Design and Implementation of Online Live Streaming System Using A 3D Engine

Abstract

With the growing demand for live video streaming, there is an increasing need for low-latency and high-quality transmission, especially with the advent of 5G networks. While 5G offers hardware-level improvements, effective software solutions for minimizing latency remain essential. Current methods, such as multi-channel streaming, fail to address latency issues fundamentally, often only adding new channels without optimizing overall performance. This thesis proposes a novel approach using a 3D engine (e.g., Unity 3D) to stream multi-input video data through a single channel with reduced latency. By leveraging 3D engine capabilities, such as World/Screen Space Cameras, 3D Canvases, and Webcam Textures, the proposed system consolidates video streams from multiple external cameras into a unified, low-latency output. The affiliated project of this thesis demonstrates the implementation of a low-latency multi-channel live video streaming system. It employs the RTSP protocol and examines video encoding techniques, alongside a client-side application based on Unity 3D. The system architecture includes a WebSocket server for persistent connections, an HTTP server for communication, a MySQL database for storage, Redis for caching, and Nginx for load balancing. Each module operates independently, ensuring flexibility and scalability in the system's design. A key innovation of this system is its use of a 3D scene to map multiple video inputs onto a virtual canvas, recorded by an in-engine camera for transmission. This design minimizes redundant data, enabling an efficient and director-guided live streaming network. The thesis concludes by discussing challenges encountered during the project and provides solutions for future improvement.