如何使用地理/坐标而不是像素坐标沿曲线提取一维轮廓？

有趣的问题！

我是 EOmaps 的开发人员，EOmaps 是一个旨在简化创建这种自定义地理数据分析小部件的库，所以我很感兴趣：-）

我快速浏览了一下您的数据...这是一个非常奇怪的 netcdf，我不知道是什么定义了您用来识别点的行，所以我只是重用您的代码来提取数据。

这是一个快速解决方案，它使用 KDTree 来识别相对于线点的最近点......它不是 100% 精确，因为它根据以度为单位定义的最大距离而不是地球上的实际距离来识别这些点，但它工作得很好，我认为如果你需要更精确的东西，你应该能够从这里拿走它......

同样重要的是要注意，此方法将向您显示实际数据，而不是由用于显示数据的算法创建的一些插值点！（所以这条线不会很好看，它只会显示最接近这条线的实际数据值）

from pathlib import Path
from itertools import chain

from scipy.spatial import KDTree
import xarray as xar
import pandas as pd
import numpy as np

from eomaps import Maps


# --------------------------------------------- read the data
p = Path(r"GOLD_data-20230319T161539Z-001\GOLD_data")

with xar.open_dataset(p / "GOLD_L1C_CHA_NI1_2018_286_23_10_v04_r01_c01.nc") as ncfile:
    lon, lat = ncfile.REFERENCE_POINT_LON.values, ncfile.REFERENCE_POINT_LAT.values
    wavelength = ncfile.WAVELENGTH.values
    radiance = ncfile.RADIANCE.values

O5s_ids = np.argwhere((134.3 <= wavelength[50,25,:]) & (wavelength[50,25,:] <= 137.7))
O5s = np.array(np.nansum(radiance[:,:,O5s_ids],axis = 2))*.04 # integrate under the peak!
O5s = np.transpose(O5s[:,:,0])
x=lon[9:-8,9:-8]
y=lat[9:-8,9:-8]
z=O5s[9:-8,9:-8].T


Mag = pd.read_csv(p / "mag_cords.csv")
X_dip10s,Y_dip10s = Mag.LON10S.to_numpy(),Mag.LAT10S.to_numpy()
line = np.column_stack((X_dip10s, Y_dip10s))


# --------------------------------------------- create a new map
m = Maps(ax=211, crs=4326)
m.set_extent((-61.8,-20.5,-1.5,24.5))
m.add_feature.preset.coastline()
m.add_feature.preset.ocean()

# plot the data
m.set_data(data=z, x=x, y=y, crs=4326)
m.set_shape.ellipses(radius=0.35)
m.set_classify.UserDefined(bins=np.linspace(0, 20, 11))
m.plot_map(cmap="viridis", vmin=0, vmax=28)
m.add_colorbar(orientation="vertical")

# add a normal matplotlib axis (and share x-limits)
ax = m.f.add_subplot(212)
ax.sharex(m.ax)
# turn off navigation on the normal matplotlib plot
ax.set_navigate(False)

# position axes 
m.apply_layout(
    {'figsize': [6.95, 4.8],
     '0_map': [0.1, 0.44678, 0.5875, 0.53552],
     '1_cb': [0.75, 0.04546, 0.2125, 0.95484],
     '1_cb_histogram_size': 0.8,
     '2_': [0.1, 0.10859, 0.5875, 0.32578]}
    )

# setup a new map-layer (will be used to draw identified points)
m2 = m.new_layer()

# --------------------------------------------- identify points along line
tree = KDTree(np.column_stack((x.ravel(), y.ravel())))

def get_lineoffset(pos, **kwargs):
    # get the offset required to shift the line to the click position
    x, y = line.T
    minxid = np.argmin(abs(pos[0] - x))
    offset = pos[1] - y[minxid]
    return offset

def identify_points(pts, max_dist):
    # identify closest points with respect to the line
    tree2 = KDTree(pts)
    q = tree2.query_ball_tree(tree, max_dist)
    q = list(chain(*[i for i in q if len(i) > 0]))
    coords = tree.data[q].T
    data = z.ravel()[q]
    return coords, data

def indicate_found_points(pts, offset=0, max_dist=1):
    # offset line
    pts = pts + [0, offset]
    # identify closest points
    coords, data = identify_points(pts, max_dist=max_dist)
    
    # plot identified points on the map (replace on each click)
    m2.set_data(data, coords[0], coords[1])
    m2.set_shape.ellipses(radius=0.2)
    m2.plot_map(set_extent=False, zorder=2, fc="r")
    
    # sort found points so we can plot a line
    sortp = np.argsort(coords[0])
    lx, ly = coords[0][sortp], data[sortp]    
    
    # plot a line on the map
    l, = m.ax.plot(*pts.T, c="r", lw=0.5)
    # remove it on next pick
    m.cb.pick.add_temporary_artist(l)
    
    # plot data on normal matpltoib plot
    l2, = ax.plot(lx, ly, lw=0.25, c="k", zorder=0)
    # color points same as in the map
    sc = ax.scatter(lx, ly, c=ly, s=10, norm=m._norm, cmap=m._cbcmap)
    # remove it on next pick
    m.cb.pick.add_temporary_artist(l2)
    m.cb.pick.add_temporary_artist(sc)

    ax.relim()
    ax.autoscale_view()
    m.redraw()

def callback(pos, max_dist=1, **kwargs):
    offset = get_lineoffset(pos)
    indicate_found_points(line, offset=offset, max_dist=max_dist)

# attach the callback to the map
m.cb.pick.attach(callback, max_dist=0.75)