Industrial image processing, or machine vision, is a multifaceted field of application for image processing and for the use of vision sensors. Machine Vision is used in production engineering and quality control, pattern recognition, enterprise resource planning, robotics, semiconductor inspection, logistics, monitoring, warehousing and many other applications. The technology is characterized by application-specific solutions, which often have a proprietary character.
Machine vision is based on capturing and analyzing photos and videos. It involves taking advantage of computers in image recognition, evaluation and analysis. Modern industrial image processing works with artificial intelligence and uses Deep Learning and Convolutional Neural Networks( CNN). The application areas include any kind of image recognition in industrial and medical applications. For example, character recognition, optical character recognition, pattern recognition and mark recognition, and furthermore the technical aids used in production techniques for position recognition, layer thickness measurement or surface inspection.
Vision systems consist of a camera, a frame grabber and a computer. In newer approaches, the frame grabber can be omitted because they are no longer needed with standard PC interfaces and specialized camera interfaces such as GigE Vision. The GigE Vision standard, which uses Gigabit Ethernet, offers some advantages because of the standardized cabling, the large distances that can be bridged and relatively high data rates, and the possibility of connecting multiple cameras.
In addition to GigE Vision, there is also CameraLink, CoaXPress and Thunderbolt for industrial image processing. In addition, the PC interfaces IEEE 1394 in the B version and USB3 Vision have also secured their place among the camera interfaces. CameraLink has established itself for real-time requirements with the shortest latency times and response times in the microsecond range.