Detecting Clouds: How Machine Learning Improves Field Health Maps

by Xiaoyuan Yang, Remote Sensing Science Lead, Climate LLC

August 6, 2019

Share:

Has this ever happened to you? You pull up a field health image and find that your scouting map is almost completely red. You go out and scout the field, expecting devastation and to your surprise, everything is perfectly fine. This is a classic example of cloud blockage - clouds are quite literally getting in the way of field health images.   

Unclouding the View

Unclouding the view
Above is an instance of cloud blockage. The image on the left is the real-life cloud image. The image on the right is the FieldView™ map prior to our current cloud detection technology. The large red spot is where the cloud is either casting a shadow or obscuring the view of the satellite taking the picture. 

When clouds obstruct the view of satellites, the technology that analyzes images to produce scouting maps and other data interprets this as a field health problem. This results in misleading field imagery. As a data scientist, the last thing I ever want to do is provide inaccurate data. So my colleagues and I set out to solve this issue by creating better, more sophisticated technology. 

"We developed a way for machine learning to detect and identify cloud blockage."

We developed a way for machine learning to detect and identify cloud blockage. First, we needed to help FieldView learn what clouds look like. Our field health imagery tool was built for analyzing crop imagery. So in a sense, we had to “teach” or “train” FieldView to identify a cloud. This process is commonly known as “machine learning.” But how do you teach technology? 

As people, we look at a picture as one object. A computer, on the other hand, will process an image as a collection of pixels. When FieldView receives a field health image, it scans every single pixel in the picture and looks for specific colors and patterns. When it sees a white cloud or dark shadow, it highlights this as a potential field health issue, when in actuality, it may not be.   

Our solution was to create what’s called a “classifier” to automatically separate the parts of the picture that show signs of cloud blockage. We taught the classifier to do this using hundreds of thousands of images of clouds of different shapes, sizes and transparency. These images were analyzed across different landscapes and times during the growing season, helping the classifier learn to identify them. This innovation enables FieldView to provide a valuable field health image even if part of the picture was compromised by clouds. 

The image above is a conceptual illustration of how cloud detection technology can improve field imagery. As you slide from left to right, you see how this machine learning model offers a more detailed and accurate field health map.

Using the classifier, FieldView creates scouting and vegetation images that show only the parts of the field that are not compromised by a cloud. Inaccurate data is eliminated, providing you with the most precise depiction of the health of your field based on the available image not affected by cloud cover. However, if the majority of the field image is obscured by cloud cover, FieldView does not provide that image.

We also understand that it’s valuable for you to be able to compare scouting and vegetation images with an actual, unchanged image of the field, so we provide an image we call “true color,” which is the real-world view of your field. You can use the true color image to confirm the cloud cover and find any small or thin clouds that may have escaped our cloud detection. 

Comparing a scouting map alongside true-color imagery allows you to see precisely where cloud cover interference has occurred.

Continuous Evolution

As a company working on the cutting-edge of digital farming, we’re always scrutinizing our methods and looking for new ways to improve. Becoming more exact, predicting more precisely and providing more value to farmers is always top of mind. There are countless variables to consider both in technology and agriculture. We’re dedicated and extremely passionate about reducing uncertainty and risk for farmers whenever we can. 

Much like science and technology at large, the FieldView platform will never truly be “complete.” In our research and development pipeline, we are working on further improvements to our field health imagery using deep learning and artificial intelligence, and we’re creating advancements in the areas of seed selection and placement, disease detection and diagnosis, and more. There will always be opportunities for improvement, and the scientists, engineers and designers behind FieldView are constantly testing, learning and refining digital tools to help you get the most out of every acre. 

 

About The Author

Xiaoyuan is the remote sensing science lead and manager of the data insights science team at Climate LLC. She and her team are specialized in generating advanced agronomic insights at a global scale with remote sensing, weather, and other geospatial datasets through the fusion of data driven and physically based models. She is passionate about earth observation, big data and building digital tools for farmers.