Recommendation Algorithms Powering Amazon, Netflix Can Also Improve Satellite Imagery

Algorithms that help consumers decide what to stream or buy online can do more than predict customer habits: They can help satellites see Earth better, according to research by Rutgers.

Optical satellites lose sight of the Earth’s surface when it’s covered in clouds, and researchers have long relied on inaccurate tools to fill in blind spots, especially along coastlines. By adapting a recommendation algorithm originally designed for Netflix, Ruo-Qian (Roger) Wangassistant professor of civil and environmental engineering at the Rutgers School of Engineering, has created a more accurate and faster system for predicting cloudscapes in coastal areas than conventional data-filling tools.

The conclusions are published in the International Journal of Applied Earth Observation and Geoinformation.

“Electronic service platforms like Alibaba and Amazon use recommender systems, which leverage large datasets to provide personalized product recommendations to help customers make decisions,” Wang said. “Interestingly, the way recommender systems process data is not unlike the process of predicting cloud-obscured coastal landscapes.”

In the open ocean, cloud-filling algorithms used in remote sensing measure continuous data — such as water temperature, color and algal content — to make predictions about what’s hidden. But these solutions fail along the coast, where “errors are amplified due to increased cloud cover, vegetation and other variables,” Wang said, adding that recommender systems “could do a better job in that regard”.

To test his hypothesis, Wang built a cloud-filling model on the work of Simon Funk, a software developer that won a Netflix recommendation tool contest. The algorithm, called Funk-SVD, plots consumer reviews on a matrix. This data is then used to predict the viewing habits of users who have not registered reviews.

It’s a similar process for cloud filling: each coordinate on a map is represented by a pixel on a photograph and that pixel can be water or land, with clouds representing unrecorded data. Wang’s adaptation of Funk-SVD makes guesses about what’s under the clouds based on other data points.

Using an image database of 258 images derived from Landsat missions in Delaware Bay, Wang formed Funk-SVD to complete the cloud fill scheme. Its solution was more accurate than the most widely used cloud population tool, Data-Interpolating Empirical Orthogonal Functions (DINEOF), and achieved similar accuracy to another popular tool, Datawig, which is powered by machine learning. . While Datawig uses a lot of computing power and can take days to process, Wang’s solution took 30 seconds.

Wang said his solution has many applications for long-term Earth observation. The method could be used to measure agricultural production, for example, or to map the rate of urbanization over large areas. It can also do it faster and cheaper than conventional methods.

“Any general land use change could be monitored using this tool,” Wang said.

/Public release. This material from the original organization/authors may be ad hoc in nature, edited for clarity, style and length. The views and opinions expressed are those of the author or authors.View Full here.

Sharon D. Cole