Band

This module contains the classes to handle bands - that is, single layer raster images, with georeference.

The Band class represents an object in the filesystem, where the raster is not read and only the handle and metadata are stored. The BandSample contains the in-memory raster data that can be operated.

Band

class aeronet.dataset.raster.Band(fp)

Bases: aeronet.dataset.raster.geoobject.GeoObject

Filesystem object Band - Rasterio DatasetReader wrapper.

The Band provides access to a georeferenced raster file placed in the filesystem. On creation the Band opens the file for reading and stores all the necessary metadata and allows to read the raster data on request.

The majority of properties are inherited from rasterio DatasetReader.

Any file format supported by GDAL drivers can be read.

Parameters:fp – full path to the raster file

crs

CRS - coordindate reference system of the band; in the file

transform

Transform matrix as the affine.Affine object. This transform maps pixel row/column coordinates to coordinates in the dataset’s coordinate reference system.

affine.identity is returned if if the file does not contain transform

nodata

Band nodata value, type depends on the image dtype; None if the nodata value is not specified

res

Spatial resolution (x_res, y_res) of the Band in X and Y directions of the georeferenced coordinate system, derived from tranaform. Normally is equal to (transform.a, - transform.e)

width

Width of the raster data object in pixels

height

Height of the raster data object in pixels

count

By design of the aeronetlib, should be always 1. A Band can be created from image of any channel count, but only the first band can be read. If you need to work with multi-channel image, use aeronet.converters.split.split() to get the one-channel images.

Returns:(int) number of the bands in the image.

shape

The raster dimension as a Tuple (height, width)

name

Name of the file associated with the Band, without extension and the directory path

bounds

Georeferenced bounds - bounding box in the CRS of the image, based on transform and shape

Returns:BoundingBox object: (left, bottom, right, top)

meta

The basic metadata of the associated rasterio DatasetReader

dtype

Numerical type of the data stored in raster, according to numpy.dtype

numpy()

Read all the raster data into memory as a numpy array

Returns:numpy array containing the whole Band raster data

same(other)

Compare if samples have same resolution, crs and shape.

This means that the samples represent the same territory (like different spectral channels of the same image) and can be processed together as collection.

Parameters:other – GeoObject to compare with Returns:True if the objects match in shape, crs, transform, False otherwise

sample(y, x, height, width, **kwargs)

Read sample of the Band to memory.

The sample is defined by its size and position in the raster, without respect to the georeference. In case if the sample coordinates spread out of the image boundaries, the image is padded with nodata value.

Parameters:

• x – pixel horizontal coordinate of left top corner of the sample
• y – pixel vertical coordinate of left top corner of the sample
• width – spatial dimension of sample in pixels
• height – spatial dimension of sample in pixels

Returns:

a new BandSample containing the specified spatial subset of the band

resample(dst_res, fp=None, interpolation='nearest')

Change spatial resolution of the band. It does not alter the existing file, and creates a new file either in the specified location or a temporary file

It is based on rasterio.warp.reproject, see for more variants of interpolation.

Parameters:

• dst_res (Tuple[float, float]) – new resoluton, georeferenced pixel size for the new band
• fp (str) – a filename for the new resampled band. If none, a temporary file is created
• interpolation – interpolation type as in rasterio, nearest, bilinear, cubic, lanzsos or others

Returns:

a new resampled Band.

reproject(dst_crs, fp=None, interpolation='nearest')

Change coordinate system (projection) of the band. It does not alter the existing file, and creates a new file either in the specified location or a temporary file.

The band ground sampling distance is not changed, however the resolution may change due to the new coordinate system It is based on rasterio.warp.reproject, see for more variants of interpolation.

Parameters:

• dst_crs – new CRS, may be in any form acceptable by rasterio, for example as EPSG code, string, CRS object; if dst_crs == utm, the appropriate UTM zone is used according to the center of the image
• fp (str) – a filename for the new resampled band. If none, a temporary file is created
• interpolation – interpolation type as in rasterio, nearest, bilinear, cubic, lanzsos or others

Returns:

a new reprojected Band

reproject_to(other: aeronet.dataset.raster.geoobject.GeoObject, fp=None, interpolation='nearest')

Reprojects and resamples the band to match exactly the other.

This function ensures that the raster size, crs and transform will be the same, allowing them to be merged into one BandCollection. If the intial raster exceeds the other in coverage, it will be cut, and if it is insufficient or displaced, it will be zero-padded.

It aims to overpass the rounding problem which may cause an image to be misaligned with itself after a different series of transforms.

If the images are far from each other, the warning will be shown, because the raster may be zero due to severe misalignment.

Parameters:

• other (GeoObject) – the Band with the parameters to fit to
• fp (str) – a filename for the new resampled band. If none, a temporary file is created
• interpolation – interpolation type as in rasterio, nearest, bilinear, cubic, lanzsos or others.

Returns:

a new reprojected and resampled Band

reproject_to_utm(fp=None, interpolation='nearest')

Alias of Band.reproject method with automatic Band utm zone determining

generate_samples(width, height)

A generator for sequential sampling of the whole band, used for the windowed reading of the raster. It allows to handle and process large files without reading them at once in the memory.

Parameters:

• width (int) – dimension of sample in pixels and step along X axis
• height (int) – dimension of sample in pixels and step along Y axis

Yields:

BandSample – sequential samples of the specified dimensions

profile

A joint representation of the main properties

Returns:{ ‘crs’: crs, ‘nodata’: nodata, ‘transform’: transform } Return type:Dict

BandSample

class aeronet.dataset.raster.BandSample(name, raster, crs, transform, nodata=0)

Bases: aeronet.dataset.raster.geoobject.GeoObject

A wrapper over numpy array representing an in-memory georeferenced raster image.

It implements all the interfaces of the GeoObject, and stores the raster data in memory

Parameters:

• name (str) – a name of the sample, which is used as a defaule name for saving to file
• raster (np.array) – the raster data
• crs – geographical coordinate reference system, as CRS or string representation
• transform (Affine) – affine transform for the
• nodata – the pixels with this value in raster should be ignored

width

Width of the raster data object in pixels

height

Height of the raster data object in pixels

count

(int) number of channels/bands.

shape

The raster dimension as a Tuple (height, width)

dtype

Data type of the associated numpy array

res

Resolution (or ground sampling distance) along X and Y axes in units of the CRS. Tuple (x_resolution, y_resolution)

transform

Transform matrix as the affine.Affine object. This transform maps pixel row/column coordinates to coordinates in the dataset’s coordinate reference system.

crs

Geographic coordinate reference system of the object Returns a rasterio.CRS

nodata

The value that should be interpreted as ‘No data’. May be None or a value within dtype range

bounds

Georeferenced bounds - bounding box in the CRS of the image, based on transform and shape

Returns:BoundingBox object: (left, bottom, right, top)

name

name of the sample, is used as a base filename when saving to file

classmethod from_file(fp)

Reads the raster data directly from the file. File must have only one channel. If you need to read multi-channel file, use aeronet.converters.split.split() first

Parameters:fp – full path to the file Returns:a new BandSample object

same(other)

Compare if samples have same resolution, crs and shape.

This means that the samples represent the same territory (like different spectral channels of the same image) and can be processed together as collection.

Parameters:other – GeoObject to compare with Returns:True if the objects match in shape, crs, transform, False otherwise

save(directory, ext='.tif', **kwargs)

Saves the raster data to a new geotiff file; the filename is derived from this BandSample name. If file exists, it will be overwritten.

Parameters:

• directory – folder to save the file
• ext – file extension; as now only GTiff driver is used, it should match tif, tiff, TIF or TIFF.
• kwargs – other keywords arguments to be passed to rasterio.open .

sample(y, x, height, width)

Subsample of the Sample with specified dimensions and position within the raster:

Parameters:

• x (int) – horizontal pixel coordinate of left top corner
• y (int) – vertical pixel coordinate of left top corner
• width (int) – spatial x-dimension of sample in pixels
• height (int) – spatial y-dimension of sample in pixels

Returns:

a new BandSample object

reproject(dst_crs, interpolation='nearest')

Change coordinate system (projection) of the band. It returns a new BandSample and does not alter the current object

It is based on rasterio.warp.reproject, see for more variants of interpolation.

Parameters:

• dst_crs – new CRS, may be in any form acceptable by rasterio, for example as EPSG code, string, CRS object; if dst_crs == utm, the appropriate UTM zone is used according to the center of the image
• interpolation – interpolation type as in rasterio, nearest, bilinear, cubic, lanzsos or others

Returns:

a new instance with changed CRS.

Return type:

BandSample

reproject_to_utm(interpolation='nearest')

Alias of BandSample.reproject method with automatic Band utm zone determining

resample(dst_res=None, dst_shape=None, interpolation='nearest')

Change spatial resolution of the sample, resizing the raster according to the new resolution. dst_res should be specified, otherwise the destination transform will be equal to the source. If dst_shape is not specified, it is calculated from dst_res, but it can be specified to override it and get the desired output shape

It is based on rasterio.warp.reproject, see for more variants of interpolation.

Parameters:

• dst_res (Tuple[float, float]) – new resoluton, georeferenced pixel size for the new band
• dst_shape – new shape of the resampled raster, can override calculated new shape
• interpolation – interpolation type as in rasterio, nearest, bilinear, cubic, lanzsos or others

Returns:

a new resampled BandSample.

numpy()

A numpy representation of the raster, without metadata

Returns: Sample’s raster data as a numpy array

generate_samples(width, height)

A generator for sequential sampling of the whole sample, similar to Band, used for the windowed processing of the raster data.

Parameters:

• width (int) – dimension of sample in pixels and step along X axis
• height (int) – dimension of sample in pixels and step along Y axis

Yields:

BandSample – sequential samples of the specified dimensions

profile

A joint representation of the main properties

Returns:{ ‘crs’: crs, ‘nodata’: nodata, ‘transform’: transform } Return type:Dict