在python np.array中对连续值进行分组

Question

I have got an array (500*490) long with data. It's actually raw data from a radar. I am planning on creating some sort of tracking, which finds showers & thunderstorms, based on their value in the array. The values range from 0 tot 75.

In order to find the showers & thunderstorms based on a treshold value, I have created an array from a raw h5 image, using h5py to read the file & then create an array out of it.

import numpy as np
from numpy import inf
import pylab as pl
import math
import h5py
import scipy
import scipy.spatial

np.set_printoptions(threshold=10)

# read data
# *********
print "Reading & converting data data (mm/u - dbz)"
f = h5py.File('test.h5','r')
data = f.get('image1/image_data')
data_as_array = np.array(data)
datadbz = 20 * np.ma.log10(data_as_array)
print "*** Data read & converted"
print "raw array data = 'data_as_array'"
print "dbz array data = 'datadbz'"
print ""

# print raw data
# **************
print "Array from h5-file"
print "******************"
print data_as_array
print "*"
print ""

datadbz[datadbz == -inf] = 0
datadbz[datadbz == 96.32946777] = 0
datadbz[datadbz < 40] = 0

print "Array in dbz"
print "************"
print datadbz
print ""

# > 40 dbz
pixelarray = np.array(zip(*np.where(datadbz > 40)))
print "Array wich cores (> 40 dbz)"
print "***************************"
print pixelarray
print "*"
print ""

# > 55 dbz
hailarray = np.array(zip(*np.where(datadbz > 55)))
print "Array with cores (> 55 dbz)"
print "***************************"
print hailarray
print "*"
print ""

The 2 arrays I need to use are the array "pixelarray", which looks like the following snippet in the python shell

Array wich cores (> 40 dbz)
***************************
[[  5 106]
 [  5 107]
 [  6 105]
 ..., 
 [440 270]
 [440 271]
 [489 151]]
*

and the other one is "hailarray", looking like this in the python shell

Array with cores (> 55 dbz)
***************************
[[  7 104]
 [  8 103]
 [ 27 194]
 ..., 
 [433 272]
 [435 272]
 [438 271]]
*

The point where I am stuck is the following...

Provided this is the radar image in question:

radarimage

The array contains clutter, which should be discarded & ignored. Those are the little splots & dots on the image. What I do need however, are the storms / showers.

In order to do that I need to figure out how to distinguish the clutter (random spots (values in the array) from the real echo's in the image (data ranges in the array)

Since the array I use looks like this, where each duo of values represent ax pixel & y pixel of the detection

[[  5 106]
 [  5 107]
 [  6 105]
 ..., 
 [440 270]
 [440 271]
 [489 151]]
*

I was thinking about grouping the pixelranges together. I know of a possibility to use "consecutive values" in an array and group them together, but I havent found a way to do that properly, because my array looks more like a matrix of some kind.

Since I am not an expert in python I would have like to be pointed in the right direction as to how I can go about this. Once I have done this It's "only" a matter of finding which groups have a larger dimension than a certain treshold value, so I can get rid of the litte clutter spots & just be left with the storms themselves.

I can then use the center of those larger groups as the center of the storms & the tracking should be complete.

So if anybody could help me set up the search for consecutive pixel ranges in the detection arrays I would be forever grateful, because I can't help the feeling of trying to invent the wheel here.

Answer 1

If you have enough memory, you can build a dense representation of the image, and use you arrays to set the corresponding pixels there. Then you can use an erosion filter on the image to eliminate the noise (like SciPy's binary_erosion )