Introduction

Geoio provides facilities to easily interact with geospatial data. The interactions that are supported include data retrieval, spectral processing, metadata handling, shapefile intersection/extraction, and retrieval of statistical information. Specific attention has been paid to accessing DigitalGlobe data and metadata, but the same facilities in this module can be used to access non-DigitalGlobe data or to build custom processing and metadata handling for other satellite platforms.

Installation

pip install geoio

Dependencies will be handled at install if possible. GDAL is not cleanly installable via pip so should be handled separately (conda, yum, apt-get, etc.). The run dependencies are: gdal, xmltodict, pytz, tzwhere, ephem, numpy, tinytools. Additionally, dgsamples is required for testing and matplotlib must be available for the plotting functions to work.

Note for MAC users: if pip fails for ephem, try installing it directly with conda within the conda virtual environment, i.e.:

conda install ephem

Imports

import geoio

imports the main classes GeoImage and DGImage to the module root.

The GeoImage class is a relatively thin wrapper around gdal that provides a pythonic interface for accessing an arbitrary geospatial image format (generally those supported by gdal plus the DigitalGlobe .TIL format). Operations supported include reading, writing, chipping, reprojecting, and meta data access. The class methods are populated with reasonable defaults and object interfaces, making image operations less painful so that you can get on with the important stuff!

The DGImage class inherits all the capabilites of GeoImage and adds DigitalGlobe meta data handling, spectral processing, and band alias data retrieval. Therefore, it requires that the input image be a valid DigitalGlobe image. This is currently either a .TIL file with the associated meta data files (.IMD and/or .XML) present in the image directory or a .TIF files with an identially named .IMD or .XML file. The metadata is read into an OrderedBunch object (inherited from the tinytools package) attached to the instantiated object.

Quick Start

The geoio classes are best used interactively from within ipython where the relevant pretty print methods can be triggered. Meta data information will be reutrned regardless of the interpreter, but the readability is currently much better in ipython.

The dgsamples repo is used below. However, all the operations below can be run on local data by replacing the dgsamples call with a string to the image location. From exmaples, instead of typing `dgsamples.wv2_longmont_1k.ms`, a local files at `/path/to/imgfile.TIF` can be used.

Using the GeoImage object:

import dgsamples

# Instantiate an image object
img = geoio.GeoImage(dgsamples.bayou_chip.extract_test)  # a TIF file

# Print useful information about the object
img.files
img.meta

# Get numpy array
data = img.get_data()

# Process data and write to new image
newdata = data*2
img.write_img_like_this('/path/to/newfile.TIF',newdata)

Using the DGImage object:

import dgsamples

# Instantiate an image object
img = geoio.DGImage(dgsamples.wv2_longmont_1k.ms)    # a TIL file
# Can also be used directly with a DigitalGlobe TIF file if an XML and/or IMD
# is available with same name as the TIF file.

# Print useful information about the object
img.files
img.meta

# Print the full IMD OrderedBunch object
img.meta_dg.IMD  # tab completeable through the OrderedBunch

# Return an ImgArr (a numpy array with band meta data handling)
data = img.get_data()

# Convert an ImgArr to a pure numpy array
npdata = np.asarray(data)

# Return a pure numpy array
data = img.get_data(meta=False)

# Get specific bands using aliases - see geoio.constants.DG_BAND_ALIASES for
# additional aliases.
data = img.get_data(bands='VIS')

# Get specific bands using band aliases
data = img.get_data(bands=['C','Y'])

# Get image data and convert to TOA reflectance
data = img.get_data(stype='toa')

Plotting

Plotting with the geoio.plotting functions:

import dgsamples

# Instantiate an image object
img = geoio.DGImage(dgsamples.wv2_longmont_1k.ms)  # a TIF file

# Plot the RGB image
geoio.plotting.imshow(img.get_data(bands='RGB'))

# Plot the near-infrared false color image
geoio.plotting.imshow(img.get_data(bands=['N1','G','B']))

# Plotting a histogram of the image bands
geoio.plotting.hist(img.get_data())

# Plotting a histogram of specific bands
geoio.plotting.hist(img.get_data(bands='VIS'))

Spatial Resampling

The geoio module has upsampling and downsmapling code that allows the user to easily resample two images to the same grid for easy multi-image proceesing.

import dgsamples

# Import wv3 images
ms = geoio.DGImage(dgsamples.wv3_longmont_1k.ms)
swir = geoio.DGImage(dgsamples.wv3_longmont_1k.swir)

# Upsample the swir image
swir.upsample_like_that(ms,method='nearest')  # default method is bilinear

# Downsample the ms image
ms.downsample_like_that(swir)  # default method is aggregation

# Or let geoio figure it out
ms.resample_like_that(swir)
swir.resample_like_that(ms)

Iterators

The geoio module also provides several iterators to allow easy access to yield based portions of a raster file.

import dgsamples
ms = geoio.DGImage(dgsamples.wv2_longmont_1k.ms)

# iterate through vector geometries
v = dgsamples.wv2_longmont_1k_vectors.poly_geojson_latlon
[x for x in ms.iter_vector(vector=v,bands='RGB',mask=True)]

# random windows from the image
[x.shape for x in ms.iter_window_random(win_size=[10,10], no_chips=20)]

# iterate through image with evenly spaced windows based on requested stride
[x.shape for x in ms.iter_window(win_size=[10,10], stride=[100,100])