How Artificial Intelligence Elevates Aerial Imagery

Artificial intelligence fueled by aerial imagery improves the life cycle of a policy from quote to claim.

By Doug Jentzsch

How Artificial Intelligence Elevates Aerial ImageryConsider an aerial image of a house. You know it’s a house because your eyes recognize the shape of a roof and the ridges and slope of various roof faces. You can clearly see a concrete driveway and landscaping that includes areas of grass and large trees. You may also be able to identify the faint outline of the property or parcel.

While you might determine some characteristics from this image, it’s unlikely you would be able to obtain many usable property details from sight alone.

Now take this image and add a million more that have many similarities but are each uniquely different. Every image has basic details—a gray driveway, a dark roof, some landscaping—that allow you to recognize it for what it is, but your human eye can’t process that many images and transform them into usable property details, at least not quickly or accurately. That’s where machine learning comes in. Computer vision technology can quickly analyze high-resolution aerial images like this and extract the relevant property information.

Machine learning, artificial intelligence, computer vision—these buzz words are tossed around so often that it can be hard to see beyond the hype. But here’s the bottom line: The practical applications of machine learning, artificial intelligence, and computer vision are endless in the world of aerial imagery and, by extension, the property/casualty insurance industry.

With machine learning, the neural network can be trained to identify elements within the aerial photo and instantly begin to derive specific property data. As the learning algorithm is developed, its understanding deepens and becomes more sophisticated, accelerating the process and increasing automation as the neural network learns to identify the correct attributes on its own.

This means the property data of any one of those aerial images can be accessed quickly with the click of a mouse.

Machine learning’s application to aerial imagery is highly valuable, especially when it comes to processing imagery covering large geographic areas. The higher the imagery resolution, the more data available. And the more often imagery is refreshed, the better trained and “smarter” the neural network becomes—allowing for even faster and more accurate data.

For insurance work, machine learning and aerial imagery go hand in hand to enhance accuracy and the speed at which information can be accessed through all phases of a policy.

Insurers underwriting a new policy can select their preferred property attributes. Artificial intelligence then scans aerial imagery to identify these exact data points—providing up-front information to determine whether a property meets insurer guidelines or might need to be flagged for a potential on-site visit.

During the initial quote, many questions are asked of the homeowner—giving plenty of opportunities to receive inaccurate self-reported data. If the homeowner can’t provide the answers, you’ll need to check public records. It’s possible—and probable—that most homeowners don’t know their roof’s material type and condition or the percentage of tree coverage over their home, and they may not be forthcoming about the size of their swimming pool or how it’s enclosed. That leaves you having to schedule an on-site inspection, which can jeopardize the possibility of your company being the one to bind the policy.

But when you have powerful property insights available, there’s a good chance you’ll need to ask the homeowner only one question: “What’s your address?”

Obtaining objective, precise information from aerial imagery leaves little room for doubt and minimizes human error. When you don’t have to depend on best guesses or assumptions, you can make more structured and informed underwriting decisions and streamline your entire process.

Property data extracted from machine learning and aerial imagery can help supplement on-site inspections or even reduce the need for them altogether. In many cases, it can help you quote a policy immediately, avoiding the delay of on-site visits and the need to search public records for additional information.

This same information can help during the renewal period. Usually, a policy profile is routinely created once and never updated again. Up-to-date imagery, however, provides you with reliable information, enabling you to assess any percentage of change detection with the touch of a button.

Should this same property incur an insurance claim in the future, you already have a head start with the data gathered at the onset of the policy. Current aerial imagery compared with the historical condition of the structure provides early assessments to triage the damage and best establish next steps for claims resolution.

Continually capturing and refreshing aerial imagery is an ongoing process, but it’s vital to advancing machine learning far beyond what’s possible today. The current and future applications for this artificial intelligence will be invaluable in helping the insurance industry keep pace with ever-changing customer expectations. 

Doug Jentzsch is director of marketing for Geomni, a Verisk (Nasdaq:VRSK) business.

You will soon be redirected to the 3E website. If the page has not redirected, please visit the 3E site here. Please visit our newsroom to learn more about this agreement: Verisk Announces Sale of 3E Business to New Mountain Capital.