Emerging technologies in machine vision
UKIVA’s Machine Vision Conference and Exhibition (www.machinevisionconference.co.uk) to be held on Wednesday 16th May 2018 at Arena MK, Milton Keynes UK
The conference and exhibition will provide visitors with the opportunity to find out about emerging machine vision technologies. The comprehensive program of technical seminars includes a presentation theatre dedicated to innovations in vision. Three of the most topical issues in the industry at the moment are embedded vision, Industry 4.0 and the IoT and deep learning.
The rapid evolution of computing power in embedded, single board computer systems is providing new, exciting possibilities for vision. Embedded vision systems are the newest variants of intelligent vision and are finding increasing use in applications where space is constrained, cost is an issue and a self-contained vision solution is required. Embedded vision is also an obvious platform for large volume solutions where economy of scale can have a real impact. Embedded vision offers great potential but with a wide variety of hardware and operating systems available careful consideration needs to be given to choosing the optimum platform for the application.
The Internet of Things (IoT) and Industry 4.0
Machine vision is already used extensively industrial quality control applications and can be directly linked into process control systems to indicate when actual measurements are nearing the allowed tolerances. Standardisation of communication is an intrinsic requirement for the integration of machine vision into the Industrial IoT-based smart factory initiative. The Open Platform Communications Unified Architecture (OPC UA) is a new interface standard for future Industry 4.0 projects. This will provide the connectivity between different layers of the automated process and machine vision systems to allow abstract results and raw data to be transmitted to a PLC, MES or ERP. These, in turn, may then be used to configure and control machine vision systems.
Deep learning is a type of machine learning in which a model learns to perform classification tasks directly from images. Deep learning is beginning to gain some real traction in machine vision by virtue of the fact that it is becoming much more widely available through commercial machine vision products. It is used in applications where it is difficult to predict the full range of image variations that might be encountered using conventional algorithms. Deep learning utilises a special kind of neural network called a convolutional network (CNN) which is taught how to categorise images by being shown a large set of example images and learning to accommodate the variations. The term ‘deep’ refers to the number of layers in the network – the more layers, the deeper the network. Traditional neural networks contain only 2 or 3 layers, while deep networks can have hundreds. Now deep learning is finding its way into the main stream of machine vision with applications in text and code reading, and classification tasks in the inspection of industrial goods or the recognition of components. This technology is now being included in industry-standard image processing libraries, making it much more readily available to end users.