Python求和地理半径内的lat / lon点和网格的总和

Question

Basically, my experimental program is trying to find the number of points that fall within a (eg, 50km) radius of a valid point at a given time. My data is structured (but I can restructure if need-be) in three separate arrays such:

1_LAT,1_LON,1_TIM

Where 1_LAT,1_LON,1_TIM all contain roughly ~250 values corresponding to Latitude, Longitude (decimal degrees), and time respectively.

I have 20 sets of these arrays (ie, 1_LAT,1_LON,1_TIM...20_LAT,20_LON,20_TIM).

Here is what I would like to accomplish:

1) Figure out the number of lat/lon sets that fall within a particular radius of each set. For example, how many points fall within a 50km radius of 1_LAT,1_LON at the valid time of 1_TIM from the other 19 sets of points. I would then like to iterate through each valid time to figure out the number of points in the valid radius at each specific point and valid time.

I have attached a picture below to help visually describe.

The black squares represent all the points in the LAT_1/LON_1 arrays. The blue squares represent all the points in the LAT_n/LAT_n arrays.

I would like to count the number of points in each radius at each valid time for each set of lat/lon arrays. The final display would be a summed raster or meshgrid of the denisty (ie, number of counts / 20) for each grid spot on a geographic basemap image.

I have a feeling that a KDEtree may be the best way to accomplish this, but I have little/no experience with such. Any ideas or suggestions would be greatly appreciated.

Answer 1

You would do something like the following... First, group your (x, y) coordinates for each group in a single points_x array:

points_1 = np.column_stack((LAT_1, LON_1))
...
points_n = np.column_stack((LAT_n, LON_n))

It may be a good idea to store them in a list of arrays:

points = [point_1, points_2, ..., points_n]

Now, make a kdTree out of each set of points:

import scipy.spatial as spsp
kdtrees = [spsp.cKdTree(p) for p in point]

And you are ready to go. If you now run the following code:

r = whatever_your_threshold_value_is
points_within_r = np.zeros((len(kdtrees), len(kdtrees)), dtype=np.int)
for j in xrange(len(kdtrees)):
    for k in xrange(j+1, len(kdtrees)):
        points_within_r[j, k] = kdtrees[j].count_neighbors(kdtrees[k], r, 2)
points_within_r = points_within_r + points_within_r.T

You should now find that points_within_r[j, k] holds how many points in points_j are within radius r of a point in points_k .

Keep in mind that distances here are the euclidean distance of the coordinates, disregarding the fact that what they measure are spherical angles.