▲ Yoo Hyung-gon, Head of Mobiltech Division (right)

Pedestrian-level data generation through aerial photography and LiDAR data fusion
Mobiltech Secures Modernity and Scalability Through Self-Developed MMS Equipment

Mobiltech shared the limitations and problems of existing digital twins and introduced 'Replica City,' a 3D city data solution that provides the latest data for the city, which changes every day.

On the 28th, e4ds hosted a webinar at the 2022 e4ds Automotive Tech Concert, inviting Mobiltech’s Head of Division Yoo Hyung-gon to speak on the topic of "Technology for Updating and Utilizing 3D Spatial Information for Realizing Immersive Smart Cities."

Mobiltech aims to present a paradigm that goes beyond spatial recording by supplying an immersive digital twin service called 'Replica City'.

The self-made equipment is mounted on a vehicle to scan the city every day, and the city object information (POI) obtained at this time is built into a DB every day.

Based on this data, it is created into a realistic 3D model through automatic modeling post-processing.

Mobiltech is resolving the data collection and processing tasks, which account for significant costs and time in mapmaking, through AI automation. By training AI based on acquired data to develop technologies for detecting changes in spatial information and automatically updating it, the company is updating 3D spatial information daily. there is.

Replica City can be used as core infrastructure for customized metaverses desired by various organizations (B2B, B2G), and can provide big data on Points of Interest (POI) for various urban objects that were not available in existing maps.

Furthermore, Mobiltech can extend high-precision maps beyond autonomous driving and digital twins to various services such as virtual production, simulation, and UAM.

Existing digital twins have limitations in that they are difficult to integrate with actual digital twins because they represent space based on photos and are mostly constructed as one-time events.

In addition, when captured from above based on aerial photography and LiDAR, pedestrian-level data such as ground areas consists of a closed structure with low detail.

Since the production process is semi-automated and based on manual work rather than automation or precision of detail, errors such as occlusion and light intensity may occur. Consequently, when visualization is implemented based on this, the absence of PBR elements and the reliance on photo textures result in imprecise details.

To overcome the problems and limitations mentioned above, Mobiltech is focusing on the most important characteristics of maps: precision, up-to-dateness, efficiency, and compatibility.

To address precision issues, this company is acquiring data by deploying a large number of its own-developed devices instead of expensive overseas equipment.bsp;

Although it was necessary to utilize expensive INS to produce smooth data for urban areas, Mobiltech fused aerial photographs and LiDAR data to generate precise pedestrian-level data and equipped itself with state-of-the-art and scalable MMS equipment developed in-house.

Mobiltech has developed its own equipment in cooperation with global LiDAR manufacturers such as Veolodyne and is operating data collection equipment specialized for AI processing, capturing more than 10 photos per second and processing them with AI.

To ensure up-to-date data directly linked to safety, we have established an ecosystem to increase collection frequency, and for efficiency and speed, we are automating processing based on AI instead of manual work.

The average driving distance of urban drivers is around 40km, making it impossible to cover the entire national road network.

As a solution to these problems, Mobiltech distinguished between in-house drivers and crowdsourced drivers.

The company focuses on acquiring high-precision data to use as a standard map for its own drivers and seeks to expand data such as pedestrian and sidewalk data through strategic partnerships with taxi and courier companies.

Mobiltech is also maximizing processing efficiency through automating correction using aerial photographs and LiDAR data, and using AI for object extraction.

To address the issue of significant time and cost consumption in field surveying when using existing MMS data, we utilized aerial photographs of Seoul and LiDAR data, and built a system capable of accurate position correction even in dense urban areas by comparing it with vehicle-collected data.

Through AI-assisted object extraction, it is possible to classify and estimate the location of road environment objects among the digital twin batch objects.

It processes over 100 types of objects repeatedly placed in road environments, such as trees and traffic facilities, with a detection rate of over 90%, and synchronizes with aIt is built to enable realistic expression by comprehensively analyzing camera and GPS data and introducing PBR (Physically based Rendering) to represent gloss, reflection, etc.

"The demand for reality-based spatial information is rapidly increasing across various industries," said Director Yoo Hyung-gon. "Although it is called immersive spatial information, it has the disadvantage of not being linked to actual data, so Mobiltech will address these shortcomings and deficiencies through Replica City."

In response to a question about whether there are correction methods based on the surrounding environment in cases where updating spatial information is difficult, such as underground roads or mountainous areas, Director Yoo stated, “Even with the best GPS, errors can occur in spaces like tunnels,” adding, “We are also utilizing methods to secure overall tunnel data by fusing data on the number of vehicle wheel rotations using a wheel sensor called DMI, or by fusing reference points such as null start and end points.”