To uncover insights that would help combat drought in a profound new way, OmniEarth Inc. needed to process and categorize massive amounts of aerial data quickly and accurately.
By feeding aerial imagery into the IBM Watson Visual Recognition service and training the software to differentiate between pools, grass, gravel and other terrain features, OmniEarth can evaluate land parcels and help water districts identify properties that use more water than necessary, then help owners adjust usage accordingly.
40x faster image processingthan was possible using manual methods
Higher capacity for analyzing terrainon a massive scale, creating new business opportunities all over the world
Deeper insight from satellite imageryon a highly granular scale
Business challenge story
Addressing extreme drought
In April 2015, 15 months after declaring a state of emergency due to a drought, the governor of California issued a proclamation: the state needed to reduce urban water consumption by a total of 25 percent over the next 10 months. This mandate presented a huge challenge to the state’s water districts. Achieving such a drastic reduction would require more than just asking people to use less water.
Scientists at OmniEarth knew how to help. If they could provide granular water consumption data, they could team with water districts, cities and counties to identify precisely which land parcels needed to reduce water consumption, and by how much.
Chelsea Minton, senior sales engineer at OmniEarth, explains, “If we can determine how much water a property needs on every given day, and then compare that to how much that property is actually using, we can calculate exactly how much potential that property or that homeowner has to save water.”
Armed with that information, OmniEarth could help water districts take a direct and effective approach to reducing water consumption, using mailers and other outreach techniques to educate property owners about their water usage. “A lot of folks didn’t think we could answer those questions,” continues Minton, “because they require such computationally heavy analysis.”
Processing images cognitively
With that concern in mind, OmniEarth’s director of products, Alistair Miller, approached the IBM® Watson™ team. “We were really looking to see how we could leverage Watson’s processing, learning and visual recognition capabilities to address the challenge of processing large amounts of imagery data,” he explains.
OmniEarth began by feeding massive amounts of aerial imagery into the IBM Watson Visual Recognition service, which understands the content of images. Taking advantage of the IBM Watson platform’s capacity for machine learning, the team customized the way the software interprets different topographical features, including pools, grass, turf, shrubs and gravel.
“We were able to very rapidly improve our water budget and efficiency analyses because of the scale of Watson and the speed with which it learned and applied various classifiers to the data we fed it,” explains Jonathan Fentzke, OmniEarth’s chief strategy officer and cofounder. “Even as a very young machine, it’s learning very quickly.”
By using the IBM Watson platform to analyze large tracts of land, OmniEarth can help water districts make specific recommendations to property owners and governments, with the ultimate goal of saving water in areas of drought-ridden terrain. Such suggestions could include replacing a patch or percentage of turf with mulch, rocks or a more drought-tolerant species, or draining and filling a swimming pool less frequently.
Fentzke describes the difference it made using the IBM Watson platform’s cognitive capabilities in visual recognition to further the company’s mission. “On one of our early test runs, we were able to identify pools in 150,000 parcels in 12 minutes. Doing that manually would have taken hours to days.”
With that kind of processing capability, Fentzke notes, “we can do this across the US and across the world to deliver our insights and analytics to any number of customers and partners in near real-time.”
And for places such as California, where saving water is paramount, the granularity of the OmniEarth approach may just change the game. “Analyzing things at the parcel-by-parcel level represents a paradigm shift for this industry,” says Fentzke. “It’s an exciting time, and the sky is the limit.”
OmniEarth is also looking for ways to integrate its solution with other devices and sensors, with the intention of helping automate water conservation efforts.
“Sprinklers are a great example. By calculating area and determining what the optimal amount of water is, we found that people who were using so-called smart sprinkler systems were actually overwatering significantly. The only information those systems were using was the amount of moisture,” explains Fentzke.
“But the fact is, the grass may be dry and you may water it for 10 minutes, but it’s not going to green up as soon as you put water on it. There must also be some data in there about the growth cycle: you use this much water to get it this far,” he continues. “Our analytics, paired with Watson, will create a cognitive context for how connected sensors make informed decisions and ultimately help keep lawns green without excessive watering.”
In addition to helping reduce water usage, says Fentzke, OmniEarth hopes to use its cognitive capabilities to help revolutionize agriculture. “We can use our understanding of the physics of growth and the decay of plants and other things to inform algorithms and create decision-making tools for smart agriculture applications. We’ll be spending less time and money and getting better yields.”
“It’s the green data revolution,” Fentzke concludes. “OmniEarth is certainly a part of it, and it’s going to be enabled by technology like Watson.”
About OmniEarth Inc.
Headquartered in Arlington, Virginia, in the US and founded by a group of scientists and entrepreneurs in 2014, OmniEarth Inc. builds scalable solutions for processing, clarifying and fusing large amounts of satellite and aerial imagery with other data sets. The results have a broad range of applications, from pipeline monitoring to precision agriculture to resource management. OmniEarth employs approximately 20 people.
Take the next step
To learn more about IBM Watson application programming interfaces (APIs), please contact your IBM representative or IBM Business Partner, or visit: ibm.com/WatsonDeveloperCloud. To see a demonstration of the IBM Watson Visual Recognition platform’s capabilities, visit: http://ibm.biz/VisualRecognition.
View more client stories or learn more about IBM Watson Platform