This week Planet announced the beta version of Planet Analytics, a tool using machine learning to process through the six terabytes of daily image data it collects from its fleet of 200 plus satellites. The new solution is part of Planet’s larger vision of a “Queryable Earth” making imagery accessible and actionable to monitor changes around the globe.
The newly announced Planet Analytics solution uses machine learning to process the company’s global, daily satellite imagery into information feeds that detect and classify objects, identify geographic features, and monitor change over time. Feeds can integrate into existing workflows and give users insights into places they care about.
Utilizing and analyzing earth observation imagery has been a painstaking task until recently, requiring a combination of image calibration processing to compensate for angle, time of day, weather and other factors before manual visual (human) analysis can start. A number of existing and startup companies are applying machine learning to satellite imagery to automate key tasks, including the ability to identify changes and find geographic features.
Planet Analytics is using machine learning to automatically identify objects of interest within specific areas, such as ships and planes, as well as extract geographic features of interest, like buildings and construction activity, road networks, and land cover. For anyone involved in moving goods, involved in urban planning, or investing in real estate, Planet Analytics will provide a new channel of information.
This week’s beta release provides five new analytical feeds over an area of interest (AOI). Building detection will identify the creation of buildings in an AOI to create geographic information layers, update maps and charts, track urban development, and track changes before and after natural disasters. Road network detection will identify the creation of road networks in an AOI while aircraft identification will monitor key areas, such as military and commercial airports, to provide insight into the movement and volume of aircraft for national security and economic activity. Marine vessel identification will monitor the ocean and identify vessels to track “pattern of life” activity and even track specific vessels for law and policy enforcement. While not as exciting, deforestation detection will monitor land cover and land change usage for tracking deforestation and validating sustainable land use commitments made by corporations and governments.
Planet’s key advantage is its intake of over six (6) terabytes of global data per day along with its existing catalog of over 6 petabytes (PB) of imagery. Over time, Planet’s machine learning models will improve and, in the future, be able to detect new categories of objects. The feeds can be plugged into existing workflows — presumably through the use of APIs — so users can layer in processed imagery and other information from Planet Analytics into in-house and third-party data to get a more integrated view of the world. Co-founder and CEO Will Marshall says Planet Analytics will eventually be able to predict changes through a machine-learning index based on historical data and trends.