Page 377 - Contributed Paper Session (CPS) - Volume 4

CPS2315 Nele C.
Vast improvements have been made in these fields. The democratization
of data through generous programs like Sentinel and Copernicus makes
satellite imagery and radar data freely available to everyone. The Internet of
Things and explosion of sensors (drones, webcams, video feeds, connected
traffic sensors, etc.) have ensured that we have access to more data than ever
before.
While the data itself is freely available, knowledge of the technology to
convert the data to usable information is not. While we can get free radar
coverage of every point on the earth every two weeks, processing that data is
still complicated. We can unwrap radar data to find minute changes in the
earth’s surface (down to a millimeter), but the knowledge of those change
detection algorithms and how to use them remain a black box – and there is
no foreseeable end to that.
Even with the technology to perform change detection on satellite
imagery, it still requires a human being to interact with the imagery and make
a map. Frequently, this process can take weeks to months to complete, and by
the time the imagery is processed, and a map is created, the information is out
of date. Because making traditional maps takes so long and is so expensive,
we try to make them do too much. Every map has to perform multiple
purposes: land cover, land use, roadway mapping, and topography, to name a
few. This compromises at least the intent of the map, if not the map’s accuracy.
Information that can be derived from the data collected by all these
sensors has great potential. What is needed is a way for domain experts to
build sophisticated, reusable algorithms that can ingest streams of sensor
data. We need a platform that allows data to be plugged in to the platform as
soon as it is collected, and then have the system pull the data through
processing steps so that analyses run automatically and generate updated
maps. These maps would not only allow end users to see the current state of
the land, but to see the entire time series so they can understand the patterns
behind the change and begin to formulate predictions. Users need to see not
only what was and what is, but also what can be.
This technology exists today. It is not a map; it is a Hexagon Smart M.App
— a dynamic information service. By moving from the static map model —
which collects data, analyzes it, and then produces a static printed, digital, or
web-based map — to a dynamic information service, we can not only
automate the process, but we can build job-specific and use-case-specific
maps. Instead of multiple departments sharing a single multi-purpose map,
each department can access its own view of the map containing information
produced from the data specifically for them. Because this Smart M.Apps are
lightweight and quickly produced, they are easy to prototype. Domain experts
can build the map, incorporate feedback from users, and then make the map
accessible to land use departments. As new data comes in, it is fed into the

366 | I S I W S C 2 0 1 9