NEC Unveils AI Technology That Creates High-Precision 3D Models from Standard Camera Footage
NEC Corporation (TSE: 6701) has announced the development of a groundbreaking artificial intelligence technology capable of generating highly detailed three-dimensional (3D) models from ordinary camera footage in as little as one minute. The proprietary solution, developed in collaboration with the Keio AI Research Center, is the first of its kind to combine rapid 3D reconstruction with the automatic removal of unnecessary objects captured in video, enabling accurate digital representations of real-world environments without requiring specialized imaging equipment.
Designed for use with footage captured by widely available devices such as smartphones and standard digital cameras, the technology represents a major step forward in digital twin development. By eliminating the need for expensive scanning systems and reducing the time required to produce high-quality 3D models, NEC aims to make digital twin technology more practical and accessible for industries including infrastructure management, construction, manufacturing, and facility maintenance.
The company plans to commercialize the technology during fiscal year 2027, with the goal of helping organizations address labor shortages, improve operational efficiency, strengthen safety measures, and support more informed decision-making through virtual representations of physical assets.
Advancing Digital Twins Through AI
Digital twins—virtual replicas of real-world environments, facilities, or assets—are becoming increasingly important across industries as organizations seek to monitor operations remotely, improve maintenance planning, and optimize business processes.
Creating accurate digital twins, however, has traditionally required expensive laser scanners, LiDAR equipment, or specialized cameras. In addition, capturing data often requires temporarily shutting down active worksites to prevent people, machinery, or moving objects from interfering with the scanning process.
Even after data collection, converting raw information into usable 3D models can take hours or even days, delaying inspections and limiting real-time decision-making.
NEC’s newly developed AI technology addresses each of these challenges by enabling organizations to create detailed virtual environments using nothing more than standard video footage recorded during normal operations.
The result is a dramatically faster, simpler, and more cost-effective approach to creating digital twins.
Addressing Industry Challenges
Infrastructure operators, utility companies, and construction firms around the world continue to face significant workforce shortages and rising operational costs.
Routine inspections often require experienced engineers to travel long distances to assess equipment, review construction progress, or investigate potential problems.
These visits consume valuable time and resources while limiting the availability of skilled personnel for other critical tasks.
To overcome these challenges, many organizations have begun sharing video footage from worksites with remote supervisors.
While video conferencing and recorded footage improve communication, they still have important limitations.
Traditional video only shows what the camera operator records, making it difficult for remote experts to examine locations from different angles or revisit areas that were not captured during filming.
If an important detail is missed, additional site visits may be required.
NEC’s AI-powered 3D reconstruction technology transforms ordinary video into fully navigable virtual environments, allowing users to inspect locations from virtually any viewpoint after filming has been completed.
Creating Detailed 3D Models in Minutes
One of the technology’s most significant advantages is speed.
Conventional 3D reconstruction methods often require extensive processing time before a usable model becomes available.
NEC’s AI significantly reduces this waiting period.
According to the company, the system can generate detailed, high-resolution 3D models in as little as one minute, enabling near real-time visualization of complex environments.
This rapid processing allows engineers, maintenance teams, and decision-makers to review site conditions almost immediately after video capture.
Such capabilities could prove invaluable during emergency situations, equipment failures, or disaster response efforts where timely information is essential.
Powered by Gaussian Splatting
At the core of NEC’s innovation is Gaussian Splatting, an advanced 3D rendering technique that has recently gained attention within the film, animation, and visual effects industries.
Unlike traditional polygon-based 3D modeling methods, Gaussian Splatting represents environments as millions of tiny, spatially distributed points.
These points collectively recreate surfaces, textures, colors, and depth with remarkable realism while requiring significantly less computational complexity than many conventional approaches.
The technology enables highly accurate rendering of complex environments, including buildings, machinery, equipment, structural components, control panels, and fine details such as buttons, gauges, and meters.
NEC combined this emerging rendering technique with proprietary AI algorithms to improve both reconstruction quality and processing speed.
Intelligent Point Optimization
A key innovation behind NEC’s solution lies in how it manages the enormous number of points used to build 3D environments.
Instead of applying the same point density throughout an entire scene, the AI automatically analyzes visual complexity across the captured environment.
Areas containing intricate machinery, structural details, or complex textures receive a higher concentration of reconstruction points.
Conversely, smoother or less detailed regions require fewer points.
By intelligently adjusting point density according to scene complexity, the system minimizes unnecessary computational workload while maintaining exceptional visual accuracy.
This optimized approach reduces the total number of points needed without compromising image quality.
Compared with conventional Gaussian Splatting methods, NEC reports that its AI reduces model generation time by approximately 90 percent.
Automatic Removal of Unwanted Objects
Another major breakthrough is the system’s ability to automatically remove temporary objects from captured footage.
Construction sites and infrastructure facilities are dynamic environments where workers, vehicles, equipment, and moving machinery frequently appear within camera views.
Traditional 3D reconstruction systems often include these temporary objects within the final model, reducing accuracy and making virtual inspections more difficult.
NEC’s AI automatically identifies these unnecessary subjects during processing and removes them from the reconstructed environment.
Even more impressively, the technology intelligently reconstructs the background hidden behind these temporary objects by analyzing surrounding video frames.
As a result, the final 3D model accurately represents the permanent site infrastructure rather than the temporary activity occurring during filming.
This capability allows organizations to capture footage during normal operations without stopping work or clearing personnel from the area.
No Specialized Equipment Required
Perhaps one of the most practical aspects of NEC’s technology is its compatibility with ordinary cameras.
Unlike many existing digital twin solutions that depend on expensive LiDAR scanners, drone systems, or dedicated 3D imaging hardware, NEC’s platform works with video captured using standard smartphones or commercially available digital cameras.
This significantly lowers the cost of adoption while making advanced 3D modeling accessible to a much broader range of organizations.
Field personnel can simply record video while performing routine inspections or site visits without changing existing workflows.
The captured footage is then processed automatically by the AI system to generate a navigable digital model.
Supporting Remote Inspections
Once generated, the 3D models can be viewed using standard computers or tablet devices without requiring specialized visualization equipment.
Users can freely navigate the virtual environment, examine equipment from multiple perspectives, and inspect areas of interest in far greater detail than conventional video allows.
This capability supports remote collaboration between field personnel and technical specialists located elsewhere.
Supervisors can review conditions without traveling to the site, reducing costs while accelerating maintenance decisions and improving response times.
In situations involving equipment failures or operational abnormalities, decision-makers gain immediate access to a detailed virtual representation of the affected area, helping them identify potential causes and determine appropriate actions more quickly.
Improving Workforce Efficiency
As skilled labor shortages continue affecting infrastructure and construction industries worldwide, organizations are increasingly seeking technologies that maximize the productivity of existing personnel.
NEC’s AI-powered 3D reconstruction platform enables experts to oversee multiple locations remotely while reducing the need for repeated site visits.
Rather than dispatching specialists to every inspection, organizations can allocate resources more efficiently by using digital models for preliminary assessments and detailed reviews.
This approach not only reduces travel expenses but also enables experienced engineers to support a greater number of projects simultaneously.
Enhancing Safety
Safety represents another important application for the technology.
Infrastructure inspections often involve hazardous environments such as power facilities, industrial plants, elevated structures, tunnels, or construction zones.
By enabling remote inspections through accurate virtual models, organizations can reduce the frequency with which personnel must enter potentially dangerous environments.
Digital records also provide valuable documentation for compliance, maintenance planning, incident investigations, and historical comparisons over time.
Collaboration with Keio AI Research Center
The technology was developed through a collaborative research effort between NEC and the Keio AI Research Center.
The project included contributions from Professor Hideo Saito, a recognized authority in visual computing and computer vision.
Combining NEC’s expertise in artificial intelligence with Keio University’s advanced research in visual computing enabled the development of a system capable of delivering both high accuracy and practical usability for real-world industrial applications.
Expanding Digital Twin Adoption
NEC believes this innovation will help accelerate digital twin adoption across numerous industries.
Potential applications extend well beyond infrastructure maintenance and construction.
Manufacturing facilities can create virtual production environments for equipment monitoring and process optimization.
Utilities can remotely inspect substations and transmission facilities.
Transportation organizations can document stations, terminals, and critical infrastructure.
Public agencies can create digital records of civic assets for maintenance and disaster preparedness.
As organizations increasingly embrace digital transformation, accurate virtual representations of physical environments are expected to become foundational components of future operational strategies.
Looking Ahead
NEC plans to commercialize the technology during fiscal year 2027 and continue enhancing its capabilities as demand for digital twins grows across global industries.
Future development is expected to focus on expanding compatibility with additional environments, improving reconstruction accuracy, supporting larger-scale facilities, and integrating the technology into broader AI-powered operational platforms.
By combining rapid AI-driven 3D reconstruction, automatic object removal, intelligent scene optimization, and compatibility with everyday cameras, NEC has introduced a solution that could significantly lower barriers to digital twin adoption.
The innovation enables organizations to create highly accurate virtual environments faster, more affordably, and with far less operational disruption than conventional approaches.
As industries continue navigating workforce shortages, aging infrastructure, and increasing demands for operational efficiency, NEC’s breakthrough technology offers a practical pathway toward smarter inspections, improved safety, and more effective digital transformation.
Source link: https://www.nec.com