Machine vision is enormously important to industry and new vision technologies are certain to cause a manufacturing revolution.
Machine vision is a fascinating field with constant innovation. New technology will generate novel techniques to utilise advanced cameras, frame grabbers, lenses, lights, and lighting controllers. Embedded systems and Industry 4.0 will become commonplace and AI techniques will dramatically enhance the boundaries of possibility.
Industry worldwide has been severely impacted by the Covid-19 pandemic during 2020. As relaxed government restrictions allows industry to rebuild, there may be enduring reluctance to employ large workforces. This will accelerate an existing trend towards automation and bring added flexibility to producers and processors.
Hyperspectral, 3D and Computational
Hyperspectral imaging, 3D techniques and computational imaging will soon become mainstream, driven by huge increases in processing power. New 3D techniques, such as laser line triangulation, stereo vision, time of flight and structured light are already generating massive interest, particularly in robot-guided applications such as pick and place,
Processing power continues to increase and the low-power processors with on-board image processing will bring embedded vision technology to more applications, such as photometric stereo. Computational and multi-shot imaging, where a sequence of images is combined into a composite image, will increasingly be used to create surface topography, glare reduction, ultra-resolution color, and extend depth of field.
Filters, Lights and Lighting Control
New sensors featuring inbuilt polarisation filters have created a generation of polarisation cameras, making it cheaper to identify surface defects, stress, and birefringence within transparent objects. Lights have seen recent innovation such as the new, thin and flexible OLED panels. Lighting controllers have become capable of driving very low or very high power in applications where backlighting is bright, and for high frequency, high power LED pulsing.
Protocols, Interfaces and AI
The Precision Time Protocol (PTP) enables exact synchronisation of Ethernet devices. PTP with GigE Vision 2.0 standard can be used to trigger components such as lights, lighting controllers, lenses, and shutters with microsecond precision. The recently-announced USB 4 interface promises even faster transfer speeds, better management of video and compatibility with Thunderbolt 3.
There is no doubt that AI will be at the forefront of future developments. Deep learning and machine learning are already gaining traction, especially in applications such as the classification of foodstuffs. Massive advances in parallel processing and huge training data sets will make deep learning easier to implement and the recent emergence of ‘inference cameras’, where trained neural networks are implemented directly on a camera processor, may bring AI to mass applications.
To read more about the future of machine vision, download the free Gardasoft paper at www.gardasoft.com/future-of-machine-vision.
