written as longitude/latitude pairs), and a list E of ~1m events (with locations written as longitude/latitude pairs), each of which occurs at a point in L. I want to tag each event in E with its corresponding location in L. But the cooordinates in L and E are rounded differently—E to five decimal places, L to thirteen—so ostensibly identical coordinates can actually differ by ~10^-5 degrees, or ~1 meter. (Points in L are separated by at least ~10 m.)

I thus need the nearest point in L to every point of E; the obvious O(|L||E|) brute-force algorithm is too slow. L is small enough compared to E that algorithms that preprocess L and amortize the preprocessing time over E are fine. Is this a well-studied problem? The links that I can find are for related but distinct problems, like finding the minimal distance between a pair of points in one set.

Possibly relevant: Voronoi diagrams, though I can't see how preprocessing L into a Voronoi diagram would save me computational time.

解决方案

From your description:

Each point in E is identical to a point in L rounded to five decimal places. Thus, each point in E deviates up to ~1 meter from its match in L.

Points in L are separated by at least ~10 meter.

Solution: Round coordinates in L to the same precision of E and match equal pairs.

Explanation: Rounding effectively maps each point to its nearest neighbor on a square grid. As long as the grid resolution (~1 meter when rounding to five decimals) is lower than half the minimal distance between two points in L (~10/2 meters), you are good to go.

For maximum performance and exploiting |L| << |E|, you might want to build a hash map over the rounded coordinates in L and match each point in E in near constant time. Total runtime would then be limited by O(|E|).