Design and Implementation of Video Analytics System Based on Edge Computing

Abstract

Real-Time video analytics, whose applications range from safety, public security to smart cities, is a typical use case of Internet of Things (IoT). However, uploading the video stream to the cloud for analytics cannot meet the requirements of low latency and efficient bandwidth usage. Edge video analytics, which uploads the stream at the edge node, is a key to solve the abovementioned problem. This paper proposes an intelligent video analytics system on edge computing platform. Combining the edge computing and video analytics, this system can analyze the video stream by face recognition, indoor positioning, and semantic analytics in real time and archive the videos automatically. Specifically, applied in conference room, the video analytics system analyzes the conference room scenario and files the conference videos, which reduces the cost of manual recording and promotes the data sharing. The implementation results prove that our system can operate smoothly on the edge computing platform to provide real-time and efficient video analytics services.