From Ruptured Road to Living 3D Map: How Drones Are Helping Rebuild Noto

The drone is the most visible part of the experiment. It rises from a box beside a scarred roadway, follows a prescribed route over earthworks and returns carrying hundreds of photographs. But the more consequential work begins after its rotors stop.

At a disaster-recovery site on the Nōetsu Expressway’s Anamizu Road in Ishikawa Prefecture, those images no longer had to be pulled onto a local computer, sorted by a technician and manually uploaded into specialized software. They moved automatically through a cloud system over KDDI’s Starlink Business connection. Analysis began without another human handoff. The output was a three-dimensional point cloud—a dense digital representation of the road, slopes and construction surface that could be viewed from the site office or from far away.

Tokyu Construction, KDDI and KDDI SmartDrone tested the chain on May 18 and 19, 2026, using a DJI Dock 3 and a remotely operated DJI Matrice 4D. Announcing the results on August 7, the companies said the automated workflow cut the relevant conventional work by about half and eliminated one to two hours of manual data-transfer and setup work after a flight.

That may sound like a modest office saving beside the physical task of rebuilding an expressway. In fact, it addresses one of construction technology’s stubborn gaps: collecting data has become easier than turning it into information people can use repeatedly.

A road built to shorten distance

The technology is easier to understand when placed against the history of the road beneath it.

The Nōetsu Expressway was conceived as a high-standard route of roughly 100 kilometers, joining Wajima in Ishikawa through Nanao to Tonami in Toyama. The 6.2-kilometer Anamizu Road section, part of National Route 470, opened in June 2006 between Noto Airport and Anamizu. It was designed to connect the airport more effectively with the peninsula and to reduce the isolation imposed by Noto’s long, mountainous geography.

On January 1, 2024, geography asserted itself violently. A magnitude-7.6 earthquake struck at 4:10 p.m. at a depth of 16 kilometers, producing the maximum intensity of 7 on Japan’s scale. Buildings collapsed, slopes failed and a tsunami reached the coast. The Japan Meteorological Agency later noted ground uplift of roughly four meters in western Wajima—an extraordinary deformation for a recent earthquake beneath Japanese land.

Roads that had been built to defeat distance were cut by the terrain. For recovery crews, that created a painful circular problem: damaged routes were essential for reaching communities, carrying machinery and moving supplies, but repairing those routes required workers to enter unstable places that were themselves difficult to reach.

The national government established the Noto Reconstruction Office on February 16, 2024, to drive work on the Nōetsu Expressway, coastal National Route 249, landslide areas, rivers and other damaged infrastructure. Road restoration was not a single repair. It became a prolonged program across a changed landscape.

Rebuilding a line through broken ground

Tokyu Construction’s contract covers multiple collapse locations on the damaged Anamizu Road. At approximately kilometer point 14.1, a major failure had distorted the road alignment and forced a 40-kilometer-per-hour speed restriction.

Embankment work began there in June 2025. By the end of November, before the severe winter, the restored alignment had opened in a form designed for 80-kilometer-per-hour travel. Tokyu Construction said the change improved conditions for heavy trucks and buses carrying volunteers—vehicles for which a narrowed or irregular recovery route can be especially difficult.

The repaired section did not end the larger job. Earthworks continue to change the shape of a site by the day. Soil is excavated, placed and compacted; slopes are cut; drainage and retaining structures alter the surface. Managers must know not only what the site looks like, but how its dimensions compare with design and with earlier surveys.

Traditionally, that knowledge has required people to enter the site, establish measurement points and move data through a sequence of separate tools. A drone can reduce exposure and cover a wide area quickly, but only if its pictures become trustworthy measurements.

The drone becomes a repeatable instrument

Tokyu Construction had already installed a DJI Dock 3 at the site. The dock is more than a storage box: it provides a known base from which the aircraft can launch, land and recharge. That turns a drone from equipment brought out for a special occasion into a device that can repeat a defined route.

The earlier operation also confronted an unusual regulatory problem. The flight needed to cross an automobile-only road, something authorities generally treat cautiously because a falling aircraft could strike moving traffic. Tokyu Construction routed the aircraft over a pedestrian bridge, using the structure below as protection against a direct fall onto the roadway, and obtained permission with application support from Boundary Administrative Scrivener Corporation and operating support from KDDI.

The company used Japan’s Level 3.5 framework, introduced in December 2023. The system permits certain beyond-visual-line-of-sight operations without placing assistants and warning signs along the entire route when licensed operation, insurance and digital confirmation of people on the ground provide the required safeguards. It does not mean “unregulated flight.” It replaces some physical watchers with a documented combination of technology, qualifications and risk controls.

Tokyu Construction reported in April that remote monitoring under this arrangement reduced the labor associated with site observation by 50 percent. At that stage, the primary use was fixed-point imagery. Photo surveying and emergency response were described as the next steps. The May test put the first of those plans into practice.

What a point cloud sees

A point cloud is neither an ordinary photograph nor a finished engineering drawing. It is a collection of points positioned in three-dimensional space. Taken together, millions of such coordinates describe the visible shape of terrain and structures.

In drone photogrammetry, the aircraft takes overlapping images from changing positions. Software identifies common features across those views and calculates their spatial relationship—a family of methods often described as Structure from Motion and multi-view stereo. The result can be used to derive surfaces, elevations, cross-sections and volume comparisons.

For road construction, that matters because a manager can compare the measured surface with a design model, estimate how much earth has moved, document progress or inspect a slope without relying only on selected photographs. Repeating the same flight can create a sequence of site states—a record of change rather than a single snapshot.

Accuracy is not automatic. Image overlap, camera calibration, positioning, ground control, vegetation, lighting and wind all influence the result. The Geospatial Information Authority of Japan maintains procedures and accuracy requirements for UAV-based public surveying, and its research has shown that flight pattern and oblique images can affect height accuracy.

For that reason, the most important sentence in the companies’ release may be the least dramatic: they verified the generated data’s position accuracy for practical construction management. The absence of a published number, however, means readers cannot independently judge what task the data was precise enough to perform.

Stage	Earlier workflow	Tested workflow
Capture	Drone collects hundreds of overlapping images	Remote Matrice 4D flies from a Dock 3
Transfer	Images downloaded to a local PC	Images move through KDDI SmartDrone’s cloud over Starlink Business
Processing	Technician uploads files and configures analysis	Transfer triggers cloud analysis automatically
Output	Specialist distributes the finished data	Point cloud is stored centrally for the site, branch and headquarters

The invisible bottleneck after landing

Construction’s digital tools often arrive as islands. One system flies the drone, another stores pictures, a third performs photogrammetry, and a fourth holds project records. A skilled employee becomes the bridge: downloading, renaming, uploading, configuring and checking.

Before the Noto integration, the companies said, a survey produced several hundred images that had to be downloaded once, then uploaded again to the three-dimensional analysis service. The transfer and setup took as much as two hours. More importantly, someone who knew the software had to be available.

The new workflow connected the islands. KDDI supplied Starlink Business as the communications infrastructure. KDDI SmartDrone supplied the platform and built the system link. Tokyu Construction provided the site, managed the trial and judged its usefulness for actual operations.

Once the aircraft completed its route, the images passed through the drone platform to the point-cloud system and analysis began. The finished dataset was managed in the cloud, where people at the field office, a branch or headquarters could inspect the same representation without emailing large files.

That distinction is central to the national government’s i-Construction 2.0 program. The Ministry of Land, Infrastructure, Transport and Tourism aims by fiscal 2040 to reduce the staffing required on construction sites by at least 30 percent—or raise productivity to 1.5 times its former level. Its three pillars are automation of construction, automation of data connections and automation of construction management. The Noto trial is a compact example of the second and third pillars meeting in one workflow.

Why satellite communications matter in Noto

Cloud automation is useful only if the site can reach the cloud. Remote civil works often sit in valleys, mountains or disaster zones where terrestrial networks are weak, damaged or temporarily overloaded. Sending hundreds of high-resolution images is far more demanding than transmitting a short message.

Starlink Business supplied the backhaul for the Noto test. The satellite connection did not turn the drone itself into a satellite-controlled aircraft; rather, it connected the remote site and its systems to the wider network needed for operation and data transfer. That distinction matters because “satellite drone” can imply a direct aircraft-to-satellite control link that was not the subject of this trial.

KDDI and its drone subsidiary have been exploring variations of the same architecture at dams, mines, tunnels and isolated communities. The recurring idea is to place sensing equipment where work happens, give it a resilient connection and move expertise away from the hazard without cutting experts off from the evidence.

A tool for scarce people—not a site without people

The companies emphasize labor savings, and construction’s aging workforce makes the argument urgent. Yet the trial should not be understood as a road project with no surveyors or engineers.

People still set the flight plan, assess weather and airspace, maintain the aircraft, verify the coordinate framework, judge accuracy and decide what the model means for construction. Automated processing can reproduce an error as efficiently as it reproduces a correct survey. A cloud point model does not decide whether a slope is safe or a layer of fill is acceptable.

What automation can do is reserve scarce expertise for those judgments. It can remove repetitive file handling, reduce the need to position a specialist at every remote site and limit some entries into unstable ground. It can also make measurement more frequent. A process that is cumbersome may be performed only when necessary; one that runs almost automatically can become routine.

That frequency may be the technology’s most important long-term effect. If a road is surveyed on a repeatable schedule, managers gain not just a model but a history. Differences between point clouds can reveal progress, movement or erosion. After an earthquake or heavy rain, a new flight can be compared with the prior state.

The distance still to travel

The Noto demonstration solved a specific chain under specific conditions. Scaling it will require dependable bandwidth, reliable docks, standardized data interfaces, strong security, clear responsibility for failed flights and proof that accuracy remains stable across seasons and terrain.

Cost will matter, especially for smaller contractors. So will weather: UAV surveying remains sensitive to rain, wind, visibility and surface conditions. Regulations and coordination do not disappear merely because the launch is remote. And disaster sites may lose power or connectivity at the moment automation is most needed.

The companies say they will refine the integration and consider deployment at more sites. The test will become significant if it stops being an impressive two-day demonstration and becomes an ordinary, audited part of construction management.

Even at this stage, the setting gives the experiment unusual weight. The Anamizu Road was built to reduce Noto’s distance from airports, cities and services. The earthquake broke that promise in physical places. Now a different network—an aircraft, a satellite link and a cloud model—is helping engineers see the road as they put it back.

The drone returns to its dock. The road continues changing. Somewhere beyond the peninsula, a model begins to take shape.