bg line bg line

HiSilicon Builds 5G Machine Vision Solution, with Industry Partners

[ShangHai, China, April 3, 2020]At a time when government measures have effectively curbed the spread of COVID-19 in China, the emergence of 5G has also helped with the construction of essential new infrastructure. Smart manufacturing is a core application for 5G infrastructure, in which 5G industrial modules breathe new life into traditional manufacturing. The 5G machine vision solution developed by HiSilicon (Shanghai), China Telecom, and Changhong AI-Link recently passed the E2E verification for application in vehicle body quality inspection, which represents an important milestone for 5G in smart manufacturing.

Traditional machine vision solutions utilize industrial high-speed cameras to collect images and transmit them to the application server in wired mode for image recognition, to support such industrial applications as quality inspection. These solutions face issues related to wired deployment and scalability of single-node application deployment. Furthermore, requirements for high bandwidth and stable transmission limit the application of wireless technologies, such as 4G. The co-developed "device-edge-cloud synergy" solution expands the scope of 5G machine vision across a range of new smart manufacturing scenarios.


On the device side, differing from the traditional CPE relay mode, industrial cameras are directly integrated with 5G industrial modules and deployed as packed 5G industrial cameras for commercial use. The new solution also applies super uplink technology in 5G industrial modules for the first time to match the 5G network, to achieve a next level of performance, featuring high-speed, low-latency, and stable transmission of high-resolution images, and demonstrating the commercial feasibility of 5G industrial cameras.


On the application side, China Telecom has deployed the 5G multi-access edge computing (MEC) connected to the production system in the equipment room of the manufacturing area. The machine vision application deployed in the MEC enables the edge AI model to detect quality in real time. This ensures better scalability of single-node deployment, data security and real-time performance. China Telecom has utilized edge-cloud synergy to send desensitized images to the e-Cloud AI training platform via the isolation security mechanism, harnessing a rich array of GPU and CPU hybrid computing capabilities to implement forward iteration of the detection AI model, for improved detection efficiency and quality across the board.

This innovative 5G machine vision solution can be widely applied in such fields as vehicles, TVs, and consumer electronics, and serves as a powerful tool in the 5G-driven upgrade of traditional industry. It has passed E2E verification under lab tests conducted by FAW Group Corporation—a Chinese automotive manufacturing company, and set to be implemented on a widespread scale.

HiSilicon, China Telecom, and Changhong AI-Link will continue to develop new 5G industrial modules for smart manufacturing, with new around the corner in 5G+AR remote guidance, 5G+automated control, 5G+AGV smart warehousing, 5G+device management, and 5G+information security management.