Manuel Robalinho: Working multispectral images

Some basic principles of the electromagnetic spectrum

The electromagnetic spectrum is composed of a range of different wavelengths or “colors” of
light energy. A spectral remote sensing instrument collects light energy within specific regions of the
electromagnetic spectrum. Each region in the spectrum is referred to as a band.

Space vs Airborne Data

Remote sensing data can be collected from the ground, the air (using airplanes or helicopters) or
from space. We can imagine that data that are collected from space are often of a lower spatial
resolution than data collected from an airplane. The tradeoff however is that data collected from a
satellite often offers better (up to global) coverage.

For example the Landsat 8 satellite has a 16 day repeat cycle for the entire globe. This means that
we can find a new image for an area, every 16 days. It takes a lot of time and financial resources to
collect airborne data. Thus data are often only available for smaller geographic areas.

Bands and Wavelengths

When talking about spectral data, we talk from both, the electromagnetic spectrum and image bands.
Spectral remote sensing data are collected by powerful camera-like instruments known as imaging
spectrometers. Imaging spectrometers collect reflected light energy in “bands.”

A raster can contain one or more bands. We can use the raster function to import one single band from
a single OR multi-band raster. Source: Colin Williams,https://www.earthdatascience.org

A band represents a segment of the electromagnetic spectrum. For example, the wavelength values
between 800 nanometers (nm) and 850 nm might be one band captured by an imaging spectrometer.
The imaging spectrometer collects reflected light energy within a pixel area on the ground.
Since an imaging spectrometer collects many different types of light — for each pixel the amount of
light energy for each type of light or band will be recorded. So, for example, a camera records the
amount of red, green and blue light for each pixel.

Often when we work with a multispectral dataset, the band information is reported as the center
wavelength value. This value represents the center point value of the wavelengths represented in that
band. Thus in a band spanning 800–850 nm, the center would be 825 nm.

Imaging spectrometers collect reflected light information within defined bands or regions of the
electromagnetic spectrum.https://www.earthdatascience.org/

Spectral Resolution

The spectral resolution of a dataset that has more than one band, refers to the spectral width of each
band in the dataset. While a general spectral resolution of the sensor is often provided, not all sensors
collect information within bands of uniform widths.
Spatial Resolution

The spatial resolution of a raster represents the area on the ground that each pixel covers. If you have
smaller pixels in a raster the data will appear more “detailed.” If you have large pixels in a raster,
the data will appear more coarse or “fuzzy.”

https://www.earthdatascience.org/

What is Multispectral Imagery?

One type of multispectral imagery that is familiar to many of us is a color image. A color image
consists of three bands: red, green, and blue. Each band represents light reflected from the red, green
or blue portions of the electromagnetic spectrum. The pixel brightness for each band, when composited
creates the colors that you see in an image. These colors are the ones your eyes can see within the
visible portion of the electromagnetic spectrum. In a Multispectral image we can plot each band of
a multi-band image individually using a grayscale color gradient.

Near infrared light reflects strongly off of vegetation.https://www.earthdatascience.org

The multispectral images are used in several fields, ranging from medicine, agriculture, maps,
definition of terrain relief, mineral exploration, or study of composition of metals, for example.
My goal is to use it in the classification of scrap, using the study of its spectrum for real time,
and using deep learning, to do its classification.
Let’s do some examples of exploring multispectral images, using python and Jupyter Colaboratory.