China News Service, Beijing, November 2 (Reporter Pang Wuji) The archway at the entrance of the town, the blocks with red roofs and white walls, the Minning Middle School with big red letters, the rotating wind turbines... these real scenes are included in the same continuous three-dimensional model.
This model is a replica of Minning Town in Ningxia. The documentary "Minning Avenue" turns an area of 23.5 square kilometers of the town into a three-dimensional scene, allowing the camera to freely travel through neighborhoods, roads, and industrial parks. This is not a special effect, but a real-life three-dimensional reconstruction. Providing technical support for this documentary is Shenzhen technology company Qiyu Innovation.
Shu Chenliang, partner and chief financial officer of the company, said in a recent exclusive interview with a reporter from China News Service that this project only required three collectors, seven days of field collection, and 14 days of data processing to complete the real-life three-dimensional construction of the entire Minning Town. In the past, the same task required dozens or even hundreds of people and took several months.
Traditional three-dimensional reconstruction is a complex project. The collectors have to carry the lidar to every corner, and then use drones to take additional high-altitude views. After returning, they have to spend several weeks doing data calibration and manual model modification.
Its domain innovation attempts to integrate lidar, cameras, gyroscopes and other acquisition capabilities into handheld devices, and then automatically process the data through software and artificial intelligence. Shu Chenliang said that it takes about 10 minutes to collect a scene of hundreds of square meters on site and about 100 minutes to reconstruct it later. Work that used to take days or even months has been compressed to the hour level. The company also recently launched a mobile app that allows users to perform 3D reconstruction through mobile phone cameras without purchasing professional equipment. Three-dimensional reconstruction is changing from "a professional job for a few people" to "a tool available to more people."
After the tools became lighter, the application scenarios began to move outward.
Film and television companies were the first to try. Shu Chenliang introduced that taking the movie "Wild Times" as an example, the large number of long shots in the film required extremely high positioning, and in the past, it had to be rehearsed repeatedly in real scenes. Now real scenes are scanned into a 3D digital environment with centimeter-level accuracy, and photographers, actors and prop teams can complete planning in the digital space in advance.
The same logic has been copied to more fields: at the ancient site of Saqqara in Egypt, the China-Egypt joint archaeological team used 3D scanning equipment to digitally collect ancient coffins and statues, and established permanent digital archives for the preservation of cultural relics; in the Qingdao Jiaozhou Bay Second Undersea Tunnel, the company's equipment was used to quickly collect point cloud data on the tunnel face, and the over- and under-excavation conditions could be measured on-site.
More importantly, a scene can be collected once and used for a long time. In addition to film and television production, it may also become a shared digital resource in fields such as simulation training and industrial applications.
In order for a robot to complete operations in a real environment, it not only needs to recognize objects, but also needs to understand the three-dimensional structure, spatial position and mutual relationships - this information can only be provided by three-dimensional data. Shu Chenliang revealed that his domain innovation is cooperating with NVIDIA. The company's independently developed LCC format has been connected with the NVIDIA Omniverse platform to explore the introduction of real spatial data into the simulation environment for the training and interaction of robots and AI. The value of 3D digitization has also been upgraded from “seeing for people” to “learning for AI”.
Founded in 2020, Qiyu Innovation has entered more than 80 countries around the world and serves more than 8,000 customers. As a sample of private enterprises developing new productive forces, the growth of its domain innovation is inseparable from the support of the domestic innovation ecosystem. The government and state-owned enterprises have opened up real needs through projects, and the company has also received investment from institutions such as Shenzhen Venture Capital and support from related industry funds.
Starting from three-dimensional reconstruction and extending to film, television, architecture, and robots, innovation in this field has taken a pragmatic path: driving research and development based on real needs, accelerating implementation through supporting industrial chains, and then leveraging the global market to open up growth space, reflecting the changes in the innovation logic of China's private technology companies.
