Vision systems: Processing at the edge, in the cloud, or both?

Computer vision technology enables computers to derive actionable results from digital images and video. By allowing computers to "see" and understand images, computer vision is being used in a wide range of applications, from the visual inspection of peanuts on a conveyor belt to the camera systems in autonomous vehicles.


Traditionally, image processing has been done in the camera itself or a nearby embedded computer, but the emergence of cloud computing has made it possible to perform image processing on a much larger scale. However, for real-time applications, the latency of sending data to the cloud and back can be too long, making it difficult to use cloud-based solutions. Additionally, there are security concerns with transmitting data over a public network.


To solve these problems, some companies have begun processing images and video at the edge, where the data is generated. This reduces latency to near zero, maintains critical data at its source, and eliminates the need for recurring cloud-based costs. Additionally, processing images and videos at the edge eliminates the need to transmit data over a public network, which reduces security risks.


To enable processing at the edge, companies are integrating high-performance computing devices into cameras and other devices. These devices, which include CPUs, GPUs, and AI processors, can perform a wide range of image processing and machine learning tasks directly on the camera. Even small, low-cost devices can now perform machine-learning tasks with a high degree of accuracy.


Designers have several options for integrating computer vision technology into their systems, including integrating hardware and application software, training computer vision models, and integrating existing models and computer vision libraries into the application. To make this process easier, some suppliers are integrating compute units so designers can use a single, unified programming interface for all types of processors.


Overall, the ability to process images and video at the edge is becoming increasingly important for real-time applications such as robotics and ADAS, as well as for maintaining data security. With the development of high-performance computing devices and easy-to-use programming interfaces, it is becoming easier to incorporate computer vision into a wide range of systems.

Comments

Post a Comment

Popular posts from this blog

Are microLEDs the next big (small) thing?

Light-emitting diode (LED) technology has been around for decades, but the introduction of micro LEDs has changed the way we think about it. As the name suggests, micro LEDs are tiny versions of the regular LEDs engineers use in embedded systems. The main application areas for micro LEDs are displays of various shapes and sizes, from very small to very big. MicroLEDs are defined as LEDs smaller than 100 micrometers (um) in size. This site is significantly smaller than traditional LEDs, and it opens up new possibilities for applications. The main advantage of micro LEDs is their increased brightness and higher power efficiency compared to OLED or LCD-based displays. This makes them ideal for microdisplays, which are defined as very small displays used for near-eye applications (i.e. not direct view). These displays are usually under 1-inch in size, with high performance (in terms of resolution, brightness, and refresh rates). The manufacturing process of micro LEDs is challenging due to...
Read more

Safety and security in autonomous driving

The integration of autonomous technology in vehicles has brought about a new set of concerns, particularly when it comes to safety and security. As the industry moves towards higher levels of automation, it becomes increasingly important to ensure that cybersecurity measures are in place to protect both the vehicles and their passengers. One key standard that is being used in the automotive industry is ISO/SAE 21434, which provides a structured process for incorporating cybersecurity considerations into automotive products. This standard requires manufacturers and suppliers to demonstrate due diligence in implementing cybersecurity engineering and ensures that cybersecurity management is applied throughout the supply chain. The development of automated driving features also brings new challenges to safety. Historically, automotive electrical and electronic systems have been regulated according to the functional safety standard ISO 26262, which assures a high level of resilience in th...
Read more

Six cool and update autonomous technologies

 EVAA prevents plane crashes with its moral compass: NASA’s Resilient Autonomy team and the Federal Aviation Administration have developed the Expandable Variable Autonomy Architecture (EVAA) to avoid aviation accidents by giving aircraft autonomous capabilities. It can assist human pilots or act on its own and has already saved 11 lives according to Mark Skoog, principal investigator for autonomy research at NASA's Armstrong Flight Research Center. Autonomous tractors increase farm productivity: Autonomous farming can increase the efficiency of large industrial farms, by using self-operating tractors and other machinery, such as John Deere's Autonomous Tractor Concept, which uses GPS and sensor technology to navigate and perform tasks on the farm. Self-driving tractor-trailers fight driver fatigue: Self-driving tractor-trailers can help reduce the risk of accidents caused by driver fatigue, by using sensors and cameras to navigate the roads, and advanced algorithms to make dec...
Read more