What is remote sensing?

Remote sensing is using information from satellite or aerial imagery to make observations about features in the images. This can include observations in our natural world, the anthropogenic world, or the atmosphere. These observations have many uses in the scientific world.

How do you get information out of images from space?

Every pixel of satellite and aerial imagery contains data values. Different sensors record data from different intervals of the wavelength spectrum. Some sensors record data in intervals, or “bands” which are in the visible wavelength spectrum, while many sensors record other information, such as thermal radiation.

Why do we care about this information?

Comparing two images from different time periods, such as before and after an earthquake, can help us detect different changes such as a bridge or building collapse. We can also evaluate post-event imagery to classify different effects of the disaster based on the image’s spectral properties. This can be important information for first responders as they prioritize evacuation routes or search for casualties.

Are data from satellite images always accurate?

There are oftentimes corrections that need to be made to satellite imagery, such as correcting for atmospheric distortions to the data. This and other corrections are known as “pre-processing” and is done before analyzing the data.

How do we get satellite images?

Satellite images, or aerial images from planes, can be obtained from a variety of sources. One source which our group has used for more than a year is the xBD dataset from the xView2 challenge, established by the Defense Innovation Unit. xView2 provided a high-resolution dataset of pre- and post-disaster imagery covering more than 45,000 sq km and 850,000 buildings for six types of natural disasters across 15 countries.

What else is in these images?

In the xView2 dataset, buildings are manually classified as no damage, minor damage, major damage, or destroyed based on a visual assessment. This scale was established to provide a consistent damage ranking methodology between disaster types and organizations. Work in our group will focus on automating the data classification system by identifying buildings in satellite images followed by classifying them into an appropriate damage category.

Are all images the same?

Pictures from different datasets will have different levels of detail, known as resolution. Spatial resolution refers to the number of pixels per square unit area of a picture. Temporal resolution refers to the frequency with which images are taken of a certain area. These images can be compared to see what changes occurred during the time period between them. Spectral resolution is the amount of detail covered by a spectral band. High spectral resolution refers to a sensor which can differentiate between wavelengths in a narrow width of a spectral band. In all cases, higher resolution generally refers to more detail. Conducting remote sensing processes on more detailed images sometimes provides better results, leading to resolution increases with technological advancements. However, higher resolution images can take up more storage and processing power and sometimes do not provide better results. Therefore, the resolution of image required will vary by project.

For more information, check out What is Remote Sensing? The Definitive Guide.