NCL feature detection

A simple feature detection algorithm written in FORTRAN with NCL wrapper.

Prerequisite	Version
gfortran	>= 4.8
NCL	tested with 6.3.0
fplus	>= 0.1

Compile

make

make will automatically download the fplus preprocessor from https://github.com/r5r3/fplus

Test case

make test

The test uses the input file etc/test-1.nc. The content is a two dimensional array containing the word "Test" and a border around this word. The background has the value 0.0 the word and the border have values around 1.0.

Input data:

Output data:

NCL API

At first, you have to load the external library:

external LIBFEATURE "./libfeature.so"

then you can call the procedure detect_features:

procedure detect_features(
    nx              : integer
    ny              : integer
    input_data  [2] : float
    output_data [2] : float
    threshold       : float
    fill_value      : float
    )

Arguments:

• nx, ny: Dimensions of input and output data
• input_data: The input data array, it is allowed to contain missing values
• output_data: The output data array.
  • it contains missing values at points where the input array contains missing values
  • is zero at grid points without detected feature
  • is 1, 2, ... for grid points that belong to a feature (e.g. all grid points that belong to feature 2 the value 2 is returned)
• threshold: grid points with larger values are considered to belong to a feature
• fill_value: the @_FillValue used for input AND output data.

Have a look at the example!