By Wang Jingyao, Wireless Development Expert at Alibaba Entertainment
Released by AliENT
Virtual reality (VR) has an increasing number of application cases in different industries. With 720-degree panoramic cameras, some manufacturers perform multifurcating tree modeling based on multi-view photos. It is a technique that has been widely used in real estate companies, such as Lianjia. In the ticketing industry, similar seat selection products have also been piloted in and outside of China; examples include TicketMaster and StubHub outside China and Moretickets in China. For large venues, the current implementation tends to adopt a 3D modeling + post-rendering approach to output ECB-based panoramic photos. The output photos are primarily used for displays on multiple big frontend clients.
2. Damai’s Solution
Damai’s intention of implementing panoramic VR was to provide assistance and an immersive experience for users’ seat selection decision-making. After Damai comprehensively evaluated the user interaction and implementation cost of the seat selection pages, it opted for modeling + panoramic rendering for large venues and VR shooting for small venues.
2.1 Implementation Strategy
- Large Venues: 3D Modeling + Rendering
- Small Venues: Scene VR
2.2 Full-Chain Modeling Process
A venue model is established in the classic Autodesk 3ds Max program through the process of outputting white mold and texture mapping.
2.3 Full-Chain Process
The whole process is divided into four major parts based on milestones: modeling/shooting > production > underlying platform of seat selection > big frontends.
Currently, the conventional scheme of 3ds Max is used for modeling, while Damai is still investigating the mass production scheme based on 3D laser radar + SLAM.
3. 3D Venues
3.1 Loading of Model Texture
An output venue model of 3ds Max has a capacity between 50 MB and 200 MB. After being compressed by the glTF model and combined with LOD multi-level detail texture, an ultra-large texture venue based on three.js can be opened more efficiently.
1. VR Engine
The overall VR solution adopts a projection scheme based on ECB spherical coordinates to render and synthesize photos into 2:1 panoramic photos. The photos undergo compression by MozJPEG, intelligent noise reduction, and super-resolution reconstruction before they are rendered by the VR engine. This also lays a solid foundation for displaying panoramic photos within seconds. Regarding big frontend VR engines, Android/iOS engines are comparatively mature on the app side, and therefore Google/Apple built-in schemes are used. On the H5 side, engines vary in terms of performance and integration. Damai finally decided on Pannellum as the top choice after studying several common engines on the market. Pannellum supports extended functions, such as compass indication, panoramic tour, panoramic annotations, and multiresolution.
2. VR View and Optimization
GVR optimized the packaging of perspective view, fisheye view, stereographic view, architectural view, Panini view, little planet view, and other views. This optimization includes but is not limited to:
- Lens Distortion Correction
- Spatial Audio
- Head Tracking
- 3D Calibration
- Side-by-Side Rendering
- Stereo Geometry Configuration
- User Input Event Handling
5. Full-Chain Mass Production
Comprehensive implementation of the VR solution requires modeling and photographing many venues, making mass production necessary. The process can be split into the following steps:
- Shooting: Professional cameras + Damai business end = Remotely controlled photo-taking + Seat information appended
- Production: Binding, FOV inclination calculation, compression, and super-resolution reconstruction
- Underlying Platform for Seat Selection: Base map protection — dark watermark and stream encryption
- Big Frontends: Three-client VR engine, progressive loading, and stream encryption
Synthesis optimization, Lanczos/HPX-based photo compression, and stream cipher encryption are performed on online photos to ensure the instant access and data security of users. The continuous development of 3D and VR technologies is an important part of building smart venues. They also provide users with visualization methods to make purchase decisions. After mass production, Damai will continue to explore cloud rendering modeling and commercial marketing capabilities based on large-scale scenarios to bring a better immersive experience to users.