Are machine vision providers facing difficult times?
Keppler: Machine vision is a key technology for the implementation of Industry 4.0 and the digital transformation. The implementation of the OPC UA standard in particular is bringing image processing and automation technology even closer together. In this respect, it is only a logical step for automation providers to become more involved with image processing. This does not mean that image processing specialists are facing difficult times. On the contrary - it will be even more important in future for users to be able to rely on advice from competent partners for this key technology in the field of Industry 4.0. Stemmer Imaging has been focusing on industrial image processing for 30 years and has played a decisive role in shaping the development of the industry.

Where do you see the limits for fully integrated systems?
Keppler: This question has been around since the introduction of the first intelligent cameras. At that time, however, a clear distinction could still be made between freely programmable, powerful PC multi-camera systems and parameterizable, compact smart cameras. However, the possibilities offered by current embedded architectures based on ARM platforms and Linux OS are increasingly blurring the boundaries. Nowadays, it is therefore no longer expedient to talk about the theoretical limits of individual systems. It is much more important to focus on the customer's application when designing the system. This is where the flexibility of the suppliers is required to find the best combination for the customer.

Until now, image processing has tended to be something for specialists and experts. With 'embedded vision', the boundaries are also softening here - future vision systems should be designed so simply that even vision amateurs can operate them. Where do you position yourself in this environment?
Keppler: For the simplest applications, these vision sensors can of course be designed in such a way that they can be operated by 'vision laymen'. However, not all tasks can be solved so easily - and I don't think it is expedient to break down a future technology to just this single aspect. In my opinion, it is much more important that, on the one hand, the usability of the systems is tailored precisely to the target user and, on the other hand, that the users are trained in image processing in general and the systems specifically. Regardless of which side you want to approach image processing from, manufacturer-independent advice on design will remain the key to success in the future.

To what extent will the topics of machine learning and artificial intelligence influence industrial image processing?
Keppler: Machine learning has been used in industrial image processing for many years with great success for a wide range of tasks. It is therefore a proven technology with thousands of successful installations in industrial environments. The digital transformation is currently leading to a dramatic expansion of the field of application for machine vision into completely new markets. Examples include the agricultural and food industries, safety and traffic technology - such as 'autonomous driving' - and numerous other applications in non-industrial environments - such as 'digital signage' - which often involve the reliable detection and classification of objects with a high degree of variability. While traditional image processing algorithms quickly reach their technological limits for these tasks, machine learning and, in the next step, artificial intelligence offer excellent prospects.