The growing number of "intelligent" devices and machines in industry and the consumer sector places new demands on the embedding of image processing functionalities and systems. Whether self-driving cars, intelligent vacuum cleaners or interacting robots as colleagues in industrial production - only fully integrated vision systems as a native component of these new applications make them useful.

Embedded vision enables visual intelligence for machines and thus teaches them to see and think. Embedded vision describes the combination of image processing within a device with digital processing and intelligent algorithms to evaluate the image and video data obtained. This enables the "seeing and thinking" machine to react to events and processes and interact with the environment. Integrated image processing is the basis for safe human-machine collaboration and the networked work of robots and machines in automated Industry 4.0.

Advantages of embedded vision platforms
Progressive miniaturization combined with increasing performance has been and continues to be important for the spread of fully integrated systems. Embedded vision leads to reduced space requirements, less weight and therefore lower energy consumption and lower unit costs. This means that a drone can also see without significantly increasing in size or weight - which is hugely important for energy consumption, maneuverability and range. The dynamics of a robot are also not affected by the lightweight and compact systems. Intelligent "recognizing" applications can thus be built into stationary as well as mobile and even portable devices.

Embedded vision modules are very compact and energy-saving. © Framos

Dr. Christopher Scheubel, Framos IP & Business Development, says: "Miniaturization makes it possible to replace external machine vision systems with integrated embedded vision. This development will continue to accelerate over the next few years, for example through new technologies such as the Intel RealSense Suite. In a few years, almost every device will be able to see and think through embedded vision systems."

As an integrated technology in devices and machines, embedded vision ensures that all industries and every type of application can benefit from image processing. Seeing robots or self-positioning lasers are industrial examples. In the consumer sector, self-driving cars and many other applications give an idea of the great potential of embedded vision.

Part of the overall system
Traditional image processing systems consisting of an external camera, a stand-alone PC and accessories often required a lot of space and were usually expensive. They were developed as proprietary systems, which made it very complicated to integrate them into an application. Embedded vision systems, on the other hand, are integrated directly into the device or machine, and the development is therefore interlinked with the device as part of the overall project right from the start. Embedded vision systems can be designed for a specific application, which leads to a significant reduction in costs per system for large quantities.

The heart of an embedded vision system is the intelligent camera. © Framos

Embedded vision systems essentially consist of three components: a miniature camera without a housing, a processing unit and a space-saving interface. The systems can be designed completely flexibly. The heart of the system is the intelligent camera, which can be a simple vision sensor or a high-end camera. Camera and lighting settings are therefore just as much a part of the overall system as the parameterization and programming of vision algorithms.

Embedded vision for Industry 4.0 and IoT
Due to the properties described above and the wide variety of designs, embedded vision systems are the basis for the intelligent automation of Industry 4.0 and modern manufacturing. The potential of seeing, thinking and therefore interactive machines is fully realized in Internet-of-Things environments and networked systems. The integration of different vision systems with different video and vision inputs and outputs as well as different imaging pipelines is a challenge for the new embedded architectures and image processing-based analyses in real time. Triggers and control pulses can be controlled synchronously with the overall system, even in dynamic systems under changing conditions. This is precisely the advantage of embedded vision and the opportunity to make new applications see and think with the help of image processing. Manufacturers need consistently precise results for quality control, robot control and more efficient production. The ever deeper integration of embedded vision and its further development are helping to ensure successful implementation.