Abstract
Industrial video surveillance systems play a pivotal role in smart industry prioritizing safety protection. Object detection resorting to deep neural networks (DNNs) is promising in achieving accurate and autonomous localization and identification of anomalies in video frames, supporting broad intelligent video surveillance applications. However, existing approaches are either computation- or communication-intensive. Limited by the constrained resources of industrial systems, they usually suffer from a high end-to-end (E2E) latency, cannot be directly applied to latency-sensitive applications. In this article, we present a lightweight edge-cloud collaborative branchy DNN, CombiNet, and customize an intelligent edge device, Edge-Vbox, to construct an effective real-time video object detection (VOD) solution. In our case study of intelligent smart grid substation operation and maintenance, experimental results using real-world data demonstrate that our approach significantly outperforms state-ofthe-art methods in E2E latency, and manages to achieve real-time VOD with negligible accuracy loss. Source code is available at https://github.com/iotdata-grid/CombiNet.
Funding Information
  • National Key Research and Development Program of China (2022YFA1004100, 2021YFB2401300, 2020YFA0713900)
  • National Natural Science Foundation of China (62172329, U1811461, U21A6005, 11690011)

This publication has 15 references indexed in Scilit: