Image Processing in the IoT Simcom Moves AI Image Analysis to Compact Wireless Modules

Related Vendors

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

If you want to analyze images in cameras or machines, you need computing power directly on the device. The component manufacturer Simcom bundles processors and wireless technology on tiny circuit boards for this purpose. The company is thus responding to the trend towards edge computing in industry.

Chinese manufacturer Simcom is launching two new components on the market that are specifically aimed at developers of industrial image processing applications. The SIM8666 and SIM8668 models presented at the Embedded World trade fair integrate an AI accelerator as well as Wi-Fi and Bluetooth interfaces in addition to computing cores and memories. The aim of the architecture is to perform complex calculations directly in the device instead of sending large amounts of data to central servers via mobile networks.

At their core, these modules resemble the architecture of a modern smartphone. They combine the control of cameras and displays with local computing power and communication interfaces. The circuit boards measure 1.69 × 1.73 inches and are just 0.13 inches thick. Inside are main processors from the manufacturer Rockchip, a graphics chip from Arm and a so-called Neural Processing Unit (NPU). The latter delivers a computing power of one tera operation per second (1 TOPS) and allows models from software frameworks such as Tensorflow or Pytorch to be executed directly on the component.

Local Evaluation Saves Bandwidth And Server Costs

The combination of local data processing and connectivity opens up new technical applications. In practice, this applies, for example, to access control via facial recognition, optical monitoring of production lines in factories or use in medical analysis devices. Instead of transmitting a high-resolution image via a network, the algorithm on the chip analyzes the pattern. The hardware processes the information in fractions of a second and only transmits the result—such as confirmation of access authorization.

To ensure the physical connection to cameras and screens, the manufacturer equips the boards with interfaces such as PCIe, USB, HDMI and various connections for camera sensors. Mads Fischer, Sales Director for Europe at Simcom, summarizes the intended use: "Our new intelligent AI modules are aimed at new IoT applications in image processing that require AI functions for positioning or identification." While developers have often relied on Android as the operating system for previous smart modules, Simcom is now seeing a clear shift towards Linux in the industrial environment.

Storage Prices Slow Down, Established Mobile Communications Dominate

Although the industry is currently investing heavily in artificial intelligence, the boom also poses economic challenges for module manufacturers and their customers. The high global demand for memory and semiconductors for large data centers is driving up component prices. As the required memory chips come from the same semiconductor factories, the compact industrial modules are also becoming more expensive. According to Fischer, components that used to cost 25 US dollars are now reaching prices of up to 45 dollars in some cases.

When the locally processed data is ultimately transmitted wirelessly, system architects currently rarely rely on the latest wireless standards. While 5G is growing in the area of broadband communication, it still plays a subordinate role for the IoT.

Instead, the established LTE standard continues to dominate, particularly in the Cat 1-Bis version. This technology offers manufacturers a good balance between module costs, low energy consumption and sufficient bandwidth for IoT devices. Simcom assumes that LTE will continue to shape the industrial IoT for well over the next decade, as upcoming standards such as 6G for pure machine communication do not offer any significant technical advantages for the time being. (mc)

Advantages of AI for analyzing the design

Benefit from AI in Tessent products from Siemens EDA