17-05-2022 | Review Display | New Technologies
Review Display Systems examines the demands and needs of embedded computing in a machine vision application with Peter Marchant, embedded division manager.
Over many industries, the usage of embedded computing technology is evolving to be more wide-ranging. Compared with conventional desktop PC hardware, the latest embedded computing systems are powerful, energy-efficient, mechanically compact, and offer an effective, inexpensive computing solution.
Intel, AMD, and ARM processors are widely used across many embedded computing platforms. The key factors determining an embedded computer's suitability for a real-world application are processing performance, memory bandwidth, expansion capability, and energy efficiency.
Migrating from PC hardware to an SBC, COM, or SOM will invariably create a smaller, more power-efficient, cost-effective embedded system. Extra memory, add-on peripherals, and expansion boards can enhance performance, functionality, and connectivity.
Machine vision is one of the fastest-growing application sectors across a diverse range of manufacturing, control and service industries that is assisting to improve productivity, develop functionality and enhance reliability.
Machine vision systems employ high-performance computing and image technology together with software algorithms to provide automatic image-based recognition, inspection, and analysis for process control, factory automation, food processing, access control, robot guidance, quality assurance, and security systems name, to but a few.
An application where machine vision is actively used is in-vehicle monitoring systems where ANPR systems are employed to control vehicle access, provide traffic control, enable site security, or monitor complex multi-lane, national motorways.
To perform intelligent vehicle tracking across an image sensor's full range of view, tracking systems must be capable of high-speed number plate recognition in fast-moving, high-density traffic environments. Image processing algorithms need to locate and recognise multiple number plates in the image sensors' field of vision, read, acquire, and process image data in real-time with a high degree of accuracy and repeatability.
Machine vision-based vehicle monitoring systems must be able to react, respond and execute image recognition algorithms with speed and precision to determine: 1. a vehicle is identified, 2. identification of a number plate, 3. character and symbol recognition, and 4. perform follow-up data analysis with the image data obtained.
A typical machine vision system for a vehicle monitoring application will need the integration of a camera/image sensor, high-performance embedded computing hardware, and appropriate power management.
Image sensors can now support multiple resolutions and enhanced frame rates while still offering excellent image quality and uniformity. Sensors are now providing resolutions of 8K (8190 x 5460 pixels) with frame rates of 60fps and above.
Of the many embedded SBCs now available, the Pico-ITX board offers an ideal hardware solution for machine vision applications in quality control, security, and transportation.
The most recent Pico-ITX boards now provide the latest 11th Generation (Tiger Lake) Intel Core i3, i5, i7, Celeron SoC processors. These processors are now manufactured utilising a 10nm process, resulting in a 15-20% performance improvement over the 14nm process of previous 9th generation processors.
The Pico-ITX provides many options for diverse levels of processor performance together with its strong expansion capability. It offers a highly capable but no less flexible and versatile embedded computing platform. The processor of choice and power efficiency can be determined by the image capture performance and data throughput needed, although naturally, this will vary across industry applications.
Several expansion modules are available that operate as dedicated hardware accelerators supporting parallel computing. The Intel Movidious VPU, for example, can be added when serial computing performance is inadequate for processor-intensive vision-based applications.
The Pico-ITX board also offers peripheral functionality to support onboard memory and SATAIII storage. The provision of dual LAN ports, USB interface connections, serial port headers, general-purpose I/O, I²C and SMBUS serial communications interfaces can help integration in a stand-alone embedded machine vision system. Support for external displays with resolutions of up to 4K is catered for with Embedded DisplayPort (eDP) and HDMI 2.0 outputs.
Although other compact SBC form factors exist, such as the 1.8" Femto-ITX, which offers dimensions of just 55mm x 84mm, this format has not yet seen widespread adoption as an industry standard. With external mechanical dimensions of 72mm x 100mm, the Pico-ITX form factor is currently the smallest, mainstream industry-standard SBC which can be readily and easily used in remote, confined spaces.
The continuing development of embedded computing and imaging technology leads to the increased use and installation of machine vision systems over various industries and applications. With the correct implementation and integration of embedded computing hardware, enhanced productivity and the delivery of fast, effective data analytics can be successfully realised.