Sponsored

Computer-on-Modules for AI-Based Industrial Vision Systems

Flexibility is Key

Vision and artificial intelligence (AI) are megatrends in automation, Industry 4.0 and cyber-virtual factories. Embedded computer technologies for such systems have to meet increasingly complex demands. What’s good enough today may be too little tomorrow – in terms of processor performance, cores, virtual machines or the number of onboard interfaces. With their high scalability, Computer-on-Modules can provide the necessary flexibility and, as convergent components, are even suitable for closed-loop engineering.

That AI based vision is an industrial market with a dynamic future becomes obvious when looking at a megatrend in a sector that’s gaining considerable ground commercially: the camera technology market for autonomous robotic vehicles. At 140%, it is growing significantly stronger than the market for autonomous vehicles at around 40%. It is therefore safe to assume that more than three times as many cameras will be installed per vehicle than before. The number will be even higher if the costs per camera fall. However, what won’t decrease is the amount of image data to be processed. On the contrary, it will increase significantly as more cameras and higher resolutions also promise safer situational awareness. Besides subsystems for data preprocessing, embedded computing cores and their GPGPUs are being used more and more for image recognition. This trend also affects industrial autonomous vehicles, as well as collaborative or cooperative stationary robotics and all other industrial vision systems.

Figure 1 — The market for autonomous robotic vehicles is growing rapidly and will also have an impact on industrial vision systems because of its embedded artificial intelligence.

Scalable Performance for a Dynamic Environment
Computer-on-Modules are a perfect basis for scaling the embedded computing core. They are standardized, come in various form factors such as COM Express, SMARC 2.0 and Qseven, and can support an extremely broad range of processors in different performance classes. COM Express leads the current high-end class of modules, which are offered with Intel Core and Xeon processors as well as AMD Ryzen and EPYC processors, and are scalable to the AMD G-Series or entry-level systems such as Intel Pentium, Celeron and Atom. SMARC 2.0 and Qseven cover the lower range of low-power embedded computing, with significantly smaller dimensions. Current configurations include Intel Atom and Celeron processors as well as the AMD G-Series up to the latest NXP i.MX 8 processors, which are available with as little as 3W TDP in normal operation. This class, in particular, was also developed for embedded computing in vehicles.

Figure 2 — Two MIPI-CSI interfaces, which are natively supported by the processor, can be implemented either with Intel Atom or NXP i.MX 8 based modules.

Spoiled For Choice
congatec , for example, offers two brand new processors from the NXP i.MX 8 series on SMARC 2.0 and Qseven modules. While both clearly target the automotive sector, they are also great for a variety of industrial automation applications. The NXP i.MX 8 QuadMax natively supports 2 MIPI-CSI interfaces and, thanks to OpenVX (vision), also provides ideal conditions for vison applications, which can also use the graphics core for parallel processing thanks to Open CL support. The new i.MX 8X, on the other hand, targets particularly energy-efficient systems with a somewhat reduced feature set. Both versions are provided on SMARC 2.0 modules.

Intel Atom, Celeron and Pentium processors are available on SMARC 2.0 as well. They can also be prepared for direct access to MIPI-CSI interfaces, as congatec’s first MIPI-CSI 2 Smart Camera Kit for vision systems at the edge of the IIoT shows. It is an application-ready kit for the evaluation and deployment of MIPI-CSI 2 based rugged smart camera analytics in harsh industrial, outdoor and automotive environments. Developers benefit from an instantly deployable, smart MIPI-CSI platform in an industrial grade design. Built with commercial, off-the-shelf available components, the new kit simplifies the development and shortens the time-to-market of smart camera analytics solutions for IIoT end devices. With SMARC 2.0, developers can therefore easily test which processor platform is better for them.

Comprehensive Services Accelerate Design-In and Reduce Costs
However, all these Computer-on-Modules not only convince with their application-ready design. They are also complemented by numerous congatec add-on services that reduce the complexity of the integration while shortening the design-in time for the fastest time-to-market. The main pillars of congatec’s premium service include personal design-in support for each OEM implementation, and individually selectable next-level support from the Technical Solutions Center. This team of specialists covers all customer-specific demands – from requirement engineering support and boot loader configuration with extended operating system support to test, validation and debugging services. High quality and personal support to simplify the use of embedded computer technologies are further congatec service characteristics. Customers benefit from fast and efficient product design-in because ‘plug & play’ is more efficient and cost effective than ‘trial & error’.

Figure 3 — Computer-on-Modules such as SMARC and Qseven are perfect for closed-loop engineering based on convergent computing cores because of their high, architecture-independent scalability.



Don't miss out on upcoming issues of Circuit Cellar. Subscribe today!

 
 
Note: We’ve made the October 2017 issue of Circuit Cellar available as a free sample issue. In it, you’ll find a rich variety of the kinds of articles and information that exemplify a typical issue of the current magazine.


Would you like to write for Circuit Cellar? We are always accepting articles/posts from the technical community. Get in touch with us and let's discuss your ideas.