Often in business, as in life, we’re told to keep the big picture in mind by not getting bogged down by minor details—don’t lose sight of the forest for the trees. This outlook can help a company set its strategic vision or empower someone to make quick decisions. But what do you do if you only have a big picture or big idea, and need smaller details to build your approach?
Premise recently partnered with a leading geospatial analytics company that has a lot of big pictures but truly needed an on-the-ground perspective. Aerial imagery and big data enable predictive artificial intelligence models that can serve purposes ranging from identifying construction trends to forecasting harvest yields. Before creating these models for the latter, however, it’s crucial to identify precisely what crops the images contain—a challenge made difficult when these photos are snapped tens of thousands of kilometers from Earth and cover an area of hundreds of thousands of square kilometers.
For this particular task, the client needed to classify crops from multiple regions of interest already identified from its field satellite imagery. These crops were limited to a select few based on the time of year and growing season, yet some were harder to distinguish from others (think bananas vs. plantains or oranges vs. grapefruits). Another obstacle was the impending harvest, which created a hard time limit on the collection period for crop identification.
Premise was an ideal partner for this program due to the strength and breadth of our Contributor network, automated quality control processes, and round-the-clock country support teams to manage collection campaigns. Contributors followed routes spanning multiple kilometers and took photos, recorded their geolocation, and even identified unique fields bisected by roads or by other less discernible details. In total, more than 1,600 observations were collected in less than two months, exceeding the project’s target goals.
The insights gained from this program are already being incorporated into other current and future Premise geospatial validation projects. Perhaps the most noteworthy lesson is that there is no idea or picture too big that can’t benefit from the value of real-time, ground-truth data to fill in the details.