IntelinAir and John Deere’s Blue River Featured in Inaugural Issue of Bay Vision Magazine

IntelinAir is a San Francisco-based artificial intelligence company that gives insight to farmers about the state of their fields and farms. They do this by processing aerial images that are taken in the multi-spectral domain and analyzing them several times over the duration of the entire season to identify potential problems and present them to the farmers in a way that is easy to digest and easy to take action.


Ara Nefian, CTO and founder of IntelinAir, tells us that they specialize in corn and soybean. He says that corn, soybean and weed together cover the vast majority of raw vegetation in the world and IntelinAir’s work is helping to solve “one of mankind’s biggest problems.” He adds that by 2050, the Earth’s population is expected to reach 9 billion people, so their method is not just efficient, but vital. There are many companies that focus on specialty crops for which the profit margins may be higher, but admirably, IntelinAir are looking at large acreage for the most commonly-planted vegetation, because then they really solve an important problem for the world. Ara says: “We’re focusing on things that in my opinion really matter and we understand that we might not have the best profit breaker.”


The IntelinAir method analyses multispectral images using visible, near infrared and thermal cameras. They create large maps that cover the areas they fly over, then for each of these maps they respond to individual fields and do their in-depth analysis. Some of the specific problems they look for are anomalous changes in the fields, anomalies in areas where vegetation didn’t emerge, the presence of weeds, and the presence of water patches before the emergence of the crop. Finding areas which need additional irrigation is also in their plan.


Their analysis is driven almost entirely by the farmers’ needs. They have a large sales and customer satisfaction team that is in constant touch with the farmers. The farmers provide feedback and IntelinAir build a product for the following season that is closer to what the farmers need. The algorithms that they develop are mostly influenced directly by the farmers because, as Ara points out,“They are our customers and they probably know much better than us what is important and what the problem is.”


In terms of computer vision, Ara tells us they do a lot of work in image segmentation, statistical modelling of aerial images, artificial intelligence, and convolutional neural networks. Surface segmentation is one of the more difficult problems the team face, particularly doing it outside with so many variations to illumination conditions – cloudy days vs clear days – instead of within a confined lab environment. Ara says their strength comes from how they manage to overcome these challenges over a long period of time. The season lasts six months and variations in illumination can happen day- to-day, from geographical area to geographical area.


How do they solve it? Ara explains: “In my opinion, the best way to solve these variations in illumination condition is by understanding the images. Start from an image formation model. The fields that we are analyzing, they’re all described by a similar set of parameters, which are the vegetation rows that are all equidistant and parallel. Rather than doing a blind image segmentation, we start from the underlying knowledge that this is what the image looks like. We start from an image formation model and we adapt the parameters of the segmentation that best fit this image model. Using our prior knowledge about the image within a statistical framework certainly helped us a lot in dealing with global illumination variations.”


Finally, Ara tells us about the excitement the whole team has for this work: “You wake up in the morning and the first thing you do, instead of checking your phone, you look for the weather in Iowa and Illinois where we fly to see if it’s a good day for flying.


Next, you talk to the flight operators to see that the pilots are all flying and there are no issues, no clouds. Then a few hours later you see the data come in and within 24 hours you are passing it to the farmers. I think that’s the thrill that comes with a beautiful product, a great problem to tackle and to solve, and certainly the excitement that you see from the satisfaction of our customers. I guess that’s the great thing about working in a start-up.”


Read the inaugural issue of Bay Vision from RSIP Vision.