Announcement

Check out these presentations from DIL postdocs for more inspiration at 4pm after class!

Reading / Tech

Please review our very own project pages!

Tentative plan

First, Catherine Burton from Urthecast will give us a brief overview of earth observation satellites, remote sensing image processing, machine learning algorithms, APIs and the advent of big data analytics and unstructured data.

Then we’ll have class review!

Then, we’ll have a lively review of the semester. If a team can’t make it for the final presentations, this is an alternative date. Teams on deck so far:

• CEGA-trace
• The Carpool Project

The carpool project

• has been working on this for a year
• charge riders to get from a to b and pay riders
• different scales
• mobile product was also designed
• hit a wall when thinking of the end-user service

Data project evolved; challenges:

• how do we enlist people and collect mobility data -> solution: development of a userguide
• app needs to be downloaded
• user needs to understand it is private, all about saving money, time and the planet!

Currently, the data collection has to be done as a 2-step process:

• people have to download the moves app

• people have to export data to local device then send it over —> Reprocessing needed to be done

• How do we figure out that people have common routes?
• configured a time window (within 10, 15 30 minute intervals)
• then subclassifications within walking distance/time to walk
• then take longest trip from subclassifications, assume they start from the same location and make the classification more specific

Overall goal is to reduce competitive ride-sharing apps prices by reducing ride costs to a fraction of the competitive prices, even if that means walking 20 minutes for the rider.

Earth observation and big data analytics (Catherine Burton)

Background in geospatial analytics “Urthecast” : They have 4 satellites, including a color live camera

Remote sensing: the process of obtaining information about an object and gathering data from a sensor from a distance

The technology started during the war, cameras were strapped to pigeons or balloons

1946: first image taken from a rocket 1960: first meteorological satellite 1960-72: spy satellite program 1972: first civilian land satellite 1999: first high resolution image satellite

Different technologies

NASA’s EO (earth observation) system

• LOTS of satellites
• images can found on earth explorer; any given location every 16 days

Aerial photography:

• lots of pictures for one location
• VERY high resolution (3 inches/pixel!)

Kite photography:

• pictures need to be patched together into a final image
• VERY cheap
• Drones are the new thing

LiDAR

• creates a 3D point cloud, flown from a plane
• uses light data

SAR (synthetic aperture radar)

• advantage: can go through clouds
• sound-based (more like a sonar)
• “holy grail” application of SAR: being able to find ships

Measuring the Earth with sensors

• Natural entities can be measured using different electromagnetic frequencies
• satellite can be trained to recognize different wavelengths of the EM spectrum
• vegetation health, water levels, melting ice caps etc. can all be measured!

how to visually interpret satellite imagery?

• each raw image has pixels, each with an intensity value
• supervised and unsupervised analysis
• tone and color
• size and shape
• texture and pattern
• location and features

The four V’s of big data:

Volume Velocity Variety Veracity

Geographic Information Systems (GIS)

• information system that integrates, stores, edits, analyzes, shares and displays geographic information
• tools that allow users to create interactive queries
• Big data on the other hand is a broad term for data sets so large or complex that traditional data processing applications are inadequate

Big data and IoT have been merging together

Pattern recognition: algorithms apply some filters to an image and apply morphological recognition Machine learning: algorithms aimed at prediction of data beyond recognition only Deep learning: algorithms extract high level complex abstractions as data representations through a hierarchical learning process