Tomohito Tomaru’s daughter thought his intense research into crows showed his fondness for the birds, but in reality, he was devising an AI system to more efficiently destroy their nests.

Tomaru, a 41-year-old employee of NTT Comware Corp., a Tokyo-based subsidiary of Nippon Telegraph and Telephone Corp., spent years developing the “crow home detection” system, which was marketed in April.

It uses an artificial intelligence image-recognition technique to help power companies deal with a persistent problem.

Crows often build nests on utility poles. If coat hangers and other metal pieces are used, they can come in contact with power cables, causing short circuits and electricity failures.

Such incidents occur repeatedly around Japan every year.

Utility workers need to visually check each pole to find the problem-causing nests for removal.

The workers also conduct patrols, but they have found it difficult to predict where the birds will build their homes.

In Tomaru’s system, the AI analyzes surrounding areas filmed by cameras mounted on vehicles, allowing the workers to concentrate on driving.

The system automatically shows photos and locations of crow nests or ones being built. The information can also be sent by email to utility workers.

Tomaru and his co-workers wanted to reach a level of precision at which even “a few branches” in the early stage of nest-building could be uncovered by the system.

The challenge was amassing enough nest photos for AI learning at such a high degree of accuracy.

The nests are often difficult to spot, and crows build their homes for their offspring only in spring.

Shikoku Electric Power Co. provided Tomaru with video reels each lasting four to five hours that showed 10 to 20 nests. He heard about approximate nest locations beforehand, but he was not familiar with the island.

Every day, Tomaru rewound and replayed the reels to pinpoint the locations of the nests.

In the end, he had collected 10,000 images of crow nests that were scanned by the AI system for learning.

Since getting such images was difficult during autumn and winter, he created a nest model himself.

Tomaru said his daughter saw him making the nest model from branches he picked up at a park.

She also watched him viewing the birds' nests during a family visit to a zoo and botanical garden.

On a stroll with his daughter, a kindergarten pupil at the time, she told him, “Daddy, there’s a crow nest there.”

“She must have believed I love the bird’s homes,” Tomaru said.

Some of his work colleagues were skeptical of the “profitability of a service that only targets crows.”

So Tomaru distributed copies of an article reporting on the difficulties that utilities face in eliminating the nests.

He also envisages a future where various infrastructure, not just utility poles, can be inspected efficiently with AI.

IDEA HATCHED AT TECHNICAL FAIR

As a division head responsible for AI-based services at NTT Comware, Tomaru led its exhibit at a technical fair held at the Tokyo Big Sight convention complex in spring 2017.

The company provides digital technologies to support communications, but it has lately focused more on services that use its AI image recognition technique to automate infrastructure inspections and equipment checks.

At the fair, NTT Comware displayed an AI technology that confirms the condition of electricity poles for inspection and maintenance purposes.

An official of a group company of Shikoku Electric Power visited the booth and talked with Tomaru. The subject of crow nests came up in their conversation.

Tomaru thought about a solution that could help power companies throughout Japan.

He began visiting the Shikoku region almost monthly.

PROMISING BUSINESS

Shikoku Electric Power Transmission & Distribution Co. adopted the system for a trial run. It put it in full operation in spring this year in Kagawa and Ehime prefectures.

Tomaru’s next goal is selling the system to other utilities. If more businesses sign up for the service, the AI accuracy will improve with the additional images available.

NTT Comware sees much business potential in AI and other digital technologies for infrastructure inspections and management.

In April, it established a brand for such services and announced plans to raise sales to 10 billion yen ($91.1 million) in five years.

The move is aimed at increasing its presence in the NTT group as the telecommunications giant continues its reorganization.

Tomaru aims to share footage from car-mounted cameras with other companies for a wide variety of infrastructure equipment checks.

“Making information collected with a single vehicle available for both utility companies and NTT will help lead to a far brighter future,” Tomaru said. “I want to design a service to achieve that.”